The potential subject area is large and the focus here is on the chronology of air pollution by human activity, identifying the main issues, their causes and the regional and global trends. Other papers in this volume, to which links are made, provide the wider context, the policies developed to address the problems and the possible futures. Show There are four rather different sources of evidence to provide the narrative for this account. These include written documents including early legislation, direct measurements of atmospheric composition, chemistry transport models, which simulate atmospheric composition changes from a knowledge of emissions, meteorology and chemical processing of pollutant gases, and, finally, remote sensing of the atmosphere from aircraft and space. The early documents are fascinating and provide hints at the underlying chemistry, but are entirely lacking in quantitative detail. Legal documents indicate the intent, but, for reasons elaborated later, did not significantly constrain the developing global issues until the later decades of the twentieth century. High-quality measurements of air pollutants are restricted to the last 150 years and numerical modelling to the last 40 years, leaving considerable scope for speculation on the early trends. There is necessarily some subjectivity in the selection of information sources used to describe the air pollution chronology outlined here, as summarized in table 1. For this purpose, we have focused on what we consider to represent major milestones based on: (i) the recognition of key aspects of air pollution, (ii) of quantitative evidence and (iii) of major points of changes in air pollution levels. Other perspectives on the topic have been provided by Colbeck [5] and Mosley [6]. The main air pollutants of interest examined here are sulfur dioxide (SO2), nitrogen oxides (NOx), ammonia (NH3), volatile organic compounds (VOCs), primary particulate matter (PM), and their reaction products, including fine particulate matter (PM2.5) and tropospheric ozone (O3).
Clear written evidence shows that early society recognized a threat to human health and the wider environment from air pollution. However, the identity of the gases and particles remained largely unknown, and there were no measurements to quantify the problem. Early attempts to regulate emissions show that the lawyers in their day clearly had the means to articulate societal desire for a cleaner environment, but the laws developed were not supported by the infrastructure necessary to make them effective. The lack of consistent language describing the underlying science also makes the early literature difficult to interpret from a twenty-first century perspective. The early history takes us to the period of elucidation of the compounds present in the atmosphere and to early measurements, mainly in the seventeenth to nineteenth centuries, following which direct measurements began in earnest. Sporadic measurements of air quality began in the late nineteenth century, especially by Robert Angus Smith in the UK [7], the first scientist to attempt multisite, multipollutant investigations of the chemical climatology of the polluted atmosphere. The early developments in understanding of air pollution were mainly by chemists, who continued their leadership of the mechanistic underpinning of the science through the twentieth century (e.g. [8,9]). Distributed sites to measure atmospheric composition gradually developed through the mid-twentieth century and by the time acid rain became a focus of scientific and political interest in the late 1960s there were networks in Europe and North America to study the composition of air and precipitation at regional scales (e.g. [10,11]). In addition, local pollution problems in industrial cities, mainly in Europe and North America, and around notable point sources, provided early measurements of large local effects by some of the main pollutants. The ground-based monitoring networks in place by the year 2000 (table 1) included regional and global air chemistry measurements. The third main source of time-series data to assess the chronology of air pollution is the application of chemistry transport models (CTMs) with global meteorological models and spatially disaggregated inventories of pollutant emissions. The final source of data is that provided by satellite remote sensing, which has developed over the last three decades, providing global concentration fields for the major air pollutant gases (SO2, NO2, NH3, CO, and O3). These complementary sources are used here to provide a summary of the development of specific air pollution issues through the late nineteenth and early twentieth centuries and in the last two decades revealing some important signs of recovery from effects of air pollution in Europe, North America and East Asia. Early humans would have been aware of at least some of the potential hazards in the air they breathed from their general discomfort in the presence of smoke and combustion gases close to open fires. The need for shelter and warmth led to fires inside shelters, and in confined structures, the exposure to potentially toxic gases and particles is considerably enhanced. Given the directly noxious properties of many combustion products (smell, and lachrymose and respiratory effects), it is surprising that so many societies had dwellings with open fires and no chimneys. The development of the chimney itself can be seen as a key milestone for indoor air quality, adopted at first in the largest houses from the twelfth century [12]. Today, indoor air pollution is an important contributor to effects on human health. All subsequent analysis here, however, is devoted to the outdoor environment. Evidence from Greece shows that the problems of polluted air outdoors were being documented at least 2400 years ago. The book Airs, waters and places attributed to Hippocrates (ca 400 BC) suggested all sorts of illness as being related to the quality of air. The worst it seems was in cities facing damp westerly winds, where the inhabitants ‘are likely to have deep, hoarse voices, because of the atmosphere, since it is usually impure and unhealthy in such places' ([13], p. 83). Writers a little later from Imperial Rome understood the probable health impacts of smoke with Seneca (ca AD 63–65) referring to the problem and Frontinus (ca AD 96) proudly declaring how his contribution to aqueducts and fountains has helped make the air purer: ‘the causes of the unwholesome atmosphere, which gave the air of the City so bad a name with the ancients, are now removed’ ([14], p. 417). As Seneca recorded of a health break from Rome: As soon as I escaped from the oppressive atmosphere of the city, and from that awful odour of reeking kitchens which, when in use, pour forth a ruinous mess of steam and soot, I perceived at once that my health was mending… So I am my old self again, feeling now no wavering languor in my system, and no sluggishness in my brain ([15], p. 193). It is notable that the reference to the brain matches an effect of ammonium-containing air pollution from naturally burning coal caves along the Silk Road in Central Asia as later recorded by the Arab geographer Al-Mas'ūdī [16]. A book by Shen Kuo (1031–1095) written during the Song Dynasty (AD 961–1279) provides further evidence of concern in China about air pollution from coal burning [17]. Other post-classical writers, especially in the Arab world, contributed observations about air pollution during the ‘Dark Ages’ when considerable learning was being lost in Europe [18]. Ultimately, however, little changed throughout the Middle Ages in the understanding of the causes of disease and possible role of air pollutants reflecting the persistence of the classical miasmatic concept that odours and other matter in air were the controlling influences for human health [19], an idea going back to the time of Hippocrates. In the seventeenth century, John Evelyn published Fumifugium or The Inconvenience of the Aer and Smoak of London [20] (figure 1). This iconic document described air pollution in London and suggested ways of reducing the scale of the problem. He proposed moving industries including brewing and lime-burning to the countryside, well outside the city. John Graunt, a contemporary of Evelyn, suggested a correlation between rates of mortality and pollution, especially in fog episodes [21]. In the absence of any chemical data, or indeed any numerical values to quantify the pollutants present, we have only the narrative, but it clearly identifies a serious problem for human health. Evelyn wrote of London in 1661: ‘that this glorious and ancient city should wrap her stately head in clouds of smoke and sulphur, so full of stink and darkness’.  Figure 1. John Evelyn and the title page of Fumifugium (1661). (Online version in colour.)
There was some recognition of the strong-smelling sulfur pollutants derived from coal or industrial processes such as discussed in Fumifugium and also in Shakespeare's observation about the reek of lime-kilns (The Merry Wives of Windsor, Act III, Scene 3). Lime-kilns were used extensively in Europe since Roman times and were a noted source of air pollution. There was little understanding of atmospheric chemistry, however, although scientific interest became more important by the mid-1600s [19], with the harmful effects of air pollutants ascribed to various components of the air by Kenelme Digby (acids), Nehemiah Grew (lead), John Evelyn (sulfur) and John Hall (antimony or mercury). The earliest legislation in England was the 1273 Smoke Abatement Act, prohibiting the use of coal as it was ‘prejudicial to health’ [22]. Some mediaeval societies approached air pollution control by keeping the sources outside the city walls, a concept found in Aristotle's Athenian Politics and in ancient Roman regulation [23]. This practice continued in mediaeval Europe, but also Asia, notably in relation to the extensive fifteenth-century Thuriang pottery kilns, which were located in the northern lee of Si Satchanali, Thailand. These early examples of what we may now like to call ‘environmental law’ include controls on the burning of sea coal and the ‘forestry laws’ (protecting the various species of game living in the forests). Most of these examples derive from the particular whims and prejudices of individual rulers, often heavily influenced by those within the upper social echelons of society. There was no modern science involved: the problem was perhaps a visual blot on the monarch's landscape, an appalling smell or a passion for hunting. In every case, the control or prohibition was imposed without any need to resort to the scientific knowledge base of the time. In the Western world, one has to look to the Renaissance and the subsequent Reformation as forming the basis from which arose our modern methodologies for scientific enquiry. In the UK, this change in methods of enquiry alongside the rise of industrialization in the latter part of the eighteenth century and its increasing pace in the nineteenth century enabled the likes of David Hume (1711–1776), Jeremy Bentham (1748–1832) and John Stuart Mill (1806–1873) to articulate social philosophies such as utilitarianism—philosophies which ultimately gave momentum to centralized regulation of what we would now describe as ‘environmental issues'. The direct intervention of the British Government by way of legislation was limited throughout this period. The Alkali Works Regulation Act 1863 and its Alkali Inspectorate were the prime example of governmental responsiveness to environmental matters during this period: necessity, driven by widespread and self-evident health and welfare problems, but enacted reluctantly. Initially, it was the ‘common law’ that was used to combat instances of polluting activity in England. The English ‘civil claims procedure’ requires a complainant, a defendant and proof on a balance of probabilities that the ‘injury’ complained of was caused by the action or inaction of the defendant by reason of breach of a standard of care—a standard established and refined by the judiciary over many decades on a case-by-case basis. This was thereby inevitably a standard that was intrinsically susceptible to circumstance and prevailing norms. This reliance on an ‘after-the-fact’ procedure, coupled with a requirement to establish a clear, legally recognizable causal link between alleged cause and supposed effect, was to permeate the UK's approach to what we now term environmental regulation. The approach was one of future prevention of what could be shown to have already been clearly unacceptable rather than a ‘precautionary approach’, the latter now underpinning much of current thinking in relation to matters of the environment and social well-being generally. A key step on the way to developing a UK legal framework for air pollution is the ‘The Law of Nuisance’. As early as 1610, William Aldred's case, as it is known, saw the courts intervene against one Thomas Benton for building a pigsty ‘so near the house of the plaintiff that the air therof was corrupted’. The court found that light and clean air were considered necessary for wholesome habitation. In discussing the issues raised, the court drew a distinction between ‘trifling inconveniences' that made life inconvenient or uncomfortable—when location could be a legitimate consideration in making such a finding—and material damage to property that diminished its value when location was largely irrelevant. The overall result was that the common law proved increasingly inadequate to address the sorts of issues that social philosophers and reformers were coming to focus on as increasing urbanization gave rise to self-evident public health and welfare issues. Between 1800 and 1850, the population in England and Wales doubled to 16 million and doubled again by 1900, accompanied by dramatic changes in the distribution and concentration of the population as industrialization drew people from rural areas to what soon became highly urbanized areas with insanitary housing, disease, noxious emissions and fossil fuels adding to the toxic mix. The ‘Royal Commission on Noxious Vapours’ of 1878 recorded many examples of the kinds of damage resulting from what at that time was uncontrolled industry. However, the UK government was slow to take remedial action because of the importance of industry to the national economy. Despite the manufacturing controls introduced in the Alkali Act 1863, and the establishment of an Alkali Inspectorate, the increasing number of alkali works (manufacturing sodium carbonate, while emitting hydrochloric acid as air pollution) meant that the UK experienced no meaningful decrease in emissions of pollutants. Legislation requirements to use ‘best practicable means', together with a piecemeal approach, exacerbated what was in any event often indifferent enforcement of the legislation. Nevertheless, laws to control air pollution that are recognizably modern did develop through the latter part of the nineteenth century, and these also reflected the sanitary reform that characterized the broad public health concerns of the time [24]. As might be expected, they were common in Europe and North America, but also followed imperial administrations across the world, so were well known in India (e.g. Smoke Nuisance Acts in Bengal 1905 and Bombay 1912) and Hong Kong. The wide range of international law was reviewed at the London Public Health Congress in 1905, often cited as the place where Henry Antoine Des Voeux coined the term ‘smog’ [25] During the early phase of the industrial revolution, beginning in the UK in the late eighteenth century and spreading through Europe and North America, a rapid growth in coal combustion in the developing cities substantially increased emissions of SO2, NO2, NH3 and smoke (e.g. [26,27]). The problem of air pollution focused in this period on human health. In part, emissions were due to industrial development and rapidly increasing emissions from short stacks. Additional sources were from domestic emissions by the rapidly growing urban population of factory workers who mostly burned coal for warmth and cooking. Ambient concentrations were not measured during the eighteenth and early nineteenth centuries, and values are a matter of speculation. Emissions from combustion were the main contributors to poor air quality, but they were not the only pollutants. It is important to mention emissions of NH3 from the large urban population of horses for transport, which would have added to NH3 released by coal combustion [16]. The quantity of horse dung on urban roads was recognized as a growing problem in the late nineteenth century, for example 100 000 horses in New York producing 1000 tonnes of manure daily (the ‘great manure crisis’ in New York and London; [28]), with a major problem projected into future decades. Prior to the twentieth century, horse populations were substantial in all major cities. The poor state of sewage treatment, especially during the rapid expansion of the eighteenth and nineteenth centuries, also contributed to emissions of NH3. The rapid replacement of horse drawn transport by motor vehicles in the early decades of the twentieth century avoided the problems forecast for cities like London and New York. Little attention has been drawn to the combination of SO2, NOx and NH3 in the urban chemical climate of the nineteenth century, perhaps due to the lack of measurements and the focus on pollutants from combustion sources. But the presence of large emissions of NH3 would have promoted the formation of particulate (NH4)2SO4 [29] and the rapid deposition of SO2 to terrestrial surfaces [30]. Among the few early urban measurements, Smith [31], recorded concentrations of NHx (NHx is the sum of gaseous NH3 and particulate NH4+) in London of 80 to greater than 1000 µg m−3, with the highest values recorded during fog. The deposition of NH3 would also have contributed to changes in species richness of plant communities in urban areas [32]. The degradation of air quality during the period 1750 into the twentieth century was primarily in urban areas or close to large industrial point sources. Most major European cities in the late nineteenth century had air quality problems. London and Edinburgh, respectively known colloquially as ‘the Smoke’ and ‘Auld Reekie’, were notable but far from unique. All major cities of the UK suffered. Popular works of English literature by Dickens and Conan Doyle contain many descriptions of dense swirling smog contributing to an air of danger and gloom in Victorian London. Likewise, the major cities throughout Europe, where coal provided the main fuel for industry and domestic heating, developed similar air quality problems. The pollutants from coal combustion included SO2, NO2, smoke and, to a lesser extent, HCl from the chlorine in coal [33,34]. Urban concentrations of SO2 and smoke in the large cities in the middle decades of the twentieth century were commonly between 50 and 100 µg m−3, and many UK cities had annual values in this range [35]. Meteorological conditions in the winter months leading to low wind speed and a cold surface air greatly reduce dispersion of pollution, and in these conditions, concentrations of smoke and SO2 could exceed 1000 µg m−3, as in the infamous 1952 London smog episode [19]. Air pollution was, until the 1950s, largely accepted as a consequence of industrial activity, with a perceived willingness to tolerate the grime, degraded visibility, erosion and blackening of valued buildings and effects on human health, agriculture and natural ecosystems. It took a major event to change the public and political perception of the problem and the need for control measures. The 1952 London smog resulted in the premature mortality of approximately 12 000 people [1]. The public and then more slowly the political reaction led to the introduction of the Clean Air Act in 1956, some 3 years after the event. It arose from a Bill to the UK Parliament initially proposed by a back-bench Member of Parliament (Sir Gerald Nabarro), and not an initiative of the Government Ministers at the time, an indication of the prevailing focus on housing, industrial growth and recovery from the effects of the Second World War. The lack of prioritization for matters of the environment was a feature of 1950s Britain, where food rationing was still in place in 1952. However, this Act of Parliament was a very important step, eventually leading to widespread reductions in emissions of smoke and SO2 in urban areas. During the three decades following the London smog, many urban power stations and other polluting industrial sources were closed, and new, larger, more efficient power stations were constructed in rural areas. These each typically produced 2000 MW output of electrical power and consumed 5 million tonnes of coal annually. UK emissions of SO2 continued to increase through the 1950s to a peak in the 1960s, mainly driven by industrial emissions and especially power generation. The new large power stations were equipped with reasonably effective controls for PM, but none of the new units had SO2 removing equipment until Drax in 1988 and Ratcliffe in 1995. The closure of the large number of smaller very polluting urban power stations and other industrial sources with short stacks further reduced emissions of SO2 and smoke in cities and contributed significantly to the improving urban air quality. Ambient concentrations of smoke and SO2 declined by 60% between 1962 and 1975 in London, nearly a quarter of a century after the event that tipped the scales in favour of effective action on urban air quality (figure 2). Figure 2. The decline in SO2 and smoke in London following the Clean Air Act (1956), including data from the ‘bubbler method’ sampling air through a peroxide solution in water and ultraviolet (UV) spectroscopy. (M. L. Williams, personal communication, 2017). (Online version in colour.)
Recognizing needs to reduce ground-level concentrations of SO2 and smoke, new power stations from the 1950s were built with increasingly high chimney stacks. The large rural power stations, many in the Trent and Ouse valleys of the English Midlands and industrial north, had larger stack heights, many at 200 m, promoting dispersion and reducing local effects [36]. Similar changes in power generation were taking place across Europe where SO2 emissions peaked in the 1980s [37]. It is notable that on a global scale, Europe and North America were responsible for most (greater than 80%) of the global SO2 emissions prior to 1970 (figure 3). Emissions in North America grew rather faster than those in Europe and with tall stacks also used in North America to disperse pollutants, to minimize local effects. Global emissions of NOx, non-methane volatile organic compound (NMVOC) and NH3 also increased rapidly during the late twentieth century (figure 3), as consequences of increased energy consumption, transport, solvents and agricultural activity. These graphs show the major shifts since 1980, where China and the Asia/Pacific region have replaced Europe and North America as the main global sources of air pollution. Figure 3. Global and regional emissions of SO2, NOx, NH3 and NMVOC between 1750 and 2010. Adapted from Hoesly et al. [37]. The dots show global estimates of an earlier study (CMIP5 [38]). (Online version in colour.)
Although there had been earlier concerns regarding ecological effects, the policies to control air pollution during the 1950s and 1960s were aimed at protecting human health, with a focus on urban air quality. But, as noted above, the gradual improvement of urban air quality took place while country-scale emissions of SO2 were close to their maximum, as a consequence of increasing emissions from large combustion plants with tall chimney stacks. During this period, annual UK emissions of SO2 reached their peak of 3 Mt S annually [39]. Annual European and North American emissions of SO2 also peaked during this period at 32 Mt-S [40] and 31 Mt-S [41], respectively. The scale and effects on the countryside from the high levels of SO2 (and NO2) were not immediately recognized as an issue. In the Pennine hills between Sheffield and Manchester, the Forestry Commission was unable to establish plantations of Scots Pine due to the ambient exposure to SO2 and large areas of central England as well as all of the major cities were devoid of lichen species sensitive to SO2 [42]. Effects on agricultural crops were substantial, with some cultivars of grass developing resistance to SO2 [43], but knowledge of these domestic problems of air pollution was insufficient for the UK Government to introduce further legislation to reduce emissions. The Clean Air Act was regarded as adequate, and there were no limits on the overall scale of SO2 emissions, just a requirement to use stack heights tall enough to minimize the concentrations downwind at the surface following the philosophy of Best Practical Means [44]. Policy priorities at this stage were firmly focused on human health. Elsewhere in continental Europe, tall stacks and large power plants located outside the cities were also regarded as effective policies to minimize effects on human health. Areas highly polluted by SO2 were extensive over England, Germany, eastern central Europe and the Low Countries (figure 4). Ecological effects were not considered sufficiently important to introduce further control measures, even though it was known that some industrial processes, especially smelting, produced striking examples of local damage from SO2 and metal deposition. For example, emissions from the Sudbury smelter in eastern Canada during the early twentieth century caused extensive areas of natural vegetation to be destroyed by the combination of exposure to very large concentrations of SO2 and large deposition rates of a range of metals [45,46]. Smelters were present in many countries globally, and large exposures to SO2 and metal deposition were common in their proximity with examples in Slovenia, Peru, Canada, USA, Russia, China, France, Poland and Zambia [47]. Figure 4. Annual mean European SO2 concentrations (µg m−3) in 1970, at around the time of peak SO2 emissions, modelled using EMEP4UK with 1970 emissions and 2012 meteorology (M. Vieno et al., personal communication, 2020). (Online version in colour.)
There were seventeenth-century analyses of rainfall, possibly the first by Ole Borch in Denmark. These became more commonly undertaken by agriculturalists during the 1800s [19] and increasingly used worldwide [10,48,49], providing early evidence of inter-country exchange of pollutants from observations of contaminated snowfall. The deposition of urban sulfate in London rainfall was determined by Robert Angus Smith in 1869–1870 [7]. The number of estimates of the concentrations of substances in air increased rapidly through the nineteenth century. Russell [50] measured total PM gravimetrically in central London at 120, 360 and 860 µg m−3 in fine, dull and foggy weather, respectively [51]. Although measurements of carbon dioxide were frequent (as listed in Callendar [52]), this occurred only occasionally for trace gases such as ammonia [31,53]. The early twentieth century saw the development of the deposit gauge that measured wet deposition and whatever else fell into the large glass bowl [54] and the use of lead peroxide candles to determine deposited sulfur dioxide [55]. The widespread occurrence of SO2 led to increasingly sophisticated methods that involved drawing air through Dreschel bottles (bubblers) containing solutions of iodine, hydrogen peroxide or disulfitomercurate [56]. Table 2 shows the dates when regional networks to measure atmospheric composition were introduced, revealing that only the European Air Chemistry Network (EACN) was in place prior to the peak in European emissions of SO2. At the global scale, monitoring of global trends in background atmospheric composition is now coordinated through the Global Atmospheric Watch (GAW) network [58], whose origins can be traced back to the Background Air Pollution Monitoring Network (BAPMoN) first established in 1974 [59].
The Norwegian playwright Henrik Ibsen's play ‘Fire’ (Brand) showed that people were aware of long-range transport of pollutants in the nineteenth century [60]: Worse times, worse sins through the night of future flashes of Britain's suffocating coal dust is slowly descending over the countryside soiling all that is green strangling all that strives to grow creeping low and mixed with poison stealing sun and light from the valley pelting down as rain of ashes. Even in the early 1950s, it was known that the majority of UK sulfur emissions were exported from the UK coastline, as Meetham [61] demonstrated using national monitoring data and a simple atmospheric mass balance. However, the magnitude of the effects of long-range transport of air pollutants from the major emitting countries in Europe on the net importing countries was not considered an important issue until the late 1960s and 1970s. The potential for transboundary transport within Europe can be readily visualized from the data in figure 4, given typical boundary layer wind-speeds of 10 m s−1 and an atmospheric e-folding lifetime for SOx of a few days. Swedish scientist Svante Oden initially advanced the idea that the long-range transport of sulfur and acidity in Europe from the major sulfur emitting countries (UK, Germany, France and Poland in particular) was responsible for widespread acidification of freshwaters and loss of fish populations in Scandinavia [62]. Monitoring data for air and precipitation established by Egner and colleagues from 1955 (the EACN) provided the vital chemical data showing both the geographical patterns and trends in concentrations of the major ions in precipitation (e.g. [10]). The EACN data showed acidity and sulfate in precipitation increasing steadily through the 1950s and 1960s [63]. Oden attracted considerable Swedish interest with these ideas, much of it critical. The ideas were new and many aspects of the atmospheric chemistry and physics of the compounds involved were poorly understood. However, the evidence was persuasive, and further analysis by Swedish colleagues provided strong support for his arguments. A United Nations conference on the Human Environment in Stockholm in 1972 advanced the wider case that the pollution of one country by another through emission, atmospheric transport and deposition was unacceptable [64], building on the evidence of long-range transport of sulfur compounds within Europe and the effects in Scandinavia. The Stockholm Conference of 1972 was a turning point in environmental science. There is little doubt that long-range transport and effects of pollutants had been taking place for decades; in fact, many had noted the inter-country transport of pollutants in Europe (e.g. [60]). It was only following Oden's analysis that international scientific and political attention was drawn to the subject and monitoring and process studies demonstrated the scale and ecological significance. The initial reaction of the major polluting countries in Europe was mixed. All recognized the need to quantify the scale of inter-country transport of the major pollutants, the underpinning atmospheric chemistry and physics, and the effects on ecosystems. However, the legal instruments needed to be developed, and the supporting monitoring and analysis tools were still lacking. The industrial nations had designed tall stacks to disperse the pollutants without considering possible effects outside their jurisdiction. The combination of the science and political background is described in detail by Grennfelt et al. [57]. An earlier description of the political background in the UK by Rose [65] provides further insights and is typical of the large number of publications on the politics of acid rain. Research and monitoring activities expanded across Europe and the essential details of the emissions, atmospheric chemistry and deposition were presented at Dubrovnik in 1977 [66]. The first major international conference on acid rain was held in Columbus, OH, USA in 1975, beginning an important series of meetings on the subject approximately every 5 years from 1975 to 2015. Table 3 provides links to this series of meetings which shows the developing global scale of air pollution issues through the latter decades of the twentieth century, beginning with acid rain. The results of the Norwegian SNSF research programme into the causes and effects of acid rain were presented at the second international Acid Rain conference in Sandefjord in 1980 [68]. Clear links between sulfur emissions in the industrial nations of Europe and long-range transport to, and deposition and effects within, Norway and more widely in Scandinavia were demonstrated. The effects were primarily on freshwaters, with large declines in fish populations in the most acidified regions. The early studies in Scandinavia did not show negative effects on forests of the long-range transport of pollutants. Discussion meetings at the Royal Society of London reported the Pathways of pollutants in the atmosphere in 1979 (Phil. Trans. A, vol. 290) and Terrestrial Effects of deposited sulfur and nitrogen compounds in 1984 (Phil. Trans. B, vol. 305).
By the late 1970s, the requirement to reduce European emissions especially of SO2, and thereby reduce deposition of acidifying compounds, in Scandinavia, was clear. This led to the establishment of the Convention on Long-range Transboundary Air Pollution (LRTAP) as a major international framework to address the problem [75]. This required extensive monitoring of the chemical composition of air and precipitation and associated meteorological variables, the development of atmospheric transport and deposition models to quantify the net transfer of pollutants between countries (the European Monitoring and assessment Programme (EMEP)), and a framework for interpretation and negotiation of the issues between the countries. The development of the LRTAP Convention has proved a very effective process to bring together the process-based science, the monitoring and the modelling (within the EMEP) and policy development, ultimately leading to international agreements to reduce emissions of SO2 and subsequently NO2, VOC and other air pollutants. Like the human–health-oriented measurement programmes instigated in the USA, the LRTAP programme recognized the need to simultaneously measure a wide range of constituents, but also recognized the need to make measurements over long time periods to overcome the considerable inter-annual variability in meteorological conditions [76]. The LRTAP programme also introduced the explicit goal of making measurements to support associated atmospheric modelling. In the early 1980s, a group of European countries considered a reduction in SO2 emissions appropriate, but in the absence of country-specific contributions to the ecological damage in Scandinavia, an arbitrary agreement to reduce emissions by 30% was proposed. Many, but not all countries, supported the measure, forming the 30% club. Arguments were presented that the costs of control were substantially greater than the benefits. For example, it was stated (incorrectly) that ‘acid deposition is a million dollar problem with a billion dollar solution’ [77]. Nevertheless, the 30% club was the basis for the first Sulphur Protocol, signed in Helsinki in 1985, which stipulated a reduction in sulfur emissions of 30% between 1980 and 1993. Freshwater acidification and a decline in fish populations were the initial focus and the evidence that the cause was long-range transport of pollutants, mainly sulfur, was compelling. There was a secondary focus on the health of forests in Scandinavia but here the evidence was not persuasive. The scientific and political interest in acid rain in Scandinavia and more widely in Europe stimulated interest elsewhere, especially North America, where similarly large increases in emissions of SO2 had occurred (figure 3). The USA had established its own National Atmospheric Deposition Program (NADP) in 1977. The combination of a large source area in the Ohio River valley and a large area downwind with geology and ecosystems sensitive to acidification soon led to the recognition of problems from long-range transport of pollutants similar to those identified in Europe. The fact that substantial areas of acid-sensitive ecosystems were located in Canada added a political dimension similar to that in Europe, with one country being responsible for ecological problems in a neighbouring territory. By 1980 acid rain, or more correctly acid deposition, recognizing the importance of both wet and dry deposition to the total input to the ground [78], was established as an international issue, and all industrial countries engaged in research and many in the development of control measures. Interest in acid deposition in Europe was greatly stimulated in the early 1980s by a decline in the health of forests in the most polluted regions [79]. The most damaged forests were those in the uplands in border regions of the Czech Republic, Poland and the German Democratic Republic where die-back of the forest was extensive. In parts of Germany, especially the Harz Mountains, the tree-line moved down the hills as damage at high elevation which was a feature of the problem progressed to lower levels. The large areas of forest decline throughout Germany (Waldsterben) became a defining environmental issue of the late twentieth century [32]. The causes of forest decline were hotly debated and contentious [79]. The main causal agents appeared to be acid deposition and ozone, but excessive nitrogen deposition and metals were also possible contributors and, at many of the sites of forest decline, exposures to large inputs of a combination of these pollutants were common. While many publications address the problem, there is no consensus to date on the proportions of the observed damage attributable to each of the pollutants and mechanisms of damage. Forest decline was an important part of the acid deposition story in North America, and a particular species, red spruce, showed winter injury that was shown to be associated with the exposure to acidic cloud-water and sulfate in the Appalachians [80]. From the mid-1980s, sulfur and acid deposition declined steadily in Europe and North America, with recovery in atmospheric composition preceding any signs of ecosystem recovery [81,82]. While there were clear links between acid deposition and forest decline in both Europe and North America, ground-level ozone was also implicated in both regions [83,84]. The broadening of the ecological focus from freshwaters to forests and the expansion of the number of pollutants implicated in effects was an important development. In expanding the range of effects and pollutants, the regional scale was also expanding considerably. Ozone is formed within the atmosphere following photolysis of oxygen in the stratosphere, and some is transferred into the troposphere and contributes to ozone at the ground level [85]. However, ozone is also produced through the photochemical degradation of carbon monoxide and VOCs in the presence of NO2, and the issue of ozone in surface air is commonly referred to as ground-level ozone to distinguish it from stratospheric ozone issues. It was first recognized as a problem for human health and vegetation in California, and especially the Los Angeles basin, where it was first described by Middleton et al. [86]. The presence of ozone concentrations that posed a risk to vegetation and human health over Europe was demonstrated in the early 1970s [87]. Over Europe, the background concentration of ozone has increased by approximately a factor of two since pre-industrial times [88], and episodes of elevated ozone were shown to be widespread in Europe in the 1980s [89]. The effects of ozone on natural vegetation and crops are discussed by Stevens et al. [32] and Emberson [90], respectively. The discovery of damaging ozone concentrations in Europe and North America greatly increased the recognition of photochemical oxidants in regional air quality issues in the 1970s and 1980s. The focus of control measures therefore broadened from sulfur and nitrogen oxides to include VOCs in many other industrial countries [57]. The relatively long lifetime of ozone in the troposphere (approx. 20 days) and photochemical production over regional scales makes ground-level ozone a continental and hemispheric scale pollutant [91]. The industrial regions of Europe and North America experienced frequent summer episodes of ozone in the 1970s and 1980s with concentrations exceeding 200 µg m−3. The control measures to date have all been country or regional in scale, and while important progress has been made in reducing peak values, especially in California, but also across much of the USA and Europe, ozone remains a substantial threat to crops, natural vegetation and human health [32,90]. Methane is playing a major role in the formation of background ozone, and there is increasing interest in taking policy actions to control methane emissions as it is both a greenhouse gas and an ozone precursor [92]. As an understanding of acid deposition developed, and ground-level ozone was recognized as an additional regional-scale air pollution issue, the importance of nitrogen compounds grew. Nitrogen compounds were always a part of acid deposition, even when deposited in reduced form as NH3 or NH4+ in precipitation, as the protons generated in soil following microbial oxidation to nitrate create acidity [93]. However, the Netherlands and the UK were first to observe widespread changes in botanical species composition of heathlands [94]. It was soon shown that the changes were being driven by nitrogen deposition from the atmosphere and by ammonia in particular. As always in ecology, the story is a little more complex, as the replacement of heather-dominated heathlands by grassland in the Netherlands was mediated by the heather beetle, but the underlying driver of change was the deposition of nitrogen compounds from the atmosphere [95]. The eutrophication of ecosystems by nitrogen deposition has been shown to reduce species richness of grasslands over regional scales in Europe [96,97]. Close to livestock sources of ammonia, the changes in flora can be substantial [98] and the form of the nitrogen deposited has been shown to be an important factor in the scale of effects, with gaseous ammonia being more damaging to heather than wet-deposited NO3− or NH4+ [16,99]. Similar effects of deposited nitrogen on ecosystems have been reported in North America and China. The scale of effects of pollutants on ecosystems quantified at the turn of the twenty-first century showed that 24% of global forests were exposed to phytotoxic exposures of ozone [100]. The development of the Critical Loads approach and integrated assessment methods proved valuable instruments in the development of policies to maximize the ecological benefits of control measures within the LRTAP Convention [57]. As noted above, the early evidence of air pollution effects were largely human health-related until the discovery of acid rain effects in Scandinavia in the late 1960s. The recognition of effects of long-range transport and the deposition of pollutants changed the scientific and, for a while, the political attention. The broadening of the science interest into ground-level ozone and eutrophication were important in the science and effects, and led to controls on the precursor pollutant emissions in Europe through LRTAP protocols. In North America, efforts to control the precursor gases followed a different control process, but achieved similar reductions in emissions over the longer term. Emissions of SO2 in Europe and North America have been reduced in 2016 by approximately 90% from their peak values in the 1970s and 1980s, respectively (figure 3). However, in the early 1990s, a publication showing associations across six US cities between human mortality and morbidity and levels of air pollutants, especially PM, changed the political and scientific focus of effects [101]. Subsequent publications on human health effects of pollutants following similar epidemiological approaches revealed the scale of effects on human health throughout the developed and developing nations. Current estimates are that outdoor concentrations of PM2.5 alone are responsible for annual burdens of 4.2 million premature deaths and 100 million disability-adjusted life-years lost globally [102]. These publications showed air pollution to be one of the major global causes of premature mortality and drew attention to the human health effects of pollutants at much smaller concentrations than had been implicated in the London smog of 1952. This refocused scientific and political attention on air pollutants back to human health. The underlying logic of the change in focus is understandable, given the large numbers of individuals and the societal costs of poor health and mortality. By comparison, effects of pollutants on natural ecosystems, which are always difficult to value, and on agricultural and forest crops are smaller in value than those on human health. For these reasons, ecosystem effects have become a secondary consideration for the policy makers. Human health has been the primary focus for the control of air pollution since the late 1990s. Clean air legislation in Europe, North America, Japan and other developed countries targets both ambient levels and emission sources. Nevertheless, the multi-impact effects of PM, NO2 and O3 on human health and managed and natural ecosystems mean that UNECE-LRTAP protocols still fulfil a crucial role [57]. The chronology presented here describes the development of air quality issues as they arose rather than providing a narrative for each pollutant. However, it is important to draw attention to PM and its role in current air quality problems. PM features in the earliest reports of air pollution, although terminology has been inconsistent and often poorly defined with terms including smoke, soot, fume, haze and dust, frequently used somewhat indiscriminately through the literature. PM, described in detail by Harrison [103], in this issue, refers to the sum of all solid and liquid particles suspended in air and is a complex mixture of size, spanning at least four orders of magnitude (1–10 000 nm) and with a large range of chemical composition. The latter reflects the wide variety of contemporary sources and very broadly comprises carbonaceous particles emitted directly from combustion, dusts from industrial processes and within-atmosphere conversions of inorganic (SO2, NOx and NH3) and organic (VOC) gases into PM. PM is the main contributor to human health effects by some margin, and it is also the form in which most of the long-range transport of sulfur and nitrogen-containing pollutants occurs. PM contributes to changes in the Earth's energy balance both by absorption (e.g. black carbon) and by dispersion and reflection of radiation. Many of the links between air quality and climate change are therefore due to interactions between PM and the radiative balance and thus climate [104]. Similarly, many of the effects of pollutants on ecosystems are due to the deposition of PM either directly by dry deposition on foliar surfaces or through occult or wet deposition [105]. Smog includes both particulate and gaseous components, but the visibility effects are dominated by PM. Given the contribution of PM to the chemical climatology of the atmosphere over the developed and rapidly developing countries and the contribution of PM to effects on human health, it is likely that PM will continue to dominate control measures for some decades to come. Emissions of most primary pollutants have declined in Europe, North America and Japan from the 1990s until the present with the greatest progress in SO2, but even NO2 and VOC emissions have decreased more than 50% from their peaks in these regions. By contrast, during the period 1990–2010, emissions have increased in East and South Asia, and elsewhere, so that reductions in global total emissions, even for SO2, are modest, with a reduction of 15% from the peak in 1990 (figure 3) [37]. For NOx emissions, the global total continued to rise and all the reductions in emissions in Europe, North America and elsewhere have been counterbalanced by increases elsewhere and mainly in Asia (figure 3). For NH3 and VOC, the case is similar to that for NOx, with the global total steadily increasing [37]. The large increases in emissions of all primary pollutants in South and East Asia have been widely reported and described by Zheng et al. [106]. Air quality in Asian megacities shows values for PM, SO2 and NO2 in episode conditions that are similar to the highly polluted atmosphere of London in the smog episodes of the 1950s, for example the Beijing ‘haze’ events in January 2012. The global burden of air pollutants has therefore continued to increase into the first two decades of the twenty-first century. The focus of political attention remains firmly on human health due to the PM and NO2 exposure in urban areas of the developed and developing world. The distribution of ambient PM2.5 concentrations experienced by different regional populations presented in figure 5 shows how the current global air pollution health burden is disproportionately borne by countries in East and South Asia, rather than the countries that were afflicted in the early stages of the Industrial Revolution. Even so, the majority of the world's population live in locations where levels of ambient PM2.5 exceed the WHO guideline value. Figure 5. Distributions of the population as a function of annual (2013) average ambient PM2.5 concentration for the world's 10 most populous countries and the rest of the world. Dashed vertical lines indicate World Health Organization Interim Targets (IT) and the Air Quality Guideline (AQG). Source: Brauer et al. [107]. (Online version in colour.)
It is important to note that this focus on human health deflects attention from the continued widespread exceedances of thresholds for effects of pollutants on managed and natural ecosystems [90]. The need to observe tropospheric pollution globally drove the development of passive and some active remote sensing techniques to measure the tropospheric burden of key pollutants trace gases and aerosol. In 1981, 1984 and twice in 1994 the MAPS (Mapping Pollution with Satellites), a gas correlation radiometer instrument flew for typically 9 day missions on the space shuttle and measured middle and upper tropospheric CO between approximately 55°N and approximately 55°S. A more advanced nadir gas correlation instrument MOPITT (Measurements of Pollution in The Troposphere) measuring in both the thermal and short-wave infrared has now made over 20 years of measurements of tropospheric CO and some CH4 from the NASA Terra, which was launched at the end of 1999. The total columns and vertical profiles of ozone, O3, data products from NASA TOMS and SBUV on Nimbus 7 and later TOMS and SAGE II data [108,109] were used to retrieve O3, O3 amounts and distributions with a focus in the tropics. From 1984, the SCIAMACHY (SCanning Imaging Absorption spectroMeter for Atmospheric CHartographY) project was developed and proposed to ESA in 1988. This led to the smaller GOME (Global Ozone Monitoring Experiment), being flown on ESA ERS-2 (1995–2011, [4]) and SCIAMACHY on ESA ENVISAT (2002–2012, [110,111]). Both satellites flew in polar orbits with equator crossing times of 10.30 and 10.00, respectively, and measured in nadir viewing geometry the upwelling radiation in the solar spectral region at the top of the atmosphere. The UV and visible nadir measurements of GOME and SCIAMACHY have been exploited to retrieve tropospheric columns of NO2, O3, SO2, HCHO, CHO.CHO, BrO, IO and H2O [4] in cloud-free regions and above clouds. The SCIAMACHY short-wave infrared spectral measurements enabled CO columns and for the first time the total dry column mixing ratios of CH4 and CO2 to be determined globally. Satellite measurements revealed the growth in emissions in Asia and the declines in Europe and North America during the period 1996–2004 [112]. Satellite remote sensing in the solar spectral region also provides global fields for tropospheric SO2, CO, HCHO and CHO.CHO. The launch of the instruments AIRS on NASA and IASI, a CNES FTIR on EUMETSAT MetOp A B, has led to the detection of NH3 (see [113,114]). A combination of data from the GOME, SCIAMACHY and OMI instruments provides clear evidence of the increase in NO2 in Eastern China between 1995 and 2010 and the subsequent decline from 2010 to 2018, as shown in figure 6. Figure 6. Trends in the tropospheric NO2 column over East China between 1995 and 2018 (A. Richter and J. P. Burrows, personal communication, 2020). (Online version in colour.)
Satellite remote sensing also provides measures of PM, aerosol optical thickness, e.g. MODIS on NASA Terra (1999–present) and Aqua (2002–present) [107]. Global emissions of sulfur have declined since the peak in 2000, and recent trends in China since 2012 show a reduction in emissions approaching 50% (figures 3 and 7). Such a reduction represents remarkable progress, relative to the time it took to reduce emissions in Europe and North America by a similar amount (approx. 20 years). Emissions of NOx in China have also declined, by approximately 25% over the last 8 years [115] and figure 6, although surface ozone has continued to increase [116]. It is therefore possible that the world has passed the point of maximum emissions of several major gaseous air pollutants as a combination of further controls in North America, Europe and East Asia drive down global totals. Climate change policies directed towards reduced use of coal and oil are expected to contribute further reductions in emissions of SO2 and NO2 over coming decades [117]. However, there are good reasons to be cautious, because emissions of ammonia, an important contributor to PM and eutrophication, continue to rise, and possible feedbacks between emissions of these gases and climate may drive overall emissions upwards [118,119]. Global emissions of CH4 and VOC also continue to rise, and in the case of biogenic emissions, it is possible that changes in climate and the widespread planting of new forests may accelerate global emissions of biogenic VOC (BVOC). Decisions over the species chosen for tree planting to increase carbon sequestration will also need to be made to simultaneously ensure that BVOC emissions do not increase. At present, it remains inconsistent in international policy that land use, land-use change and forestry are recognized as areas to count as carbon credits in the UN Framework Convention on Climate Change, but when it comes to the revised Gothenburg Protocol under the LRTAP Convention, the accompanying BVOC emissions are considered ‘natural’ and are excluded from the emissions commitments. Both the benefits for carbon and the possible disbenefits for BVOC will need to be recognized in future international agreements. Figure 7. Annual emissions of (a) SO2, (b) NOx and (c) NMVOC in China between 2010 and 2017 (adapted from Zheng et al. [115]). (Online version in colour.)
Despite the widespread elevated levels of PM2.5 illustrated in figure 5, data from the global burden of disease project (figure 8) indicate that globally the world may now be on a downward trend of death rates from outdoor PM2.5 and from ground-level ozone. Figure 8. Annual death rates attributed to outdoor PM2.5, outdoor ground-level ozone and indoor pollution from solid fuels 1990–2017. Source: www.ourworldindata.org/air-pollution/ based on data from the Global Burden of Disease project. (Online version in colour.)
Given the current scale of effects of air pollution on human health and ecosystems and uncertainties in measurements and modelling, it is premature to celebrate the downturn in global emissions of two of the most important air pollutants (SO2 and NOx). However, the temporal pattern in emissions of pollutants displayed in the Environmental Kuznets Curve [120], with increasing efforts to control emissions as economies mature, continues to be consistent with observations as parts of Asia now show substantial reductions in emissions, at least for pollution arising from combustion sources. Furthermore, there is a reasonable expectation that measures to combat climate change and increase the use of renewable energy in developing regions, especially Africa, may substantially mitigate emissions of air pollutants as their economies develop. In providing a chronology of what has become a very large and complex field, this narrative has, of necessity, been selective. It has also been excessively brief on developments over the last two decades during which the range of issues, geographical scale and very different trends in different areas of the world obscures the wider picture. The suggestion that the world has passed the peak in air pollution problems is a strong statement, and may prove to be incorrect. However, the evidence from SO2 and NOx emissions is persuasive for the major emitting countries in Europe, North America and also for East Asia. If similarly strong controls were applied to ammonia emissions, which are certainly possible technically, the current problems with the nitrogen cycle could also be addressed [121]. It is less clear when or how global VOC emissions may be controlled, but these emissions will become less important if the world moves towards a lower NOx chemical climate. The Discussion meeting at the Royal Society in London 11th and 12th November 2019 took place just before the first case of the Sars-COVID-19 was reported in China on the 17th November 2019. By June 2020, 6.3 million cases had been reported across 188 countries and territories resulting in 376 000 deaths. Lockdown measures have led to major effects on industrial and transport activities and reduced emissions of many of the primary pollutants contributing to poor air quality. While it is too soon to provide a detailed analysis, there are many preliminary reports, including surface measurements from monitoring networks and satellite remote sensing. In the major cities, reduced combustion-related emissions are revealed by CO2 flux measurements, with reductions of 55% in central London [122]. The reductions in urban NO2 in the UK during the first weeks of lockdown of 20–30% [122] are similar to reductions in other major cities across the developed world. Similar reductions have been observed in column NO2 satellite remote sensing (J.P. Burrows, personal communication, 2020). The effects on PM2,5 are much smaller and more variable than effects on NO2, with some COVID-19-affected cities in China reporting reductions in PM10 similar in scale to reductions in NO2 but for a shorter period [123]. The analysis for the UK during lockdown suggested reductions in personal exposure in London to PM2.5 in the range 5–25%, depending on the mode of travel, but effects on ambient PM are small and very variable. The global scale of the pandemic produced a clear effect on global emissions of combustion-related emissions of pollutants, with expected health and environmental benefits due mainly to reduced NOx emissions. Whether these benefits lead to longer-term reductions in emissions is much less clear as transport and industrial emissions grow following the widespread population lockdown. It seems likely that an effect of COVID-19 will be to reduce net acidity and increase the gaseous alkaline fraction [16] as transport and combustion emission are reduced, but with little anticipated reduction in NH3 emissions from agriculture. While this may be associated with health benefits, additional adverse effects of ‘alkaline air’ on ecosystems will also need to be considered. The bulk of the paper is text and review, but there are data provided for this paper which have not been published. The data are shown in figures 2, 4 and 6. The sources for the data are listed in the legends for each figure and the data may be obtained from the named sources. D.F., M.R.H., A.J. and P.B. wrote the core sections of the manuscript with contributions and critical revisions provided by P.G., D.S.S., E.N., M.C., M.H.U., M.A.S., X.L., Y.C., Z.K. and G.W.F. M.V. provided the EMEP4UK modelling and J.B. provided the satellite remote sensing analysis. We declare we have no competing interests. We received no funding for this study. The authors gratefully acknowledge constructive comments from anonymous reviewers. The contributions by M.C., M.A.S., E.N. and M.V. were supported by the UK Natural Environment Research Council (NERC) National Capability award NE/R016429/1, UK-SCAPE. We gratefully acknowledge support from UKRI for support of the Global Challenges Research Fund (GCRF) South Asian Nitrogen Hub (M.A.S., E.N., M.V. and D.S.S.) and the GEF/UNEP project ‘Towards the International Nitrogen Management System’ (M.A.S. and M.V.). FootnotesOne contribution of 17 to a discussion meeting issue ‘Air quality, past present and future’. © 2020 The Authors. Published by the Royal Society under the terms of the Creative Commons Attribution License http://creativecommons.org/licenses/by/4.0/, which permits unrestricted use, provided the original author and source are credited. References
Page 2
Page 3Discussion meeting issue ‘Air quality, past present and future’ organised and edited by David Fowler, John Pyle, Mark Sutton and Martin Williams
Page 4Airborne particulate matter (PM) is a pollutant of great importance which presents many challenges. Its significance lies particularly within two areas. Firstly, it is the pollutant having by far the largest impact upon public health. This is clearly elaborated by other contributions to this volume, and according to the Global Burden of Disease study [1], it ranks very highly among the avoidable causes of non-communicable diseases. PM is also important because it both absorbs and reflects solar radiation and therefore affects climate [2]. Absorption of incoming solar radiation by components such as black carbon, which are strongly absorptive, causes local heating of the atmosphere while more reflective particles such as ammonium sulfate reflect sunlight back to space and have a net cooling effect at the surface. However, there are also secondary effects concerned with cloud formation, and the number density of particles referred to as cloud condensation nuclei has a profound influence upon the albedo of clouds and hence upon surface air temperatures [3]. Such effects are not the focus of this paper which will address primarily issues concerned with the size distribution, chemical composition and sources of airborne PM. It will consider emerging contributors to PM concentrations before focussing on airborne nanoparticles and considering possible impacts upon future concentrations. Airborne particles present great complexity because unlike atmospheric trace gases which have the same chemical and physical properties wherever they occur, airborne particles are in reality a suite of pollutants varying in particle size and chemical composition on a range of temporal and spatial scales. Airborne particles can be both directly emitted, referred to as primary, or formed within the atmosphere from the condensation of trace gases, referred to as secondary particles. The smallest such particles arising from gas-to-particle conversion processes are 1–2 nm in diameter. The largest airborne particles are in excess of 100 µm in diameter but have a rather short atmospheric lifetime due to high gravitational settling speeds. The air quality guidelines and standards applicable to PM are framed in terms of two health-relevant fractions. The first is referred to as PM2.5, which describes particles measured by mass which pass a sampling inlet with a 50% cut-off efficiency at 2.5 µm. They are, therefore, in effect all particles smaller than 2.5 µm. The other metric, PM10, describes particles measured by mass passing a sampler inlet with a 50% efficiency cut-off at 10 µm. It therefore includes the mass of all particles smaller than 10 µm and consequently includes PM2.5. The other relevant definition is that of ultrafine particles, generally defined as particles with one dimension smaller than 100 nm (0.1 µm). These are often also referred to as nanoparticles because of their nanometre dimensions. Harrison et al. [4] have described the size distribution of particles and how the characteristics of the size distribution can look very different according to whether it is expressed in terms of particle number, surface area or volume/mass. Most airborne particles by number are typically smaller than 100 nm diameter and hence the peak abundance of particles in the urban atmosphere is often at a size of around 20–30 nm, and in European and North American cities, typically 80–90% of particles by number are smaller than 100 nm. However, when the size distribution is transformed into a surface area distribution, the majority of the surface area is associated with particles in the 0.1–1 µm size range referred to as the accumulation mode. Nanoparticles contribute only a relatively small amount of the total surface area. When the distribution is converted to a volume distribution or a mass distribution (if the density is known), two modes typically appear; one in the 0.1–1 µm range referred to as the fine mode and one in the 1–10 µm range referred to as the coarse mode. The minimum between the two most typically lies at around 1 µm, but the sub-division at 2.5 µm used by regulatory agencies to define the PM2.5 metric provides an approximate sub-division of the fine fraction. As a general observation, fine fraction particles arise predominantly from gas-to-particle conversion processes within the atmosphere or from emissions from high-temperature processes such as vehicle exhaust or industrial combustion. On the other hand, coarse fraction particles are more typically associated with mechanical break-up through abrasion or wind-driven processes such as soil resuspension or the creation of sea spray by breaking waves. Figure 1 shows average particle number size distributions measured by a Scanning Mobility Particle Sizer, which separates particles on the basis of giving them charge and then measuring their mobility in an electric field, from air sampling campaigns in London, Beijing and Delhi. These are winter campaign data which do not reflect the major seasonal changes in particle concentrations which are seen in both Delhi and Beijing, and rather less in London. The lowest concentrations are seen at the London North Kensington site which is an urban background location in central London. The Marylebone Road site is at roadside and shows a broadly similar size distribution but an appreciably higher average concentration. The particles from Beijing are greater in both number and size while those in Delhi are far greater in number with a much larger modal diameter close to 100 nm. It, therefore, appears that the number concentration and modal diameter tend to scale with the degree of pollution of the city, with Delhi showing by far the highest pollution levels and London the lowest. The trend in diameter may reflect different predominant emission sources or may be the result of particle growth in the atmosphere. In more polluted atmospheres, particles grow more rapidly by coagulation which depends upon particle number concentrations and by condensational growth largely due to atmospheric oxidation processes creating species of low volatility which condense onto existing particles causing them to grow [5]. Figure 2 shows typical diurnal variations of particle number counts from London, Beijing and Delhi. Just discernable in the Marylebone Road and North Kensington data are the influences of road traffic emissions, and most notably the morning rush hour, on number concentrations, while in Beijing, the highest concentrations are seen in the early afternoon. This is due to two major differences from London. Firstly, the light-duty vehicle fleet in Beijing is wholly gasoline fuelled which leads to much lower emissions of particles than the diesels which contribute a large part of the light-duty fleet in London. In London, heavy-duty vehicles can move within the city at all times of day and night, whereas in Beijing the heavy-duty vehicles are restricted to the nighttime hours. The early afternoon peak in Beijing is almost certainly the result of new particle formation through regional nucleation processes referred to later in this article [6,7]. The diurnal variation in Delhi is suggestive of a major influence of road traffic especially at nighttime when the heavy-duty vehicles enter the city. There is, however, also a strong diurnal variation of atmospheric boundary layer mixing depths in Delhi with a much shallower mixed layer at nighttime which is no doubt a contributor to the very high concentrations seen at this time of the day. Figure 1. Average particle number size distributions measured in London (North Kensington and Marylebone Road), Beijing and Delhi. The shading represents one standard deviation. (Online version in colour.)
Figure 2. Average diurnal variation of particle number counts at the sites in London (North Kensington and Marylebone Road), Beijing and Delhi. The error bars represent one standard deviation. (Online version in colour.)
Particles sampled over open ocean areas show a very major contribution from sea salt as well as oxidation products of trace gases such as dimethylsulfide which are released from the oceans. However, over land there tends to be a rather different composition dominated by primary and secondary components deriving from anthropogenic emissions [8,9]. Table 1 indicates some of the major categories of particles together with their main chemical components and predominant sources. Airborne PM is hugely diverse so this list is by no means exhaustive and there are many trace element components which are not listed in the table. Road traffic exhaust has for many years been a major contributor, although this is now declining in most of the developed world [10]. Road traffic exhaust often dominates the particle number concentrations and figure 1 shows a clear difference between the roadside location at Marylebone Road and the background North Kensington which is attributable almost wholly to emissions from traffic on Marylebone Road. The particles typically arise predominantly from older diesel vehicles, although with advances in technology, the contributions from gasoline are becoming more notable in cities like London, and in North America gasoline is dominant [11]. The vehicle exhaust particles are comprised very largely of elemental carbon, referred to as black carbon and organic compounds, which derive both from unburnt fuel and from lubricating oil vapourized within the engine [12]. In London, diesel exhaust is the main source of elemental carbon, while in Beijing coal burning is a major source of particles comprised of elemental carbon and organic compounds [13]. Sea salt, with major components sodium, magnesium and chloride, makes major contributions at coastal locations but can still be seen hundreds of kilometres inland. Chloride ion also arises from neutralization of HCl vapour generated in the combustion of coal and some plastics, and makes a notable contribution in Delhi [14]. The ammonium ion derives from ammonia gas which comes largely from agriculture and is the main neutralizing species for HCl, and for nitric and sulfuric acids which arise, respectively, from the oxidation of nitrogen dioxide and sulfur dioxide, with the production of ammonium nitrate and ammonium sulfate which are major constituents of airborne particles. There are numerous sources which emit organic matter, most notable among them are biomass burning, both of wood and crop residues, and cooking [15]. Oxidation of organic vapours in the atmosphere leads to the formation of secondary organic matter which can be difficult to differentiate from primary emissions but appears to contribute a large proportion of atmospheric organic matter, amounting to 19% in a London winter campaign [16], and 16–65% in datasets from China [15]. Within this, there is a large secondary contribution from the oxidation of biogenic volatile organic compounds emitted by terrestrial vegetation, especially in summer when emissions of BVOC are highest. Differentiation of the contributions of anthropogenic and biogenic precursor VOC is challenging, but estimates of the annual average contribution of biogenic sources to SOA in China are 35%, and for southern China, 65–85% [15]. Changes in NOx emissions will impact upon SOA formation, as SOA yields are higher as NOx concentrations decline [17] which needs to be accounted for in developing policy.
The atmosphere also contains wind-blown dusts and soils which tend to reflect local geological conditions, with major components of silicon, aluminium and calcium typically, although these dusts can become contaminated with trace metals, for example as in road dust. As can be seen from figure 3, the major component composition of PM2.5 particles in London, Beijing and Delhi is not greatly different. There is far greater difference in the average mass concentrations, all of which were measured in winter sampling campaigns in these cities and do not represent the annual means which are somewhat lower. These pie charts clearly illustrate the huge importance of the organic matter and ammonium nitrate and sulfate as typically dominant constituents. Figure 3. Major chemical component composition of PM2.5 collected during winter campaigns in London (North Kensington), Beijing and Delhi. (Online version in colour.)
Chemical composition in the form of that in figure 3 does not reveal very full information on the sources of airborne particles. However, if a larger number of chemical constituents are measured on a large number of individual air samples, receptor modelling methods can be used to infer the sources of particles [18]. The term receptor modelling refers to the use of air quality data to infer the sources responsible for measured pollution levels and is the complement of dispersion modelling and chemistry-transport modelling which take the known emissions and disperse and chemically react them in the atmosphere to predict airborne concentrations. Receptor modelling of airborne particles depends upon an assumption of mass conservation as in the equation below: Ci=∑ ifi,jgj, where Ci, airborne concentration of a component, i; fi,j, mass fraction of component i in particles from source, j; gj, mass of particles from source j in an air sample.There are two main approaches used for receptor modelling. Multivariate statistical methods such as Positive Matrix Factorization make no a priori assumptions about sources and give a quantitative identification of those constituents which covary in time generating the chemical profiles of source-related factors, which with suitable intuition and knowledge, can be used to infer the sources. The other approach of Chemical Mass Balance modelling approaches the problem from the other end. It uses chemical profiles of known sources as an input and fits the measured chemical data with the best linear combination of source profiles so as to explain the measured concentration of each chemical constituent. This can in theory be carried out on a single air sample, but in practice better results are obtained by the inclusion of multiple samples. An example of the application of the Chemical Mass Balance model to PM2.5 from the London North Kensington site [16] is shown in figure 4. A notable weakness of all receptor modelling methods is the dependence upon the assumption that the chemical profiles of sources are conserved between source and receptor points. In practice, such profiles are subject to chemical change, although this is less of an issue for urban samples collected close to sources than for more remote sites. Figure 4. Source contributions to PM2.5 at North Kensington (%) derived from the application of a Chemical Mass Balance model (data from [16]). (Online version in colour.)
In less developed countries, airborne particles tend to be dominated by very familiar sources. Older and poorly maintained road vehicles are a major source of exhaust emissions and uncontrolled or poorly controlled combustion of fossil and biomass fuels as well as open burning of refuse are typically major contributors, both of primary particles and precursor gases such as HCl and VOC. Many less developed countries are also in drier parts of the world and consequently wind-blown dusts often of largely natural origin can make major contributions to PM concentrations [19]. However, in the more developed world, such sources are generally under far better control, and as they have become more controlled, so other less familiar sources are beginning to emerge as important. Four such sources will now be considered. Road vehicles marketed in Europe and many other parts of the world have to meet emissions standards referred to as the Euro standards, or national equivalents of the same. Since the introduction of the Euro 5 standard for light-duty vehicles and Euro 6 for heavy-duty vehicles, the requirements for very low particle number emissions can only be met by the fitting of diesel particle filters. More recent regulations are also requiring the use of gasoline particle filters on gasoline direct injection engines. As a consequence, there has been a marked decline in vehicle-emitted particles from road traffic both in terms of number and mass [10]. In such a situation, non-exhaust emissions become a bigger proportion of the total emissions from road traffic, and according to estimates from the UK National Atmospheric Emissions Inventory, the non-exhaust emissions now well exceed the exhaust emissions both in the case of PM10 and of PM2.5 [20]. Non-exhaust emissions are made up predominantly from abrasion particles of brake dust deriving both from the attrition of the disc and the pad, tyre dust, road surface abrasion particles, and also resuspended road surface dusts which are not currently included in the UK inventory [21,22]. The latter arise predominantly as a result of shear forces at the road surface created by wheels passing over the road and also resuspension due to turbulence occurring in the wake of the passing vehicle. A recent study from Delhi attributes a very large proportion (70%) of road traffic emissions of PM10 to particle resuspension [23] although the algorithm used to estimate resuspension is controversial [24], and the contribution to PM2.5 mass is probably very much smaller. Currently, none of these particle sources is subject to legislative control, but there is a good deal of research on the control of brake wear particles which is probably the most tractable of the emission problems. The regular cleaning of road surfaces can reduce the particle resuspension problem but has a very limited time duration of effect, and there are also dust suppressant materials which can be sprayed onto the road surface to limit the ability of particles to enter the atmosphere through resuspension [25]. Major advances in understanding the sources of airborne particles have arisen as a result of the high time resolution data generated by aerosol mass spectrometers (AMS) [26]. The AMS measures the mass spectral properties of non-refractory sub-micrometre airborne particles sampled directly from the atmosphere. Application of Positive Matrix Factorization to the mass spectral data allows the identification of source-related components of the particles and has revealed a component with a chemical signature quite close to that of cooking oils and a diurnal variation with a small peak around midday and a large evening peak [27]. This is attributed to cooking organic aerosol which has also been quantified through chemical mass balance modelling using chemical tracers for cooking [16]. There is evidence that many past studies using the AMS method may have over-estimated airborne concentrations of cooking organic aerosol [28], but even allowing for this over-estimation, concentrations in the atmosphere are appreciable, and work in the USA has shown highly elevated concentrations within the vicinity of major commercial restaurants [29]. Some western countries saw an increase in the burning of solid fuels and especially biomass as a result of fuel poverty caused by the economic recession. However, in the UK, an increased use of biomass fuels, and especially wood, has resulted from a fashion for installing wood-burning stoves or using open fireplaces within homes. This appears to be associated more with the aesthetic pleasures of a fire rather than as a primary means of heating. According to the UK National Atmospheric Emissions Inventory, this has caused a marked upward trend in emissions from domestic wood combustion especially when expressed as a percentage of the total primary emissions of PM2.5. The estimate for 2012 is that biomass sources, of which domestic wood burning is the largest, contributed 25% of total primary PM2.5 emissions [30]. The UK Air Quality Expert Group, a government advisory committee, compared emissions of PM2.5 from woodstoves operating at the limits set by the Clean Air Act and the EU Eco-Design Directive with emissions from diesel vehicles running at their upper limit, and the emissions from a single woodstove far exceed those from a modern diesel passenger car or heavy goods vehicle [30]. This appears to be a widespread problem affecting all areas, even including cities where clean air legislation attempts to limit the use of fuels such as wood. In the UK, the perception that biomass fuels are renewable has led to their incentivisation through the Renewable Heat Incentive, which has thus far influenced mainly the installation of larger combustion plant rather than domestic stoves, but seems likely to impact adversely upon local air quality. In China and India, the burning of crop residues can be a very major source of PM [31] which is subject to regional transport, affecting cities at some distance from the location of the combustion. This source has been very much reduced in the Beijing area as part of the 5 year Clean Air Action Plan which has led to significant improvements in urban air quality [32]. A further point noted by AQEG [30] is the large semi-volatile organic content of particles from sources such as wood burning, which are often not adequately accounted for in emissions inventories due to poorly designed sampling protocols. These contribute both to the mass of primary particles and to subsequent secondary particle formation [33]. Emissions of sulfur dioxide have decreased hugely in western countries over the past decades and airborne concentrations of sulfur dioxide have reduced in a proportionate manner. On the contrary, concentrations of sulfate appear to be nonlinearly related to SO2 emissions and have not fallen proportionately in Europe [34], or North America [35]. Oxides of nitrogen emissions have also reduced but far less than those of sulfur dioxide largely because of poor controls applied to the road vehicle fleet which have only recently started to impact on NOx emissions. Nitrate has not shown a commensurate reduction [35]. Consequently, secondary nitrates typically represent the largest single component of PM2.5 in countries such as the UK [36]. A factor in the resistance of ammonium nitrate concentrations to respond to mitigation measures is the fact that emissions of ammonia, largely from agriculture, have been reduced little if at all in recent decades and high ammonia concentrations favour the formation of ammonium nitrate particles which otherwise would be liable to dissociate into ammonia and nitric acid gases which would be subject to more efficient dry deposition processes [37]. While pollution by nitrates and sulfates has been well understood for many years, it still presents significant difficulties in chemistry-transport models largely because of a proliferation of mechanisms which are quite hard to differentiate using atmospheric measurements. However, among emerging pollutants, secondary organic aerosol (SOA) is now receiving significant attention. There are no wholly reliable methods of differentiating between primary and secondary organic aerosol, but current estimates suggest that secondary aerosol comprises a major proportion of organic particles. There has been even more difficulty in linking back secondary organic aerosol to specific chemical precursors, although there is fairly good knowledge of the contribution of biogenic precursors such as isoprene and α-pinene as well as anthropogenic emissions of compounds such as toluene to the production of secondary organic particles [16,38]. There have been major reductions in VOC emissions in the UK over past decades but it is unclear whether these have been reflected in a reduction of anthropogenic secondary aerosol. The biogenic precursors are known to make a significant contribution and there seems little prospect of these reducing in the near future. It appears that non-traffic related VOC arising from domestic emissions of solvents and personal care products now contribute substantially to secondary organic aerosol [39]. As mentioned earlier, ultrafine are particles usually defined as those with one dimension less than 100 nm. They can be measured by mass and are referred to as PM0.1, but this is technically quite difficult. They are far more often measured by number, and since ultrafine particles dominate the number count in most atmospheres (figure 1), the total particle number count is typically used as a surrogate for the ultrafine particle concentration. Emissions inventories have been constructed both in terms of mass and of number. The UK atmospheric emissions inventory uses a simple method to calculate a mass-based inventory by taking an inventory of particles in a larger size range such as PM2.5 and using an estimated percentage from each source sector to estimate the PM0.1 emissions. On the other hand, TNO in the Netherlands has generated an inventory of ultrafine particle emissions within Europe based upon particle number [40]. Within this inventory, the transport sector contributions about 75% of total particle number emissions with international shipping and diesel road traffic making by far the major contributions. As a result, a map of emissions serves to highlight the major shipping and road traffic routes within Europe. Projections for future years from a 2005 baseline show major reductions for 2020, mainly delivered by reductions in emissions from the transport sector [40]. This is due to the fact that particle number emissions from combustion sources are highly sensitive to the sulfur content of the fuel, and motor fuels have steadily reduced their sulfur content, now less than 10 parts per million, and shipping fuels are progressively reducing sulfur but not yet to the same degree. A notable exception is emissions from jet aircraft for which fuels still contain several hundred parts per million of sulfur, and emissions from major airports are detectable in the atmosphere in a number of major European cities at a considerable distance from the main point of emission [41,42]. Ultrafine particles from road traffic comprise both a soot mode of primarily graphitic carbon with a lesser amount of associated organic matter, and a nucleation mode which is primarily semi-volatile organic compounds condensed on the surface of a very small nucleus of sulfuric acid or inorganic ash derived from engine emissions [5]. The nucleated component has tended in the past to dominate particle number emissions, but the reduction in the sulfur content of road vehicle fuels, which took place in late 2007 in the UK and at similar times elsewhere in Europe, led to a very major reduction in particle number concentrations at roadside locations such as Marylebone Road [43]. In addition to those particles arising from combustion processes, ultrafine particles also arise in the atmosphere from homogeneous nucleation processes. These generally involve the initial formation of sulfuric acid vapour which condenses along with ammonia, amines and water to form new particles which subsequently grow by condensation of organic matter [7]. Mechanisms have also been demonstrated in which both biogenic [44] and anthropogenic hydrocarbons [45] are oxidized to form highly oxygenated molecules (HOMs) which can condense either alone, or in combination with sulfuric acid to form new particles. Such particles can appear over quite large geographical regions simultaneously and are hence often referred to as arising from regional nucleation. The initial formation of sulfuric acid vapour or HOMs depends upon photochemistry, and an analysis of average diurnal profiles of particle number count and of black carbon used as a sensitive tracer for diesel emissions shows marked differences between northern and southern Europe [46]. At sites in northern Europe, there is generally a strong correlation between the diurnal variation of particle number and of black carbon suggesting diesel emissions as being the major source of both constituents. In southern Europe, however, there is typically a further large peak in particle number count in the middle of the day which is not reflected in the black carbon data. This is the result of regional photochemical nucleation and can make a major contribution to airborne concentrations of ultrafine particles [47]. Much of the interest in ultrafine particles arises from the suggestion that they may have enhanced toxicity per unit mass compared to larger particle size fractions [48]. Currently, however, evidence on health impacts is sparse and lacks overall consistency [49], and the relative health impacts of primary nanoparticles from road traffic as compared to secondary particles from regional nucleation is not well understood. Road traffic has traditionally been considered as the main culprit for high concentrations of PM2.5. However, in cities with modern and well-maintained vehicle fleets, the contribution of vehicle exhaust to PM2.5 concentrations is decreasing rapidly due to the use of particle traps on modern vehicles. Although there are minor emissions during trap regeneration, these traps are almost 100% efficient in removing particles from engine exhaust. Consequently, the non-exhaust particles which now dominate the emissions are becoming a much greater concern and centre of attention. Other sources outlined above such as non-exhaust particles from road traffic, domestic wood burning, cooking and secondary organic aerosol are now seen as making important contributions to airborne concentrations of PM2.5. Sulfate concentrations have fallen in major western countries and are likely to continue to fall unless there is a reversal of emissions controls on sulfur dioxide, but future declines will be hard won due to the nonlinearity of the relationship between sulfur dioxide and sulfate concentrations. Airborne nitrate is currently the largest single component of PM2.5 in many developed countries and work in the UK has shown that its relative contribution increases during air pollution episodes, in this case, represented by 24-h periods where PM10 concentration exceeded the EU daily Limit Value of 50 µg m−3. Because of the complex formation mechanisms of nitrate, the nonlinearity between NOx emissions and nitrogen dioxide concentrations and the impacts of ammonia upon ammonium nitrate formation, the reductions in NOx emissions which are largely due to better controls on road traffic are most unlikely to have a substantial impact on atmospheric nitrate levels which will be far harder to control. The key may well lie in reductions of ammonia emissions, but historically these have been subject to lesser control than the other primary pollutants and there need to be major changes in policy if ammonia concentrations are to reduce to a meaningful degree. Secondary organic aerosol has a large contribution from biogenic precursors which will not change rapidly over time and will only reduce if there is clear attention to possibilities favouring low emitting species of shrubs and trees over the higher emitting species. The links between anthropogenic VOCs and SOA are in general far less clear, and SOA seems to have responded in a relatively minor way to major reductions in VOC emissions which have occurred in the UK. This suggests SOA is unlikely to reduce rapidly in the future. As a consequence of these various changes, there is a huge challenge for countries such as the UK to meet the current WHO air quality guideline for PM2.5 of 10 µg m−3 as an annual mean. Data supporting this publication are openly available from the UBIRA eData repository at https://doi.org/10.25500/edata.bham.00000435. R.M.H. conceived the study, designed the study, coordinated the study and prepared the first draft of the manuscript. I declare I have no competing interests. This research is supported by the Natural Environment Research Council through the AIRPOLL-Beijing project within the APHH Programme (NE/N007190/1), the Clean Air for London (ClearfLo) project (NE/H003142/1) and the NERC-MRC Air Pollution and Human Health Programme (NE/P016499/1). The author expresses thanks to colleagues who contributed to the collection of the data from Beijing (James Brean, David Beddows, Tuan Vu, Zongbo Shi) and Delhi (William Bloss, Salim Alam, Leigh Crilley) and others, and to Ulku Alver Sahin for data processing. FootnotesOne contribution of 17 to a discussion meeting issue ‘Air quality, past present and future’. †Also at: Department of Environmental Sciences/Center of Excellence in Environmental Studies, King Abdulaziz University, PO Box 80203, Jeddah 21589, Saudi Arabia. © 2020 The Authors. Published by the Royal Society under the terms of the Creative Commons Attribution License http://creativecommons.org/licenses/by/4.0/, which permits unrestricted use, provided the original author and source are credited. References
Page 5Air pollution is considered a key environmental threat to human health. The World Health Organisation (WHO) attributes some 7 million premature deaths to ambient and indoor air pollution annually, with many of these occurring in the urban centres of the developing world [1,2]. Few countries achieve WHO air quality guideline values, recommending, e.g. PM2.5 concentrations <10 µg m−3 and NO2 <40 µg m−3 at the annual mean, and instead, interim air quality policy targets have been adopted that provide a realistic medium-term ambition and promote progress towards cleaner air. For example, as emissions have decreased across Europe, successive editions of the European Air Quality Directive [3] and national implementations have set increasingly stringent Air Quality (AQ) objectives that have continued to drive improvements in air quality and motivated the implementation of increasingly stricter emission standards to achieve these objectives. However, despite air quality improving in many developed countries in the past decades [4], both developing and developed countries still struggle to adhere to the AQ objectives whether set by themselves or in international negotiations, based on the WHO recommendations. Over the past years, there have been setbacks in the attempts to reduce emissions. A prominent example is the failure of many diesel vehicles complying with EURO5 and early EURO6 European emission standards to achieve the required reduction in NOx emissions under real-world driving conditions [5]. In other cases, especially in the developing world, gains of increased emission control have been offset by increased activity, e.g. car ownership and annual distance travelled (e.g. [6]). As legal AQ targets are being missed through emission control, national, regional and local governments are forced to look for alternative interventions for reducing air concentrations and human exposure. This includes the introduction of low emission zones and increasingly also the targeted use of vegetation to lower concentration levels. In the UK, several local authorities have embarked on efforts to increase urban forest cover in recent years, with air pollution reduction being one of the motivators. At the same time, there is an increasing interest in valuing the air pollution removal by vegetation [7,8] as part of the overall ecosystem service it provides and to include this valuation in national accounts. In this paper, we investigate the effectiveness of existing vegetation and potential green infrastructure interventions in relation to their spatial scale and critically assess the cost/benefit of micro-interventions by considering the basic physical constraints on pollutant uptake. The paper quantifies the effect of UK-wide vegetation on lowering air pollution levels and assesses the current role of UK urban vegetation as a whole. This work was done as part of the UK national accounts compiled by the UK Office for National Statistics [9]. The paper continues, by looking at the potential effects of large-scale city-wide interventions, converting significant fractions (25 and 50%) of the available urban open greenspace into urban woodland, again for the UK. It then scales down the results to discuss the potential of smaller green infrastructure interventions and the effect they could possibly have on air pollution levels, with focus on the urban environment. Several approaches have been used in the past to quantify pollutant removal by extensive vegetation areas and to value this ecosystem service. Most of these value the benefit of pollutant removal via the equivalent emission damage costs, i.e. the same damage cost that governments ascribe to emissions of pollutants can be used to valued air pollution removal (as a negative emission). These emission damage costs have been tabulated from offline calculations using atmospheric chemistry and transport models (ACTMs) to assess the consequences of the emissions in terms of population exposure and outcomes [10]. They often vary by degree of urbanization reflecting the proximity and size of exposed population. The pollutant removal is then quantified with models such as the Urban Forest Effects Model (UFORE) [11] and its descendant i-Tree Eco [12], which is based on modelling the uptake of individual trees. While computationally simple, there are several disadvantages to this method: (i) the approach is indirect in its evaluation as it relies on modelling of emission–exposure relationships to quantify the underlying emission damage costs in the first place, (ii) it is not mass conserved in the sense that it is driven with prescribed concentrations with no feedback of the deposition on concentrations, and (iii) it, therefore, does not take other interactions and processes into account that respond to concentration changes such as atmospheric chemistry and wet deposition. By contrast, one advantage is that the deposition-based approach allows for a detailed description of the vegetation, in the case of i-Tree Eco the trees or tree canopy. Full application of a three-dimensional ACTM overcomes the shortcomings described above, but in turn tends to rely on a simplified representation of the vegetation during the analysis and its spatial resolution is limited to the model grid size, in this case approximately 5 × 6 km2. An ACTM tracks the pollutant emissions, their chemical transformations and deposition in a three-dimensional meteorological (flow) field. Rather than prescribing the pollutant concentrations, they are predicted by the model as influenced by the dry deposition to the underlying landcover in the model. With this modelling framework, the effect of vegetation can be quantified by comparing two model runs with modified landcover description. This approach not only quantifies the additional dry deposition induced by the vegetation, but also directly the influence on the concentration. From this concentration field, human exposure and health impacts can be derived more directly than via the emission damage costs. In addition, this approach accounts for the full range of chemical and spatial interaction. For example, the enhanced removal of PM precursor gases (NH3, SO2, NO2, VOCs) and also oxidants (e.g. O3) by vegetation reduces their contribution to secondary PM formation and thus reduces PM2.5 through a route that is additional to enhanced PM2.5 dry deposition. A crucial additional step for this approach is the decision on the baseline scenario: removal-based approaches such as iTrees Eco normally derive the benefit from the quantification of the dry deposition to the vegetation, without consideration that some, although reduced, dry deposition would also occur to a non-vegetated surface. In the ACTM approach, the non-vegetated landcover needs to be prescribed explicitly and the choice affects the results. This paper uses this ACTM-based approach to evaluate the effect of vegetation and in doing so also quantifies explicitly some of the interactions in the model that would be missed with the deposition-based approach. In this study, an implementation of the European EMEP Eulerian ACTM [13] was used with a nested higher resolution UK domain at a scale of approximately 5 × 6 km2 (EMEP4UK) [14–17]. The current baseline reference (UKBASE) UK landcover definition was derived by remapping the UKCEH Landcover Map 2007 (LCM2007) [18] to the seven existing landcover classes of the EMEP model (deciduous forest, coniferous forest, crops, semi-natural land, water, desert and urban). For the urban scenario runs, LCM2015 was used and three new landcover classes (urban forest, urban open greenspace, urban water) were derived from the Ordnance Survey MasterMap for areas lying within the urban morphology layer of the UK as described in detail by Jones et al. [8]. Briefly, first land was classified as ‘urban’ based on the existing urban morphology layer from the UK Office for National Statistics, supplemented by a variable buffer. In a second step, this urban layer was intersected with the OSMasterMap ‘natural surface’ category, and of the resulting urban vegetation map all OSMasterMap objects with the term ‘trees’ or ‘woodland’ in the main descriptor were assigned to the ‘urban forest’ class and the remainder to the ‘urban open greenspace’ class, which thus contains e.g. open parkland, gardens and playing fields. The EMEP4UK model uses a tiled approach, in which the landcover fractions of the seven types are specified for each grid cell. The meteorological input was generated with the community weather research and forecasting (WRF) model v. 3.7.1 (www.wrf-model.org) which included data assimilation (Newtonian nudging) of the numerical weather prediction (NWP) model to the meteorological reanalysis from the US National Center for Environmental Protection (NCEP)/National Center for Atmospheric Research (NCAR) Global Forecast System (GFS) at 1° resolution [19] at 6-hourly intervals. The performance of this model combination (WRF-EMEP4UK) for the UK has been thoroughly established elsewhere [14–17]. Alternative landcover maps were created: in the first, no-vegetation (NoVEG) scenario, all vegetation landcover (i.e. deciduous, coniferous, crops and semi-natural; urban and non-urban) within the UK was replaced by desert, representing bare soil. In addition, a no-urban-vegetation (NoUrbanVEG) landcover map was created by replacing only urban vegetation (urban forest and open urban greenspace) with desert, and two tree planting scenarios were created by converting 25 and 50% of the open urban greenspace category into urban forest, referred to as 25OGSC and 50OGSC (open green space conversion), respectively. The total land area of the urban landcover classes in the various scenarios is summarized in table 1. Together, the urban landcover types represent 7.1% of the UK land area.
The effect of the choice of the baseline scenario/non-vegetated landcover has implications for the results. Theoretical choices could include sealed land (concrete/asphalt) or bare soil. In the EMEP modelling system, only two terrestrial non-vegetated terrestrial landcover types exist: urban (with the aerodynamic roughness of the built-up environment) and desert (with the properties of sand). Under the assumption that desert comes closest to bare soil in terms of aerodynamic properties and its affinity for pollutant uptake, but that its grain size distribution results in large resuspension, a decision was made to use desert as a reference, but to completely discount the contribution of desert dust to PM2.5 in all runs and/or to report it separately. It should be borne in mind, however, that some of the effect of increased resuspension is real. The model set-up and landcover scenario use is summarized in table 2. The UK vegetation simulations were carried out with EMEP4UK implementation of EMEP rv4.10, while the urban vegetation simulations were based on the more recent version EMEP rv4.17. Differences between the two versions are small, but one additional significant difference needs to be considered: in the UK vegetation simulations, soil NO emissions were calculated as a function of landcover and changed with vegetation cover, while in the urban vegetation simulations, NO emissions were prescribed and fixed between scenarios. Two different current-vegetation reference runs (UKBASE and UrbanBASE) were used to match the model set-ups of each set of simulations. All runs were performed with the same driving meteorology for the year in question, based on status quo landcover, i.e. WRF was not rerun for the various landcover scenarios, the rationale being that (i) the impact of landcover on the meteorology which drives advection (bottom layer is at 45 m) is a second-order effect and (ii) the modelled meteorology would then become incompatible with the observation-derived dataset against which it is constrained. In the tiled approach deployed for deposition in the EMEP model, the windspeed at a reference height is then extrapolated individually to the different landcover types, depending on landcover-specific roughness height and heat flux, weighted by the landcover scenarios. The effect of current UK total vegetation cover on annual total pollutant deposition, annual average concentration fields and annual average human exposure was then derived by comparing an annual model run using the UKBASE or UrbanBASE landcover to a run based on the modified landcover scenarios. Effects were assessed for total PM2.5, SO2, NO2, O3 and NH3. UK national runs were performed for a range of meteorological years (2007, 2011 and 2015). For these baseline, gridded emissions of the UK National Atmospheric Emissions Inventory (NAEI) for 2014 were used for the 2015 run, and emissions for 2007 and 2011 were created by rescaling the 2014 gridded emissions according to the NAEI totals. All these UK runs were based on 2007 landcover. The urban runs were based on 2015 meteorology and the NAEI gridded emissions for 2015. Here, landcover was based on 2015.
The total changes in PM concentrations are summarized for 2015 in table 3, with other years (2007 and 2011) shown in electronic supplementary material, table S1. Concentration and changes in concentrations were averaged both over the entire UK and over the predominantly urban grid cells (see the next section for details). In these summaries, the effect on the PM contribution from wind-blown dust has been listed individually. As mentioned above, due to the choice of desert as the reference surface, vegetation not only has the effect of capturing PM and its precursors, but also suppresses desert sand resuspension. The two effects are similar in magnitude for PM2.5 and the dust suppression is the larger effect for PM10, bearing in mind that resuspension shows large inter-annual variability, presumably linked to the statistics of extreme wind speeds and rainfall. Although vegetation does suppress resuspension from bare soil, desert sand is likely a poor proxy for UK soils when it comes to resuspension and the effect is likely overestimated by the model. For the non-dust PM components, vegetation lowers concentrations by about 10% for PM2.5 and 6% for PM10 as a UK spatial average, with the values being slightly smaller if averaged by population density. The benefit is predicted to have decreased with time both in absolute and relative terms (electronic supplementary material, table S1). Possible reasons are changes in the contribution of UK versus non-UK or primary versus secondary sources as emissions evolve over time. Biogenic secondary organic aerosol (BSOA) is the only component of the PM that is increased through the presence of vegetation. As a UK spatial average, about 55% of the BSOA is due to BVOC emissions from UK vegetation, the remainder presumably originating from BVOC sources outside the UK as the ACTM does not ascribe any BVOC emission to desert soil. As explained in §2a, values of PM2.5 and PM10 provided throughout the remainder of this paper do not include the wind-blown dust component.
The estimates of the change in dry deposition and surface concentration caused by the UK's total vegetation are shown in figure 1 for 2015 for PM2.5 (i.e. now without wind-blown dust) as an example. Urban areas show up in the deposition field of the no-vegetation run (figure 1a) not only because here PM2.5 concentrations are locally elevated, but also because urban buildings are aerodynamically rough and capture PM2.5 more effectively than smooth desert soils. The relative change in dry deposition between the two scenarios is particularly pronounced in the areas with woodland vegetation and elevated wind speeds because the dry deposition velocity (Vd) of PM scales with wind speed in addition to being enhanced to forest. Figure 1. Model simulations for PM2.5 for 2015, showing (a) the annual total PM2.5 dry deposition to a vegetation-less UK (NoVEG) (mg m−2), together with the (b) absolute (mg m−2) and (c) relative (%) changes in deposition caused by the vegetation (UKBASE-NoVEG), with positive (red) values indicating an increase in deposition to vegetation compared with no vegetation. (d) The annual average PM2.5 surface concentration for a vegetation-less UK (µg m−3), together with its (e) absolute and (f) relative change due to vegetation (UKBASE-NoVEG), with negative (blue) values indicating a decrease in concentration above vegetation compared with no vegetation. The dust component is not included in these figures (see text).
The effect of vegetation on the concentrations of the gaseous pollutants is summarized in table 4 for 2015 with additional years in electronic supplementary material, table S2. Spatially averaged relative reductions are largest for SO2 (30% overall), followed by NH3 (25%) and O3 (13%), with somewhat smaller reductions if weighted by population and for the urban grid cells. Surprisingly, the UK average effect is very small for NO2; the spatial pattern of figure 2 reveals that vegetation suppresses NO2 in source areas which is approximately balanced by increases in NO2 in rural areas, especially in areas with large forest cover. This is partly due to the fact that forest soils are associated with larger NO emissions than desert sand, but other interactions also contribute as also found in the urban landcover scenarios described below, in which NO emissions do not respond to landcover change and the reason is explored in §3c(iii). Figure 2. Model simulations for NO2 for 2015, showing (a) the annual average NO2 surface concentration for a vegetation-less UK (µg m−3), together with its (b) absolute (µg m−3) and (c) relative (%) change due to UK vegetation (UKBASE-NoVEG), with red (blue) values indicating an increase (decrease) in concentration above vegetation compared with no vegetation.
Table 5 summarizes the effect of all current (2015) urban vegetation on annual average UK concentrations and the additional effect large-scale conversion from open urban greenspace to urban woodland could have. Also shown are the reductions averaged over all grid cells that are dominated by urban landcover, where most of the effect is expected and at which such intervention would be targeted. These ‘urban’ concentrations were very similar to UK population-weighted and urban population-weighted concentration changes which were explored as a further metric (electronic supplementary material, table S3). For PM2.5, the current urban vegetation is calculated to be responsible for reductions in PM2.5 of −0.95%, compared with −0.84 and −0.93%, depending on the averaging method. Overall, the model results indicate that current urban vegetation (5100 km2) reduces average urban PM2.5 concentration by about 1%, with a similar additional reduction (0.8%) expected from large-scale conversion of open urban greenspaces to additional woodland. The effect of existing vegetation is somewhat larger for the PM precursors SO2 and NH3, but the conversion to woodland has a comparably smaller effect than on PM because the deposition rate of the gases is less sensitive to the vegetation type. Interestingly, for both O3 and NO2, the current urban vegetation cover decreases concentrations as may be expected due to the additional dry deposition, but the additional conversion of open urban greenspace to woodland is forecast to increase concentrations in urban areas, although slight reductions are seen away from sources (figure 3). This would imply that tree planting, aimed at reducing NO2 emissions, may in fact increase urban NO2 concentrations through processes that are explored in §3c(iii). Figure 3. Maps for NO2 for the urban vegetation runs for 2015, showing the absolute (a) and relative (d) change in NO2 caused by current urban vegetation (UrbanBASE–NoUrbanVEG), as well as the associated results for the 25% (b,e) and 50% (c,f) urban woodland conversions (e.g. 25OGSC–UrbanBASE). The blue values in the current urban vegetation run (a,d) show the decrease in the concentrations caused by the present vegetation (relative to no vegetation). The red (blue) values in the tree planting scenario runs (b, c, e and f) indicate an increase (decrease) in the concentration due to the additional urban vegetation (relative to current urban vegetation).
The modelling of the pollutant capture by vegetation through the ACTM approach allows some interesting interactions to be identified and quantified that may not be entirely obvious initially. The reduction in pollutant concentration due to the enhanced dry deposition to vegetation results in a reduction in wet deposition, which to some extent counteracts the benefit (table 6). The relative effect is largest for compounds that are primarily deposited via wet deposition such as PM. In the national vegetation scenarios, for example, an increase in PM2.5 dry deposition of 19.9 kt yr−1 is accompanied by a decrease in wet deposition by 10.4 kt yr−1. Thus, the net removal is actually only half of what would be estimated by considering dry deposition in isolation, e.g. through the i-Tree approach. For PM10, the net removal is only two-thirds of what would be expected. The interaction with wet deposition is to some extent accounted for in the calculation of the Emission Damage Costs that underpin the deposition-based evaluation. This is derived by modelling the increase in concentration and exposure for changes in concentrations and wet deposition is presumably considered in this evaluation. However, precipitation varies between years both in magnitude and spatial distribution, and the importance of the interaction, therefore, highlights an important uncertainty in the deposition-based approach.
An increase in vegetation cover is associated with an increase in emissions of biogenic volatile organic compounds (BVOC) which themselves are a precursor for ozone and PM2.5 formation. This is a recognized negative side effect of tree planting [20–22]. However, vegetated soils also tend to emit more NO than bare soil and this contributes to the net effect of vegetation on NO2 concentrations being negligible. In a world devoid of vegetation, soil resuspension would be increased. This effect adds to the benefits of vegetation for PM2.5 and PM10. Chemical interactions occur at various places in the modelled atmospheric system. Changes in the chemical pollution climate affect the dry deposition rates of pollutants. This tends to be only crudely represented by current deposition modelling approaches. The dry deposition of SO2 and NH3 in the EMEP model depends on the annual average NH3/SO2 ratio [23,24], i.e. in highly alkaline environments, SO2 deposition is enhanced and NH3 deposition is limited and vice versa [13]. Thus, chemical interactions not only occur in the atmosphere, but also on leaf surfaces. Another route of chemical interactions is through air chemistry. For example, because vegetation enhances the terrestrial sink for all pollutants (except BVOCs and soil NO), many secondary pollutants tend to be decreased not only through their increased removal, but also due to the increased removal of their chemical precursors, exacerbating the vegetation's overall effect. The response of NO2 to vegetation changes seen in the simulations is less intuitive and requires further consideration: UK vegetation decreases NO2 in urban areas, but increases it in more remote regions. For the urban runs, current urban vegetation again decreases NO2 in urban areas, but additional conversion to urban forest is simulated to increase NO2. As mentioned before, at the UK scale, conversion to forest in the EMEP4UK model was configured to increase the soil NO emission in the model, and, although this increased total NOx emission by only a small amount, this is a possible reason why NO2 increases in the rural environment by vegetation. However, in the urban runs, this increase in soil NO emission was deactivated and still additional tree planting increased NO2. In addition to sources and deposition sinks, NO2 concentrations depend on the photo-stationary equilibrium between NO, O3 and NO2. For the urban scenarios, current total urban vegetation has the effect of reducing O3, while the conversion of open urban greenspace to urban woodland is simulated to increase O3 everywhere in the UK (electronic supplementary material, figure S1; table 5). The reason is found in the relative magnitudes of the increased terrestrial sink compared with the increase in O3 production due to larger BVOC emissions from trees. Total urban vegetation includes trees that emit BVOCs, but the increase in the dry deposition sink has the larger impact on O3 concentrations. Additional conversion of open urban greenspace to urban forest increases O3 production (through increased BVOC emissions) more than it reduces O3 through more efficient dry deposition. The increased O3 concentration in the tree planting scenarios results in increased titration of NO as can be seen in reduced NO concentrations across the country (electronic supplementary material, figure S2) and this is partly responsible for the increase in the NO2 concentration. However, the maps of total NOx (=NO + NO2) (electronic supplementary material, figure S3) show almost exactly the same pattern as NO2, suggesting that the re-partitioning between NO and NO2 via O3 interaction cannot be the sole or even main reason for the increase in NO2. An additional two model scenario runs were performed to shed light on the underlying processes: these were a current vegetation base run and a 50% open urban greenspace conversion run in both of which BVOC emissions were switched off. The results show that without the change in BVOC emissions associated with tree planting, the increase in urban tree cover causes a very small (less than 0.3%) reduction in NOx concentrations in urban areas, which is due to the only small difference in deposition rates between grassland and forest in the model (electronic supplementary material, table S4). At the same time, the NO/NO2 partitioning changes slightly in response to changes in O3 deposition and concentration. Thus, it is the additional BVOC emission in the urban tree planting scenarios which causes the increase in NO2 in the model through two additive effects: firstly, the BVOC emissions increase NOx concentrations (presumably by competing for OH and reducing NOx oxidation to HNO3), and secondly, they also increase O3, which in turn modifies the NO/NO2 partitioning in favour of NO2. The scope for pollution removal by vegetation is of particular interest for PM, which tends to dominate the health impacts of air pollution, and for O3, which causes health impacts in particular in polluted, warmer regions. While the current health burden of NO2 is estimated to be lower, this compound is currently in the spotlight because many cities across Europe are out of compliance with target values. Several studies have attempted to quantify the removal of air pollution by vegetation and, in particular, urban trees and vegetation, including in the UK, and estimates differ widely. Here, we show that this is in large part a question of scale. The (UK) national vegetation as a whole reduces air pollution levels significantly, estimated here at 10% for PM2.5, 30% for SO2, 24% for NH3 and 13% for O3. Importantly, this includes the cumulative effect of pollutant removal during transport from source to receptor. For example, the concentration in London is lowered also by the uptake of some of the pollutants transported long range from continental Europe by vegetation encountered en route, e.g. in Kent. It also includes the secondary effect of reducing precursor gases involved in secondary PM and O3 formation. By contrast, the effect of localized vegetation on local concentrations is limited: the overall effect of urban vegetation on urban PM2.5 is a reduction of the order of 1% and a conversion of 50% of available open urban greenspace to urban forest would add a further reduction of similar magnitude. By comparison, applying a simpler model in which total PM had to be scaled up from the primary fraction to two example UK conurbations, McDonald et al. [25] estimated that a 50% planting scenario might lower PM10 deposition by 4% in the English Midlands and 0.7% in Glasgow, which compares to an average urban reduction of 0.6% derived here. In the earlier study, the urban cells were defined via the city boundary and in the Midlands scenario included agricultural areas situated between the cities of Birmingham and Coventry, with an increased potential for planting, and this explains the higher gain. Litschke & Kuttler [26] similarly estimated a typical PM10 reduction by 1% due to trees in urban areas, while Nowak [27] estimated that current vegetation removed less than 1% of PM10 and NO2 in Chicago and would only remove less than 5% if the city was fully covered by trees. The new study is also consistent with conclusions of the UK's Air Quality Expert Group that, while beneficial, urban tree planting is not a solution for reducing urban air pollution at the city scale [28]. Many alternative studies of valuing air pollutant removal by vegetation have been based on quantifying the benefit via the quantification of dry deposition rather than through modelling the effect on the concentration itself, including, for example, for the vegetation in London [29,30], but few studies appear to have taken the ACTM approach adopted here. The assessment of the human health benefit is not the primary topic of this paper, but these can be estimated using the results of an assessment based on the total urban vegetation by Jones et al. [8]. According to that study, existing UK urban woodland removes 0.7 kt PM2.5 yr−1, reducing the health burden from PM2.5 by about 1900 life years lost/year, with a similar gain achievable through the 50% planting action. The PM2.5 removal of a single mature tree would then equate to 1.7 life hours/year saved. Although the pollutant collection properties of small green infrastructure interventions depend on the exact set-up and location, the country-level model results enable some approximate downscaling to investigate the efficacy that may be expected. Commercial forests are typically planted at a density of 1000–2000 trees ha−1, but a fully mature woodland has a final density of closer to 100 trees ha−1 [31]. Tree surveys for London and other cities suggest that most urban trees have a diameter at breast height (DBH) of less than 20 cm [30]. At this DBH, one might expect a higher planting density, more like 400 trees ha−1, for a closed forest stand. The current UK urban woodland extent was estimated here to be 976 km2 and to take up 0.70 kt PM2.5 yr−1 (electronic supplementary material, table S5) [8]. Assuming full maturity, this equates to a removal of 71 g PM2.5 mature tree−1 yr−1, about 64 g yr−1 more than if the same area were covered by grassland. To put this into context, the UK National Atmospheric Emissions Inventory implies a total (i.e. exhaust plus non-exhaust) fleet average combined (all road types) emission factor for diesel cars of 0.023 g PM2.5 km−1 [32]. At a typical annual travel distance of about 15 000 km yr−1 [33], a car emits 350 g PM2.5 yr−1, probably somewhat more under urban driving conditions, similar to the uptake of five properly mature trees. A taxi on average covers almost three times this mileage (14 mature tree equivalents). For buses, the total fleet average emission is 0.094 g PM2.5 km−1. In London, 490 million km are covered annually by local buses with a total fleet of about 10 000 vehicles [34]. Thus, based on these average figures, each London bus travels on average about 50 000 km yr−1, thus generating an emission of 4.7 kg PM2.5 yr−1, the offsetting of which would require 73 mature trees, or about 3/4 ha of mature urban woodland. The emission from domestic solid fuel burning has recently come under increased scrutiny. In the UK, larger biomass boilers have a strict emission limit of 30 g PM/GJ in order to be eligible for government subsidy under the UK's Renewable Heat Incentive (RHI) [35]. At a domestic heat requirement of 5000–30 000 kWh (18–108 GJ), such appliances might emit up to 0.54–3.2 kg PM yr−1, mainly as PM2.5. Older appliances and those not covered by the RHI may emit significantly more [36]. Room heaters are generally less efficient: even the EC Ecodesign Directive (2009/125/EC) allows a maximum emission of 5 g PM kg−1 dry matter burnt for log stoves [37]. Assuming a room heater burns 1 tonne of dry wood a year, it would cause an emission of up to 5 kg PM yr−1, mainly in the form of PM2.5, similar to the average bus above, but emitted over a much smaller fraction of the year (and concentrated on individual days of the week and hours of the day). This emission is again equivalent to the PM removal of some 78 trees or 0.78 ha of woodland. Older appliances and open fireplaces not adhering to the Ecodesign standard are likely to emit significantly more, and so does the use of non-ideal fuels (e.g. wood with high moisture content) [38]. A number of solutions have been developed as green infrastructure solutions to reducing air pollution. These range from passive uptake such as green walls or green lampposts to active filtering devices such as the commercial CityTree [39,40]. According to the advertising material on the company website, the CityTree may be run with an airflow rate of 5.5 m3 min−1 and removes 15 ± 5, 23 or 26–64% of PM2.5, depending on the study. The absolute amount removed from the air scales with air concentration. In the UK, the modelled annual average urban concentration of PM2.5 was 8.8 µg m−3. Even at a significantly more polluted UK location with an average PM2.5 concentration of 25 µg m−3, the device would remove just over 20 g PM2.5 yr−1, based on an average efficiency value of 28%, about one-quarter of what is estimated here for a single mature tree. Of course, in a highly polluted location, e.g. in some Asian cities, the concentrations are much larger and the CityTree would remove much more material, but then so would a tree. Because it would be lacking the mechanical aspiration of the CityTree, a passively collecting green infrastructure element of similar size, such as a green wall or a green lamppost, would likely remove significantly fewer pollutants. Turbulent transport limitations would hamper the pollutant uptake. Green walls and roof gardens have been suggested as features for large-scale introduction of green infrastructure into urban areas. In China, first cities are planned as green cities from inception. The 50% urban woodland simulation has shown that conversion of 12% of the total urban area (table 1) from open urban greenspace into urban woodland would reduce urban emissions of PM2.5 by about 0.8% on average (table 5). Given that roads, car parks and other sealed areas that cannot be converted to green infrastructure make up a significant fraction of the urban space, even a large-scale implementation of green infrastructure would be unlikely to make a much larger impact than the 50% tree planting scenario and this assumes that the PM2.5 capture efficiency of the green infrastructure matches that of mature trees. It should be noted, however, that the pollutant removal by green infrastructure depends on the exact positioning of the infrastructure and that ACTM modelling cannot simulate the intricacies between individual surface elements and ground-level concentration. Using an atmospheric chemistry model expanded with an extension to simulate street canyon mixing and dry deposition, Pugh et al. [41] suggested that continuous green wall cover within street canyons in particular may be effective in reducing street-level concentrations of PM by up to 60% while highlighting the need to get the design right to avoid concentration increases. This is a very extreme value for 100% vegetation cover, assuming deep canyons, very low wind speeds (1 m s−1 at roof level) and relatively large deposition velocities that are likely to be inconsistent with these low wind speeds. Furthermore, this green infrastructure would ameliorate mainly the concentration build-up deriving from in-canyon sources rather than the contribution from background concentrations and will, therefore, mainly aid the reduction in roadside increments. Nevertheless, that study highlights that certain infrastructure options may offer opportunities to reduce local source contributions. Consistent with our study, Pugh et al. [41] found that green rooftops were relatively ineffective in reducing street-level concentrations. Overall, our findings mirror those of Vos et al. [42] who argue that localized urban vegetation cannot be expected to be a solution for alleviating local air pollution hot-spots. This study reveals important interactions and feedbacks that can be quantified by the full ACTM approach, but which are not taken into account when the pollutant removal benefit is quantified via quantifying the deposition via approaches based on prescribed concentrations that do not account for deposition-concentration feedbacks. The negative side effect of BVOC emissions through tree planting on air pollution has been widely recognized through their role in increasing O3 concentrations [20]. The present study highlights an additional, secondary effect on NO2 levels and shows that if tree species are not carefully selected, urban tree planting could even increase concentrations of NO2 and, therefore, one of the key pollutants they are often meant to reduce. This implies a greater than 100% error in static deposition-based approaches which would predict a benefit while the real net effect may actually be negative. Clearly, the exact magnitude of the chemical effect on urban NO2 depends on the additional BVOC emission associated with the landcover change. In the EMEP model, the additional urban tree cover is ascribed the same isoprene and monoterpene emission factors as is thought to be representative for the average existing UK vegetation. If trees were selected to minimize BVOC emissions (e.g. [20,21,43]), it may be possible to control the effect on O3 and NO2. Furthermore, the results highlight the magnitude of the counteracting effects on wet deposition. As increased dry deposition removal lowers air concentration, this in turn reduces wet deposition. This offsetting effect is here found to be large (50%) for slowly depositing compounds such as PM2.5, and introduces considerable uncertainty when quantifying the benefit via the deposition term and emission damage costs. Clearly, the relative magnitude depends on the amount of precipitation in the area and is likely to be larger in the UK than in some other parts of the world. This is an issue that does not appear to have so far been fully recognized in deposition-based approaches like i-Tree Eco. For local-scale interventions, the enhancement in dry deposition and reduction in wet deposition become spatially decoupled, adding to the uncertainty in assigning geographically varying damage costs to deposition increases. Overall, the scope of reducing pollution levels through urban vegetation is shown to be very limited. Nevertheless, it may be worthwhile considering and valuing this effect within the context of the multiple other (ecosystem) services provided by vegetation in general and urban vegetation in particular. The benefit of vegetation for air pollution removal, even if modest, might provide one of several elements for the decision to go ahead with green infrastructure measures. The overall benefits and disbenefits of tree planting have been summarized in several review papers [22,44,45]. For example, the i-Tree Eco model also ascribes a value to the role of urban trees for carbon sequestration, their assistance in water management by reducing storm-water run-off and their effect on heating/cooling [12,30]. This latter effect can be beneficial or detrimental: trees generally reduce the heat-island effect, which in summer potentially increases labour productivity and reduces adverse health outcomes (morbidity and mortality) and/or reduces air conditioning needs; but trees may increase heating bills in winter [46]. Similarly, at the local level, trees can shelter houses from wind or they can shade buildings with contrasting effects on energy requirements in winter. Green spaces reduce noise, promote physical activity and contribute significantly to the wellbeing of the society and provide long-term benefits to physical and mental health [47,48], which have also been speculated to improve the immune response to air pollution [48,49] and may, therefore, be beneficial in combating air pollution effects through a second route. Semi-natural ecosystems and especially also urban vegetation such as gardens support a large number of plant and animal species. However, the benefits of vegetation for biodiversity are more difficult to value economically as there is little consensus on the value society ascribes to biodiversity per se, beyond its role in ecosystem provisioning [50]. Where urban vegetation is targeted at reducing air pollution, negative side effects also need to be considered as well as the positives: Other pollutant emissions. Some pollutant emissions associated with vegetation (e.g. pollen) and its management are not considered in this study: wounding of leaves and branches due to grass cutting or pruning emits leaf alcohols and other VOCs with potentially large ozone and BSOA forming potential, even for plant species that are generally low VOC emitters [51]. Fertilizer application may result in increased emissions of ammonia (NH3) which is involved in secondary aerosol formation and has detrimental effects on biodiversity and plant species adapted to low-nitrogen growing conditions, nitric oxide (NO) as well as nitrous oxide (N2O), a potent greenhouse gas [52]. Potential application of fungicides and pesticides has its own environmental impacts and an increase in the emissions of allergens like pollen need to be considered when evaluating the overall health impact of vegetation for the population [53]. Safety. There are costs involved in the management of vegetation to reduce dangers to the public. Road safety can be a significant concern when it comes to green infrastructure: if poorly managed, leaf fall can render roads, pavements and paths slippery and increase the risk of road accidents and falls. Falling trees and branches can cause injury to passersby, property and residents, and are one of the main cause of storm-related deaths [54]. In addition, UK city councils have expressed their concerns about the potential effect the increase of tree cover in urban parks might have on crime and personal safety, and leaf fall from deciduous trees can clog up drainage of urban run-off. Thoughtful species selection and careful management can reduce many of these disbenefits. In addition to taking up air pollutants, trees can affect local turbulence and wind speed, which can also have adverse effects on concentrations. Above forests, the increased aerodynamic roughness of trees compared with short vegetation and asphalt/concrete tends to increase vertical mixing and dispersion of pollutants and this tends to reduce surface concentrations. The situation is very different within the tree canopy and within the built-up environment: street canyon trees can significantly lower the air flow and dispersion of local emissions, depending on local conditions and meteorology [55]. As a result, urban trees can increase ground-level concentrations and human exposure to air pollutants at the same time as promoting their removal through enhanced dry deposition. For example, using CFD simulations, Jeanjean et al. [56] found for a 4 km2 city area in Leicester, UK, that dispersion had a negative effect that exceeded the enhanced deposition by an order of magnitude at low wind speed, while at higher wind speeds, it had a positive effect that again exceeded the positive effect of dry deposition. Trees can, at the same time, increase the accumulation of pollutants on one side of the street canyon, while lowering concentration on the other [22,45]. Trees are sometimes planted to protect individual receptor sites, e.g. at school carparks. Here, again, the effect can be positive or negative, depending on geometry and meteorology [42]. The literature is more consistent about the use of trees for separating people from sources, e.g. through hedge-like features between roads and pavements [21,22]. In this situation, the trees increase the line of travel of pollutants from source to receptor and promote dispersion and dilution en route. While such interventions might increase concentrations at the source side, they are highly likely to reduce concentration at the receptor side, independent of meteorology [22]. The use of vegetation to spatially separate people from sources appears to be their least controversial and most effective use. However, it should be mentioned that non-vegetation features such as walls would physically have the same effect. Uncertainties in the modelling work are significant and difficult to quantify. Both the ACTM- and deposition-based approaches are sensitive to the parametrization of the dry deposition process used. These are uncertain and vary significantly. For example, Flechard et al. [57] compared the results from four deposition models, including the EMEP parametrization, to derive dry deposition of reactive nitrogen compounds from time-integrated concentration measurements across 55 European sites. This exercise indicated an intra-model variability in the deposition velocities (Vd) of a factor of 3 for NH3 and 5 for NO2. For fine aerosol components (NH4+, NO3−), relative good consistency (within a factor of 2) was found for grass and semi-natural short vegetation (shrubland), but values varied by more than a factor of 10 for forest. This variability reflected not only the parametrization of the deposition mechanism, but also the model assumptions on tree height and leaf area index. The EMEP model tends to fall into the middle of the range for most compounds, but it is one of two models that predicts significantly lower aerosol Vd to forest than the other two models, whose values look unreasonably large in the light of more recent reviews of aerosol dry deposition [58]. Additional uncertainty arises from the limitations of the chemistry scheme in the model. In particular, many aspects of the mechanisms leading to BSOA formation are uncertain: as is typical for ACSMs the EMEP emissions model and chemical scheme only treats some key BVOCs. In particular, very rapid BSOA formation from sesquiterpenes and some monoterpenes may be missing [59]. In addition, as mentioned before, the simulations here assume the same average tree species mix as currently implemented in the EMEP model, while in the urban scenarios, in particular the BVOC emissions will depend on the tree species that exist or are planted in the urban context. Quantification of the benefits of vegetation appears to be an emotive subject and estimates of the efficacy of vegetation and in particular trees to combat air pollution vary greatly in the literature. This study shows that this is at least in part a question of scale of the vegetation extent. While total national vegetation together significantly reduces pollution levels (e.g. about 10% for PM2.5), urban vegetation is estimated to account for only a small reduction in pollution levels and even very large-scale conversion of available open urban greenspace to urban forest would reduce urban air concentrations by only about 1% overall. The impact of small-scale green infrastructure implementations on air quality is very small, except where the vegetation acts as a barrier between source and receptor, and in most cases far less economic than implementing measures to reduce emissions in the first place. Thus, the benefit of urban tree planting for air pollution should in general only be considered as one of multiple benefits within the natural capital approach. The use of an ACTM to estimate reductions in concentrations, exposure and associated health impacts, while more computationally cumbersome, has key advantages over the more commonly used deposition approach in which air pollution removal by vegetation is valued via the damage cost of the equivalent emission (e.g. i-Tree Eco): it is mass conserved and accounts for physical and chemical interactions. The UK-wide simulations identified that a purely dry-deposition-based approach would overestimate the PM2.5 uptake by UK vegetation by a factor of 2, if the associated reduction in wet deposition is ignored. Increased natural BVOC and soil NO emission from forest ecosystems further counteract the benefit of the vegetation for PM2.5 (via BSOA formation), NOx and also O3. The simulations suggest that vegetation only has a small effect on NO2 (and NOx) at the national scale and that conversion of urban green space to urban forest may even increase NO2 levels. This could be traced to the combined impact of increased BVOC emissions on increasing NOx levels and, by also increasing O3, sustaining more of the NOx as NO2. Thus, the magnitude of the increase depends on the BVOC emission potential of the tree species selected, but even for non-emitters, the potential of NO2 removal through urban forests was found to be negligible. The modelled annual average concentration and deposition fields from which the statistics presented here are derived are available from the NERC Environmental Informatics Data Centre (EIDC) at https://eidc.ac.uk/ at https://catalogue.ceh.ac.uk/documents/bad6721c-574b-4229-b023-c7b13ae4c099. The data on pollution removal by UK vegetation are also available via an ONS data selector at https://www.ons.gov.uk/economy/environmentalaccounts/articles/ukairpollutionremovalhowmuchpollutiondoesvegetationremoveinyourarea/2018-07-30. E.N. directed the modelling work, conducted much of the data analysis and drafted the manuscript. E.C. produced the modified landcover datasets. M.V. and C.S. conducted the EMEP model runs. L.J. designed and led the original ONS study and coordinated the data flow. A.F. processed spatial datasets and automated scripts. R.D.M. calculated the urban extent and urban landcover classes. M.H. conducted the health evaluation. D.C. contributed to the discussion of dispersion aspects. All authors commented on the manuscript. We declare we have no competing interests. The modelling work to quantify the air pollution removal by vegetation was funded as part of the National Accounts by the UK Office for National Statistics (ONS) and Defra. The city-scale modelling and the write-up of the paper was supported by the Natural Environment Research Council award number NE/R016429/1 as part of the UK-SCAPE programme delivering National Capability. FootnotesOne contribution of 17 to a discussion meeting issue ‘Air quality, past present and future’. Electronic supplementary material is available online at https://doi.org/10.6084/m9.figshare.c.5092135. © 2020 The Authors. Published by the Royal Society under the terms of the Creative Commons Attribution License http://creativecommons.org/licenses/by/4.0/, which permits unrestricted use, provided the original author and source are credited. References
Page 6Discussion meeting issue ‘Air quality, past present and future’ organised and edited by David Fowler, John Pyle, Mark Sutton and Martin Williams Page 7Particulate air pollution has been the focus of a global research effort for several decades. The aims have been not only to understand and describe associations between exposure and capability to adversely affect human health, but also to identify the plausible biological mechanisms that could explain and support these associations. The exceptional achievements began with the seminal epidemiological studies in the 1990s showing a clear association between increased respiratory and cardiovascular mortality and acute and chronic exposures to ambient particulate air pollution [1,2]. These findings have subsequently been substantiated in epidemiological studies conducted outside of the USA [3] and numerous attempts have followed to quantify the global annual burden of mortality due to particulate matter (PM) less than 2.5 µm in diameter (PM2.5). The current estimates approach 9 million [4]. Epidemiological investigations have also successfully delineated associations of particulate air pollution exposure with increases in respiratory and cardiovascular morbidity [5]. Evidence is particularly strong for reduced lung function, heightened severity of symptoms in individuals with asthma and chronic obstructive pulmonary disease [6] and ischaemic heart disease [6,7] and, in 2012, particulates in diesel fumes were classified as carcinogenic [8]. Data also link exposure with atherosclerosis [9] and a host of childhood respiratory conditions including an increased susceptibility to infection [10] and symptoms of low lung function [11]. The epidemiological work has been matched by a considerable toxicological research effort to define the underlying mechanistic pathways of toxicity elicited by airborne PM, and again the lungs and cardiovascular system have been particularly well studied. One such successful approach to investigate effects on the airways exposed human volunteers (healthy and/or mildly asthmatic) for 1–2 h to whole diesel exhaust (DE; particulates plus the associated gas phase) from an idling engine at concentrations ranging from environmentally relevant (PM with a diameter less than 10 µm [PM10] 100 µg m−3, nitrogen dioxide [NO2] 0.7 ppm) to those occasionally experienced in exceptionally busy diesel-dominated traffic environments (PM10 300 µg m−2, NO2 1.6 ppm). By performing blood, bronchoalveloar lavage and bronchial mucosal biopsy sampling after exposure, these studies have been instrumental in uncovering systemic and pulmonary inflammatory cascades following the stimulation of antioxidant defences and redox-sensitive signalling pathways [12–18]. Exposure to NO2 alone, at similar or higher concentrations for a longer duration, failed to elicit adhesion molecule upregulation or significant changes in inflammatory cells in the bronchial mucosa sampled, suggesting that the PM content of DE was the responsible pollutant [19]. In addition to the large number of studies on the inflammatory effects on airway epithelium and immune cells, a capacity of DE particles (DEP) to directly interact with airway nerve fibres responsible for respiratory symptoms has also been demonstrated [20]. In 2010, Brook and colleagues presented persuasive evidence that oxidative stress is also a critically important cause and consequence of PM-mediated cardiovascular effects. The latter are manifested through several, likely overlapping, pathways including at the functional level, endothelial dysfunction, atherosclerosis, pro-coagulant changes, alterations in autonomic nervous system balance and changes in blood pressure [7]. At the molecular level, principal pathways include (i) the instigation of pulmonary and systemic inflammation [21], (ii) the translocation of ultrafine and nanosize particles and/or particle constituents (organic compounds, metals) across the alveolar membrane into the systemic circulation possibly enabling interaction and localized toxicity within the vascular endothelium and/or cardiac tissue [22], and (iii) the activation of airway-sensitive receptors or nerves and subsequent autonomic nervous system imbalance [23]. At numerous points within each of these functional and molecular pathways there is potential for cellular oxidative imbalances to occur, as has been demonstrated in human experimental studies, healthy and diseased animal models, isolated organs and cell cultures [24]. The past decades of epidemiological and toxicological research have taught us a great deal about the health effects of PM and particularly, the cardiorespiratory effects of roadside PM (primarily DEPs). Knowledge is, however, still lacking in many areas, two of which are discussed in this brief review. First, the differential toxicity of airborne PM, particularly in the light of the likely shift in composition in response to global pressure to reduce combustion emissions, and second what appears to be a growing range of disease outcomes that may ultimately be associated with exposure to airborne particulates. Health studies describing robust associations between ambient PM and ill health have contributed to the World Health Organization Air Quality Guidelines (WHO AQG) and national air quality standards that, owing to the technical limitations and costs of stationary monitoring networks, use the mass concentration of PM2.5 or PM10 as the metric. As a consequence, all particles are treated as equally toxic, without regard to their source and chemical composition. It is unlikely, however, that every component within the overall ambient PM mix is equally harmful to the exposed population. There has, therefore, been an enormous drive to identify which component(s)/source(s) of ambient PM, and/or which of their physical and chemical characteristic(s) are most harmful to health to facilitate reappraisal of air quality guidelines/standards and prioritize targeted PM control strategies to more effectively protect public health. Epidemiological and toxicological research findings have indeed shown that PM mass comprises fractions and sources with varying types and degrees of health effects but, despite this, the question of differential PM toxicity represents one of the most challenging areas of environmental health research [25]. Rather than constituting a single entity, ambient particulate pollution is a complex, heterogeneous mixture that can exist in the atmosphere as solids or liquids. Primary PM is directly emitted from source, while secondary particles are formed following chemical reactions with other pollutants. The mix includes emissions from man-made activities as well as natural sources. The organic particulate mix is particularly complex and constitutes around 104–105 different compounds in today's atmosphere [26]. These materials can be classified in several different ways: with respect to volatility, i.e. volatile organic compounds, semi-volatile organic compounds or condensed-phase compounds; primary or secondary condensed-phase organic compounds; carbonaceous particulate materials existing in the elemental carbon or organic carbon fraction. Since the latter is comprised of a very large number of individual compounds, observations of epidemiological associations unfortunately do not tell us about the identity and source(s) of the individual compound(s) that may be driving health effects. The major anthropogenic sources of organic materials in the atmosphere include internal combustion engines, wood and biomass burning, fuel oil combustion, natural gas combustion, biogenic emissions, resuspended road dust, tyre and brake wear and cooking emissions. Airborne particles vary in chemical composition (and hence solubility and reactivity), mass, size, number, shape and surface area depending upon source and atmospheric processing. All of these properties have the potential to influence health effects. For example, with respect to size, particles can vary from a few nanometres to tens of micrometres. Particles with an aerodynamic diameter smaller than 0.1 µm (PM0.1), PM2.5 and between 2.5 and 10 µm (PM10-2.5) are termed ultrafine (UFPs), fine and coarse particles, respectively. The smaller particles, particularly UFPs, have a greater capacity to (i) penetrate the lung and probably translocate to extrapulmonary sites and (ii) adsorb toxic chemicals owing to a larger surface area to volume ratio. UFPs are, however, challenging to study in epidemiological settings that rely on central site measurements owing to their high spatial variability and high correlation with other combustion-related pollutants. Toxicological studies can also be challenging because of the rapid agglomeration of such particles. Attempts to identify specific effects of components/sources are complicated further since PM can vary in space and time as a consequence of atmospheric chemistry and weather conditions, as well as complex interactions with gaseous air pollutants (e.g. ozone and NO2) that share biologically plausible associations with various health endpoints. Unsurprising, the current database of experimental and epidemiological studies does not allow individual PM characteristics or sources to be definitely identified as being closely related to specific health effects [5,27,28]. It appears that the strengths of associations between effects and individual chemical components of the ambient aerosol vary from effect to effect and that the situation is further complicated by components being associated with certain effects in some locations, but not in others. The ambitious US NPACT studies—that used coordinated toxicology, epidemiology and exposure research to examine and compare the toxicity of PM components, gases and sources—concluded that ‘the studies do not provide compelling evidence that any specific source, component or size class of PM may be excluded as a possible contributor to PM toxicity’ [28]. In the context of organic PM, the findings of other large, multi-year large toxicological and epidemiological research programmes including SPHERES (Secondary Particle Health Effects Research), NERC (National Environmental Respiratory Center), ACES (Advanced Collaborative Emissions Study) and TERESA (Toxicological Evaluation of Realistic Emissions of Source Aerosols) have recently been reviewed [29]. Despite clear health impacts from emissions containing carbon-containing PM, difficulty remains in apportioning responses to certain groupings of carbonaceous materials, such as organic and elemental carbon, condensed and gas phases, and primary and secondary material. Another illustration of the variety of results reported in the literature is a systematic review of the findings of epidemiology, controlled human exposure and toxicology studies that used apportionment methods to relate sources of PM with human health outcomes [30]. Among the 29 studies reviewed, soil, sea salt, secondary sulphate, motor vehicle emissions, coal burning, wood smoke, biomass combustion, Cu smelter emissions, residual oil combustion and incinerator emissions were found to be associated with health outcomes. Another noteworthy quote on the subject from Krall and colleagues states ‘Associations with a given PM2.5 chemical component should be considered as potentially indicative of associations with another component or set of components with similar sources' [31]. It is in fact a belief of various commentators in the field that the literature suggests that various complex mixtures may be involved, and that the capability of PM to induce disease may be the result of multiple components acting through different physiological mechanisms [32,33]. Efforts continue to tease out which components and sources of PM are most harmful since to identify regulation targets can better protect human health. Epidemiological studies of sources are, however, challenging since source-specific exposures (e.g. PM from road transport) often are not directly measured and must be estimated by applying source apportionment models [34–36]. Using data from the Atlanta Commuting Exposure studies, Krall et al. found that exposures related to crustal and secondary pollution were associated with decreased lung function among asthmatic commuters [37]. Findings from another recent study that adopted source-apportioned PM2.5 concentrations suggest a role for emissions from spark ignition, and diesel vehicles, tyre and brake wear, residual oil combustion emissions from large building heating and nitrate particles in the triggering of acute cardiovascular events [38]. A recent toxicological study incorporating comparative data from different road traffic sources focused on brake abrasion dust (BAD) and DEPs [39]. This is highly relevant in that while there is an extensive literature on the health effects of engine emissions [40], the toxicity of non-exhaust PM—from brake wear, tyre wear, road surface wear and resuspended road dust—has not been extensively studied despite becoming a significant component of urban air pollution. Furthermore, this currently unregulated component of traffic emissions is expected to become proportionately more important, as vehicle exhaust PM emissions from road transport are expected to decrease over the coming years. The in vitro study by Selley and co-workers compared the relative toxicity that BAD and DEPs exert on airway macrophages to investigate whether the marked compositional difference between these particle species is reflected in their ability to perturb cell function [39]. Although BAD contained considerably more metals/metalloids than DEP, similar toxicological profiles were observed in U937 monocyte-derived macrophages at minimally cytotoxic doses (4–25 µg ml−1; 24 h). Responses to the particles included transient, dose- and metal-dependent increases in secretion of IL-8, IL-10 and TNF-α and decreased phagocytosis of Staphylococcus aureus and, for both particles, the metal chelation restored bacterial uptake to levels comparable with the particle-free control (figure 1). Figure 1. Quantities of S. aureus ingested by U937 cells over a 2 h period subsequent to 24 h incubation with (a) BAD, (b) DEP (SRM-2975), (c) BAD (±METAL CHELATOR DFO) and (d) DEP (SRM-2975) (±DFO). BAD, brake abrasion dust; CFU, colony-forming units; DFO, desferroxamine mesylate; SRM-2975, standard reference material 2975. [39]. Published by The Royal Society of Chemistry.
Microplastics (less than or equal to 5 µm particles and fibres produced from the breakdown of larger items such as clothing, car tyres and mismanaged urban waste) are another particulate pollutant of increasing environmental concern owing to the astonishing global mass production of plastic plus its persistent nature in the environment [42] and synthetic physiological fluids [43]. While occurrence, sources and fate of airborne microplastics are still poorly documented, in part owing to the technological challenge associated with detection and identification, progress is being made in assessing atmospheric deposition in both indoor [44] and outdoor [45–48] environments. Of the latter, airborne microplastics have been measured in the major population centres of Paris [45], Dongguan [46] and London [48] as well as at a remote and pristine site in the French Pyrenees [47]. In central London, microplastics of various shapes, but primarily fibrous, were detected in all samples tested, with deposition rates ranging from 575 to 1008 per m2 per day. Across all samples, 15 different petrochemical-based polymers were identified (figure 2). Figure 2. Time-series of deposition rates (n/m2/d) for fibrous, non-fibrous and total microplastics (a); proportional distribution of identified petro-chemical-based fibrous microplastics (b); proportional distribution of identified petro-chemical-based non-fibrous microplastics (c). PAN, polyacrylonitrile; PES, polyester; PA, polyamide; PP, polypropylene; PVC, polyvinylchloride; PE, polyethylene; PET, polyethylene terephthalate; PS, polystyrene; PUR, polyurethane; Pol. Petr. Res, polymerized petroleum resin [48]. (Online version in colour.)
Evidence of a novel component of air pollution is, therefore, emerging while complementary existing fields indicate potential hazards. For example, evidence exists that people working in the textile industry experience respiratory symptoms and develop interstitial lung disease following exposure to nylon flock [49] and similar symptoms have been described among workers in facilities manufacturing polyethylene and polypropylene flock [50,51]. Although the toxicology of microplastics is an emerging field, several potential mechanisms exist through which harm to human health could occur. These may include the activation of pathways in response to the particle/fibre per se and/or leaching of adhered chemicals (owing to the surface area:volume ratio of microplastics, plus their surface hydrophobicity), from both additives incorporated during manufacture and contaminants accumulated from the environment. For example, an acute inflammatory response has been demonstrated in the lungs of rats after intratracheal instillation of nylon fibres of a respirable size (2 µm diameter, 14 µm length on average) [52]. Furthermore, inflammation induced by granular and spherical particles (polyethylene/polyethylene terephthalate) from abraded plastic prosthetic implants has also been reported [53,54]. With respect to the leaching of plastic-derived chemicals, potentially with reproductive, carcinogenic and mutagenic effects, there is no information on human tissues but transfer of plastic-derived chemicals from ingested waste to the tissues of marine-based organisms has been described [55]. Much of our understanding of the mechanisms by which particulate air pollution elicits ill health has been gleaned from traditional, hypothesis-driven approaches that focus on a priori defined clinical parameters employing a limited number of endpoint assays [56,57]. These have been fundamental to epidemiologists in providing the means to support causal inference by providing associations between physiological endpoints (e.g. respiratory or cardiovascular symptoms) and underlying molecular events. The limitation lies in the inability to uncover the multiple molecular targets and novel pathways that are undoubtedly behind the toxicological response to complex environmental exposures. Current and more informative mechanistic studies, which heavily focus on questions rather than hypotheses, have a greater likelihood of unveiling unexpected relationships and generating novel insights that in turn can lead to hypothesis generation [58,59]. Advances in analytical technology, in the form of metabolomics, involves the simultaneous measurement, by mass spectrometry or nuclear magnetic resonance, of multiple small metabolites (less than 1 kDa) arising from specific cellular processes, such as energy production and storage, signal transduction and apoptosis [60]. Since metabolites are the terminal products of gene expression, i.e. the final consequence of biological function, their profiles in biological samples report on actual functional status. The staggeringly large amounts of information of such global analyses should not, however, be underestimated and gaining biologically relevant conclusions from a given metabolomics dataset requires a specialized data analysis. Notwithstanding such challenges, identifying metabolite perturbations caused by air pollution exposure is a particularly relevant and promising approach in characterizing the interactions of living organisms with their environment by identifying disregulated molecular pathways and predicting health endpoints [61]. It is not surprising, therefore, that to address issues such as differential toxicity, workers have adopted such an approach in both toxicological and epidemiological studies [62–68]. Experimental studies have investigated shifts in the metabolite profiles of bronchial wash (BW) and bronchoalveolar lavage (BAL) of healthy volunteers following exposure to biodiesel exhaust (BDE) compared with filtered air [62,63]. This approach greatly enhanced the number of metabolites that were detected and, in turn, novel pathways including alterations in energy metabolism and degradation of cell membrane lipids associated with BDE exposure. Notably, BDE-induced shifts in metabolite profiles of the BW versus BAL fluids differed appreciably, and a stronger response was detectable for peripheral regions of the lungs. Incorporation into epidemiological research has also been demonstrated to sensitively detect internal metabolic perturbations in healthy subjects, pregnant women and people with asthma following complicated exposures such as those present in urban environments [64,65,67–69]. A key finding from these studies has been the identification of several oxidative stress and inflammation-related pathways (including leukotriene, cytochrome P450, vitamin E, tyrosine, methionine and tryptophan metabolism) that were consistently associated with elevated pollution exposures. For example, in an analysis of healthy college students living close to a major urban highway, leukotriene, vitamin E, cytochrome P450 and methionine metabolic pathways were linked to longer-term (over 3 months) exposure to elevated traffic-related air pollution (TRAP), including black carbon (BC) and PM2.5 [65]. As illustrated in figure 3, Liang et al. [69] detected numerous significant metabolic perturbations associated with in-vehicle exposures during commuting, validated metabolites that were closely linked to several inflammatory and redox pathways and collectively implicated these mechanisms as part of the impact of TRAP toxicity in asthmatic individuals. Figure 3. Potential molecular mechanisms underlying the effects of TRAP toxicity on individuals with asthma elucidated using untargeted high resolution metabolomics on the study participants. IL-4, interleukin 4; IL-10, interleukin 10; NOS, nitric oxide synthase; ROS, reactive oxygen species; TNF-α, tumour necrosis factor alpha; TRAP, traffic-related air pollution; XOR, xanthine oxidoreductase [69]. (Online version in colour.)
Emerging epidemiological and experimental data from a growing number of studies suggest that particle exposure may exert a wider threat to human health, beyond the cardiorespiratory systems, by negatively influencing a broader number of diseases including adverse birth outcomes [71–73], slower rates of cognitive development in children [74,75] and accelerated cognitive decline in adults [76,77]. There remain substantial gaps in our knowledge as to possible causal pathways for these relatively new scientific observations. This could, however, be due to the translocation of the smallest particles in the overall mix into the target organs as well as more indirect pathways acting through inflammatory mediators produced in response to the particles. Animal studies have documented the ability of small inhaled particles to reach the brain, with evidence suggesting that this occurs following deposition in either the nasal epithelium (via the olfactory nerve) or the alveolar epithelium by entering into the systemic circulation and eventually crossing the blood-brain barrier [78]. Importantly, the translocation of airborne particles to the brain is potentially supported by human evidence as Maher et al. reported the presence in postmortem brain samples of magnetite nanoparticles, consistent with those formed by combustion and/or friction-derived heating [79]. The numerous investigations into whether nanoparticles can cross the placenta show a dependency on size, shape and surface charge [80], while Valentino et al. [81] demonstrated ‘nanoparticle-like’ aggregates in the cytoplasm of placental trophoblastic cells of rabbits following exposure to aerosolized DEPs. These experimental data are also supported by human evidence with Bove et al. [82] reporting the presence of BC particles in placenta at both the maternal and fetal side. Such findings confirm the translocation of ambient PM directly to the fetus and represent a potentially novel mechanism to explain adverse health effects from early life onwards. The quest to uncover associations between ambient PM and all possible diseases, including prevalent but rarely studied ones, has recently been tackled by using a hypothesis-free analysis of a large dataset [83]. This study analysed more than 95 million hospital admissions of Medicare beneficiaries plus PM2.5 concentrations on the day before presentation over 13 years. In addition to confirming previously established associations between short-term PM2.5 concentration and cardiorespiratory disease, diabetes mellitus and Parkinson's disease, the researchers found that each 1 µg m−3 increase in PM2.5 was associated with 2050 extra hospital admissions as a consequence of previously unassociated diseases (figure 4). The latter included fluid and electrolyte disorders, acute and unspecified renal failure, septicaemia, intestinal obstruction without hernia and urinary tract infections. Moreover, associations remained consistent when restricted to days when daily PM2.5 concentrations fell below the WHO 24 h AQG. Figure 4. Analysis showing absolute increases in risk of hospital admission, ordered from highest to lowest, associated with each 1 µg m−3 increase in lag 0–1 PM2.5 [83]. CCS, Clinical Classification Software code. * Indicates newly identified disease groups. Reproduced from Short term exposure to fine particulate matter and hospital admission risks and costs in the Medicare population: time stratified, case crossover study, Wei et al. [83] with permission from BMJ Publishing Group Ltd. (Online version in colour.)
Recent decades of epidemiological and toxicological research have taught us a great deal about the health effects and underlying toxicology of PM, particularly the cardiorespiratory effects of roadside, combustion-related particles. To keep abreast of the substantial challenges that air pollution continues to throw at us requires yet more strides to be achieved. The multi-disciplinary efforts to identify the most toxic components/sources and the range of associated disease outcomes must continue. Closure of some of today's knowledge gaps will be fundamental to inform the revision of air quality guidelines and standards, set effective mitigation strategies and assist their ultimate remit of protecting human health. This is particularly relevant owing to predicted shifts in PM composition in response to global pressure to reduce combustion emissions. Particles arising from the non-exhaust component of traffic emissions are prime examples of current challenges on which to focus on. These have become a significant component of traffic-related PM and are projected to become a more dominant source. Added to this, particle characteristics with respect to both (small) size and (metallic) composition suggest that this currently unregulated source that is concentrated in highly populous urban environments may be a particularly hazardous one. While recent research suggests that the capacity of BAD to harm pulmonary cells is equivalent to that of DEP [39], unambiguous toxicological data are currently lacking. For example, for the most part, toxicological studies of tyre wear have employed pulverized or size-fractionated material, rather than that formed from ‘real world’ friction between tyres and roadways. In addition no direct epidemiological studies on non-exhaust PM at the roadside have been undertaken. New epidemiological and toxicological research, incorporating future trends (e.g. potential benefits of regenerative braking versus added risks associated with increased tyre and road wear from heavier electric/self-driving vehicles), should be undertaken to further understand the potential health risk of this aspect of vehicular pollution. Without such research, policy changes to control emission sources and benefit human health will be difficult. The contribution of microplastics to the risks that airborne PM inflicts on human health is another timely research field that has emerged. Early work presenting evidence of airborne plastics in indoor air [44] and outdoors, in populous urban environments [45,46,48], raises concern for public health, especially with a predicted increase in plastic use, particularly in the textile sector, pointing to a proportional increment in airborne microplastic concentrations. A robust evidence base characterizing population exposure is, therefore, required. We also need to establish the toxic characteristics of microplastics, their behaviour in the body, and what, if any, constitutes a safe threshold for exposure when plastics are inhaled. Environmental metabolomics has emerged as a means to provide a broad-spectrum of measurements of human metabolism that may reveal biological effects and associated toxicological mechanisms associated with an exposure to particulate air pollution. Detecting and monitoring markers of adverse outcomes following exposure to different air pollutants in large human cohorts is limited at present [84,85], but could divulge differential toxicities and thereby help to target regulatory efforts to those pollutants that pose the greatest risk to public health. Continued development of this field, in combination with complementary ‘omics’ technologies such as genomics and proteomics, as well as traditional hypothesis-led research will be crucial to help strengthen the causal basis for the epidemiological findings that associate air pollution with an ever growing number of diseases. Indeed, research that has used hypothesis-free analysis and predicted epidemiology via toxicology suggests that current figures for PM2.5-associated morbidity, which focus on established disease associations, might be considerable underestimates [83,86]. This again calls upon more epidemiological research to investigate newly reported associations, and particle toxicology to provide plausible biological mechanisms that could explain and support these associations. As the burden of disease associated with particulate air pollution becomes more apparent, it is ever more clear that there is much still to learn. Previous work, strengthening the evidence for both the adverse effects and benefits of intervention [11,87,88] tell us that the sooner we act to close knowledge gaps, increase awareness and develop creative solutions, the sooner we can reduce the public health burden attributable to this complex and insidious environmental pollutant. Data presented in this review can be accessed through previously published articles, a list of which is provided. F.J.K. and J.C.F. conceived the review. J.C.F. wrote the text to which F.J.K. contributed. F.J.K. and J.C.F. gave final approval for publication and agree to be held accountable for the work performed therein. We declare we have no competing interests. This work was funded by the National Institute for Health Research (NIHR 200 880) Health Protection Research Unit in Environmental Exposures and Health, a partnership between Public Health England and Imperial College London. The views expressed are those of the author(s) and not necessarily those of the NIHR, Public Health England or the Department of Health and Social Care. FootnotesOne contribution of 17 to a discussion meeting issue ‘Air quality, past present and future’. © 2020 The Authors. Published by the Royal Society under the terms of the Creative Commons Attribution License http://creativecommons.org/licenses/by/4.0/, which permits unrestricted use, provided the original author and source are credited. References
Page 8Discussion meeting issue ‘Air quality, past present and future’ organised and edited by David Fowler, John Pyle, Mark Sutton and Martin Williams
Page 9Over recent decades ammonia (NH3) has often seemed like the Cinderella of air pollution, as it has been given much less attention than other pollutants, such as sulfur dioxide (SO2), nitrogen oxides (NOx), ozone (O3) and particulate matter (PM). In the 1980s, research focused on ‘acid rain’, especially in the light of SO2 and NOx emissions [1–3] with only a few researchers at that time examining the possible effects of NH3 and ammonium (NH4+) on the environment, including threats to soils, biodiversity and forest health [4–6]. The same can be said for European air pollution policy, with successive international protocols on SO2 and NOx emissions [7,8], preceding the multi-pollutant, multi-effect Gothenburg Protocol [9], which included NH3 for the first time. Even then, the commitments for NH3 were much less ambitious than for other air pollutants, requiring that little action be taken by most countries. The situation is similar with the 2020 ceilings of the revised Gothenburg Protocol of 2012. With insufficient measures implemented, several countries are unlikely to meet their legally binding NH3 ceilings for 2020, while overall Europe-wide NH3 emissions have actually been increasing since 2013 [10]. The barriers appear to be primarily political, as The Netherlands and Denmark have shown that it is possible to reduce NH3 emissions substantially. With this perspective in mind, it is appropriate to take stock of what ammonia has meant to people in the past, what it means today, and what it might mean in the future. We rapidly discover that NH3 and NH4+ were historically far from insignificant, fulfilling several important roles. Whereas recent efforts have focused on reducing NH3 emissions from agriculture, with the main sources being livestock excreta and fertilizers, the historical picture helps to raise awareness of the multi-dimensional relevance of ammonia for environment and society. Considering the present, across much of Europe and North America we now inherit a world where substantial emission controls have already been achieved for SO2 and NOx. We consider in detail the implications of the changed ratio of NH3 to the acid gases, especially for some of the most sensitive ecological receptors. Instead of acid rain, we now face challenges from ‘alkaline air’, which was the original name given by Joseph Priestley [11] for gaseous ammonia. Today, we may also define alkaline air more generally as air where alkaline gases (primarily NH3, but in principle also including volatile amines) dominate over those that are acidic in nature. Finally, we consider what might be expected for the future. What are the implications of current legislation, of the slightly more ambitious emission reductions for 2030 under the revised EU National Emissions Ceilings Directive (2016/2284/EU)? We conclude by placing NH3 mitigation in the context of the circular economy for nitrogen and United Nations actions on nitrogen to help meet multiple Sustainable Development Goals (SDGs). In the following sections, we show how a broad approach linking past, present and future could help raise awareness about the importance of ammonia and nitrogen as a contribution to catalysing action on the SDGs. We juxtapose the historical value of ammonium in international trade and alchemy with current development of the nitrogen circular economy. The analysis is underpinned with a more detailed examination of ecological datasets for epiphytic lichens and bog ecosystems which together emphasize the emerging importance of alkaline air. While the popular historical narrative ascribes the discovery of ammonia to Priestley [11], his achievement needs to be set in the context of at least two millennia of human exploration and investigation into ammonia and ammonium. By the start of the Tang Dynasty (AD 618–907), ammonium salts for use in metallurgy, medicine and food were already being traded as a luxury product along the Silk Road in Central Asia [12]. Spontaneous combustion of near-surface coal deposits explains the development of fire caves, some of which burn for hundreds of years. Nitrogen (N) in the burnt coal volatilizes as NH3, reacting with co-emitted hydrochloric acid (HCl), sulfuric acid (H2SO4) and nitric acid (HNO3) to form a mix of ammonium chloride, sulfate and nitrate salts [13]. Ammonium chloride tends to dominate in the collected sublimate (also known as nūshādir, nao sha, sal ammoniac, the Eagle (nasr) and a wealth of other names), presumably because it is more volatile than ammonium sulfate, while ammonium nitrate formation may be limited by low HNO3 concentrations relative to NOx (ammonium nitrate is also decomposed to N2, N2O and water at high temperatures). Along with many other point sources, NH3 emissions from such fire caves can now be detected from space [14], such as at Jharia in India [10] (figure 1). Figure 1. Fire cave at Jharia, India, where spontaneous combustion of surface coal deposits has resulted in burning at this location for over a century. Sites such as this across Central Asia, together with volcanic fumaroles, represent the earliest recorded sources of traded ammonium salts (Photo © Johnny Haglund). (Online version in colour.)
The historical collection of the sal ammoniac sublimate around the cooler edge of fire caves, as well as from a range of volcanic fumaroles (from Etna in Sicily to Mount Damavand in Iran), allowed it to become a key commodity of long-distance trade up to the early nineteenth century [12]. The importance and the stability of the sal ammoniac market can be illustrated by comparing prices from AD 620 (Central Asia) with those from AD 1000–1140 (Mediterranean trade) as shown in table 1. The estimates for Central Asia are based on transactions recorded in tax records discovered near Turfan, in present-day Xinjiang province of China [15]. These values are compared with documentary records recovered from a geniza or document repository, as uniquely preserved in Cairo [16].
Table 1 shows impressive similarity for the prices of sal ammoniac and spice from these independent datasets, while the price of silk and slaves had increased substantially in Cairo compared with Central Asia. With today's perspective, it is shocking to note that just 6–8 kg N would purchase a human being. This reflects both a high price of nitrogen and a low value of human life compared with the present. Relative to changing gold and silver prices (electronic supplementary material, §2), N compounds are today around three orders of magnitude cheaper, with prices decreasing rapidly during the twentieth century as large-scale manufacture, mainly through the Haber–Bosch Process [17], has increased their availability. The burning coal caves of Central Asia also provide the first recorded example of ammoniacal air pollution. It appears that locals would encourage the natural coal burning specifically to harvest sal ammoniac, as recorded by ibn-Hauqal: Over the spot whence the vapour issues, they have erected a house, the doors and windows of which are kept so closely shut and plastered over with clay that none of the vapour can escape. On the upper part of this house the nūshādir rests. When the doors are to be opened, a swiftly running man is chosen, who having his body covered over with clay, opens the door; takes as much as he can of the nūshādir, and runs off; if he should delay, he would be burnt (translated by Ouseley [18], p. 264, who renders nūshādir as ‘copperas’). Further details of the pollution threat are given by al-Mas'ūdī: Travellers in summer take their road from Khorāsān to China by this mountain; for there is a valley through it, which is forty or fifty miles long. At the entrance of the valley wait some men who offer themselves to carry the baggage, if they are well paid. They use sticks to drive the passengers on their journey; for any stoppage or rest would be fatal to the traveller, in consequence of the irritation which the ammoniacal vapours of this valley produce on the brain, and on account of the heat. The way becomes more and more narrow till the travellers come to the end of their perilous passage. Here are pits with water, in which they throw themselves, to obtain relief … When travellers arrive in the Chinese territories, they are beaten as in passing (to counteract the congestion of blood in the brain) (translated by Sprenger [19], pp. 359–360). Caution is needed with regard to the comment of al-Mas’ūdī about effects of nūshādir on the brain. This may reflect the fact that the Chinese term, nao sha, includes a component referring to the brain, so that nao sha was sometimes termed brain salt (see [20], pp. 446–447), for which there are several possible explanations.While the above examples illustrate the historic importance of ammonium in trade and air pollution, these were probably not the earliest applications. Pliny the Elder (Natural History 28: 19, 149) was already familiar with use of the fumes of deer horn and hair to make people breathe naturally when choking with hysteria. This use is directly analogous to the eighteenth century popularity of ammonium carbonate as ‘smelling salts’; these liberate gaseous NH3, which acts as a vasodilator in the airways. Ammonia and ammonium were also known in scientific circles, if not always openly. In particular, they were at the heart of alchemy, well-known as a ‘reserved’ science (i.e. unspoken, secret, limited to the few), making it extremely difficult to trace how they were used. One of the most clear alchemical writers on nūshādir was the Persian physician al-Rāzī. It has often been stated that earlier Greek alchemy used exclusively metals and other minerals, while Islamic alchemy introduced the use of organic materials (e.g. [20], p. 435, [21]). The following illustrates the methods of al-Rāzī: Take of black cleaned hair, distil its water and oil and calcine its residue according to what is [explained] further above, and put away each part of it separately … Then tie it [the solidified oil] up in a linen cloth and hang it into distilled urine in a clay container on the hook of the blind [cucurbit]. Place it on a small oven under which burns a fire of a lamp. Leave it for 24 h, that the urine becomes red. Then pour it off and renew the urine. Repeat this operation until all the colour is extracted. Then gather all and distil. Distil white urine, but its redness remains. Then mix that what remained from the oil in a batch with the distilled juice of a lemon and treat it with the urine with the help of the operation… Then convert it into a hard state in a blind [cucurbit]; it solidifies it into white nuqra like crystal … But if you want, that it [transmutes] into the red [into gold], thus put in it before it solidifies, the red [residue] … it solidifies, transmuting into red nuqra, a dirham of which transmutes 1800 dirham of any metal whichever you want into pure gold (trans. by G. Fischer from [22], pp. 109–110). Special caution is needed here, as al-Rāzī uses so-called ‘cover names’ (Decknamen), referring to the ammoniacal distillates as ‘urine’ (because of how it comes out of the alembic) or ‘lemon juice’ (because of its sharpness). Considering these processes, the Arab/Persian alchemist Jābir refers to alchemy as a mesocosm or middle-world, which links understanding of the macrocosm (universe) with the microcosm (humans) (cf. [23], p. 74). In experiments like this, ammonia and ammonium were key to early experimental philosophy. As to the gold, even more caution is needed. In the margin of one manuscript, one of al-Rāzī's readers commented: ‘Truly I have looked into this book … Do not occupy yourself with them [the essences of Arsenic and Sulphur] unless you already know the secret of the process … Only if you know the secret, God willing, will you accomplish the work’ (translated by Heym [24], p. 191).In fact, it looks likely that earlier Greek alchemists were already familiar with ammonia and ammonium salts. For example, characteristic steps from the al-Rāzī process given above can be found in writings of the Greek alchemist Zosimus (e.g. [25], pp. 30–33; [26], pp. 486–492) and in those attributed to Democritus (e.g. [27], p. S91; electronic supplementary material, §3). There is also a question about the oldest name for sal ammoniac. The term nūshādir appears earliest in its Chinese rendering as nao sha, but has a well-established Iranian etymology, meaning ‘immortal fire’ [12]. It is a name that matches just as well to the macrocosmic fire caves as to the processing of ‘elements’ in the mesocosmic analysis of earlier Greek alchemy, leaving open the question of its origin. Obscure as these beginnings may seem, they form the foundations on which modern science was built. This is no more apparent than with Isaac Newton, who experimented and wrote extensively on alchemy, but deliberately kept his findings secret (e.g. [28], p. 159) and encouraged others to do so. Newton thus wrote to Henry Oldenburg, the Secretary of the Royal Society, encouraging Robert Boyle not to reveal alchemical secrets: [It] may possibly be an inlet to something more noble, not to be communicated without immense dammage to ye world if there should be any verity in ye Hermetick writers, therefore I question not but that ye great wisdom of ye noble Authour [Boyle] will sway him to high silence till he shall be resolved of what consequence ye thing may be … there being other things beside ye transmutation of metals … which none but they [the alchemists] understand … but pray keep this letter private to your self [29]. The message was the traditional one of many alchemists over the centuries: not to reveal the secrets of alchemy, which could otherwise lead to the destruction of society (cf. al-Jildakī [30], p. 49). While Boyle may have engaged in the practice of advertising secrecy [31], Newton appears to have recognized the ethical dilemma concerning open explanation of alchemy.Ultimately, the scientific community turned away from the secrecy of alchemy, pushing towards openness of scientific publication for practical benefit. As the experimentalist Stephen Hales wrote in the year that Newton died: If those who unhappily spent their time and substance in search after an imaginary production, that was to reduce all things to gold, had, instead of that fruitless pursuit, bestowed their labour in searching after this much neglected volatile Hermes, who has so often escaped thro’ their burst receivers, in the disguise of a subtile spirit, a mere explosive matter; they would then instead of reaping vanity, have found their researches rewarded with very considerable and useful discoveries ([32], p. 180). Hales’ experiments were to be decisive as a prelude to the scientific discovery of ammonia. His work introduced the idea of ‘pneumatic chemistry’, distilling all sorts of products and then collecting the resulting gases in an inverted vessel over a trough of water. In the case of ammonia distilled from blood or harts-horn, this first filled the vessel, but then gradually dissolved in the water, leaving Hales with no ammonia to collect (e.g. [32], p. 95, Experiment XLIX). Continuing these kinds of experiments 50 years later, Joseph Priestly instead filled his pneumatic trough with mercury in which the ammonia would not dissolve. This enabled him to isolate and characterize pure ammonia gas [11]. Priestley's first report was in a private letter to Benjamin Franklin in September 1773, later presenting his findings to the Royal Society ([33], pp. 93–99).It was only in the 1790s that Priestley's alkaline air started to become known as ‘ammonia pura’, given its relationship to sal ammoniac. Subsequent chemical discoveries came quickly, with Scheele [34] showing that it was present in the atmosphere, and Berthellot [35] demonstrating that it consisted of one part nitrogen to three parts hydrogen. The reminder of ammonia as alkaline air is highly relevant to the present, as emissions of SO2 and NOx have decreased greatly over the last 30 years, leaving European and North American atmospheres increasingly rich in NH3. This can be illustrated by the temporal evolution of emissions, gas and aerosol concentrations and rainfall acidity across the UK. While SO2 emissions have been almost entirely abated (97% reduction since 1970) and NOx emissions reduced by 70%, estimated NH3 emissions increased substantially up to 1990, decreased by 18% (1990–2013), and then increased 9% (2013–2017; figure 2a). National mean NH3 concentrations have not changed significantly since the National Ammonia Monitoring Network [37] was started in 1997 (though increasing in remote areas), while aerosol NH4+ concentrations have decreased significantly, consistent with declining SO2 and HNO3 (figure 2b). This has led to less formation of ammonium sulfate and ammonium nitrate, which will have also helped maintain gaseous NH3 levels [38,39]. Figure 2. (a) Emissions of SO2, NOx and NH3 from the UK relative to 1970, comparing the Defra National Atmospheric Emissions Inventory (NAEI) including estimates from the Long-Term Large-Scale (LTLS) model for earlier years [36]. (b) Annual mean concentrations of gaseous NH3, SO2 and HNO3 and of aerosol NH4+ (for 12 sites), from the UK monitoring network (for further details and error analysis, see Tang et al. [37,38], compared with the earlier trend for five sites (SO2(a)), normalized to the UK mean for 1999–2001. (c) Volume-weighted mean pH of precipitation across the UK based on spatial interpolation of measured values.
As a consequence, acid rain is now a thing of the past for UK conditions. Since 1986, volume-weighted rain pH has increased from 4.62 to 5.48 (figure 2c), now being close to the value of 5.6 due to dissolution of atmospheric CO2. Together these changes demonstrate how alkaline air is becoming increasingly important across the UK countryside, in a pattern that is reflected across much of Europe and North America [40,41]. A corresponding trend is now occurring in China, following implementation of SO2 emission controls from 2012 [42], while in India, NOx emissions have been increasing even faster than NH3 emissions [43]. The gaseous alkaline fraction (expressed as NH3 divided by the sum of NH3, 2SO2, HNO3 and HCl) is now at 88% in the UK (electronic supplementary material, §4), while estimated global variation is shown in figure 3. In many areas of the world, the gaseous alkaline fraction is over 60% (including NOx) or 80% (excluding NOx). Figure 3. Global distribution of the gaseous alkaline fraction for 2010 as estimated by the EMEP-WRF global model [44], here calculated based on surface atmosphere mixing ratios (ppbv/ppbv) as NH3 / (NH3 + HNO3 + 2SO2 + NOx): (a) including NOx, (b) excluding NOx, since it is unclear to what extent NOx concentrations influence leaf surface acidity (see electronic supplementary material, §4).
The net result of these changes is that NH4+ is now making an increasing relative contribution to the composition of airborne particulate matter, relevant for effects on human health [45]. In parallel, the increasingly alkaline, NH3-rich atmosphere is having substantial consequences for the natural environment, as examined in detail below for lichens and other sensitive plants. While lichens are well known to be sensitive to SO2 concentrations, here we emphasize that NH3 is now the primary air pollution driver of lichen distributions in many areas of Europe. To understand the dynamics, we first consider a local-scale transect from Scotland [46] that shows how lichens can change in the vicinity of a poultry farm emitting NH3. Lichens on tree trunks of both Scots pine (Pinus sylvestris) and Sitka spruce (Picea sitchensis), and on branches of birch (Betula pubescens), which are all naturally acid-barked trees, were scored according to a standard methodology [47,48]. In this approach, lichen species are categorized as ‘acidophytes’ (e.g. Usnea, Hypogymnia, Pseudevernia, Bryoria), preferring naturally acidic bark, and ‘nitrophytes’ (e.g. Xanthoria, Physcia), favouring higher levels of nitrogen air pollution (electronic supplementary material, §3) [49]. Using this approach, frequency-based lichen indices for acidophytes (LA) and nitrophytes (LN) were calculated (see electronic supplementary material, §5), where the difference (LAN = LA–LN) distinguishes bark dominated by acidophytes (+ value) or nitrophytes (− value). Findings from the local transect are summarized in figure 4, showing how acidophyte species were gradually eradicated between mean NH3 concentrations of 1 and 12 µg m−3, with acidophytes on twigs being more sensitive to NH3 than those on trunks. Acidophytes on both twigs and trunks were already significantly reduced at the third cleanest location (approximately 1.7 µg m−3, two-sample t-test, two-tail assuming unequal variance, trunks: p < 0.001; twigs: p < 0.01), where the first nitrophytes on twigs were also recorded. Highest nitrophyte occurrence was recorded at 30 µg m−3, with a significant reduction at 70 µg m−3 for both trunks (p < 0.001) and twigs (p = 0.01). Figure 4d shows that there was also a significant relationship between LAN and measured bark pH. This effect can be largely explained by NH3 increasing bark pH nearer the farm (see electronic supplementary material, §5). It is notable that there is no significant difference in the relationship between LAN and bark pH for twigs versus trunks (figure 4d). This indicates that the greater sensitivity of acidophyte lichens on twigs is consistent with the differences in bark chemistry between twigs and trunks. One of the advantages of the local study shown in figure 4 is that it covers a wide range of pollution levels from 0.3 to 70 µg m−3 demonstrating its wide relevance for different pollution conditions. Figure 4. Response of epiphytic lichens on trunks and twigs of Betula pubescens to increasing NH3 concentrations in a farm transect in South Scotland. (a) trunks and (b) twigs, for a cover index of lichen acidophytes (LA) and nitrophytes (LN). (c) Results for the joint index LAN = LA – LN. (d) Combined relationship between LAN and bark pH for twigs and trunks: LAN = − 5.73 (bark pH) + 32.1, with R2 = 0.91. Error bars are ±1 standard error for five replicate trees at each location. (Online version in colour.)
The lichen methodology was subsequently applied at 30 sites across the UK [50]. It must be recognized that different tree species also have naturally different bark pH, and therefore the analysis distinguished lichen communities on naturally acid-barked oak (Quercus robor, Q. petraea, recorded where available) from communities on other tree species. LAN was generally not found to be correlated with SO2 concentrations (except for a weak relationship for oak trunks, p = 0.04, n = 11), with a lack of relationship with SO2 also found in a later survey [51]. At the UK-scale, trunks and twigs both show reducing LAN score with higher NH3 and with higher bark pH, demonstrating the broad relevance of these relationships (figure 5). Substantial scatter can be seen between NH3 concentration and LAN score, which is expected based on natural variation in bark pH, even under clean conditions. In addition, variations with climate may have introduced some scatter. For example, precipitation was found to have a weakly significant effect for lichens on twigs (p < 0.05: R2 = 0.15 (all data), R2 = 0.35 (oak); electronic supplementary material, figure S4), but was not significant for lichens on trunks. Figure 5. Relationship between epiphytic lichens on trunks and twigs of oak and other tree species to ambient NH3 concentrations (a,c) and bark pH (b,d) from 30 sites across the UK for trunks (a,b) and twigs (c,d). Results are shown for the joint index LAN = LA – LN, where LA is the cover score for acidophyte lichens and LN the score for nitrophytes. (Online version in colour.)
However, an even higher correlation was found for the UK by relating the LAN scores to a combined index of NH3 (μg m−3) + 4 (bark pH) (figure 6). This points to NH3 as potentially having two effects. Firstly, NH3 has an alkaline effect, shown by its increasing bark pH (electronic supplementary material, figure S3A). Secondly, there appears to be an effect that is not explained by the changes in bark pH. If NH3 only had its effect by altering bark pH, then this would not explain why a combined index of NH3 and bark pH gives an improved relationship with LAN than with bark pH alone. Other diversity and nitrogen indicators are illustrated in electronic supplementary material, §5 (electronic supplementary material, figure S3), with further statistical comparisons and the full UK dataset given in electronic supplementary material, figure S4 and table S6, respectively. Figure 6. Relationship between the lichen index (LAN) and a combination of NH3 concentration and bark pH. LAN = LA – LN, where LA and LN are abundance indices for acidophyte and nitrophyte lichen species, respectively. Combining the trunk and twig data, the R2 values are 0.72 for lichens on oak (n = 25, p < 0.001), 0.56 for other trees (n = 25, p = 0.001) and 0.64 for all data (n = 50, p < 0.001), where LAN = − 1.8771 [NH3 + 4 (bark pH)] + 44.752. (For further comparisons see electronic supplementary material, figure S4.) (Online version in colour.)
Comparable results have been recorded for The Netherlands [52], showing in particular a long-term decline of acidophyte lichen species from 1991 to 2016, consistent with differences in bark pH [49]. Since these datasets focus on the principles of lichen responses to acid and alkaline gases, similar relationships can be expected in other parts of the world. For example, extremely high levels of NH3 in the Indo-Gangetic Plain [14] can be expected to be adversely affecting acidophyte epiphytic lichens in the oak forests of the Himalayan foothills, which have significant economic importance, being traded to Arabic speaking countries to make valued-added products like perfumes [53]. While data on NH3 responses have not been available until now, first analysis as part of the GCRF South Asian Nitrogen Hub [54] shows major gradients in modelled NH3 and N wet deposition both N-S and W-E across sub-Himalayan forests and at levels that greatly exceed the known impact thresholds for lichens in temperate biomes. Differential sensitivity of vegetation according to the form of N air pollution is also indicated by the results of a long-term pollution manipulation experiment at Whim Bog in Southern Scotland [55,56]. In this globally unique experiment, the effects of gaseous NH3, assessed using free-air enrichment from a line-source of NH3, are compared with the effects of wet deposited NH4+ and NO3−, as delivered to replicated mesocosms (12.8 m2; four replicate plots for each N level/form combination) through a misting system (see electronic supplementary material, §6). Treatments at this site have been continuing for 18 years since 2002, allowing examination of the long-term effects of different N forms. Background deposition to the site is estimated at 8 kg N ha−1 yr−1, with treatments achieving total inputs of 16, 32 and 64 kg N ha−1 yr−1. Assessment of plant species composition is based on recording at three permanent quadrats for each experimental plot, with each quadrat divided into 16 sub-quadrats (see electronic supplementary material, §6). The outcomes for changes in plant cover of the main species sensitive to N pollution are summarized in figure 7. This provides an analysis of ‘Eradication Time 50’ (ET50), which is defined here as the time taken to reduce species cover by 50% relative to cover at the start of the experiment. This is shown together with the ‘Eradication Dose 50’ (ED50), the cumulative N deposition over the period associated with a relative 50% reduction in cover of each species. The values of ED50 are calculated as the product of ET50 and the annual nitrogen inputs for 2002–2019 (see electronic supplementary material, figure S5 and table S6). Changes for other species included hare's-tail cotton grass (Eriophorum vaginatum), which benefited from NH3 relative to other, more-sensitive species [56]. Figure 7. Response of bog vegetation to exposure of gaseous NH3 and wet deposited NH4+ and NO3− expressed as the time taken to reduce cover of each plant species by 50% of initial values (Eradication Time 50%, ET50), and the Eradication Dose 50% (ED50) representing the total accumulated N dose that led to a halving of cover. The smallest values are most robust, with large values most uncertain, since these depend on extrapolation. (Online version in colour.)
Considering the species shown in figure 7, reindeer lichen (Cladonia portentosa) was found to be overall most sensitive, followed by red-stemmed feather-moss (Pleurozium schreberi) and red bog-moss (Sphagnum capillifolium). ET50 for NH3 at 32 kg N ha−1 yr−1 for these species were all estimated in the range of 0.6–1.6 years. By contrast, common heather (Calluna vulgaris) and heath plait-moss (Hypnum jutlandicum) were less sensitive to NH3, with ET50 values of 3.8–4.9 years, for the same N dose. Expressed as ED50, the most sensitive species had values of 19–51, while Calluna and Hypnum had values of 122 and 157 kg N ha−1 (electronic supplementary material, table S7). Overall, it is expected that ET50 values are larger at lower annual N dose rates, as shown by the left side of figure 7. By contrast, it is expected that ED50 should be independent of annual N dose rate, so the extent to which this expectation is not met indicates that the ecological response is not directly proportional to accumulated N inputs. As expected, ET50 values were larger at low N inputs and smaller at high N inputs. This is shown for impacts of NH3 (all species) and for impacts of wet deposited NH4+ and NO3− for the most sensitive species (Cladonia, Pleurozium). By contrast, significant scatter is seen for Calluna and Hypnum in response to wet deposited N, reflective of longer and more uncertain ET50 estimates, with values greater than 17 years based on extrapolation of linear fits (electronic supplementary material, figure S5). Data for Cladonia, Pleurozium and to some extent Sphagnum give the best evidence for the utility of the ED50 indicator, as the plots for these species all show little difference between N dose rate, as expected compared with the ET50 indicator. By comparison, lower ED50 values for Hypnum, Sphagnum and Calluna for the NH3 treatments at 32 and 64 kg N ha−1 yr−1, as compared with the 16 kg N ha−1 yr−1 NH3 treatment, suggest that these species are more-than-proportionately vulnerable at the higher N rates. This could point to a toxic effect as a result of higher NH3concentrations rather than dose. The most important observation from this dataset for the present analysis is that it further demonstrates the higher sensitivity to NH3 compared with wet deposited NH4+ and NO3−. This is most clearly seen by calculating statistics based on normalizing the data as 1/ED50, which also allows inclusion of all treatments, with the data then plotted as ED50 (figure 8). Overall, it can be seen that the reductions in cover of these five species occur three and five times faster for gaseous NH3 than for the same N dose of wet deposited NH4+ and NO3−, respectively. Comparable differences were also revealed by an independent assessment of physiological response for lichen transplants (electronic supplementary material, figure S6). Figure 8. Differential sensitivity of bog vegetation to gaseous NH3 and wet deposited NH4+ and NO3−, expressed as the average Eradication Dose 50 (ED50). Mean and standard errors for five plant species at three N input levels (n = 15, figure 7 and electronic supplementary material, table S7). Different letters show statistical significance with p < 0.05 (two-tail); (a,c) are significantly different at p < 0.01, based on paired t-tests of the reciprocal values.
Although it is widely assumed that effects on plant nutrition are governed by total N inputs, our data thus show that reality is more complex, otherwise there would be no difference between the wet NH4+ and NO3− treatments, as well as between the wet N and dry NH3 treatments. Other experimental studies have mainly been conducted using increased wet deposited N or fertilizer addition [57,58]. Had such studies included NH3 enhancement, they may therefore have found even larger adverse effects per unit N added. This is not to suggest that effects of wet deposited N are unimportant, but rather to emphasize the need to consider N form. For example, in agricultural areas subject to high NH3 concentrations, our results show that remaining bog habitats will be more than proportionately at risk (based on estimated N deposition). Conversely, wet deposited N may lead to larger effects on an area basis, since the largest areas of bogs are far from agricultural sources, where wet N deposition dominates total N inputs [56]. The preceding examples, showing higher sensitivity of plants to NH3 compared with NH4+ and NO3−, raise the question of whether there are also different rates of recovery following reductions in N pollution. It has been suggested that recovery in species composition and certain N pools may take several decades after a reduction in wet deposition of N [59]. This may be especially the case in slow-growing naturally acidic ecosystems, where N pools change slowly due to lack of removal by harvests and inhibition of denitrification [60,61]. There are no known experimental studies of the simultaneous reduction in NH3, NH4+ and NO3− pollution for any ecosystem. However, field data from a site in Northern Ireland illustrate the potential for surprisingly rapid recovery following a reduction in gaseous NH3 concentrations. In this case study, a poultry farm 50 m west of Moninea Bog Special Area of Conservation (SAC) led to greatly increased concentrations of NH3 (10–40 µg m−3) compared with local and regional background values of 1.5 and 0.5 µg m−3. The lichens Cladonia portentosa, C. uncialis and Sphagnum species were largely eradicated on the bog within 400 m of the farm, with excessive growth of algae on trunks of Betula pubescens (50–100 m from the farm) replacing the natural acidophyte lichen flora [62]. As a consequence of legal requirements for the protection of Moninea Bog, the poultry farm ceased operation in 2010, allowing examination of ecosystem recovery. Observations in 2017 showed mean atmospheric NH3 concentrations of 1.5 µg m−3, with substantial recovery of Cladonia portentosa and Sphagnum spp. As all large clumps (ca 200–400 mm diameter) of C. portentosa had been eradicated, only uniformly small specimens (ca 50–70 mm) were found in 2017 in the zone where they had been previously eradicated. The extent of Cladonia and Sphagnum growth indicated that recovery must have started within 2–4 years of the reduction in NH3 concentrations. Conversely, it was not obvious whether the condition (i.e. overall health) of Calluna had improved, while residual algae levels on Betula less than 100 m from the farm indicated only partial recovery. An illustration of a potential ‘alternative stable state’ [59] was also found (figure 9a). This showed continued colonization of a Sphagnum hummock by algae, 7 years after NH3 concentrations reduced. It suggests ongoing competition, where a coating of gelatinous algal slime restricts gas exchange and growth of the Sphagnum until the latter can manage to grow over the algae. Monitoring across Moninea Bog by the Northern Ireland Environment Agency confirmed the recovery of Sphagnum after 2010, with no evidence of any recovery in Calluna (figure 9b). The latter effect may be age-dependent, where recovery of old Calluna plants (weakened or dead) is limited, while ultimately conditions may favour recolonization by young Calluna plants. Figure 9. (a) Hummock of Sphagnum moss on Moninea Bog photographed in 2017, seven years after reduction in NH3 concentrations, showing a hummock still covered by algal slime characteristic of high NH3 levels (insert: partial cross-section). (b) Statutory monitoring showed an overall tendency for recovery in Sphagnum populations at Moninea Bog, but not yet of Calluna nor a return to previously lower levels of graminoids. These data confirm independent expert examination of the site by the authors in 2007 and 2017. (Online version in colour.)
The examples presented highlight the increased sensitivity of lichens and bog vegetation to gaseous NH3 compared with wet deposited NH4+ and NO3−, which appears to be at least partly related to the alkaline effect of NH3. In this way, higher correlations were found between LAN and NH3 in the UK-scale lichen survey than with total N deposition. In fact, the highest single-factor correlation was found with the ratio of N to S deposition (electronic supplementary material, figure S4). This indicator is also closely correlated with NH3 concentrations (R2 = 0.87), and, like NH3, also provides an indication of acid-base balance. Both the national and local-scale lichen surveys showed that bark pH is positively correlated with NH3 concentration, while LAN score is negatively correlated with bark pH. Hence, one of the ways in which NH3 appears to affect epiphytic lichens is by increasing substrate pH. That this is one of the driving variables is also shown by the differences between tree species, with nitrophytes found to be more prevalent on trees with naturally higher bark pH. Natural differences in bark pH can similarly explain differences in lichen communities between twigs and trunks (figures 4d and 6). This raises the question of whether the NH3 effect on lichens is entirely mediated by its effect on surface pH. Examination of the UK-scale data suggests a more complex interaction. If NH3 has its effect solely through changing bark pH, then differences in pH should fully explain the variation in LAN with NH3. This would therefore not explain why a combined indicator of NH3 + 4 (bark pH) gives a better relationship with LAN than with bark pH alone (figure 6; electronic supplementary material, figure S4). It suggests that NH3 may be affecting lichens by both a pH effect and another effect, such as that related to nutrient N (as more usually considered to drive N effects on ecosystems [57]). Possible relationships are summarized in figure 10: NH3 deposited on bark increases NH4+ on the lichen thallus, uptake of which will be under control of cell membranes. Such altered nutrient supply may affect competition between species. In parallel, the deposition adds NH3 to the bark/thallus surface, which increases bark pH because of the alkaline nature of NH3. The bark pH is also affected by tree species and bark age (twig versus trunk). With NH3 increasing bark pH, the chemical equilibrium favours NH3 rather than NH4+, which further increases NH3 levels on the thallus. Two subsequent effects may then be expected. Firstly, that changed bark pH affects apoplast pH of the lichen thallus (cf. electronic supplementary material, figure S7), which could affect lichen health (e.g. by affecting the buffering systems of lichen acids characteristic of different species). Secondly, NH3 may have a direct toxic effect, including that mediated by the passive diffusion of NH3 across cell membranes, leading to disturbance of symplastic pH. That plant sensitivity to NH3 and other forms of N deposition is partly due to different abilities of species to manage cell pH homeostasis has been argued by Pearson & Soares [63], who elsewhere demonstrated a positive correlation between leaf buffering capacity index and nitrate reductase (NR) activity across 18 plant species [64]. This would offer another reason why acidophyte lichens, adapted to NH4+ nutrition (with low NR activity expected), would be more vulnerable to atmospheric NH3. Figure 10. Possible mechanisms by which atmospheric NH3 pollution affects epiphytic lichens, including both positive (+) and negative (−) effects. Solid lines indicate observed relationships or those directly implied by physico-chemistry. Dashed lined indicate hypothesized relationships. The toxic and pH effects apply especially to acidophyte lichens, but may also apply to nitrophyte lichens at high levels of NH3 exposure (figure 4).
Effects of NH3 on lichen pH are also seen at Whim Bog. Electronic supplementary material, figure S8 shows that NH3 increased the pH of transplanted Cladonia portentosa thalli, confirming the thallus pH effect (figure 10), with responses seen within one month of transplantation. By contrast, the surface pH of live Sphagnum capillifolium growing in-situ remained unaffected, which may reflect a greater water-holding and buffering capacity of Sphagnum. The importance of such pH effects may also explain the rapid recovery of Cladonia portentosa and Sphagnum spp. following reduction in NH3 levels at Moninea Bog. Even though the peat might still contain high N levels, reduced alkalinity from less NH3 would be expected to allow rather rapid re-adjustment of surfaces, allowing colonization of acidophyte species. While uncertainties remain over the exact mechanisms, the higher sensitivity to NH3 compared with wet deposited NH4+ observed at Whim Bog tends to support this picture. Based on the values of ED50 (with NH3 being three times more damaging than NH4+), this suggests that ¾ of the NH3 effect on peatland vegetation could be related to pH effects, while ¼ of the NH3 effect is attributable to the common effect of increased nutrient N supply. One of the implications of our findings is therefore to pay more attention to the ‘critical level’ for NH3 concentrations for which the UNECE has adopted a value of 1 µg m−3 for lichens, bryophytes and associated habitats [65,66]. The extent to which such relationships can be generalized between species, habitats and world regions remains an important question for further work. Each species responds individually according to its nitrogen and pH preferences, sensitivity to NH3 toxicity and ability to compete with other species for light and other resources. For example, investigations on Cladonia portentosa from Whim Bog showed that different N forms affect different metabolic pathways [67,68], which may have varying importance between species. It is also possible to identify useful functional groups, as illustrated by the nitrophyte/acidophyte lichen groupings. Calluna vulgaris offers another illustration as this is found to be more sensitive to NH3 at Whim Bog than Cross-Leaved Heath (Erica tetralix) [56]. If it could be shown (according to [64]) that this reflects a lower apoplastic buffering capacity of Calluna than Erica, then this would encourage further use of buffering capacity as a predictive indicator. In the same way, species/group differences in characteristic lichen acids may also point towards predictive capability with global relevance, which may be tested by the GCRF South Asian Nitrogen Hub. The higher sensitivity of vegetation to gaseous NH3 compared with wet deposited NH4+ and NO3− has direct implications for the success of past SO2 and NOx emission reductions in protecting ecosystems. While the acid rain problem has now been addressed in the UK and most of Europe, the modest reductions in NH3 emissions mean that alkaline air is emerging as a new ecological challenge. The data presented here focus on naturally acidophyte species, which appear to be especially vulnerable to alkaline air. It remains to be tested whether naturally basic habitats, such as chalk grasslands, would be less vulnerable to ammonia. Already there are indications that NH3 concentrations are actually increasing in some parts of Europe rather than decreasing. While this is partly related to reduced SO2 and NOx concentrations leading to increased NH3 lifetimes, as reflected in NH3 monitoring for remote areas [38], there is also concern about climate change impacts on NH3 concentrations. As most NH3 globally results from volatilization processes, climate warming will increase NH3 emissions [69,70]. Strategies to address alkaline air therefore need to include measures that both reduce NH3 emissions directly [71] and minimize climate change drivers. In addition to control of CO2 and CH4 emissions, decreasing losses of all N compounds (including N2O, NO and N2 to air, and NO3− losses to water) becomes critical to increasing economy-wide nitrogen use efficiency, with multiple benefits for climate, air quality, water quality, biodiversity and stratospheric ozone protection [54,72]. Such a perspective could help transform current efforts to meet the EU National Emission Ceilings commitments for 2030, as well as many other policy goals. This takes us closer to developing the big idea whereby ammonia becomes a key focus in an emerging international strategy to manage the global nitrogen cycle. This is why the historical perspective of §2 is so important, in raising awareness about ammonia. One of the lessons of history is that ammonia has always been of significant societal importance. From its role as part of the alchemists' objective to prepare Gold and the Elixir of Life, to its economically high value as a luxury product of international trade, ammonia continues today to be important in sustaining humanity through nitrogen fertilizers and biological nitrogen fixation. If society is to learn to manage nitrogen better, then these stories can help by raising wider awareness. Ultimately, it may be the economic value of nitrogen that counts most. It has been estimated that global N losses to the environment amount to around 200 million tonnes [73,74]. This means that at a nominal market price of US$1 per kg N, a goal to ‘halve nitrogen waste’ from all sources by 2030 would offer a circular economy opportunity worth US$100 billion per year, amounting to an annual saving of approximately 12 kg N per person (cf. §2a). These issues have recently been recognized in the first Resolution on Sustainable Nitrogen Management adopted at the UN Environment Assembly (UNEP/EA.4/Res.14), with the ambition to halve nitrogen waste adopted in the Colombo Declaration [75]. The follow-up to these activities is bringing ammonia and air pollution together as part of the global nitrogen challenge, by working to establish an Interconvention Nitrogen Coordination Mechanism (INCOM), with targeted science support through the International Nitrogen Management System (INMS) [54,72]. Together these activities can be expected to emphasize how ammonia and the wider nitrogen cycle must be at the heart of the solutions needed for both environment and economy in working towards the UN Sustainable Development Goals. Data associated with this paper are included in the electronic supplementary material. The article was conceived and written by M.A.S. with text contributions from N.v.D., M.R.J., L.J.S., D.F., M.I.N., S.M. and P.A.W. The air quality measurements were made by Y.S.T., A.S., S.L., I.D.L. and N.v.D. and coordinated by C.F.B. with input from M.A.S. and D.F. Emission data were prepared by U.D., while M.V. performed the global analysis of gaseous alkaline fraction. Measurements at Whim Bog were made by M.R.J., N.v.D., S.L., I.D.L., L.J.S., M.I.N., S.M., with data analysis and interpretation by P.E.L., N.v.D., L.J.S., S.M. and M.A.S. The South Asian element is contributed by M.V., S.C., C.J.E., M.J., M.I.N., A.M., C.E.S., M.A.S. and other authors. The lichen surveys were coordinated by M.A.S., I.D.L., N.v.D. and P.A.W.; the analysis at Moninea Bog was led by M.A.S., I.D.L., N.v.D. and P.C., with input from S.M. and Y.S.T. The historical perspective was prepared by M.A.S. and the policy/future perspective prepared with input from M.A.S., C.M.H. and D.F. We declare we have no competing interests. This study supported by the UK Natural Environment Research Council (NERC, including grant no. NE/R016429/1 and NE/R000131/1 as part of the UK-SCAPE and SUNRISE programmes delivering National Capability), the Department for Environment Food and Rural Affairs, the Northern Ireland Environment Agency (NIEA), the UK Joint Nature Conservation Committee, the NEWS India-UK Virtual Joint Centre on Agricultural Nitrogen (supported through the Newton-Bhabha Fund, by the UKRI and the Indian Department of Biotechnology), the UKRI Global Challenges Research Fund (South Asian Nitrogen Hub), the EU NitroPortugal project and the ‘Towards INMS’ project of the Global Environment Facility (GEF) and UNEP. We gratefully acknowledge funding from the UK Natural Environment Research Council (NERC, including NE/R016429/1 and NE/R000131/1 as part of the UK-SCAPE and SUNRISE programmes delivering National Capability), the Department for Environment Food and Rural Affairs, the Northern Ireland Environment Agency (NIEA), the UK Joint Nature Conservation Committee, the NEWS India-UK Virtual Joint Centre on Agricultural Nitrogen (supported through the Newton-Bhabha Fund, by the UKRI and the Indian Department of Biotechnology), the UKRI Global Challenges Research Fund (South Asian Nitrogen Hub), the EU NitroPortugal project and the ‘Towards INMS’ project of the Global Environment Facility (GEF) and UNEP. We thank Kate Mason for literature support, Geertje Fischer for translations from Karimov (1957), Tony Simcock of the History of Science Museum, Oxford, and UK site operators, including those listed in electronic supplementary material, table S6. FootnotesOne contribution of 17 to a discussion meeting issue ‘Air quality, past present and future’. Electronic supplementary material is available online at https://doi.org/10.6084/m9.figshare.c.5099293. © 2020 The Authors. Published by the Royal Society under the terms of the Creative Commons Attribution License http://creativecommons.org/licenses/by/4.0/, which permits unrestricted use, provided the original author and source are credited. References
Page 10Discussion meeting issue ‘Air quality, past present and future’ organised and edited by David Fowler, John Pyle, Mark Sutton and Martin Williams
Page 11Discussion meeting issue ‘Air quality, past present and future’ organised and edited by David Fowler, John Pyle, Mark Sutton and Martin Williams
Page 12The formation of tropospheric ozone from the reactions of volatile organic compounds (VOCs) and NOx in the presence of sunlight is very well-established science that dates back to atmospheric chemistry research in the 1950s (e.g. [1]). Sunlight in the near-UV either directly destroys VOCs or initiates reactions that generate free radicals that can lead to the oxidation of VOCs, and the generation of peroxy radicals that can convert NO to NO2, and thus create a route to the net photo-chemical production of ozone from NO2 photolysis. Ozone at the planetary surface has impacts on health that include an exacerbation of asthma [2] and increased risk of death from respiratory causes [3]. More recent advances in the atmospheric chemistry of VOCs have included insight that the oxidation of certain species [4] leads to chemical by-products that can create new particles [5], add mass to the existing particulate matter (PM) [6] or change other properties of PM [7]. There is now a large body of observations confirming the ubiquitous presence of secondary organic aerosol (SOA) in virtually all environments [8]. A consequence is that the environmental motivations for controlling primary anthropogenic VOC emissions now go beyond the established science of ozone formation and form part of PM2.5 reduction strategies [9]. The term VOC, used frequently and interchangeably with the longer abbreviation non-methane VOCs (NMVOCs), is a catch-all term for any organic compound found as a gas in the atmosphere. In urban air, NMVOCs can encompass many thousands of different organic compounds including non-methane hydrocarbons (NMHCs), oxygenated, nitrated and halogenated species [10]. Since the term VOC is so broad, for regulatory and policy purposes, more specific definitions are used. One such technical definition is given in official guidelines issued by the European Monitoring and Evaluation Programme (EMEP) and the parallel United Nations Economic Commission for Europe (UNECE) Convention on Long Range Transport of Air Pollution (CLRTAP) for national inventory reporting of emissions. The same definition is also used in the European Commission (EC) Directive 1999/13/EC (Solvent Emissions Directive) [11] and EC National Emissions Ceiling Directive (NECD) [12]. NMVOCs comprise all organic compounds except methane which at 293.15 K show a vapour pressure of at least 0.01 kPa (i.e. 10 Pa) or which show a comparable volatility under the given application conditions. A slightly different definition of VOCs is given within the 2004/42/EC Paints Directive [13], which refers to a VOC as ‘an organic species with a boiling point less than 250°C at a standard pressure of 101.3 kPa’. The practical realization of both definitions is that most simple NMHCs with a carbon number falling within the range C2 to C14 are thought of as VOCs. Longer-chain hydrocarbons (>nC14) may fall outside of the definition, as may more highly functionalized organic compounds such as organic acids or organic peroxides. The majority of persistent organic pollutants (POPs) have lower vapour pressures than this definition, although some two and three ring polycyclic aromatic hydrocarbons such as naphthalene are considered as VOCs as well as POPs. The EC Air Quality Directive 2008/50/EC [14] includes within it a somewhat broader definition related to an ability to form ozone: ‘volatile organic compounds’ (VOC) shall mean organic compounds from anthropogenic and biogenic sources, other than methane, that are capable of producing photochemical oxidants by reactions with nitrogen oxides in the presence of sunlight The reporting of emissions as set out in treaties such as NECD and CLRTAP, in principle, takes into account all species that fall within the relevant technical definition. In the case of the UK, this is done through the evaluation of hundreds of different sources where statistical data must be collated for the purposes of reporting a single annual VOC emission total. The majority of countries report totals of emissions from various different sectors. The UK is unusual in then mapping the sectoral emissions onto nearly 700 different species. Pragmatic decisions are obviously needed with respect to those VOCs that might be measured as part of any verification or regulatory activities. Measurements may be needed to inform directly on attainment of specific health-related VOC concentration targets (for example, for benzene and 1,3-butadiene), or those VOCs that may act as an independent check that estimated emissions changes are being reflected in the ambient atmosphere. Since all VOC analytical methods have boundaries on specificity and sensitivity, a very limited range of VOCs are typically monitored in ambient air, compared with those actually emitted. Ideally, the species that are routinely monitored in air should reflect those that are indicative of the most significant emissions, and those that cause the greatest harm to health. Across Europe routine monitoring of VOCs has followed guidelines in Annex X of the Directive on Cleaner Air for Europe 2008/50/EC. This recommends the measurement of 31 individual VOCs, all NMHCs, ranging in molecular weight from ethane, through 1,3,5-trimethylbenzene. From these recommendations have followed a range of commercial monitoring instruments based around thermal desorption and the development of traceable standards for calibration to mole and kilogram [15]. The choices around which species have been measured historically has been a mixture of practically (what can be measured) informed by which were the most significant VOCs being emitted when routine monitoring became established in Europe in the 1990s. In the case of the UK, continuous VOC monitoring began around 1992. It is noteworthy, however, that unlike virtually all other air quality parameters covered by the NECD, there is no standard European reference method for measurement, although there have been efforts at standardization associated with calibration and reference materials [16]. This paper evaluates how VOC emissions have changed over the last three decades in terms of both absolute amount and VOC speciation, using the UK as a test case for a high-income country, and the possible implications for future observational VOC networks used to track progress towards emissions targets in 2030 and beyond. The National Atmospheric Emissions Inventory (NAEI) estimates UK VOC emissions from anthropogenic sources following methods in the EMEP/EEA Emissions Inventory Guidebook [17] for submission under the revised EU Directive 2016/2284/EU on National Emissions Ceilings (NECD) and the reporting framework of CLRTAP.1 Both the NECD and CLRTAP set emissions ceilings with milestone targets for particular dates. For example, the NECD sets a ceiling for 2020 that requires a 32% reduction in total UK emissions relative to 2005 levels, excluding agricultural sources. This equates to 724 kt in 2020 based on the latest NAEI estimates for 2005. The NECD then requires a 39% reduction in emissions relative to 2005 levels by 2030, equivalent to 649 kt. The NECD and CLRTAP define those VOC sources to be included and excluded from the national inventory (e.g. emissions of NMVOCs from biogenic sources are not included) and the technical definition (see earlier). The Guidebook provides estimation methodologies and default emission factors for each source category, although countries can use country-specific emission factors where these are deemed relevant. Key requirements for inventory reporting are Transparency, Completeness, Consistency, Comparability and Accuracy, and in this respect, it is important to provide a national inventory with time-series consistency going back to at least 1990 and forward to 2030. The NAEI uses a combination of emission factors from the EMEP/EEA Guidebook and emission rates provided by industry and regulators in the UK. The emission factors represent either (i) the total mass of all individual VOCs when added together or (ii) a metric defined as total hydrocarbons (THCs), a non-speciated mass of VOCs, used, for example, in the case of tailpipe road transport emissions. The speciation of the total emissions into individual VOCs is undertaken separately. Where THC factors are used, the methane emissions calculated separately are subtracted out. The NAEI uses three basic approaches for estimating total VOC emissions—top down, point source and industry-reported. In the first case, an emission factor approach (top-down) combined with relevant activity statistics is used for many combustion sources including crude oil refineries, other industrial sites and in residential buildings. Factors are also used for transport sources, as well as for some processes in the food and drink industry (for example, bread baking and whisky production) and for some uses of solvents,2 including many consumer products. A point-source approach (bottom-up) can be used where the sum of emissions is estimated/measured and reported by process operators for each emitting site within a sector. The UK estimate is generated simply by summing the emissions reported for all of the sites within a given sector. This approach is used for refinery processes, chemicals industry, oil and gas production and certain types of solvent use in industry such as for printing of flexible packaging, and coating of road vehicles. Key sources for data on point-source emissions are the regulators in England, Scotland, Wales and Northern Ireland who maintain inventories for the processes they regulate, and the BEIS Environmental & Emissions Monitoring System (EEMS) for the offshore oil and gas sector. In the case of solvent use, the third approach is used for UK emission estimates taking data provided directly by industry. However, this may come with little additional information on emission factors or activity data. Data are often supplied on an ad hoc basis and usually cover a limited number of years, and thus, a time-series may need to be generated by splicing together various datasets, possibly from different data providers. Therefore, there is a risk that a time-series might not be fully consistent and estimates for different years may be subject to different levels of uncertainty. However, industry estimates are essential for the VOC inventory: they provide data for sources where the lack of public domain data means that emission factor or point-source approaches cannot be used. Further details of the methods used in the NAEI are provided in the UK's national inventory report submitted annually to the UNECE and NECD. The latest version of the inventory is for years up to 2017 (the 2017 NAEI) as submitted in early 2019 (NAEI, 2019). Figure 1 shows the trends in anthropogenic VOC emissions from 1990 to 2017 estimated in the NAEI, grouped into 10 major source categories, plus projections for emissions in 2020 and 2030. National emissions are estimated to have decreased from a peak of 2,837 kt in 1990 to 807 kt in 2017, a fall of 72% and a consequence of major reductions from road transport and fugitive emissions from fuels. The decrease in road transport emissions has been mainly due to the introduction of more stringent vehicle emission standards such that by 2017 this sector contributed only approximately 4% of total UK VOC emissions, compared with 30% in 1990. Figure 1. UK emissions of VOCs from anthropogenic sources 1990–2017 and projections for 2020 and 2030. The solid black marker lines represent the NECD ceiling for that time period. The 2020–2029 ceiling is applicable to all the sectors included in the series minus emissions from agriculture (light blue bar). The dotted lines indicate the 2020–2029 ceiling of the revised Gothenburg Protocol and is applicable to all sectors, including agriculture. (Online version in colour.)
The largest contribution to VOC emissions in 2017 was from solvents. Overall emissions have decreased for this sector, but by significantly smaller amounts than road transport and fugitive emissions. There have also been relatively smaller reductions in emissions from industrial processes such that this sector contributed 15% of total emissions in 2017; these emissions now mostly come from the food and drink industry which have been increasing since 1990. The period since around 2000 has seen a substantial re-ordering of the relative contributions from each of the major source sectors. Figure 2 shows six sectoral emissions as an annual percentage contribution made to national VOC emissions. There has been a steep decline in the road transport contribution and steady growth since ca. 2000 in the contribution from solvent use, particularly ethanol. Based on the 2017 version of the inventory, the UK met the 2010 NECD emission target of 1,200 kt VOCs and then did so for all subsequent years to 2017 (the last year reported here). While a small further reduction in emissions is anticipated to occur up to 2020, emissions then remain fairly constant up to 2030. This situation arises because emissions from large sources important in the past, such as road transport, have already been reduced significantly, and there remains little scope for further reductions. For example, there are currently no further VOC emission reductions planned beyond the vehicle class of Euro 6/VI. Emissions from sources such as solvent use, food and drink and industrial processes remain fairly flat or increase slightly because emission factors are assumed to remain constant while activities are predicted to remain static or increase slightly (often following population change). Figure 2. Trends in sectoral contributions to national emissions of VOCs as a percentage of the overall annual national total, 1970–2017, data from uk-air.gov.uk and the National Atmospheric Emissions Inventory. (Online version in colour.)
Overall, a small exceedance of the 649 kt NECD target for 2030 is predicted for the UK unless there are further actions to reduce emissions. The size of additional reduction needed to meet the 2030 target is estimated to be approximately 30 kt yr−1 by 2030 based on current projections. Meeting those targets is likely to require an increased policy and regulatory focus on those VOCs from the solvent and industrial usage sectors, and so the next section examines which VOC species are significant contributors. There is no statutory requirement to report national inventories of individual VOC species to the NECD or CLRTAP, but a comprehensive speciated inventory is very valuable for other purposes. Given the different chemical reactivities of each component, a speciated inventory is essential for atmospheric models of ozone and SOA formation. It is also necessary for the interpretation of ambient measurements of VOCs and how such information can be used to verify inventory trends. The UK is one of the few countries to have developed a comprehensive speciated inventory for VOC emissions, and this is widely used in atmospheric modelling. The speciated inventory was first developed in the mid-1990s, but the lack of significant new data sources means that development essentially finished in the early 2000s when the methodology was published [18]. The NAEI's VOC inventory is broken down into 664 species with unique VOC profiles for each emission source. These are mostly as chemically unique species, although occasionally these are expressed as aggregate groups of VOCs, for example, ‘C13 aromatics’. The information used in the development of the species profiles came from various sources, for example industry trade associations in the 1990s provided some speciated estimates including the Solvent Industry Association, the British Coatings Federation and the British Aerosol Manufacturers Association. The solvent industry also helped with speciation of white spirit and other hydrocarbon mixtures. The Environment Agency's Pollution Inventory and similar inventories compiled by other UK regulators include some details of speciation, although the amount of detail is now much less than was the case in the early 1990s. Some analysis was undertaken of fugitive emissions at petrol stations, and species profiles were also provided by the refinery sector. The profiles for vehicle exhaust emissions were taken from the EMEP/EEA Emissions Inventory Guidebook. An important source for other sectors was the US EPA SPECIATE database. As much of the data used to develop the speciation profiles were gathered during a short period in the late 1990s and early 2000s, and since more recent data are not available, it is assumed that the species emitted by a particular source are the same in all years. While this is likely to be true for sources such as bread baking or gasoline distribution, it is probably not true of technologically evolving sources such as industrial coating processes, chemicals manufacture or the formulation of consumer and household products. There are approximately 360 different individual VOC source sectors or processes included within the NAEI, and each of these has a representative chemical speciation associated with it, so it is possible to interrogate the NAEI for annual amounts and therefore emission trends of individual VOCs. Although in total the NAEI contains data on nearly 700 VOCs, a much smaller subset of approximately 40 VOCs represents typically approximately 70% of the total mass of emissions. The 10 most significant individual VOCs in terms of mass emitted represented 45.3% of UK national emissions based on 2017 estimates. Ethanol was the most important VOC comprising 16.8% of all emissions. The recent trends in the 10 highest VOCs by mass emission are shown in figure 3. Figure 3. Estimated trends (1990–2017) in the UK emissions of (by rank order in 2017). 1. Ethanol, 2. n-butane, 3. methanol, 4. ethane, 5. propane, 6. n-pentane, 7. ethylene, 8. m-xylene, 9. benzene and 10. toluene. (Online version in colour.)
Figure 3 highlights a national trend since 1990 of decreasing emissions of simple NMHCs associated with natural gas leakage (ethane), evaporative loss of fuels (e.g. pentane, butane and toluene) and reductions in VOCs from tailpipe emissions (e.g. benzene and ethene). Over the same period, there have been increases in emissions of ethanol and methanol. The increase in ethanol in the NAEI is due to increased reported emissions from the whisky industries and in estimated domestic use of ethanol as a solvent, for example contained within personal care, car care and household products. The more recent addition of bio-ethanol to gasoline is included in the inventory, but due to the relatively low level of contemporary fugitive emissions and exhaust emission, this has not contributed significantly to this upward trend. Even for simple alkanes there have been some notable changes in the major contributing sources. n-butane, for example, is currently the second most abundant VOC in the UK inventory; in 1990, n-butane was emitted overwhelmingly from gasoline extraction and fugitive distribution losses (139.8 kt yr−1). However, by 2017, the largest anthropogenic source of n-butane in the inventory was from its use as an aerosol propellant (25.5 kt yr−1) with the gasoline/fugitive losses having been reduced to 23.3 kt yr−1. Figure 3 shows the multi-year trends in the estimated emissions of some of the most significant VOCs from a mass-emitted perspective, but it is possible to look more generally at trends in the types of VOCs being emitted, by categorizing the very large number that are individually speciated in the NAEI and then grouping by chemical functional groups. Figure 4a shows the estimated emission trends of 12 VOC functional group types, plus a further ‘other’ category for all other minor functionalized VOCs that do not fall within these 12 classes. Figure 4b shows the same data but expressed as a group contribution to the annual emissions in percentage terms for each year. The most striking feature is the increased significance of alcohols generally, with significant contributions to 2017 emissions made by 1-propanol, 2-propanol and 1-butanol as well from methanol and ethanol highlighted in the previous figure. Figure 4. (a) Trends in estimated national emissions of functional group classes of VOCs. Contribution of each functional group class to the overall annual national total, 1970 to 2017. (b) Contribution of each functional group class expressed as the percentage of annual emissions. Legend common to both plots. (Online version in colour.)
Understanding the exact location- and condition-specific impacts of a change in VOC speciation on the tropospheric ozone can only be done fully using explicit modelling and is beyond what we report here. It is possible, however, to evaluate how a change in VOC speciation may impact on the ozone-forming potential of national emissions in the atmosphere by considering the photochemical ozone creation potential (POCP) [19,20] of the VOC mixture. Using the overall POCP of the ensemble of VOCs emitted, it is possible to examine at a bulk level whether the changes in speciation might have led to a change in overall ozone-forming potential, per tonne of ‘average’ UK VOC emissions. Figure 5a shows the trends in total VOC emissions for the UK (by mass) and the normalized POCP for each year derived from the top 40 emitted VOCs in that year using Derwent et al. [21] POCP values. The 40 VOCs for which the calculation is made represent approximately 70% of overall emissions by mass and assumes that NOx and all other relevant photochemical parameters are held constant. Figure 5. Normalized POCP per average UK unit of VOCs emitted 1990–2017 (left-hand y-axis) and total mass of VOCs emitted (right-hand y-axis).
Over the 1990–2017 period, the normalized POCP of the UK VOC mixture declined slightly, by approximately 4%. A complex set of changes lie behind this; the largest change in any single VOC is the reduction in ethane emissions (a low POCP compound) and the growth in ethanol (a species of intermediate POCP reactivity). However, there are also concurrent reductions in a range of other high POCP alkenes and aromatic compounds that offset the growth in ethanol. The net change is to an average emission mixture that is very slightly lower in POCP per unit national emissions in 2017 than the mixture being emitted in 1990, but likely too small a change to materially affect ozone formation rates. Continuous measurements have been made of a range of non-methane hydrocarbons in the UK Automated Hydrocarbon Network since 1992. The shape of the network has changed over the years, although the methodology has always been based around thermal desorption—GC-FID. Further details can be found on the UK-AIR online resource.3 Since 2001, there have been four continuous GC-FID systems measuring NMHCs in the UK and of these the monitoring station at Marylebone Road in Central London has the most complete data record. Although Marylebone Road is formally a roadside location, it does experience the full range of urban emissions given its central position within the city. Other publications have reported data from the UK up to around 2008, for example Von Schneidemesser et al. [22]. Here, we show some of the more recent trends in ambient NMHCs as measured at Marylebone Road and then compare to changes indicated in the NAEI. Figure 6 shows trends for 25 different species. Each hydrocarbon has a unique behaviour over time, although three distinct trends are observable. For ethane and propane, derived largely from natural gas leakage, there have been only modest reductions in ambient concentrations in central London since 2000. For a range of hydrocarbons derived predominantly from sources such as gasoline evaporation and incomplete combustion, for example xylenes, toluene and 1,3-butadiene, there is a broadly log-linear decrease of approximately two orders of magnitude with reductions being seen up to the most recent year of observations. For some species like butenes, pentenes and 1,3-butadiene, ambient concentrations are now close to instrumental detection limits and there is considerable scatter in the data. There is a third type of behaviour for VOCs that showed initial falls in ambient concentrations in the early 2000s but that have plateaued more recently. These include i- and n-butane, ethene, benzene, ethyne and propene. This is potentially rationalized as arising from initial falls in concentrations through the reduction of emissions of these species from their road transport source, but that other urban sources also exist that have not declined and that now dominate ambient concentrations. Figure 6. Trends in selected ambient NMHCs measured at the Marylebone Road automated hydrocarbon network station in the centre of London. (Online version in colour.)
Ambient concentrations measured at the roadside cannot be expected to directly reflect inventory changes, since roadside monitoring sites of this kind are skewed to detecting changes in the locally dominant road transport source. However, it is still instructive to examine the general scale of reductions seen and the degree of agreement between observation and inventory. Table 1 shows the annual mean concentrations for a range of VOCs and the estimated change in emissions over a similar period. The nature of the Marylebone Road monitoring makes it particularly sensitive to road transport emissions, and for many VOCs, the reductions in roadside concentrations have been greater than those estimated as a percentage of overall national emissions, reflecting particular success in reducing emissions from this sector. In the case of natural gas-derived VOCs like ethane and propane, urban concentrations reported through UK-AIR have changed little in recent years, while national emissions of ethane are estimated to have fallen by 66% between 2000 and 2017 for all sources and by 52% for natural gas leakage according to the NAEI. This mismatch between trends in emission estimates and concentrations may, in part, be rationalized due to the inventory reductions in emissions occurring in remote offshore extraction industries and distribution networks, which are displaced from monitoring sites. It may also be the case that in London, gas leakage rates have been higher than the UK average which the NAEI is based on. It is also important to consider that the lifetimes of these species are long and significant hemispheric background concentrations are present. Observations in Europe have shown similar ambient trends to those reported here, with declines across several European roadside locations [23].
Making direct like-for-like trend comparisons between observations and inventories is generally not appropriate, given multiple contributing sources that define the ambient concentrations of any given VOC. Urban monitoring stations can be sensitive to a particular subset of sources along with reflecting broader patterns of long lifetime VOCs which are influenced by the regional transport. However, 1,3-butadiene is one VOC where some direct comparisons can potentially be made. The lifetime of butadiene during daytime with respect to hydroxyl radical oxidation is very short, around 30 min, and so its measured concentration essentially reflects local sources only. The only major urban source of 1,3-butadiene is thought to be road transport tailpipe emissions. In figure 7, we show the trend in roadside concentrations of 1,3-butadiene (Marylebone Road) against the inventory estimated emissions of 1,3-butadiene from the road transport sector (i.e. excluding all other sources such as refineries and fossil fuel extraction). There is a remarkable degree of agreement in trends between the estimated emissions and ambient roadside observations, with some evidence that for a period, ambient concentrations reduced faster than reported in the inventory, potentially due to overperformance of emission control technologies in the early and mid-2000s. The trends in 1,3-butadiene can be contrasted with benzene, a much longer-lived VOC, which has significantly greater diversity of emission sources beyond road transport tailpipe emissions. Figure 7. Comparison of trends in roadside 1,3-butadiene and benzene concentrations at Marylebone Road in central London with the National Atmospheric Emissions Inventory estimate of total 1,3-butadiene emissions from the road transport sector (solid black line). (Online version in colour.)
Previous sections have shown how at a national level, current emissions in the UK are now significantly influenced by VOCs released from solvent use, both industrial and domestic. As further reductions in emissions are required across Europe, and most reductions from the fossil fuel and transport sectors have likely already been achieved, reducing solvent emissions appear the most feasible route to meet future NECD objectives. Specific policies and interventions that might achieve these aims are beyond this paper, but there will likely be an overarching requirement for ambient observations to continue to provide some external verification that changes in emissions as determined by inventory reporting processes have occurred. This situation creates a measurement challenge since the solvent sector is heavily influenced by oxygenated compounds, rather than NMHCs which current online monitoring infrastructure is generally best configured to detect [24]. There are then further questions raised about where geographically representative measurements should be made if transport sources are no longer significant. The current speciation of VOCs from the class of ‘solvents and related products' in the NAEI is likely imperfect due to lack of up to date speciation information from manufacturers, but it provides some guide to the key species that are released. Figure 8 shows the fractional contributions for the most abundant 43 VOCs in this solvent emission class in the NAEI for 2017, dominated again by ethanol and methanol, but with significant contributions from ketones like acetone and 2-butanone and halocarbons including trichloroethene and dichloromethane. The figure is annotated by those species that are currently routinely monitored (in green) with those that are not (in red). Figure 8. Percentage contribution to the overall emission of VOCs from the ‘solvents and related products' class of emissions in the National Atmospheric Emissions Inventory (2017 edition). Red species are not routinely measured, and green species are included in the UK Defra automated hydrocarbon network. (Online version in colour.)
An alternative way of visualizing the impacts of changing emission speciation on monitoring strategies is to consider the most significant VOCs (either individual species or inventory lumped groups) emitted each year and then whether there is coverage in terms of ambient air measurements. We evaluate this by taking a time slice at 5-year intervals from the VOC speciation in the NAEI and rank order the top 40 VOCs emitted in each year. VOCs in table 2 with a white background are currently measured as part of the UK Automated Hydrocarbon Network, and this is mirrored widely in terms of analytical methodology across Europe and elsewhere. Those species shaded are not currently monitored routinely. It is clear that when the UK/EMEP networks for online monitoring were first designed there was very good alignment between the major VOCs being emitted and the VOCs being monitored. Of the top 20 unique VOCs emitted in 1990, 19 were captured by the Automated Hydrocarbon Network (this excludes a NAEI lumped group of >C13). Over time, however, the observational coverage has declined, to 2017 when only 13 of the top 20 VOCs were routinely monitored, including omission of two of the top three most significant VOCs emitted by mass.
It is also instructive to note those VOCs which are not as significant as they once were. Ethyne is one of the most challenging VOCs to measure using thermal desorption techniques due to its very low boiling point, and this can be a defining factor for much of the analytical instrumentation for analysis (for example, breakthrough volume). By 2017, ethyne had fallen to 82nd when ranked by mass of emissions from a position in 1990 of 21st. An argument can also be made that the various isomers of butenes and pentenes that were once important as contributors to both total mass emissions and to reactivity and ozone creation potential in the 1990s have been reduced so significantly that the effort expended on their measurement in networks may not be particularly productive. Since there is no standardized international requirement to report speciated VOCs as part of emissions control treaties, it is difficult to make direct comparisons between the UK-specific conclusions we report here and other global locations. We make some attempt to reality-check whether the influence of oxygenated VOCs is as great as suggested by UK inventories by examining the most significant VOCs detected in ambient air in the UK and in other countries, using data collected from short periods of research observations and process studies. Ambient data are, of course, not directly comparable to emission inventories at a national scale, since large point sources may not be detected at a given measurement location and there may be important seasonal factors not captured in short-term measurements that are reflected in annual averages. Many oxygenated VOCs are also produced through atmospheric oxidation, so they may be elevated in ambient air not solely because of direct anthropogenic emission, but also as a by-product of the degradation of other VOCs. This is particularly important for acetaldehyde and formaldehyde which are common degradation products, including also from natural emissions such as isoprene. We exclude those from this analysis. From the recent literature, we note the work of MacDonald et al. [25] which shows ethanol, iso-propanol and acetone as three of the four most abundant individual VOCs measured in ambient air in Pasadena, USA. From our own observations made using two-column GC-FID in various locations, we find a good degree of consistency between the top VOCs ranked by ambient concentrations in different global locations. Table 3 shows examples of the most abundant VOCs in ambient air in London, Beijing (measured during wintertime) and Delhi (during the post-monsoon period), all part of research intensive observation periods. We use wintertime data since this is a closer reflection of source emission profiles since oxidation losses are lower than in summer. The concentration of VOCs found in each location differs very significantly, with order magnitude differences between London (lowest) and Delhi (highest). However, for all three urban locations, ethanol is the most abundant VOC with methanol and acetone also found in the top 10 species in all three locations.
We note that since research observations exist for many oxygenated VOCs, it demonstrates that there are no insurmountable technical obstacles to improving routine measurement coverage [28]. This paper does not make recommendations for how best a broadened set of VOCs might be measured in automated networks, but there are several proven analytical approaches available—widely reported in the literature are GC-FID [29], GC–MS [30] and online chemical ionization mass spectrometry [31]. Indeed, the World Meteorological Organization Global Atmosphere Watch (GAW) programme has already supported a set of calibration infrastructure and measurement guidelines for some oxygenated VOCs [32], although measurements remain sparse at GAW background stations [33]. Inventory estimates of total VOC emissions in the UK show significant decreases over the past 30 years from a peak around 1990, which is mirrored in ambient measurements of some NMHCs. The most significant sources in the 1990s were NMHCs related to fossil fuel usage, in particular from natural gas extraction and distribution, gasoline tailpipe emissions and fugitive fuel losses. Policies and regulations across multiple sectors to reduce emissions have been very effective, particularly from road transport, which now contributes only a small amount (approx. 4%) to total UK VOC emissions. There has been little change in overall emissions from industrial and domestic solvent usage, and this has resulted in a growth of this sector as a proportion of national emissions. Along with this change has been a change in both the relative and absolute amounts of individual VOCs emitted. Ethanol is now the most abundant VOC emitted in the UK and overall, short-chain alcohols are the most important functional group, measured by mass. The shift in functional groups has not, however, had an appreciable impact on overall average POCP (a 4% decline) since the growth in ethanol has been balanced by losses in reactivity in other species, notably alkenes and mono-aromatics. If future changes in emissions are to be independently tested against external atmospheric monitoring, then a revised analytical strategy is needed, both in terms of species quantified and locations of the measurement themselves. The current online VOC networks used across Europe predominantly focus on the measurement of NMHCs, with only rather limited coverage of certain functionalized VOCs using off-line methods, for example to measure aldehydes. This results in many of the major oxygenated VOCs emissions going undetected and skewed in geography towards VOCs emitted from transport sources. Without some change to this position, it will not be possible to evaluate how successful future policies and technical interventions have been in reducing solvent emissions, noting that reductions in VOCs are needed in many countries to meet NECD and CLRTAP obligations for 2030. The addition of ethanol, methanol, formaldehyde, acetone, 2-butanone and 2-propanol to the existing suite of NMHC measurements would provide for a full observational coverage of the 20 most significant VOCs emitted on an annual mass basis. It may be possible for a change in future observational strategy to be brought closer to cost-neutral in operational terms by ceasing observations of certain analytically challenging non-methane hydrocarbons, such as acetylene, butenes and pentenes, which have seen their emissions fall very significantly and that are now often close to or below instrumental detection limits in UK ambient air. All data in this study are available for free download from public repositories. Data on atmospheric VOC emissions are included in the National Atmospheric Emissions Inventory available at the uk-air.gov.uk website. Ambient monitoring data from the Defra Automatic Hydrocarbon network are also available for download from uk-air. Data on which mean concentrations of VOC in London and Beijing (table 3) are based are available from the Centre for Environment Data and Analysis: www.ceda.ac.uk. Data on VOCs in Delhi are currently subject to embargo but will be available on CEDA in 2021. A.C.L. conceived the research, performed the analysis of VOC emission trends and subsequent impacts on observations. J.R.H., J.F.H., G.S., B.S.N., J.R.H. and A.C.L. contributed the observations and data for VOCs in London, Delhi and Beijing. D.C.C. provided the analysis of atmospheric trends in VOCs from the Defra Network. J.D., T.M. and N.P. undertook the development of the underlying emissions inventory and supported observations made in the Defra Automated Hydrocarbon Network. All authors contributed to the writing of the paper. We declare we have no competing interests. The authors acknowledge funding from NERC (grant nos. NE/R011531/1, NE/T00197/1 and NE/P016502/10). The authors thank the members of the Defra Air Quality Expert Group (AQEG) for their contributions to the recent analysis of VOCs in the UK, work which has greatly informed the conclusions of this paper. A.C.L. and J.R.H. acknowledge funding from NERC NC LTSS and from NERC grant nos. NE/R011532/1 and NE/T001917/1. This paper draws on data from the version of the National Atmospheric Emissions Inventory programme published in 2019 (the 2017 NAEI) and prepared by the Ricardo Energy and Environment under contract to the Department for Business, Energy and Industrial Strategy. Trends in ambient non-methane hydrocarbons measured at Marylebone Road are drawn from the UK-AIR website operated by the Ricardo Energy and Environment under contract to the Department for Environment Food and Rural Affairs. The Delhi measurements were taken as part of the DELHI-FLUX project, grant no. NE/P016502/1. We acknowledge the logistic support of Ranu Gadi and Shivani (Indira Gandhi Delhi Technical University for Women), Eiko Nemitz and Neil Mulligan (Centre for Ecology and Hydrology) and Tuhin Mandal (CSIR-National Physical Laboratory). Footnotes1 See https://www.ceip.at/reporting-instructions for reporting requirements of estimating and reporting emissions data under the CLRTAP. 2 The term ‘solvents’ is rather imprecise and could be possibly interpreted as defining some specific chemical functionality or property, although there is no universally agreed definition. In this paper, when we refer to solvents we are referring to those VOCs included in that emissions class for the purposes of international emissions reporting. In a chemistry laboratory setting, the term solvents bring to mind VOCs such as acetone, ethanol, toluene and so on, and they are indeed major species in solvent inventories. For emissions reporting purposes, a wider range of VOCs are included under the definition including aerosol propellants such as butane and some long-chain aliphatic compounds. 3 See https://uk-air.defra.gov.uk/networks/network-info?view=hc for further details of the UK Automatic Hydrocarbon Network. One contribution of 17 to a discussion meeting issue ‘Air quality, past present and future’. © 2020 The Authors. Published by the Royal Society under the terms of the Creative Commons Attribution License http://creativecommons.org/licenses/by/4.0/, which permits unrestricted use, provided the original author and source are credited. References
Page 13The unprecedented societal response to the ongoing COVID-19 pandemic has led to significantly reduced economic activities in the Northern Hemisphere since late winter and spring of 2020. Lower levels of air pollution were reported throughout the period as a consequence of the shutdown of numerous activities and shifted or halted mobility and working patterns. Among the decreasing pollutants, NOx, the sum of nitrogen oxide (NO) and nitrogen dioxide (NO2), is the most important precursor of tropospheric ozone (O3), that in turn is toxic to crops, (semi-) natural vegetation and humans. At mid- and high-latitude regions of the Northern Hemisphere, O3 photochemical production is low in winter due to low sunlight conditions and temperatures, but increases rapidly in spring and summer. The lock-down has caused a reduction in the NO2 column by up to 30% in Europe and North America and by up to 50% in parts of Asia during spring 2020, as shown by satellite imagery (§2). Although O3 is not expected to decrease by the same proportion, such an abatement of NOx will considerably reduce ground level O3 concentrations (§3) and O3 impacts on ecosystems, and potentially improve the productivity of crop, forests and grasslands. Extensive evidence of O3 impacts on crops has been collected through controlled experiments during the past four decades [1]. These experiments have been used to develop exposure response relationships (ERRs). Application of these ERRs in risk assessment studies suggests that ambient levels of O3 across important agricultural regions cause yield losses to staple crops (wheat, rice, maize and soya bean). In Europe, this scientific evidence has supported the UNECE's Convention on Long Range Transboundary Air Pollution (CLRTAP) to establish critical levels for O3, which are essentially air quality targets for air pollution emission reduction policies. Despite the reduction of NOx emissions by as much as 40% since 1990 in Europe and North America, these critical levels are still frequently exceeded. For instance, the recent CLRTAP assessment report [2] estimates current wheat yield losses due to O3 in Europe of the order of 13%. A range of O3 and associated ERRs metrics exists to estimate crop losses [1]. In §3, we use the simple concentration-based AOT40 metric to demonstrate the potential benefits for crops of emission reductions during the COVID-19 lock-down. In §4, we explore the opportunity to gain additional insight into the validity of other concentration or flux-based metrics that have been developed to assess O3 damage [1], as well as more recently developed crop growth models that incorporate O3 effects. The use of metrics to perform national and international O3 risk assessments stems mainly from the air quality impact research community and has not been mainstreamed into agronomic sciences. For instance, to our knowledge, no crop model used for operational crop yield assessments or crop forecasts incorporates the interaction between O3 and plant physiology. It remains to be determined whether the decrease in O3 exposure [2], as a consequence of the reductions of its precursor emissions in Europe, has led to increasing crop yields in recent decades. One of the main challenges is to isolate the overall benefits of O3 reduction on crop yields from other factors such as weather variability and management factors. This paper suggests that the unprecedented and unintended COVID-19 lock-down in 2020 provides scientifically relevant information to quantify the actual O3 impact on crops. These unusual conditions caused an in vivo atmospheric experiment, whose magnitude could have generated sizeable reductions in surface O3 levels, and resulting increases in crop production in 2020. The subsequent analysis of agricultural statistics and application of O3 risk assessment and crop modelling will allow a comparison of the predictive ability of different methodologies to estimate regional-scale crop yield loss due to O3. The best near-real-time information on emission changes is available for NOx. Emission changes of other O3 precursors are more difficult to derive from observations. The European Environment Agency (EEA) reports declining NO2 concentrations in several cities in Europe [3], as a consequence of the reduced activities associated with the COVID-19 outbreak. The data show consistent decreases in concentrations registered at road-side and background (sub-)urban monitoring stations from March to May 2020. The Copernicus Atmosphere Services (CAMS) also report reductions in NO2 concentrations [4], but caution that the use of highly variable time series of less than one month may lead to spurious conclusions on emission changes. Therefore, we focus on average values for March, April and May (MAM), with May the latest month available to us during the revision of this publication. Data from TropOMI/Sentinel5P (figure 1) show that persistent NO2 reductions in Europe were not confined to cities alone. We provide in the electronic supplementary material maps similar to figure 1, but for separate months to provide further insight in the temporal development of the NO2 reductions. Comparing MAM average NO2 columns in 2020 with 2019, large reductions are visible over extended regions of Europe, amounting to ca. 20% in Germany and the Benelux, 15% in Italy, 10–15% in Spain, France, the United Kingdom, Poland and Czech Republic, and 8% in Romania. Regions of emission reduction largely overlap with regions with extensive wheat production. Urban NO2 column reductions in Brussels, Dusseldorf and Paris are a few per cent higher than countrywide decreases, Milano's reduction of 27% is 10% higher than for Italy, and the 33% reductions in Madrid are markedly higher than the average 13% for Spain. There is significant uncertainty in estimating NOx emission changes from 2020 to 2019 NO2 column changes, related to uncertainties in the satellite retrievals [5], the photochemical conditions of the atmosphere, but also due to inter-annual variability related to weather-related transport patterns. However, the changes in NO2 column in urban conglomerations and entire countries between 2020 and 2019 are clearly attributable to lock-down-related emissions variations, while smaller changes in cleaner areas can display residual inter-annual variability, which may obscure changes related to the COVID-19 lock-down. Figure 1. (a) TropOMI/Sentinel5P NO2 tropospheric column [µmol m−2] average for March–April–May 2019 over Europe (b) for 2020 over Europe (c) the difference of March–April–May 2020–2019 over Europe. The green areas indicate soft and durum wheat areas of 500 ha and larger (d) the difference of March–April–May 2020–2019 over Asia. (e) the difference of March–April–May over North America. (Online version in colour.)
In Asia, significant declines in MAM NO2 column by more than 50 µmol m−2 are found over several urban conglomerations (figure 1), corroborated by a similar analysis by ESA [6]. The largest declines in Asia are in March, with levels in April and May more similar to 2019 levels (electronic supplementary material S2). The MAM average emission reductions between 2020 and 2019 are of ca. 15% in Wuhan/Shanghai, 20% in Macao/Guangzhou, 20% in Tokyo, 18% in Beijing and Seoul—in March reductions of over 50% were seen. In North America significant MAM averaged declines of 10–15% are also found over the Great Lakes, East and West Coast areas. In March, declines (up to 30%) were higher than in April and May (ca. 10%). In Europe, these results can be compared to an earlier analysis of a step-wise emission decline of 20% in 2010 [7], which was at least in part due to a temporary reduction in emissions, resulting from the global economic recession in 2008–2009. Therefore, we note that following this initial survey, which includes data up to the end of May 2020, further analysis over longer periods is needed to corroborate these column and related emission changes, and how these will affect ground level O3 concentrations. However, the compelling observational evidence of strong emission declines motivates our call to the wider research community to collect and analyse data on all related aspects of emissions, air quality and crop production. Surface O3 concentration depends on the magnitude and ratio of the emissions of precursor gases (e.g. NOx and VOCs), photochemical reactions, atmospheric conditions (weather), removal processes at the Earth's surface and hence on local, regional and seasonal factors. In most regions, O3 declines with decreasing NOx emissions; however, in some traffic-intensive urban regions dominated by high-NOx emissions, the O3 response to declining NOx emissions may be initially positive, as reduced NO concentrations also reduce O3 titration close to sources, but as the plume of pollutants is transported away from urban areas net O3 production begins [8]. Detailed atmospheric chemistry transport models are generally used to evaluate the variety of regional responses to reductions in the mix of emissions. O3 can also be transported between regions within Europe and over longer intercontinental distances [9–13]. For instance, Jonson et al. [13] estimate that for a scenario assuming 20% reduction in global anthropenic emissions, up to 60% change in Phytotoxic Ozone Dose over forests (POD1) in Europe is due to changes in other regions. Therefore, it is likely that strong emission reductions in Asia and North America may also have influenced O3 in Europe. To provide a gross estimate of the effects of the COVID-19 emission reductions on O3 air quality and its effects on crop production, we develop six illustrative scenarios. In scenario S1, NOx and NMVOC emissions from transport, energy and industrial sectors are reduced by 30% in Europe. In scenario S2, the same reductions are applied worldwide. Scenarios S3 and S4 assume 50% lower emissions in Europe and the world, respectively. Scenario S4 is very similar to a recent O3-carbon cycle impact study [14], in which an emission reduction of 50% in these sectors was identified to dominate the overall positive impacts on Gross Primary Production of crops with a C3 metabolism (e.g. wheat). Scenarios S5 and S6 consider that emissions of international shipping were reduced by 30 and 50%, respectively, while international aviation emissions were down by 80% in both S5 and S6. As the exact timing of emission reductions is not known, for simplicity we assume in the scenarios year-round reductions, bearing in mind that for wheat the most O3-sensitive period is approximately during the grain-filling period (approximately end of May and June). We note that emission reductions of this magnitude or even higher are projected by 2050 if aggressive air pollution and climate polices are implemented [15,16], and changes during the COVID-19 lock-down may be indicative of the benefits of emission reductions projected over a much longer timescale. After the submission of this paper some more informal and peer-reviewed analyses on emission reductions have become available, which are now worth mentioning at the revision stage of this paper. The worldwide number of flights was sharply cut by 25% in March and by 60% in April and May 2020 [17]. Commercial shipping trade analysis [18] shows that in the five weeks after 12 of March 2020 the number of ship departures from major hubs declined by 6%, of which tanker traffic was mostly affected. Due to the slow response of the shipping sector to changing demand conditions these reductions are likely to become higher in the next months. A recent estimate [19] of impacts of the COVID-19 lock-down conditions on regional and global CO2 emissions showed that in April 2020 about 80% of the world's CO2 was emitted in regions affected by lock-down conditions, and the highest daily decline of global CO2 emissions was estimated for 7 April to range from −11% to −25% (central value −17%), with similar declines during the remainder of April 2020. Apart from residential emissions, all sectors analysed in [19] showed declines ranging from −7% for power generation to 60% for aviation, in agreement with the scenarios presented here. Depending on confinement levels (social distancing only, medium or stringent measures), surface transport emissions were reported down [19] by −10% to −50%, although numerous web reports also suggest larger reductions in selected cities. An overall annual impact on CO2 emission in 2020 was estimated [19] to range from −3% to −13%, with a central value of −7%, assuming a return to pre-COVID conditions in the middle of June 2020. Overall, the reported emission reductions support the choice of sensitivity studies presented in this publication, including the assumption that effects could very likely extend through June and July 2020. By a direct comparison with a variety of sources of information available at the time of the revision of this paper, we can conclude that scenarios S3, S4 and S6 are probably upper limits for the real-world impacts. Some additional simplifying assumptions have been made in this study. For instance, we have not considered CO emission reductions, which could have some further minor impact on O3 formation [10]. Although the emissions of methane (CH4), another important O3 precursor, are probably also affected by the lock-down, and the reduction of air pollutants can influence the CH4 lifetime, the overall impact on CH4 and O3 concentrations is not a-priori clear. As any effect will play out on a timescale of 10 years (CH4's lifetime), it will not likely be discernible within 2020, but may become substantial in the following years. To estimate possible impacts of such emission reductions on crop yields, we use the TM5-FASST global source-receptor model [20]. In summary, TM5-FASST considers a set of 56 global regions and two global sectors (aviation and shipping), to determine the regional (grid-level) response to emission reduction of air pollutant precursor emissions. For each region and sector, the response of hourly O3 changes and corresponding impacts on crops is calculated using ERR available in the literature [20]. In this publication, we estimate impacts on crops using AOT40, which is a metric based on the cumulated concentration of hourly surface O3 above 40 ppb to which crops are exposed during a three-month period in the crop growing season. Comparison of TM5-FASST results with other studies described in [20] show coherent results among models and ERR methods, but the limitations of AOT40 and other ERRs should be noted, and are further discussed in [1]. Focussing on the scenario S4 + S6, the global 50% lower emission scenario, seasonal O3 changes in May–June–July (figure 2), range from 10 to 12 ppb O3 decreases in China and other parts of Asia, 2 to 6 ppb in North America, and less than 2 ppb in north-west Europe to up to 6 ppb in southern Europe. In particular, in northern Europe, and some other urban conglomerations in North America and Asia, emission reductions caused increases in O3. Accurate determination of O3 responses in such regions would need high-resolution models as well as high-resolution information on emission declines in those regions. For the 30% global emission reduction case (S2 + S5), O3 decreases were ca. 7 ppb in Asia, 1 to 4 ppb in North America and less than 1.5 ppb in Europe, respectively. Figure 2. (a) Ozone responses (ppbv) calculated by TM5-FASST in Europe and Northern Africa, (b) North and middle America and (c) Asia (c) to global emission reductions by 50% in the industry, energy, transport, shipping and aviation (–80%) sectors, i.e. the sum of scenario S4 and S6. Isolated white regions correspond to near-zero or negative ozone responses due to declining emissions, which can occur in regions with high ratios of NOx to VOC emissions. (Online version in colour.)
Impacts on wheat yields for the six scenarios are presented in figure 3. Overall improvements of wheat yields range from 1–4% in case of worldwide emission reductions of 30% (S2 + S5), and 2–7% for reductions of 50% (S4 + S6). The contribution of European emission reductions to yield improvements in a set of European countries is relatively small, ranging from approximately 0–0.5% in northern European countries to 3% in Italy (S1, S3). Shipping and aviation contribute up to 1–3% for scenario S6. Yield improvements of up to 7% were calculated for Asia and North America (S4 + S6). Figure 3. AOT40-based wheat yield increases (%) in selected European countries, USA, China and South Korea due to emission reductions by 30% and 50% in the energy, industry and transport sectors in Europe (S1, S3, green), world (S2, S4, blue) and international shipping + aircraft sectors (S5, S6 grey). The upper/lower part of the stacked bar represents the 50% and 30% emission reduction scenario, respectively, while aviation emissions were down by 80%. The total yield increase (blue) is the sum of the world and ship/aviation. Reference emissions were taken from the ECLIPSEv5a emission database [21] for the CLE-2020 scenario. Energy, industry and transport emissions amount to 56.6 and 6.2 Tg NO2 yr−1 for the world and Europe, respectively. International shipping and aviation emissions are 23.0 and 3.4 Tg NO2 yr−1, respectively. (Online version in colour.)
Over the past few decades, a wide variety of concentration-based and flux-based O3 metrics have been used to develop ERRs for use in risk assessment studies. These studies have explored exceedance of critical levels, as well-estimated relative and absolute crop yield losses from anthropogenic O3 concentrations [1]. However, substantial differences in estimates of the magnitude as well as of the spatial distribution of yield losses have been found using different methods. For example, differences of up to a factor of two have been found when estimating yield losses using AOT40 versus POD metrics [22]. In addition, there are some long-term (5 to 10 years) statistical studies that try to identify the O3 signal in agricultural yield statistics by performing regression analysis of meteorological and O3 data. The results are not always consistent with empirically based risk assessment models [1]. This has led to uncertainty on the actual size of effect and spatial distribution of O3 on crop yields. To address these uncertainties in risk assessment modelling, the Agricultural Model Intercomparison and Improvement Project (AgMIP) has started an activity to evaluate and enhance crop models with an O3 component. Some recent examples of the use of such crop models [23,24] show good potential to replicate O3 impacts found in field studies. However, crop growth models used for operational agronomic analyses do not include O3 impacts. For example, the WOFOST model, currently used by the European Commission Joint Research Centre to provide operational analysis of crop growth development and yield forecasts [25], does not explicitly consider the effect of O3 on crop phenology and growth. In spite of this, the yield forecasts for which it is used can usually achieve an accuracy within ±3% [26]. This does not rule out the existence of an effect of O3 on yields, since such a signal is likely indirectly hidden in other climatic factors, e.g. air temperature and solar radiation [27], and can be removed during the post-calibration of the model results against a reference of historical yield data. If this is the case, we might expect that the explicit inclusion of O3 effects on crop development and yield in WOFOST and other crop models would further improve their performance for operational assessments, especially for those locations and years where O3 impacts can be high and vary between years. Understanding the combined impacts of future climate and air pollution projections further requires inclusion of O3 impacts in crop models. COVID-19 has led to a myriad of societal consequences, including a strong decrease of economic activities, with grave impacts on livelihoods and society as a whole. Nonetheless, if the unintended in-vivo atmospheric experiments during COVID-19 result in substantial reductions in O3 and subsequent increases in crop productivity, it will allow us to evaluate and compare the different O3 metrics, risk assessment methods and crop growth models that have been developed. The new insights gained will support future development of operational agronomic analysis. Such analysis would need to be performed on careful consideration of other COVID-19 related factors (i.e. management decisions in response to expected returns) that may co-determine yields. Europe-wide, preliminary information [28] does not provide evidence of large-scale socio-economic responses by farmers, but this information needs to be corroborated at the end of the season. Specifically, we see the following research opportunities and steps to take related to reduced emissions during the COVID-19 lock-down in 2020 and O3 impacts:
Following the methods developed in the UNECE CLRTAP Task Force on Hemispheric Transport of Air Pollution, and ICP Vegetation regarding hemispheric O3 modelling and impacts of ozone on vegetation and AgMIP modelling of ozone impacts on crops, coordinated modelling activities at the end of the 2020 cropping season can improve process understanding and model quality, ensuring the representation of the variety of modelling methods that currently exist. The TM5-FASST model is accessible at https://tm5-fasst.jrc.ec.europa.eu/. The Sentinel 5P L2 products are converted to L3 by the Google Earth Engine. Further data processing (monthly mean, monthly differences etc.) has been made using GEE APIs and code editor. Google Earth Engine was used in full compliance with the terms and conditions imposed by Google Inc. on usage rights. F.D., L.E., S.G., G.C., A.I. and M.v.d.B. contributed to design and interpretation of results and drafting the paper. R.V.D. performed TM5-FASST simulations, and D.M. retrieved and processed satellite data. L.E., F.D. and M.v.d.B. can be contacted for inquiries on follow up work in the context of AgMIP O3. We declare no competing interests. A.I. and D.M. acknowledge partial support by a grant of the Romanian National Authority for Scientific Research and Innovation, CCCDI-UEFISCDI, project number COFUND-SUSCROP-SUSCAP-2, within PNCDI III. L.E., G.C., A.I. and D.M. received funding from the ERA-NET SUSCrop project SUSCAP. This is a contribution to the AgMIP O3 activity. We acknowledge discussions with Dr Rob Maas and Prof. John Burrows. We thank the editor for suggestions and guiding this paper. The submitted version of this paper was made available on the pre-print server https://eartharxiv.org/de9fs/ FootnotesOne contribution of 17 to a discussion meeting issue ‘Air quality, past present and future’. Electronic supplementary material is available online at https://doi.org/10.6084/m9.figshare.c.5056751. © 2020 The Authors. Published by the Royal Society under the terms of the Creative Commons Attribution License http://creativecommons.org/licenses/by/4.0/, which permits unrestricted use, provided the original author and source are credited. References
Page 14Discussion meeting issue ‘Air quality, past present and future’ organised and edited by David Fowler, John Pyle, Mark Sutton and Martin Williams Page 15Air quality and climate change are inextricably linked. Not only do some air pollutants have a direct effect on radiative forcing (e.g. ozone (O3) and particulate matter (PM), including black carbon) and thereby climate change, but changing climate can affect air quality. Furthermore, given the commonality in emissions sources, mitigation options will likely affect both air quality and climate change. For example, fossil fuel burning—whether for energy, transport, or otherwise—emits both the greenhouse gas carbon dioxide (CO2), as well as air pollutants such as nitrogen oxides, sulfur dioxide, mercury and particulate matter. Agriculture is a significant source of the greenhouse gas methane, as well as air pollutant emissions of particulate matter and reduced nitrogen. Emissions of reduced nitrogen, which include ammonia, ammonium and organic nitrogen, are increasing and are projected to continue to increase in importance globally, due to decreases in emissions of nitrogen oxides coupled with increases in agricultural and other sources of reduced nitrogen [1]. Reduced nitrogen can be an important component of PM, but also comprises an increasingly larger fraction of atmospheric nitrogen deposition, with implications for croplands and ecosystems. Mercury, an important hazardous air pollutant, has multiple emission sources in addition to fossil fuels (mostly coal). The most important of these are artisanal and small scale gold mining, cement production and non-ferrous metal [2]. Atmospheric processes also link air quality and climate. For example, concentrations of methane affect the concentrations of ozone in the troposphere, which serves as not only an air pollutant, but also a greenhouse gas. Furthermore, the lifetime of the greenhouse gases and air pollutants in the atmosphere determines the temporal and spatial scale of their effects. Changes in the composition of the atmosphere are generally the result of long-term processes such as the slow evolution in emissions associated with increases in population; increases in fuel, food and material consumption; changes in technology or the gradual implementation of air pollution control policies. However, extreme events can cause abrupt changes in air quality, such as volcanic eruptions or the recent (temporary) decrease in emissions associated with the global COVID-19 pandemic. These linkages demonstrate some potential co-benefits and trade-offs that should be considered when formulating air quality and climate change policies [3–5]. A detailed overview of chemistry–climate interactions is provided in Archibald et al. [6], or previous review articles [7–10]. In this paper, we synthesize the current understanding of the influence of climate change on future air pollution effects. Of particular relevance is the concept of a ‘climate penalty’, whereby future climatic conditions in a warming world will exacerbate the challenge in reaching air quality targets or standards. Greater emissions reductions will be necessary to attain the same air quality target in a warmer climate in comparison to a stationary climate scenario [11,12]. The climate penalty has been widely established in the literature for ozone, due to the role of photochemistry (sunlight) in ozone production, whereby increases in air temperatures enhance ozone formation [13–16]. Climate change effects are less understood for PM, due to the diversity of particulate matter components, formation and removal mechanisms, and the role of phenomena such as wildfires which are intensified under climate change [17–21]. Changing climate not only influences atmospheric processes, but also emissions. For example, higher air temperatures will increase evaporative emissions of anthropogenic volatile organic compounds, as well as emissions of biogenic volatile organic compounds [22,23] and ammonia. Furthermore, emissions of methane from wetland ecosystems, permafrost environments and other critical ecosystems are likely highly sensitive to climate change [24]. Finally, policies implemented to mitigate greenhouse gas emissions have considerable potential to improve future air quality. There is considerable potential for co-benefits whereby either climate or air quality policies could greatly improve environmental quality if interactive effects are considered [3–5]. The air quality co-benefits from climate policies may in some cases even exceed the direct climate benefits (e.g. [25]). Understanding these feedbacks and interactions sufficiently to accurately depict them in models and thereby improve future projections is a critical need to inform and guide coupled management of air quality and climate. Future projections are most often investigated through the use of scenarios. Alternatively, case studies of extreme events that might depict future conditions are also used for understanding and quantifying possible future effects (e.g. [26–29]). In the current context of the Anthropocene, human activity has a dominating influence on climate and the environment [30]. Here, we review the state of science that aims to quantify how expected changes in future air quality affect human health, ecosystems and food security. Climate change is an important driver of future air quality. However, additional, often less considered factors include changes in population demographics, adaptation measures and feedbacks across the sectors of human health, agricultural systems and ecosystems. A variety of climate scenarios are commonly used, especially in the context of changes in air quality and effects on human health and crop yields. Among the most common are the representative concentration pathways (RCPs) [31]. These four scenarios were designed for the climate modelling community to span the range of plausible emissions futures and the associated changes in radiative forcing in support of the Fifth Intergovernmental Panel on Climate Change (IPCC) Assessment Report [32]. The scenarios include land use changes and emissions of air pollutants and greenhouse gases extending to 2100, with the endpoints of the four scenarios associated with the amount of radiative forcing. The scenarios range from RCP2.6 (2.6 W m−2) which represents a scenario that would limit global temperature increase consistent with a 2°C target, and requires greater nuclear and renewable energy, as well as carbon capture and storage, to RCP8.5 (8.5 W m−2) which represents a business as usual scenario with continued reliance on fossil fuels and no specific policies to limit greenhouse gas emissions, with intermediate scenarios for RCP4.5 (4.5 W m−2) and RCP6.0 (6.0 W m−2). While the scenarios include air pollutant emissions, they do not represent the possible range of air pollution pathways, instead generally assuming stricter emission controls on air pollutants over time in all four scenarios [33]. The main difference among the RCPs is the pathway for methane emissions which increases under RCP8.5, due to the projected increase in livestock and rice production, and will, therefore, have the largest impacts on the changes in the temporal and spatial pattern of concentrations of tropospheric ozone [34]. Nevertheless, these scenarios are used in many of the studies evaluating developments in future air quality effects on human health. Additional sets of scenarios used by some studies included in this review are the Special Report on Emissions Scenarios (SRES) published in 2000 by the IPCC for future assessments of climate change and possible response strategies [35], or the shared socioeconomic pathways (SSPs) that provide a set of alternative reference assumptions about future socioeconomic development in the absence of climate policies or climate change, that complement the radiative forcing pathways provided by the RCPs [36]. A number of studies evaluate emissions scenarios for current legislation (CLE) and maximum technically feasible reduction (MFR) as provided in the ECLIPSE v5a (Evaluating the Climate and Air Quality Impacts of Short-Lived Pollutants) emissions inventory [37]. For both the SSPs and the scenarios from ECLIPSE, this information has only become available recently, and therefore the literature published using these scenarios is limited. Finally, a handful of studies also evaluate policies or policy proposals that are linked to local or regional legislation. Although future air quality will likely have adverse effects on human health, crops and ecosystems that will be exacerbated under climate change, few studies have examined the linkages of impacts across multiple sectors (figure 1). Studies tend to evaluate these broad types of effects in isolation. There is substantial literature on projected estimates of the effect of future air quality on human health, from both global and regional perspectives. The papers included in this review for the human health effects reflect the past 6 years, extending back through 2014. To this end, we performed an indexed search in the Web of Science (index terms: air pollution and health, air quality and health, ozone and health, PM and health). Further, we included studies not identified directly by the Web of Science search but cited in studies emerging in the identified literature. For an overview of earlier multi-model studies and reviews, the reader might see [38–41]. Given the impact focus, the studies included here explicitly focus on changes in future health burdens (such as changes in mortality, disease cases or monetized effects). The rich body of literature focusing on future air quality but not addressing changing health burdens has been omitted. Similarly, those studies addressing changes in health burdens linked only to changes in climate or other factors in which air quality changes are not quantified are also not included. There is far less literature on projections on effects of future air quality on crop yields, thus studies as far back as 2009 are included. All studies available through reasonable search which quantifiably related air quality to future crop production were included. The existing studies tend to focus on four main crops: wheat, maize, rice and soya beans. Both sets of literature that evaluate the effects of future air quality on human health and crops tend to use scenarios as outlined in the previous section, with the majority of studies based on the RCPs. By contrast, studies that project changes in the structure and function of ecosystems under future air quality and climate generally have not employed these types of scenarios (but see [42–46]), and those that have generally focus on climate impacts rather than air quality. There are a number of reasons for this. First, there are a broad range of ecosystem effects of air pollution, such as soil and freshwater acidification, mercury contamination, eutrophication of terrestrial and marine ecosystems, visibility, among others, each requiring separate modelling efforts. Ecosystem impacts often require simulation of linked hydrologic, biogeochemical and biotic processes involving the dynamics of multiple elements within a watershed or water body to project responses (e.g. [47,48]). Second, ecosystem disturbance by air pollution generally impacts large spatial regions. Generally, dynamic models of ecosystem effects of air pollution have not been well validated across the heterogeneous landscapes, which encompass a range of topographic, meteorological, hydrologic, edaphic and vegetation conditions. Finally, the developed future climate scenarios do not include some of the key air pollutants critical to ecosystems, such as mercury. Given this, we provide observations on the scope of air pollution effects on ecosystems and how these responses are altered by climate, as well as three case studies that demonstrate the scope of these effects. Figure 1. Conceptual diagram showing sources of air pollution and potential effects on human health, agroecosystem function and ecosystem structure and function. Human health is impacted by fine particulate matter and ozone. Crops are largely affected by ozone, but also light scattering from particulate matter and nutrient transport. Ecosystem effects include uplands, freshwaters and coastal and marine waters and involve ozone impacts, acidification, eutrophication and mercury effects. Note it also envisions linkages among human health, agriculture and ecosystem sectors. (Online version in colour.)
An overview of the 47 studies that examine future projections of changes in air quality related health burdens and mortality, as well as other human health indicators, between 2014 and early 2020 is given in table 1. Here, we provide a brief overview and some highlights from these studies, starting with global assessments. A couple of global studies focused solely on aerosol related health burdens [51,56], while most evaluated health effects of changes in both ozone and PM (generally PM2.5 rather than PM10, particles with an aerodynamic diameter of 2.5 µm or less and 10 µm or less, respectively, owing to the greater relevance for adverse health effects). The majority of global scale studies also provide detailed regional information for the USA, Europe and/or Asia. The overview of literature with a global perspective is then complemented by brief insights into regionally focused projections, including examples highlighted from each region. Global projections that include regional detail are not included in the regional sections. No regional studies were identified that examined scenarios of air quality and the associated health effects in Central and South America or Africa. In a nutshell, studies agreed that emission trajectories attaining 2°C climate targets, clean energy/transportation scenarios, and low to moderate emission scenarios among the RCPs/SSPs will lead to improved ambient air quality and reduced health burdens, while air quality will further deteriorate and increase health burdens under business as usual trajectories and high emission RCPs/SSPs. Studies also found that the monetized health gains of reduced PM and ozone burdens exceeded the implementation costs of climate protection and emission reduction measures.
A broad range of health metrics and health endpoints are used in air quality impact studies. Among the most prominent are excess deaths, defined as the increase in the mortality rate (number of deaths) due to the environmental factor of interest, or conversely avoided deaths, defined as the mortality rate (number of premature or ‘untimely’ deaths of people younger than a defined age threshold) avoided by interventions (e.g. [92]). Besides those metrics, modified mortality measures taking an individual's remaining life expectancy into account are frequently used. These include e.g. the number of years of life saved, defined as the remaining life expectancy at the point of each averted death, and conversely, years of life lost, defined as the lifespan lost due to premature mortality attributable to an environmental effect (e.g. [93]). Besides these mortality indices long-term health effects described through diverse morbidity metrics are frequently used in impact studies, including cardiovascular, respiratory or neurological endpoints, hospitalization rates, cancer, diabetes and negative pregnancy/birth outcomes (e.g. [94]). Mortality (and morbidity) resulting from ambient air pollution is also frequently monetized to quantify the benefits added by a new policy, with the most frequent economic metric used being the value of statistical life (with the valuation extended to incorporate morbidity effects), placing a monetary value on reductions in the risk of a fatality (e.g. [95]). Here, we focus results on mortality in terms of excess or avoided deaths whenever possible as it is one of the most common metrics used across studies. To provide a broader context of how air pollution and the global burden of disease might develop in the future, the work of Lelieveld et al. [50] provides a good starting point. These authors investigated the contribution of outdoor air pollution sources to premature mortality on a global scale for the recent past and first half of the twenty-first century considering a business as usual scenario, assuming only currently agreed legislation (CLE) would be implemented to affect future air pollution emissions. Their estimate of global premature mortality attributed to air pollution in 2010 was 3.30 million deaths (142 000 from ozone, the remainder from PM2.5), considering a global population of 6.8 billion people. They projected an increase in air pollution attributed premature mortality to 6.6 million deaths in 2050 (358 000 from ozone, all others from PM2.5) [50]. While these premature deaths were dominated by the fraction occurring in Asia, increases were also projected for Europe and the Americas, largely in urban areas. Finally, it was estimated that the per capita mortality rate, which was 50% higher in urban versus rural environments in 2010, is expected to increase to nearly 90% higher in urban versus rural environments by 2050 [50]. One caveat to note with this study (and other global studies) is the mismatch in the spatial resolution of the atmospheric model used to calculate pollutant burdens and the resolution of present/future population datasets used along with these burdens in the exposure response functions determining mortality. This offset can be partially accounted for by aggregating the anthropogenic emission data to the model grid and by updating the exposure response functions for the use of annual mean concentrations to estimate mortalities [96]. By comparison, Likhvar et al. [49] also evaluated a scenario of currently agreed legislation, as well as a scenario of maximum feasible reductions (MFR) [97] to quantify premature deaths attributed to changes in PM2.5 and ozone air pollution. In their evaluation, the per cent of the global population that is exposed to concentrations of PM2.5 above the WHO guideline value of 10 µg m−3 annually was 34% in 2010, and would increase to 42% under the CLE scenario, and decrease to 1% under the MFR scenario by 2030, with strong regional differences in the magnitude of the effects. These changes in PM2.5 under the MFR scenario in 2030 could avoid up to 1.5 million premature deaths annually [49]. A multi-model evaluation of future climate and health impacts based on the RCP scenarios showed that by 2030, the range of scenarios resulted in 289 000 premature deaths per year avoided (RCP 4.5) to 17 200 excess premature deaths per year (RCP8.5) for PM2.5 relative to the baseline value from 2000 [34]. Although the baseline years are different, the CLE scenario results from Likhvar et al. [49] was similar to the RCP8.5 scenario from Silva et al. [34]. By 2100 substantial reductions in PM2.5 result in global premature deaths avoided across all scenarios, ranging from −1.31 million (RCP8.5) to −2.39 million deaths per year (RCP4.5) for the multi-model average [34]. In a follow up study, the premature mortality from changes in air pollution attributable to climate change was isolated from the changes in air pollution attributable to changes in emissions [54]. Changes in air pollution associated mortality owing to climate change are projected to increase under RCP8.5. For PM2.5 the multi-model estimate was 55 600 premature deaths in 2030 and 215 000 premature deaths in 2100. This increase counters the overall global decrease in premature mortality expected under RCP8.5 in 2100 by 16%. For ozone, of the 264 000 (316 000) premature deaths projected by the multi-model average in 2030 (2100) under RCP8.5, an estimated 3340 (43 600) were attributed to the effect of climate change [34,54]. Partanen et al. [56] evaluated premature mortality using three RCP scenarios and a high and low aerosol emission scenario based on RCP4.5. Differences in mortality estimates from this study to those provided by Lelieveld et al. [50] and Silva et al. [34] could be largely explained by differences in baseline mortality rates. The low aerosol emission variant of Partanen et al. [56] yielded very similar avoided deaths by 2030 as obtained for MFR in Likhvar et al. [49]. Conversely, the authors found by 2030 the highest mortality rate (2 371 800 deaths per year) under the high aerosol emission variant. A number of papers evaluated changes based on scenarios linked to the context of the Paris Agreement and the associated 2°C target [53,55,57]. These studies evaluated climate change mitigation scenarios, but also the associated changes in air quality and co-benefits for human health. The two papers by Shindell et al. [53,57] used a similar methodology, but one evaluated the effect of US emissions reductions for two scenarios—clean energy and clean transportation—consistent with the 2°C target on global health, while the other investigated the health benefits of accelerated CO2 reduction policies globally. Cumulatively from 2020 to 2100, accelerated CO2 reductions, would prevent 153 ± 43 million premature deaths (figure 2 [57]). Markandya et al. [55] evaluated four scenarios, including scenarios in which stabilization at 2°C and 1.5°C warming were achieved. In all cases, the climate mitigation policies had substantial co-benefits for reducing air pollution emissions and human health. A comparison of the mitigation costs compared to the health co-benefits indicated that the benefits strongly outweighed the costs, with ratios of benefits:costs ranging from 1.4 to 2.45 for the 1.5°C and 2°C stabilization scenarios [55], and even greater by a factor of 5–10 for the US emission reduction scenarios [53]. Vandyck et al. [58] evaluated two scenarios, a 2°C stabilization scenario and a Nationally Determined Contributions (NDCs) scenario, documenting that the latter could lead to 71 000–99 000 avoided premature deaths globally in the year 2030 compared to current climate policies. The authors estimated that a more stringent 2°C pathway could avoid 178 000–346 000 premature deaths annually in 2030 and 0.7–1.5 million by 2050, which are similar cumulative health co-benefits as found in the Shindell et al. [57] and Markandya et al. [55] studies discussed above. Figure 2. Reduction in annual premature deaths due to PM2.5 and ozone over the period 2020–2100 from co-emissions accompanying accelerated CO2 emissions reductions, depicted as regional highlights. Values are all-cause per 0.5° × 0.5° area (approx. 50 km × 50 km at mid-latitudes) without low exposure thresholds. Note different ranges in the panels. Adapted from Shindell et al. [57]. (Online version in colour.)
Two papers evaluated sector specific scenarios, both for transportation. Morita et al. [51] evaluated three future scenarios, including a reference scenario assuming unconstrained growth and a factor of 4.8 increase in fuel consumed relative to the base year of 2006. The reference scenario resulted in an increase in excess mortality of approximately 5500 deaths from the aviation sector in 2050, relative to approximately 400 excess deaths in 2006 associated with changes in PM2.5. The two improved fuel policy scenarios (technology and operations improvements yielding smaller increases in fuel combustion compared to 2006, and conversion to fully sustainable fuels) still resulted in increased excess deaths in 2050 relative to 2006, but were substantially less than the reference scenario, with approximately 2300 and approximately 1600 excess deaths attributed, respectively [51]. Anenberg et al. [52] analysed the health impacts of mitigating excess diesel-related NOx in 11 major vehicle markets. The study found that implementing Euro 6/VI standards where they are not yet adopted, specifically for heavy-duty diesel vehicles, could avoid 104 000 premature deaths in 2040. Beyond Euro 6/VI, further adopting and enforcing next-generation standards that include more stringent real-driving emissions programs would avoid 174 000 premature deaths in 2040 [52]. The differences in premature deaths that could be avoided for these sectors is likely linked to, among other factors, the location of where the emissions occur, with vehicular emissions more prevalent in high population areas. More than 20 papers were published since 2013 that evaluated the health effects of future air quality burdens on human health in the USA (table 1). The majority of the studies examined projections at the national scale, however, a number of studies also focused on certain regions, states or metropolitan areas within the USA. The scenarios evaluated ranged from the globally relevant RCPs to those that specifically considered US regional energy policies. Kim et al. [71] investigated the changes in ozone-related mortality in the USA between 2001–2004 and 2057–2059 under a low-to-medium emission scenario, RCP4.5, and a fossil fuel intensive emission scenario, RCP8.5. They also evaluated the sensitivity of the mortality estimates to the input variables. To derive excess mortality due to ozone, Kim et al. [71] based concentration-response functions (CRFs) on the association between non-accidental, all-cause mortality and short-term exposure to maximum daily 8-h average (MDA8) O3 (following [98]) and maximum daily 1-h average (MDA1) O3 (following [98,99]) and calculated excess mortality on county level. Their results showed, compared to the base period (averaged across two different CRF coefficients, and four population scenarios), approximately 1300 additional premature deaths under RCP8.5, while approximately 2100 premature deaths could be annually avoided under RCP4.5, with significant variation spatially. The largest amount of uncertainty in the estimates was associated with the RCP emission pathways [71]. Martinich & Crimmins [87] also estimated premature mortality associated with health effects of surface ozone based on RCP4.5 and RCP8.5 projections. This was part of a broader evaluation of projections of climate damages and adaptation potential across diverse sectors of the USA for a broad range of endpoints, not only human health, but also infrastructure, agriculture and ecosystems. All other effects were, however, limited to changes in climate and not air quality and no adaptation measures were included in the sectoral impact analysis for air quality. Five hundred and fifty excess deaths annually were estimated under RCP4.5 and 790 excess deaths annually under RCP8.5 in 2050, with these values increasing to 1200 and 1700 excess deaths annually, respectively, in 2090 [87]. In comparison to the Kim et al. [71] study, these values were similar for RCP8.5 but do not show the same avoidance of excess deaths under the RCP4.5 scenario. A novel study by Abel et al. [82] evaluated the air quality related health outcomes due to heat-driven adaptation of increased air conditioning demand for buildings for the Eastern USA. The study compared representative present-day (2011) and mid-century (2069) climate scenarios (based on the IPCC A2 scenario) with and without exacerbated power sector emissions from adaptation in building energy use from increased air conditioning. The impact of climate change alone resulted in a projected increase in summer air pollution related premature mortality of roughly 13 000 deaths due to PM2.5 and 3000 deaths due to O3. Air conditioning adaptation accounted for 645 and 315 of these PM and O3 related annual excess deaths, respectively [82]. The role of such adaptation measures and feedback effects is not quantified in most studies. This shows, however, that such adaptation measures could have substantial impacts on future projections related to air quality and health impacts. In summary, the recent USA focused literature shows diverse futures for health burdens attributable to PM or ozone. While ozone health burdens are generally projected to increase under business as usual or planned legislation scenarios, reduced adverse health effects emerge regionally (particularly the Eastern USA) under scenarios considering more stringent precursor emission controls. Under RCP8.5 ozone increases are projected for the Western USA, while decreasing ozone burdens are projected for the Eastern USA [100]. For PM, future projections indicate reduced adverse health effects from emission reductions, but additional adverse health effects from population growth, with larger populations being exposed to unhealthy PM levels at the regional level. Studies considering health effects of both ozone and PM generally indicate health benefits are dominated by reductions in PM. A few more than 10 studies explicitly evaluated the health effects of future air quality burdens for Asia and Australia. Almost all studies focused on single countries, and in some cases, urban areas. The majority of the studies conducted projections for China. The studies by Qin et al. [62] and Yang et al. [63] investigated future air quality and health benefits for specific policy developments in China using the ECLIPSE_V5a_CLE reference scenario as a baseline. Qin et al. [62] evaluated the implications of China's synthetic natural gas (SNG) development strategy for 2020, whereby all planned production of SNG is deployed to replace coal in the power, industrial or residential sectors. Yang et al. [63] investigated the effect of fulfilling the 400GW national photovoltaics (PVs) capacity in 2030 as per the China Renewable Energy Roadmap 2050, which is intended to reduce emissions from coal fired power plants. The results vary depending where the PV capacity is installed and thereby the type of power plants being replaced, as well as the amount of inter-provincial PV electricity transmission. The benefit from the SNG scenarios was greatest if deployed to replace coal use in the residential sector, with approximately 32 000 premature deaths avoided annually in 2020 from outdoor air pollution alone. If indoor air pollution was included, values would be much higher. If the SNG were deployed to replace coal use in the industrial or power sectors, the air pollution associated avoided premature deaths would be much lower, 3100 or 560 annually, respectively, nationwide in China in 2020 [62]. The results from the PV scenarios showed that installation in eastern China with inter-provincial transmission has the largest benefits as it displaces the dirtiest coal power plants. The projections estimated a reduction in premature deaths associated with PM2.5 air pollution of 10 000 by 2030 relative to the base case [63]. Both studies explored the health effects of replacing coal-based energy with alternative energy forms in China, and agreed that energy transition is most promising in the residential sector. Qin et al. [62] reported larger health gains for 2020 than Yang et al. [63] for 2030. This difference can be explained by the full regional consideration of China's coal-based SNG development plan by Qin et al. [62], while Yang et al. [63] compared the efficiency of different regional solar PV deployment and utilization scenarios and did not quantify effects of full PV deployment. A recent study evaluated air quality and health co-benefits of China's climate policy for PM2.5 and ozone nationally and for three populous countries downwind of China (South Korea, Japan and the USA) [69]. The authors considered policy scenarios with CO2 reductions of 3%, 4% or 5% per year between 2015 and 2030. All scenarios yielded substantial health gains due to associated PM2.5 and ozone reductions, with the amount of net gain related to the ambition in CO2 reductions (figure 3). For example, Li et al. [69] reported detailed health gains for the 4% CO2 reduction path. Here, the largest effects emerged with 95 200 PM2.5-related and 54 300 ozone-related avoided premature deaths locally in China relative to the base year. The study also found substantial health benefits for downwind countries with 600 (the USA) to 2000 (Japan) PM2.5-related and 300 (South Korea) to 1500 (Japan) ozone-related avoided premature deaths in 2030. Generally across all CO2 reduction scenarios benefits due to reduced PM2.5 were larger than those for ozone, and cumulative health benefits in China exceed those in downwind countries by about a factor of 50–100 (figure 3). Figure 3. Avoided PM2.5- and ozone-related premature deaths under three climate policy scenarios relative to No Policy in China (a) and three downwind countries (b)–(d) in 2030. Ozone-related deaths are calculated using CRF in Turner et al. [101]. Note different scale for panels (b)–(d). From Li et al. [69]. (Online version in colour.)
A study of 104 cities across China used a combination of the RCPs (RCP4.5 and RCP8.5) and six population change scenarios from the SSPs, to evaluate not only excess ozone-related future mortality from changes in emissions and climate, but also how the ozone-related excess mortality rates changed when considering and adjusting for population change and ageing [65]. Compared to the estimates for the historical time period (2013–2015) for which roughly 13 900 deaths were attributed to short-term exposure to ambient ozone air pollution in the 104 Chinese cities, a decrease in mortality (approx. −3300 deaths) was observed for the RCP4.5 scenario annually in 2053–2055, and increase observed under RCP8.5 (approx. 1500 deaths annually), considering no population change. Using age-group-specific concentration-response functions and baseline mortality, rather than all-age values, increased the historical mortality estimate to approximately 21 600 deaths. This indicates the significant influence such choices can have on the magnitude of the derived mortality estimates. Population ageing emerged as such an important factor that it offset the decrease in excess deaths projected under RCP4.5 and dominated the net increase in excess deaths under RCP8.5, despite projected decreases in overall population size and expected decreases in age-group-specific mortality rates, as in figure 4 [65]. Changes in population demographics, and specifically ageing, are not factors that are typically considered in many of the studies investigating changes in the effects of future air quality. This study, however, highlights the important role such changes may play, likely resulting in substantial underestimates of future air pollution related excess mortality. Figure 4. Changes in ozone-related mortality according to climate and population changes from 2013 ± 2015 to 2053 ± 2055. Population changes include both population size changes and population ageing. Mortality rate indicates age-group-specific baseline mortality rate changes. Future changes (%) of annual ozone-related mortality for the population aged 5 years and above in 2053 ± 2055 were calculated relative to the historical period 2013 ± 2015. RCP4.5 and RCP8.5 represent moderate and high global warming and emission scenarios, respectively. SSP1 ± 5 represent five population change scenarios under different shared socioeconomic pathways. From Chen et al. [65]. (Online version in colour.)
Overall, the studies evaluating changes in Asia and Australia show that the projected changes in health effects are strongly dependent on the level of ambition, with scenarios that follow currently planned legislation and/or business as usual scenarios projecting an increase in adverse health effects, while reductions in adverse health effects are observed for scenarios and policies that require more ambitious emission reductions or greater technology changes. Furthermore, a number of the studies agree that there will be a substantial climate penalty for ozone-related adverse health effects, requiring greater reductions to offset the effects of climate on ozone concentrations, with the largest impact in areas with the highest population densities and ozone concentrations. In addition to the global studies that include a regional investigation for Europe, over the last 6 years three studies have focused on the effects of future air quality on the health of the European domain. All studies included both ozone and PM, while one study specifically isolated the effect of climate change on future air quality associated health impacts [91]. Tarín-Carrasco et al. [91] isolated the effects of climate change on air-pollution-related-pathologies in central and southern Europe by holding emission levels and population density constant at present-day (1996–2015) levels while considering future climatic conditions under RCP8.5 (2071–2100). The results showed an increase in premature deaths, attributable to climate driven air quality degradation of 94 900 cases annually (valued at an additional EUR 27 billion in external costs) (figure 5). This represents a 17% increase compared to present day air quality-related premature mortality (418 700 cases; EUR 158 billion in external costs). Figure 5. (a) Present cases of premature deaths (PD) and (b) associated costs, in millions of euros. (c) Changes projected in PD and (d) changes in costs (millions of euros) under the RCP8.5 scenario (2071–2100). From Tarín-Carrasco et al. [91]. (Online version in colour.)
Schucht et al. [89] investigated two Global Energy Assessment (GEA) scenarios, a reference scenario with no climate policy (consistent with RCP8.5) and a climate mitigation scenario limiting global temperature increase to 2°C by the end of the twenty-first century (consistent with RCP2.6). Both scenarios, however, include an updated representation for air pollutant emissions considering all currently legislated air quality policies until 2030. The results showed that air quality policies alone (no climate policy) effectively reduce the adverse health impacts associated with PM2.5, while the ozone-related adverse health impacts are much more dependent on climate policy [89]. Specifically, premature deaths from acute exposure to ozone, estimated to be 31 000 in 2005, were projected to increase to 48 000 in 2050 for the scenario with no climate policy (with the increase being largely attributed to a growing and ageing population in Europe over this time period). Under the climate mitigation scenario, the premature deaths from acute exposure to ozone across Europe were reduced to 7000 in 2050. In comparison, the health impacts of PM2.5 in these scenarios showed further improvements due to climate policy. From 2005 to 2050, the number of years of life lost (YOLL) already decreased from over 4.6 million to 1 million following the implementation of the air quality policy even in the absence of any climate policy. The additional implementation of an ambitious climate policy further reduced the YOLL to 0.3 million in 2050 [89]. Overall, the results from the European studies agree well with those for the USA or Asia and Australia. Recent research indicates larger health benefits from reductions in PM than ozone. Health benefits emerge on local and regional scale following more stringent air pollution policies but penalties also emerge, particularly for ozone, in the absence of effective climate policies. For simulations focusing on air quality impacts no community standard or strategic protocol exists, as has been developed e.g. for climate change simulations developed within the Coupled Model Intercomparison Project or simulations focusing on chemistry–climate connections developed within the Atmospheric Chemistry and Climate Model Intercomparison Project (ACCMIP) or Chemistry Climate Model Initiative (CCMI). The available studies vary strongly in the design of the modelling chain and the scenarios, regions, time periods and mortality and morbidity measures considered. The diversity of approaches and scenarios can be viewed as a strength in understanding the scope of possible future effects, as well as identifying consistency in projected effects. However, coordination of quantified effect metrics, as well as transparency in model assumptions and treatment of processes would support comparisons. Thus a number of important considerations and limitations can be identified across the studies included here regardless of global or regional focus. A selection of these factors has been summarized in table 2.
For one, as is demonstrated by those publications that are multi-model studies, the underlying representation of atmospheric processes, chemistry–climate interactions and treatment of natural emissions in the different model set-ups can result in large differences for the same scenario. For example, Silva et al. [34] found that for most cases in the multi-model study, uncertainty in modelled air pollutant concentrations was the greatest contributor to uncertainty in mortality estimates. Health impact studies are based on CRFs and the choice of these functions for estimating air pollution associated mortality or morbidity can have a substantial impact/lead to great uncertainty [90], as well as assumptions about the toxicity of particles [50]. A recent meta-analysis combining multiple appropriate epidemiological studies offers a first possible solution to this end [102]. Due to natural variability, the years simulated will influence the outcome of the study results. A review paper focused on the climate change penalty for ozone air quality found that for model projections, a minimum of 15-year averaging windows are required to smooth out the ‘noise’ due to natural climate variabilities and to distinguish signals forced by anthropogenic climate warming [13], while another study found that the precise number of simulation years required for a given degree of accuracy will depend on the timing and strength of the policy in question [85]. Changes in temperature and oxidative capacity will influence secondary organic aerosol, but the physical-chemistry processes are not well understood and therefore any future effects are generally not included in modelling studies [90]. In addition, few studies take interactions between temperature and air pollution on health effects into account, but recently studies have emerged that show, generally, air pollution related mortality is greater for higher temperature or warmer days [10,102]. Changes in atmospheric stagnation events, precipitation and heat waves will affect future concentrations of PM and ozone. While changes in stagnation will effect both PM and ozone, heat waves have a larger effect on ozone concentrations and changes in precipitation frequency and amount will have a larger influence on PM. Taken together these climate extremes may represent an important mechanism through which climate change will affect future air quality [8–11]. Many studies acknowledge that changes in population totals and demographics will substantially affect mortality estimates (see also the summary of Chen et al. [65] included above), but do not include such changes in the projections conducted. For those studies that have included such changes in the projections, consideration of these factors typically shows substantial increases in mortality (e.g. [65,68,90]). In addition, using future baseline mortality rate projections, rather than maintaining constant conditions will also influence the magnitude of the results, generally reducing the magnitude of the benefits [49]. Including only a limited age range, e.g. no mortality included for people under 25, will also underestimate mortality. In addition, related factors such as changes in healthcare system(s), population vulnerabilities, increasing urbanization and/or changes in the composition of the future air pollution mix are generally not being considered but would influence projections [49,103]. Finally, imbedded in some of these factors is the differential impacts of air pollution on demographic groups, with lower socioeconomic conditions generally linked to higher air pollution exposures [104]. Changes in socioeconomic conditions and the differential effects on various demographic groups are also not being considered in these studies and will likely continue to evolve over time and influence projections. An important consideration in future air quality management would be to minimize differential exposure of different socioeconomic populations to air pollutants. Compared to the impacts on human health, fewer recent papers investigated the impacts of future air quality on crop production. Most crop studies are global in scope and presented additional regional or country-level insights. These works focused largely on the major crops of the developed world (maize, rice, soya bean and wheat) but vary in future scenario, quantification metric and process complexity. An overview of the studies presented here is listed in table 3. The description and discussion below is organized by study type and topic and generally proceeds from more well-studied to more uncertain.
The impacts of air quality on crops are most frequently described in terms of relative yield (RY), which refers to the area-normalized production (yield) given a certain amount of pollutant compared to the yield in the absence of that pollutant entirely or below a certain threshold. The change in RY caused by that pollutant is then described as the RY loss or gain. When describing impacts on a global or regional scale, total or relative production values can be used to account for the land area used for each crop. Several studies further translate impacts on crop yield/production into unifying units such as commodity pricing or calorie content, although this can add additional uncertainty to future projections. The largest body of literature focuses on the decrease in crop yield associated with ozone damage using CRFs. In these studies, future scenarios are for precursor emissions only with constant future meteorology. This isolates the impact of possible future emissions regulations, rather than confounding with additionally uncertain climate projections. Van Dingenen et al. [105] found global RY loss for 2000 due to ozone to be 3–5% for maize, 3–4% for rice, 6–16% for soya beans and 7–12% for wheat. Under a CLE scenario for 2030, additional losses were projected to be 1–2% for rice and 2–6% for wheat, with less than 1% change for both maize and soya beans. This response translates into an additional loss of $14–26 billion. Ranges here reflect different available ozone exposure metrics used in CR functions. Using the same approach, Shindell et al. [107] found similar, but slightly lower, present day global RY loss for all four crops. In their analysis, imposition of ‘tight-controls’ for on-road vehicle emissions in 2030 resulted in a decrease in global wheat RY loss of 1.1% and savings of 6.1–19.7 million metric tons of crop losses worth over $1.5 billion. The A2 (high emission) scenario employed by Averny et al. [106] showed an additional global RY loss of 2–3% for maize, 1–11% for soya bean and 2–10% for wheat by 2030 compared to 2000, with the B1 (low emission) scenario contributing to less RY loss overall, a maximum 2% additional loss for wheat. The value of additional crop losses in 2030 was found to be $6–17 billion for A2 emissions and only $1–3 billion for B1 emissions. More recently, implementation of the RCP scenarios found that changes in global ozone-related crop (maize, rice, soya bean and wheat) damage led to a production gain of 3.1% (0.22 × 1015 kcal) for RCP4.5 and production loss of 3.6% (0.26 × 1015 kcal) for RCP8.5 by 2050 compared to 2000 [109]. Similarly, Schiferl & Heald [113] saw a global increase (decrease) in production of 1.4% (0.5%), 1.0% (0.1%) and 4.4% (3.0%) for maize, rice and wheat, respectively, under RCP4.5 (RCP8.5) in 2050 compared to 2010 due to the projected crop area- and growing season-weighted decrease (increase) in ozone concentrations (figure 6). Considering recent policy implications, Vandyck et al. [58] found that more stringent emissions goals limiting warming to 2°C could double the gains in global crop yields made by reducing ozone concentrations compared to those limits set by NDCs. This limited emissions policy would increase maize, rice, soya bean and wheat productivity by 0.8–1.5%, 0.2–0.8%, 1.8–2.7% and 0.9–1.7%, respectively, in 2030 compared to a reference case without climate change mitigation. Figure 6. For both RCP4.5 (left) and RCP8.5 (right) emissions scenarios: regional relative change in crop production due to surface ozone (red bars/leftmost bars in clusters), PM with maximum diffuse effect (blue bars/middle bars in clusters), and both ozone and PM (grey bars/rightmost bars in clusters). Change from 2010 to 2050 for (a) maize, (b) wheat and (c) rice. Error bars indicate range of production from 0 to –10 ppb surface ozone concentration correction, from minimum to maximum diffuse PM effect, and from both effects, respectively. Regions with a base production lower than 5% of the global total are not shown. Relative change calculated from 2010 base production. From Schiferl & Heald [113]. (Online version in colour.)
While the projected increase in ozone damage on global total crop production was estimated to be less than approximately 10% by 2030/2050, and there is the potential for global total loss reduction under more restrictive emissions scenarios, these projections vary strongly on the national and regional level. In East Asia, Amin et al. [108] evaluated a ‘policy failure’ case (higher emissions) which resulted in an increase in RY loss due to ozone of 9–10% for maize, 6–12% for rice, 13–33% for soya bean and 5–27% for wheat in 2020 compared to a baseline in 2000 (which is 17–35% RY loss for wheat). This case increased value lost from ozone damage from $14 billion in 2000 to $27 billion in 2020 in this region. Using an RCP6.0-type high emissions scenario, Chuwah et al. [110] estimated about 10% increase in RY loss in maize and rice between 2005 and 2050 in the Middle East, India and China. With more stringent policy controls, RY loss for these crops would decrease by 2050 throughout the USA, Europe and Middle East, but would remain significant (greater than 15%) in parts of China and India. Such regional differences are consistent with results found by Schiferl & Heald [113] (figure 6). Under the RCP8.5 emission scenario, wheat production is expected to decrease by 10% in India between 2010 and 2050 due to ozone damage but less than 3% in the USA + Canada, Europe, and China + South East (SE) Asia regions. Emission controls on ozone in the RCP4.5 scenario increased wheat production in China + SE Asia by over 15% and nearly 5% in USA + Canada and Europe, while India's production loss was reduced by about 5% in 2050 compared to 2010. In the USA, Capps et al. [111] implemented versions of the Clean Power Plan and found that such regulations would reduce potential productivity loss (PPL) due to ozone by up to 16% for maize (although with very low reference PPL) and 8% for soya bean in 2020 compared to a reference case. Extending the RCP8.5 pathway to 2100 for Portugal, Miranda et al. [114] suggested approximately 15% increase in productivity and quality for wine grapes due to emissions reductions, compared to approximately 35% decrease that would be expected under constant current emissions. Tai & Val Martin [112] derived new relationships using partial derivative-linear regression (PDLR) for the impact of ozone on crops using observed crop yield data and ambient ozone measurements in the USA and Europe. This method allowed for spatial differentiation in the response of crops, unlike the CR equations, which are usually applied broadly to all areas. Their work highlights the possible development of stronger ozone tolerance by crops in ozone-stressed areas, with crop-producing areas away from ozone hotspots more sensitive to ozone stress than those located near urban regions. Using the PDLR relationships, RCP4.5 emission reductions in 2050 led to a 25% increase in US wheat production compared to 2000 and only 4% increase in Europe (figure 7). Since US production areas are more remote than those in Europe, this distinction is larger than that found by Tai et al. [109] using uniform CR equations. Applying these PDLR relationships (from USA and Europe) to India and China using RCP8.5 emissions for 2050 compared to 2000 showed ozone may have a larger impact than found when using the CR equations. While these impacts were quantified, the relationship is better qualified, as the authors stress that the uncertainty in cross-continental application is large. Figure 7. Projected 2000–2050 percentage changes in total production for wheat, maize and soya bean for the USA and Europe following RCP4.5 and RCP8.5 scenarios under individual (blue and red/left and middle bars) and combined (purple/right bars) effects of ozone pollution and warming. Bars indicate the mean changes and the notches indicate the 90% confidence intervals as estimated from Monte Carlo method, denoting a ‘very likely’ change if the interval does not overlap with zero. From Tai & Val Martin [112]. (Online version in colour.)
Little work has been done to quantify the future impacts of PM on crops. Through scattering, PM decreases the total, but increases the diffuse portion of light intercepted by a plant. This diffuse light has an enhancing effect by more evenly distributing light photons throughout the plant leaf canopy. Indeed, PM can be beneficial to plants by increasing their light use efficiency, if not offset by the negative effects of the reduction in total light. Therefore, unlike ozone, PM can cause either a net increase or net decrease in crop yield/production depending on the PM concentration and composition in the atmosphere. Using the maximum impact from various uncertain relationships, Schiferl & Heald [113] found changes in PM (largely reductions) could cause a global decrease in production of 1–2% for maize, rice and wheat under RCP4.5 in 2050 compared to 2010, with less impact for RCP8.5 (figure 6). The authors pointed out the need for more physiological studies to establish the current impacts of PM, which could be important as particles are removed from the atmosphere and reduce the net enhancing impact on crops. This was especially clear in the RCP4.5 scenario, where nearly 3% of maize and rice production could be lost in China + SE Asia due to lower PM levels. In this case, the PM effects have the potential to offset improvements from ozone reduction. By contrast, higher levels of PM in India under RCP8.5 could lead to an increase in wheat production, but this increase would be overwritten by the large loss in production due to ozone. Overall, there is consensus that ozone damage has a significant negative impact on global crop production. Future projections indicate that this damage will be lessened and crop production enhanced in areas that implement ozone-reducing policies, which seems most likely to be outside of Asia. Future reductions in PM through air quality controls may have a slight negative impact on crop production, but this is highly uncertain and motivates further study. The higher temperatures associated with a warmer climate both increase ozone production and directly impact the productivity of crops. For example, studies comparing temperature effects to ozone damage indicate future crop production could be dominated by the negative impacts of higher temperatures, but air quality policy offers a significant opportunity for mitigation [109,112]. This pattern is especially relevant for more restrictive emissions scenarios such as RCP4.5 (figure 7). PM also affects cloud formation and net radiation reaching the surface, which in turn impacts significant factors for crops including hydrology and temperature. In the event of future geoengineering, where PM is introduced to the stratosphere, a study by Proctor et al. [115] shows this would have little net impact on crops in 2050, where the increase in global crop yield due to lower temperatures is largely offset by the decrease in available light. These interactions are not considered in this review, but are important for a more holistic understanding of the impacts of future climate and air quality on global crop production. The studies in this review also do not take into account future changes in land use (e.g. conversion of forest into cropland), the distribution of crops (e.g. substitution for ozone- or drought-resistant varieties) or planting/harvesting calendars (e.g. longer summers) as development advances and the climate changes. Although atmospheric deposition of nitrogen and sulfur can have positive impacts on crop production as essential limiting nutrients to plants [116,117], studies do not quantify their impacts on future crop production. Therefore, policies regarding water and nutrient availability (e.g. those that reduce the atmospheric burden of nitrogen and sulfur) are important to consider for additional study with regards to changing food production in the future. These policies may impact the ability of crop production to reach its full potential when enabled by crop-enhancing air quality policy, such as the reduction in ozone pollution, as supplemental application can be prohibitively costly or be environmentally impactful in other ways. Finally, studies of future air quality impacts on crop production largely focus on the crops of the Northern Hemisphere. However, the air quality in Africa is expected to decline in combination with projected population growth and increased combustion emissions [118,119], and there is little mention in current studies of the impacts of air quality on the dominant crops (e.g. sorghum, millet, pulses) of this region. Additional studies of these crops would help to understand to what extent future ozone and PM levels could significantly impact Africa's already vulnerable food system. Effects of air pollution on ecosystems have been studied and established for decades. These effects are manifested through a host of drivers such as impaired visibility from PM, phytotoxicity due to elevated O3, soil and freshwater acidification of forest and aquatic ecosystems associated with atmospheric deposition of acid and acidifying compounds (SO2, NOx, NH3) [120,121], eutrophication of terrestrial and marine ecosystems associated with atmospheric deposition of nitrogen (oxidized and reduced nitrogen) [122,123], and contamination of terrestrial, freshwater and marine ecosystems due to atmospheric deposition of mercury [124] (figure 1). With the exception of visibility and plant effects of ozone, which occur over the growing season cycle, ecosystem effects (acidification, eutrophication, mercury contamination) generally are the result of atmospheric deposition. Atmospheric deposition occurs as wet deposition, which includes rain, snow, sleet or hail; as dry deposition, which includes atmospheric particles or gases; or as cloud or fog deposition, which is more common at high elevations and in coastal areas. Wet deposition is fairly well characterized by monitoring at more than 270 National Atmospheric Deposition Programs (NADP; http://nadp.sws.uiuc.edu) in the USA, 135 sites in European Monitoring and Evaluation Programme (EMEP; www.emep.int), National Acid Rain Programs and Routine Monitoring Network on Acid Rain in China and Acid Deposition Monitoring Network in East Asia (EANET; www.eanet.asia) at 54 sites in 13 participating countries in Asia, and Global Atmospheric Watch by the World Meteorology Organization (WMO-GAW). Although wet deposition is relatively intensively monitored, it can be highly variable spatially. Dry deposition involves the deposition of gaseous and aerosol species and is less intensively monitored than wet deposition. It is highly dependent on topography, meteorological conditions and vegetation characteristics, which can vary markedly over short distances in complex terrains [125,126]. As a result, dry deposition is poorly characterized and highly uncertain. Meteorological conditions and climate change affect the quantity and distribution of wet and dry deposition. Wet deposition can occur as removal within a cloud known as ‘rainout’ or scavenging from the atmosphere by precipitation called ‘washout’ [127]. Changes in the quantity and the intensity of precipitation will alter the quantity of wet deposition and the distribution between wet and dry deposition. Changes in air temperature and air and soil moisture will alter emissions, partitioning and deposition of gases and particulate matter and therefore dry deposition. Due to high leaf area index, changes in forest canopy structure, exposure and function can greatly influence dry deposition. Coniferous vegetation is more effective than hardwood vegetation in scavenging air pollutants from the atmosphere (e.g. [128]) so shifts in vegetation type alter dry deposition. Changes in canopy biomass and stomatal conductance associated with meteorological conditions or atmospheric CO2 will also alter dry deposition [129]. The time scale of changes in atmospheric deposition is typically on the order of many decades to a century [47], except for extreme events. As a result, the changes in plant and soil processes within watersheds in response to atmospheric deposition occur over long time scales, again typically many decades to centuries. These long-term changes are largely due to relatively large soil pools which innately buffer soil solutions and mediate hydrologic and biogeochemical processes within watersheds. The extent of soil pools vary with soil mineralogy, depth and age as well as climate and vegetation. For example, inputs of sulfur, nitrogen and mercury to watersheds are largely retained in soil before export from the watershed. For acidification, inputs of strong acids (sulfuric, nitric acid) deplete soil pools of available nutrient cations (calcium, magnesium) or mobilize dissolved inorganic aluminium, which is toxic to plants and fish. The long time scale of ecosystem response to air pollution is in stark contrast to the relatively rapid response of human health, visibility or crops to changes in ambient air quality. These long time scale changes in atmospheric deposition and associated ecosystem processes are consistent with the long time scales of changes in meteorology, hydrology, and plant soil and microbial processes that occur in watersheds in response to climate variability and change [130,131]). Because watershed losses for elements largely occur by fluvial processes or land-atmosphere exchange and because microbial and plant processes are highly sensitive to temperature and soil moisture, ecosystem impacts of air pollution are closely coupled with meteorological and therefore climatic conditions. Because of the complexity of ecosystem response to air pollution and the different responses for different air pollutants in different ecosystem types, we give three case studies to illustrate the interplay between climate and air pollution effects. The first example is a contemporary observation involving effects of acid rain and tree health and demonstrates the complexity of capturing interactions between air quality and climate with organism response in future projections. In the 1990s, it was demonstrated that acid deposition was impairing the health of red spruce (Picea rubens). Red spruce is generally found in montane areas and is native to eastern North America, ranging from eastern Quebec and Nova Scotia, west to the Adirondack Mountains and south through New England along the Appalachians to western North Carolina. The mechanism affecting tree health was the loss of membrane-bound calcium from needles from elevated leaching by acid deposition, which decreased the cold tolerance of red spruce [132]. Under this condition cold winter temperatures caused damage to the current year needles. If needle damage and loss occurs over multiple years it results in the mortality of trees. In recent years, damage to red spruce has been almost completely mitigated due to decreases in acid deposition associated with air quality management in North America, coupled with marked decreases in very cold winter temperatures due to climate change [133]. The second example involves climate change effects on mercury. Although there are sites of localized industrial contamination, most ecosystems derive mercury inputs from atmospheric deposition [124]. As mercury is a global pollutant, virtually all ecosystems are subject to mercury contamination by atmospheric deposition. Within reducing environments (e.g. wetlands, sediments) inputs of mercury can be converted to monomethylmercury, which is readily bioaccumulated and biomagnified through food chains. Elevated exposure to humans and wildlife (songbirds, fish piscivorous birds and mammals) typically occurs through consumption of foods elevated in methylmercury, which is a neurotoxin that can affect a number of health endpoints [134–136]. All components of the transport and processing of mercury in the environment, atmospheric transport and deposition, evasion, terrestrial and aquatic processing, methylation, and trophic transfer, are affected by climatic and hydrological conditions and therefore are strongly impacted by climate change [129]. For many developed countries, the dominant human exposure to mercury occurs by consumption of marine fish with elevated concentrations of methylmercury [137]. In marine food chains, methylmercury can bioaccumulate by a factor to 10–100 million in top predators. Schartup et al. [27] used a food web bioaccumulation model to examine the mechanisms driving temporal trends in concentrations of methyl mercury in Atlantic bluefin tuna (Thunnus thynnus) in the Gulf of Maine. They found that recent and projected future increases in the methylmercury concentrations in Atlantic bluefin tuna were driven by increases in sea surface temperature rather than changes in mercury emissions and deposition. Climate change is likely to aggravate human exposure to methylmercury through marine fisheries, suggesting that more aggressive controls of mercury emissions and releases will be needed to protect ecosystem services and human health. Finally, the occurrence of wildfires has increased over recent decades, and with these events a deterioration in air quality, largely due to increases in fine PM [18,21]. Many factors influence the susceptibility of forests to wildfire, including increases in local human population and associated development adjacent to forest lands, fire suppression, and warming and drought. Increasing fire frequency and severity are linked to climate change [138]. In addition to these factors, research has demonstrated a linkage between air quality and forest predisposition to wildfires, which has been manifested in southern California [139]. Increases in atmospheric nitrogen deposition and ozone shift the processing of water, carbon and nitrogen in forest ecosystems, resulting in a cascade of synergetic effects, which make trees more prone to disease, pest invasion, drought and ultimately wildfire (figure 8). These air pollutants increase leaf turnover and litter mass, and decrease the decomposition of litter [141]. As a result, mixed conifer forests of southern California that are impacted by nitrogen and ozone pollution develop deep litter layers. Under high ozone and nitrogen, trees also shift their allocation of biomass away from foliage and roots. Elevated ozone decreases plant control of water loss increasing transpiration, which when coupled with loss of root mass increases the susceptibility of trees to drought stress. Ozone and drought stress and enrichment of nitrogen makes trees particularly vulnerable to attack by bark beetles [142]. The enhanced fuel load from tree decline and litter accumulation driven by nitrogen and ozone pollution coupled with development pressures and climate have resulted in catastrophic fires in southern California in recent years, exacerbating already severe air quality and health impacts. Such feedbacks will continue to influence wildfires, and thereby accelerate air quality and the associated impacts in future. Figure 8. Conceptual diagram showing the interactions, effects and feedbacks of human management and ozone and atmospheric nitrogen deposition on forest processes and susceptibility to wildfire and production of airborne particulate matter. Modified after Grulke et al. [140].
It is important to recognize that impacts of air pollution and climate on ecosystems are not always negative, indeed the few studies that have evaluated ecosystem responses to air pollution and climate change have often shown both positive and negative effects. For example, the first case study discussed above showed that climate warming coupled with decreases in acid deposition have mitigated air pollution damage to red spruce in eastern North America. Moreover, modelling studies have projected increases in net ecosystem production associated with a combination of increases in air temperature that increase the duration of the growing season and atmospheric CO2 and elevated atmospheric nitrogen deposition [45,143] which result in fertilization effects on vegetation production. These beneficial impacts may be offset by loss of soil carbon due to increases in mineralization associated with increases in temperature, sometimes increases in nitrate leaching and nitrous oxide production and changes in soil moisture and watershed hydrology. An important approach used to inform and guide air quality management of ecosystems is critical loads. A critical load is the amount of atmospheric deposition of acidity, nutrients or contaminants (usually expressed as mass per area on an annual basis) below which there is no known ecological harm under future steady-state conditions based on current scientific knowledge [144]. Similar to a critical load, a target load can be determined for a particular timeframe by which the specified level of protection will be attained. Critical and target loads have been widely used for management of air pollution effects in Europe and North America using empirical relationships and steady-state and dynamic models [145,146]. Given the impacts of climate change on ecosystem structure and function, future critical load and target load assessments should incorporate climate change effects into the determination of critical/target loads. We (as a research community) address air pollution effects traditionally by sector. Holistic assessments of actual projected costs and/or feedback effects of air pollution across ecosystems, agriculture and human health have to date not been conducted. A joint protocol for a multi-model, multi-effects project and/or a holistic assessment report could help close this gap. The benefits of air pollution controls documented in the scientific literature are overwhelmingly focused on the mitigation of health outcomes. But recent efforts have started to address the co-benefits of improved air quality for health and agriculture, finding that the combined air quality co-benefits for human health and agriculture will counterbalance the implementation costs of measures required to meet the Paris Agreement pledges [58]. One difficulty limiting more holistic considerations is that ecosystem benefits and damages are difficult to monetize. Nevertheless, there are compelling reasons to link the assessment of air pollution and climate stress across sectors. As we and others have suggested previously, air quality and climate management/policy should be linked or at least considered in a holistic way, and should also be viewed as a multi-media problem. Under a changing climate the pathways of air pollution on human health are shifting from predominately direct atmospheric exposure to growing recognition of the importance of indirect multi-media influences from a variety of interconnected environmental factors. This is probably most prominent for mercury (see the example above). Human exposure to mercury as methylmercury largely occurs through the consumption of fish [147] or rice [148]. The transport, methylation and trophic transfer of mercury will all likely be enhanced under climate change, particularly in wet ecosystems [129,149]. There is a climate penalty for ozone, but this climate penalty will likely be accelerated due to temperature-enhanced increases in biogenic VOC emissions [16]. While exposure to ozone reduces wheat yield, the protein proportion of wheat under such conditions has been shown to increase, a possible trade-off implication for nutrition and human health [150]. As discussed above, the contribution of wildfires to atmospheric particulate matter concentrations is increasingly recognized as a component of the air quality climate penalty exacerbated by local air pollution. Increases in atmospheric deposition of nutrients could be beneficial for crop production [116]. However, it is increasingly recognized that harmful algal blooms from cyanobacteria and red tides that release liver and neurotoxins are becoming increasingly prevalent with increases in air and water temperature and nutrient inputs to lake and coastal marine ecosystems [151]. Chapra et al. [152] modelled the future impacts of harmful algal blooms under RCP 4.5 and RCP 8.5 for the coterminous USA, finding significant increases largely due to temperature changes but also nutrients associated with increases in population and increases in runoff. The number of days per year and areal occurrence of harmful algal blooms are projected with increase from RCP 4.5 to RCP 8.5 scenarios and through this century. The greatest increases in occurrence are projected for the northeastern USA, while the greatest costs due to impaired recreation occur in the southeast. Atmospheric deposition of nutrients (nitrogen, and to a lesser extent phosphorus), in addition to land runoff, fuel harmful algal blooms [153,154]. In addition, emission controls on sulfur to improve air quality and reduce acid rain have led to reduced atmospheric sulfur deposition and increased the need for sulfur fertilizer application [117]. Going forward, models of changes in atmospheric deposition under changing climate should be coupled with ecosystem models to project the simultaneous impacts of these long-term changes. The largest body of literature that evaluates the impact of future air quality projections on human and natural resources in a systematic way is focused on human health. While published research generally addresses these sectors in isolation, there have been initial studies examining effects of future air quality on both human health and crops (e.g. [58]). For these two sectors, there is significant overlap in models and scenarios used to evaluate future projections. This is, however, not the case for ecosystems. While many studies evaluate scenarios of future air quality effects on ecosystems [47,48], fewer have coupled these with future climate scenarios [155,156]. Given the similar time scale of responses of ecosystems to atmospheric deposition and climate change and the marked responses of ecosystem hydrology, biogeochemistry, and plants and animals to changing climate, it is critical that climate impacts be integrated into assessments of the impact of air pollution on ecosystems. Of the literature evaluating projected effects of future air quality on human health, studies generally show that the level of health effects projected are strongly dependent on the level of ambition. Substantial health benefits for moderate to low emission RCPs and more stringent policy scenarios (e.g. MFR, attainment of the 2°C target) are projected. Health benefits from reduced PM burdens are projected to outweigh those from reduced ozone globally by more than a factor of 10 (e.g. [50,54]) and global health benefits are driven largely by changes in Asian emissions. Health benefits emerging following reduced Asian emissions manifest strongly at the local/regional level but are also documented to propagate to neighbouring countries and hemispherically [69]. Recent work has also highlighted substantial health benefits emerging from cap and trade policies and/or technological innovations/transitions in the transportation and energy sector (e.g. [62,63,157]). For those studies that consider factors such as population and demographics, they indicate that population growth and an ageing population will exacerbate health effects with larger populations being exposed to unhealthy air pollution at the regional level. Whether global or regional studies, almost all studies indicated that the cost of the policy implementation would be largely compensated for by or in most cases far exceeded by the monetary health benefits (on the order of millions to trillions of USD depending on the geographical scope of the study and level of ambition in the scenario). Consistent with global health benefits being driven by changes in Asian emissions, the associated monetary benefits are also larger in Asia (e.g. [55,58,64,107]). Globally, ozone damage to crops is expected to result in approximately 10% or less additional production loss by 2030/2050 under high emissions (RCP8.5/6.0) scenarios. Reduced emissions (RCP4.5, 2°C target) scenarios can lead to small global crop production gains of a few per cent. Changes to future production are expected to be larger in several regions, including India and China, but these are also more widely varied due to uncertain emissions trajectories. The impacts of PM on crop production are rarely studied and highly uncertain, but the future reduction of PM under reduced emissions scenarios could lead to global production loss of several per cent. The potential decrease in crop production owing to a decrease in PM is, however, much less significant than the likely improvement in human health associated with these reductions. As such, projected effects should be considered holistically and not in silos. Recent work points toward the use of large observational datasets to derive spatially varying relationships, rather than static CR functions, to better represent processes such as regional adaptation to pollution. The few studies that quantified the economic costs of crop losses globally under A2, B1 and CLE scenarios, comparing 2000 to projections for 2030, were relatively consistent, indicating costs ranging from 12–35 billion USD annually [105,106]. These costs are large enough to offset a significant portion of expected GDP growth rates, especially for those countries that are largely agriculturally based. Climate is fundamental to the structure and function of ecosystems. Air pollution involves a host of contaminants (e.g. ozone, sulfur, nitrogen, mercury) and can acidify, eutrophy and/or increase toxic conditions of ecosystems through direct or indirect effects. The effects of climate change on ecosystem structure and function can either mitigate or exacerbate air pollution effects. The response of ecosystems to air pollution and climate change is long-term, complex and varies widely across biomes and over space and time. Although models have been effectively used to project the long-term response to changes in emissions and atmospheric deposition or air quality, few studies have investigated the future response of ecosystems to air quality changes in conjunction with climate change, such as the RCP scenarios. Given the multi-decadal or century long time frame of air pollution and climate impacts, future assessments of air pollution effects on ecosystem structure and function need to recognize and accommodate the interactive effects of these disturbance regimes. The complex, interlinked nature of air pollution effects on diverse endpoints argues for a more holistic assessment of these effects, not only to improve our understanding of how (potential) policy decisions affect these different endpoints, but also to better capitalize on synergies. One possibility for a holistic assessment across sectors would be a joint protocol for a multi-model, multi-effects project, established as an explicit interdisciplinary effort. Since expertize across multiple sectors is rare, an interdisciplinary collaboration would allow for an appropriate application of methods and associated quantification of uncertainty, as well as an effort to standardize expression of the quantified effects. The final step to affect quantification could be the use of a common unit of output, such as economic cost, to facilitate comparison across sectors and allow for a comprehensive cost-benefit analysis, recognizing that monetizing certain types of benefits may be difficult and associated with substantial uncertainty. Given that studies of individual sectors show that the benefits of mitigation policies generally outweigh the costs, a comprehensive analysis would likely show even greater benefits. An envisioned outcome would be to foster the implementation of more ambitious mitigation policies that more clearly characterize and quantify co-benefits associated with reducing adverse effects of air pollution and climate change across human health, agriculture and ecosystems in the future. This article has no additional data. E.v.S. developed the concept and coordinated the review. All authors contributed to writing, editing, revision and gave final approval for publication. We declare we have no competing interests. We received no funding for this study. We appreciate the help of Mark Fenn for providing perspective on air pollution impacts on susceptibility of trees to fire and to Kim Driscoll and Drew Shindell for help with figures. The air pollution research by C.T.D. is supported by the JPB Foundation, New York State Energy Research and Development Authority and the National Science Foundation through the Long-Term Ecological Research Program and the National Science Foundation Research Training program. The research of E.v.S. is supported by IASS Potsdam, with financial support provided by the Federal Ministry of Education and Research of Germany (BMBF) and the Ministry for Science, Research and Culture of the State of Brandenburg (MWFK). The air pollution research by H.E.R. is supported by the University of Natural Resources and Life Sciences, Vienna and the Austrian Climate and Energy Fund through the Austrian Climate Research Program (ACRP11). L.D.S. is supported by research funding from the Department of Earth and Environmental Sciences at Columbia University. FootnotesOne contribution of 17 to a discussion meeting issue ‘Air quality, past present and future’. © 2020 The Authors. Published by the Royal Society under the terms of the Creative Commons Attribution License http://creativecommons.org/licenses/by/4.0/, which permits unrestricted use, provided the original author and source are credited. References
Page 16The current exposure to air pollution in ambient air has been identified as the worldwide largest environmental risk factor for human health [1]. Anthropogenic activities emerge as the main drivers for emissions of air pollutants and add to pre-existing sources of natural emissions (soil dust, sea salt, vegetation, etc.) [2]. Human development affects emissions along multiple pathways: increasing pressure from population growth, industrialization and modern lifestyles is counteracted by technological progress, structural changes in the economy and targeted pollution control efforts. The interplay of these factors is changing over time; while pollution levels in industrialized countries have decreased, the developing world witnesses unprecedented levels of pollution today. The World Health Organization (WHO) estimates that currently about 90% of the people living in cities are exposed to PM2.5 levels above the WHO guideline value of 10 µg m−³, and globally between 3 and 9 million cases of premature deaths annually have been attributed to exposure to ambient air pollution [1,3,4]. Given the dynamics of these factors and their complex interplay, what could be expected for future air quality around the world, and which determinants will be dominating? To answer this question, this paper identifies key factors that contributed to historic air pollution trends in different world regions, outlines conceivable ranges of their future development and examines their interplay on global air quality in the next decades. In particular, the paper provides a fresh perspective on how ambitious policy interventions could achieve clean air worldwide. The observed increase and subsequent decline in SO2 emissions in many industrialized countries during the second half of the twentieth century inspired the ‘environmental Kuznets curve’ hypothesis, suggesting that environmental degradation tends to get worse as modern economic growth occurs until average income reaches a certain point [5,6]. However, the explanatory power and general validity of this hypothesis has been strongly contested [7–12], inter alia, because similar turning points have not yet been observed for other substances, including agricultural ammonia (NH3) and greenhouse gas emissions. Also, the role of (environmental) policy interventions is not explicitly recognized but implicitly subsumed as an autonomous concomitant of economic growth. Rafaj et al. [13] identified three key factors that contributed to the observed decoupling between economic activity and air pollution in Europe between 1960 and 2010: (a) economic structural change, i.e. conversion to less energy-intensive economic activities, (b) energy policy, i.e. phase-out of oil and coal, and (c) dedicated air pollution control policies requiring efficient end-of-pipe cleaning devices (e.g. flue gas desulfurization, catalytic converters). As a result, during a period in which GDP quadrupled, SO2 emissions in Europe declined by 90% and NOx emissions returned to their 1960 levels after an increase by a factor of three. Similar findings are reported by [14] for Europe and by [15] for North America. Most recently, decoupling of economic development trends and SO2 and NOx emissions and the critical role of policy interventions have been observed in China [16–18], but not yet in India [19]. A range of studies in the scientific literature explored the implications of these findings on future emissions and air quality. For a long time, future global air pollutant trends were mainly modelled in the context of long-term greenhouse gas emission scenarios [20]. The early global studies on air pollutant emissions, notably the scenarios developed for the ‘Special Report on Emissions Scenarios' [21] and the ‘Representative Concentration Pathways’ [22,23] that have been prepared for the Intergovernmental Panel on Climate Change (IPCC) proposed declining trends of (energy-related) air pollutants, due to autonomous technological progress and assumed pollution control policies along the environmental Kuznets hypothesis. Later, the improved understanding of the importance of targeted air quality policy interventions motivated a more differentiated approach to projections of air pollutant emissions, resulting in a wider range of air pollutant trajectories than in previous global scenarios [24–26]. At the same time, the climate community addressed the interactions between decarbonization strategies and air pollutant emissions, both with the interest to reveal health benefits from low carbon policies [24,27–31]) and to explore the combined impacts of long-lived greenhouse gases and short-lived air pollutants (e.g. SO2 and black carbon) on radiative forcing and temperature increase [32–34]. In general, the literature reveals strong impacts of ambitious decarbonization strategies on energy-related air pollutants SO2, NOx and PM, due to the phase-out of fossil fuels and the containment of all flue gases connected with carbon capture and storage. However, enhanced use of biomass as a greenhouse gas policy measure may lead to higher PM emissions [35–38]. Compared to the climate-focused analyses that deal mainly with energy-related emissions and the role of climate policy interventions, only a few studies addressed the longer-term prospects for air pollution from a health- and ecosystems perspective [4,39,40]. These studies take full account of other sources that also contribute substantially to population exposure to harmful air pollution, such as agricultural activities, waste management and materials handling. Also, they developed a more holistic approach towards the understanding of future trends in nitrogen emissions and their health and environmental impacts. Not so many studies address air pollution at the global scale [26,41,42]. This is not surprising as, due to its physical features, air pollution is often considered as a local/regional and short-term problem, even if it is of universal nature, i.e. occurring as a concomitant of development in most industrialized areas throughout the world. However, a global perspective is of interest due to the intercontinental transport of pollution [43,44], the serious global health burden of air pollution [1] and the intimate interactions between clean air policies and efforts to achieve the UN global sustainable development goals [45–47]. For the second half of the twenty-first century, most long-term studies are in general agreement about declining energy-related air pollutant emissions; greenhouse gas mitigation strategies would accelerate the decline. However, projections differ for the next few decades, and particularly for the timing of global emission peaks. To explore the conceivable range of future air quality and, in particular, of population exposure to PM2.5 which has been associated with the most harmful health impacts [48], this paper develops a series of alternative emission scenarios up to 2040. Building on a widely accepted economic growth path with its structural economic changes, these scenarios combine different assumptions on the key policy areas that have been identified as critical for air pollution trends in the past, i.e. (a) energy/climate policy, (b) agricultural policies and (c) dedicated pollution control policies. The analysis employs the greenhouse gas-air pollution interactions and synergies (GAINS) model [49]—see electronic supplementary material. For exogenous projections of future emission-generating human activities (i.e. energy, transport, industrial production, agricultural activities and waste volumes), GAINS computes emissions of 10 air pollutants and six greenhouse gases (see electronic supplementary material) considering the current emission characteristics in 180 regions. For the future, GAINS takes account of the penetration of already decided control measures and estimates the potential for additional emission reductions that is offered by several hundreds of control measures. The resulting emissions of all precursor emissions of PM2.5 in ambient air, i.e. primary PM2.5, SO2, NOx, NH3 and volatile organic compounds (VOCs), are then fed into an atmospheric dispersion model to compute annual mean concentrations of PM2.5 across the globe. GAINS employs reduced-form source-receptor relationships that have been derived from the EMEP atmospheric chemistry-transport model [50] with a spatial resolution of 0.125° × 0.0625° (approx. 7 km × 7 km) in Europe [51] and 0.1° × 0.1° (approx. 5–10 km × 5–10 km, depending on latitude) outside Europe, distinguishing about 6000 individual cities with more than 100 000 inhabitants. Resulting concentration fields are then compared with air quality standards, and corresponding population exposure is computed for the population distribution assumed in the socio-economic projection. More detail is given in the electronic supplementary material. Over the last 25 years, SO2 emissions from anthropogenic activities declined by 40% at the global level. Primary emissions of PM2.5 and of VOC stabilized, NOx grew by about 10% and NH3 by one-third (figure 1). Obviously, these global trends emerge as an aggregate of sometimes diverging regional trends (see §3c(iii)). Figure 1. Trends of global GDP, population and precursor emissions of PM2.5 from anthropogenic sources, 1990–2015 (relative to 1990) (source GAINS). (Online version in colour.)
During the same period, GDP doubled, and world population grew by more than a third, i.e. faster than most of the air pollutant emissions. Thus, except for NH3, emissions on a per capita basis declined globally in these 25 years, in line with the earlier observations that led to the formulation of the environmental Kuznets hypothesis [5,6] (figure 2). It is noteworthy that after 1990 clear peaks in per capita emissions could only be observed for SO2 and NOx in China, while all other world regions exhibit constant or declining trends following the increase in per capita income. Also, per capita emissions of the six world regions that showed striking differences in the past reduced over time despite large disparities in income levels. Figure 2. Per capita emissions of PM precursors versus per capita income from 1990 to 2015 (in 5-year steps), for the six world regions (source GAINS). (Online version in colour.)
For energy-related emissions, a global level decomposition analysis apportions about half of the past divergences of economic and pollution trends to environmental policies, while the remainder occurred from economic structural changes and energy policies (figure 3). By contrast, for global agricultural NH3 emissions no impacts from structural changes nor from pollution control policies can be identified. However, trends differ across world regions. Figure 3. Factors contributing to the decoupling between the GDP trend and the evolution of global precursor emissions of PM2.5 from anthropogenic sources between 1990 and 2015 (source GAINS). (Online version in colour.)
For 2015, calculations confirm large differences in PM2.5 concentrations across the world (figure 4), with patterns rather similar to other studies that have been derived from a limited set of ground-level monitoring data [1] or from interpretation of remote sensing products [52,53]. A comparison of ground-based monitoring data [54,55] with model calculations is presented in figure 5. Figure 4. PM2.5 concentrations in 2015 as computed for this paper with the GAINS model, including contributions from natural sources. (Online version in colour.)
Figure 5. Comparison of monitoring data (annual mean concentrations in 2015) for PM2.5 with GAINS model results for 2015. Monitoring data taken from [54,55]. (Online version in colour.)
The main sources of PM2.5 in ambient air vary over world regions [56]. While in many densely populated or highly industrialized areas the major share originates from anthropogenic sources, natural sources (soil dust, sea salt and vegetation) are important in other areas, with soil dust dominating in arid regions in Asia, Africa, Latin America and Australia. Human activities might contribute only indirectly to these natural emissions, e.g. through desertification, land use changes and climate change. In general, such emissions do not appear to be controllable at the decadal time scale, as this would require geo-engineering interventions at the large scale. With a focus on future policy interventions, PM2.5 concentration fields that emerge from anthropogenic emission sources only are of special interest. For 2015, the highest hot spots caused by human activities are computed for eastern China and the Indo-Gangetic plain; concentrations exceeded 10 µg m−³ in large areas in China, Korea, Indonesia, throughout India, in central and eastern Europe, the Po valley and in the Benelux area, as well as in the Gulf States, the lower Nile valley, Nigeria, and around Johannesburg in South Africa (figure 6). Figure 6. Computed PM2.5 concentrations in 2015 from anthropogenic emission sources only (contributions from natural sources are excluded). (Online version in colour.)
While there is little evidence to suggest a threshold below which no adverse health effects would be anticipated, the WHO has issued a global air quality guideline value for PM2.5 in ambient air of 10 µg m−³ as annual mean, together with a series of interim targets of 15, 25 and 35 µg m−³, respectively [57]. For 2015, it is estimated that globally more than 90% of the city dwellers [54] and more than 80% of the total population (this study) were exposed to PM2.5 levels above the WHO guideline, especially in Asia, Africa and the Middle East. Focusing on the exposure that can be influenced by policy decisions, more than 60% of the global population (85–90% in Asia) lived in areas where PM2.5 concentrations from anthropogenic sources exceeded 10 µg m−³ (figure 7). Figure 7. Distribution of population exposure to PM2.5 from anthropogenic sources in 2015. (Online version in colour.)
In the past, changes in air pollutant emissions have been driven by economic development and policy choices [13,19]. There is general agreement in the scientific literature that in the long run, i.e. in the second half of the twenty-first century, continuing growth in economic wealth, technological development and hypothesized further policy interventions should lead to significant declines in energy- and poverty-related emissions [24]. However, the prospects for the first half of the twenty-first century are less clear, especially when and at what level emissions are likely to peak in the various world regions. Although comparably less work has addressed future agricultural emissions [58], there are hardly any studies that would indicate a possible peaking, e.g. of NH3 emissions, at the global scale [59]. To shed light on the question of emission peaking and the role of policy interventions, this paper examines the prospects of air pollutant emissions, air quality and population exposure for a range of alternative policy assumptions. The analysis extends to 2040. Although there is some uncertainty about future socio-economic development trends, the current scientific literature sees only moderate differences for the next few decades (e.g. until 2040) due to the long time scales of fundamental transformations of the social and economic systems. Thus, this paper adopts the macro-economic development assumptions of the World Energy Outlook 2018 of the International Energy Agency (IEA) [60], which up to 2040 are rather similar to the middle range of the long-term climate scenarios developed for the CMIP6 Coupled Model Intercomparison Project [28,61]. In total, the trends in 25 world regions of the IEA scenario correspond to a 25% growth in world population between 2015 and 2040. Global GDP increases by about 90%, so that global per capita income rises by 50%. Three scenarios explore the potential impact of policy interventions on future air quality:
The different assumptions on policy interventions deliver a wide range of future emission trends for PM2.5 precursors (figure 8; electronic supplementary material). By 2040, without policies and measures introduced in the last decades, energy-related emissions could have been up to 120% higher than in 2015. Current policies, if effectively implemented and enforced, would reduce global SO2 emissions by about one-third and NOx by about 10%, and stabilize global PM2.5 emissions. By contrast, with adequate political will, global PM2.5 and SO2 could be cut by about 90% below today's level, and NOx by about 70%. However, current policies will do little to decouple global NH3 emission trends from population growth, while ambitious interventions have a potential to enable 60% lower emissions. However, as this analysis does not consider potential impacts of climate change on meteorological conditions, changes in emissions from natural sources, e.g. desert dust and biogenic VOCs, are not excluded from the analysis. These could potentially enhance wind-blown PM2.5 and enhance ground-level ozone. Figure 8. The range of future global emissions (2015–2100) resulting from alternative assumptions on policy interventions. The grey area indicates the emission reductions from the policies and measures that have been adopted and implemented until 2018, while the blue areas outline the scope for additional policy interventions. The dark blue line indicates the ‘2018 legislation’ scenario which assumes that all air pollution controls that have been decided by 2018 will be timely and effectively implemented and enforced. For comparison, the dashed grey lines illustrate the range of the long-term emission scenarios used for the CMIP6 climate model calculations. Note that CMIP6 does not provide PM2.5 emissions. (Online version in colour.)
In general, the range of future emissions developed for this paper is consistent with the scenarios of the CMIP6 Coupled Model Intercomparison Project [28,61,70] of the World Climate Research Program. However, the ‘2018 legislation’ scenario indicates stabilizing or even rebounding emissions towards 2040 after full implementation of current policies, while emission declines prevail in CMIP6. Also, the global trend masks diverging developments in some world regions. Especially, in many Asian and African countries, the current pollution control policies seem insufficient to counteract pressure on NOx, PM2.5 and NH3 emissions from economic growth (figure 9). Figure 9. Emission trends 1990–2040 assuming effective implementation and enforcement of all pollution controls that were decided by 2018 (Scenario: 2018 legislation). For comparison, the global trends for the ‘Without air pollution policies' and the ‘Clean Air’ scenarios are provided. (Online version in colour.)
Changes in emissions, prominently determined by policy decisions, will impact future air quality and population exposure. By 2040, effective implementation of the 2018 legislation would bring PM2.5 concentrations from anthropogenic sources below their respective 2015 levels in Europe, North America, and North and East Asia. In other regions, concentrations would grow further, notably in South Asia and Africa (+25%). Global population-weighted annual mean concentrations would increase by 10% to about 35 µg m−³ (including contributions from natural sources) (figure 10). At the same time, due to the improvements in industrialized countries, the share of the global population living in areas where PM2.5 complies with the current WHO guideline would increase to 23% compared to 18% in 2015. Figure 10. Reductions in population-weighted mean exposure to PM2.5 in the six world regions. (Online version in colour.)
Highest PM2.5 concentrations would then occur in Africa and the Middle East, where concentrations would reach the levels currently prevailing in Asia (figure 11a). Anthropogenic sources would make the largest contributions in eastern China and the Indo-Gangetic plain and increase by 60% in Africa and 30% in the Middle East (figure 11b). Figure 11. Modelled PM2.5 concentrations in 2040 for the 2018 legislation case. Upper panel: total PM2.5 including anthropogenic and natural sources; lower panel: contributions from anthropogenic sources only. (Online version in colour.)
By contrast, the ambitious policies and measures of the Clean Air scenario would deliver significant cuts in PM2.5 levels throughout the world. In Europe and the Americas, population-weighted concentrations could be brought down to about 5 µg m−³ (including natural sources), in Asia to about 15 µg m−³ and in Africa and Middle East to 20 and 30 µg m−³, respectively (figure 12a). Fifty-six per cent of the global population would then live in areas where PM2.5 complies with the current WHO guideline. Figure 12. Modelled PM2.5 concentrations in 2040 for the Clean Air scenario. A: total PM2.5 including anthropogenic sources; B: contributions from anthropogenic sources only. (Online version in colour.)
Natural sources will continue to make sizeable contributions to population exposure, especially in Africa, Latin America and the Middle East. In these regions, the Clean Air scenario would bring contributions from anthropogenic sources to below 3 µg m−³ (figure 12b). The significant exposure reductions of the visionary Clean Air scenario emerge from a range of policies and measures that have been grouped into four policy packages: (i) environmental pollution control policies, (ii) energy and climate policies, (iii) agricultural policies, and (iv) food policies (figure 13). Further environmental pollution controls offer the largest improvements in population exposure. However, even full implementation according to the assumptions in this paper would still leave about 2.4 billion people exposed to more than 10 µg m−³ PM2.5 from anthropogenic sources. Beyond this, energy and climate measures could reduce exposure below this level for another 820 million people, changes in agricultural practices for 670 million people and the assumed food policies for an additional 500 million people. Thus, further air pollution control policies would exhaust only 60% of the total potential offered by the full range of policy interventions. Vice versa, the notion that other sectoral policies such as ambitious energy and climate policies would resolve air pollution problems on their own, e.g. [23], is not supported by this analysis, and a clear role for dedicated environmental air quality policies will remain in the future. Figure 13. Scope of the policy packages to protect people from PM2.5 exposure from anthropogenic sources in 2040. (Online version in colour.)
Some of the measures of the Clean Air scenario will not only reduce emissions of PM2.5 precursors but simultaneously also other substances that contribute to temperature increase. As a consequence of the deep energy system restructuring in the Clean Air scenario, CO2 emissions in this scenario in 2040 are 40% lower than in the 2018 legislation reference case, CH4 33% and black carbon 90% lower (figure 14). Quantification of the temperature impact of these cuts in long-lived greenhouse gases and short-lived climate pollutants is beyond the scope of this paper. We also did not quantify reductions of N2O emissions, which can be expected as a result of reduced application of mineral fertilizer and more efficient use of manure nutrients. Figure 14. Emission reductions from the policy scenarios. 2018 legislation includes all controls that have been decided by 2018. (Online version in colour.)
As this paper focuses on health impacts from PM2.5, the Clean Air scenario does not target substances that contribute to other air pollution problems. SO2, NOx and NH3 emissions, which affect acidification and eutrophication of ecosystems, are included as PM precursors. For ground-level ozone, NOx and VOC emissions are considered as PM precursors as well. However, CH4 and CO emissions are not addressed here, as their contribution to PM levels is considered marginal. A comprehensive clean air policy that would also aim at ground-level ozone would need to include CH4 and CO mitigation in its portfolio, especially in view of the past increases in hemispheric ozone levels that have been attributed to growing CH4 emissions [71]. From such a perspective, measures that reduce only CH4 but would not affect PM precursor emissions could cut global CH4 by more than half in 2040 [72]. The Clean Air scenario achieves a substantial reduction in exposure to PM2.5, bringing about significant progress in moving towards the WHO air quality guideline value for PM2.5, which is currently set at 10 µg m−³ including contributions from natural sources. In large areas of the world, the Clean Air scenario would reduce anthropogenic PM2.5 below 5 µg m−³ and thereby leave space for contributions from natural sources. These improvements will directly contribute to SDG 3 (Improve human health and wellbeing), particularly in urban areas (SDG 11—sustainable cities and communities). In addition, the measures assumed in the Clean Air scenario with the focus on health impacts from PM would deliver a host of co-benefits in other policy areas through several pathways.
This paper develops an initial global vision on policy interventions that could bring PM2.5 levels close to or below the WHO air quality guideline value in the next few decades. The analysis takes a holistic perspective, connecting past trends of economic and technological development, atmospheric dispersion and the role of policy interventions with a focus on improving human wellbeing through better global air quality. The wide range of aspects included in the analysis and the pioneering character of this paper leave a number of issues that deserve further attention. First, the paper focuses on population exposure to PM2.5 in ambient air, as associated health impacts have been found to dominate other negative consequences of air pollution in economic terms [74–76]. Thereby, the paper does not explicitly address health impacts from ground-level ozone [77], vegetation damage from ozone including economically important losses in crop production [78,79], biodiversity impacts of excess nitrogen deposition [80] and acid deposition that exceeds the absorption capacity of ecosystems [81]. For the later issues, i.e. excess nitrogen and acid deposition, the Clean Air scenario delivers significant reductions in precursor emissions as these contribute to PM2.5 formation too. By 2040, NOx would be lower by three-quarters, NH3 by two-thirds and SO2 by 80% than compared to the 2018 legislation projection. These reductions will alleviate pressure on ecosystems, but the benefits have not been quantified for this paper. Similarly, the Clean Air scenario, even while aiming at ambient PM2.5, would lead to much lower precursor emissions of ground-level ozone. Global NOx emissions would decline by about three-quarters and global VOC emissions by about two-thirds compared to the 2018 legislation case, because of their role in the formation of secondary particles. In addition, the measures would also reduce the other precursors of ground-level ozone that are often neglected in local and regional pollution control strategies. Global CH4 emissions, which have an important impact on hemispheric background levels of ozone [82], would decline by 33% as a side-effect of the changes in fossil fuel production, agricultural practices and food demand that are motivated by the interest to reduce PM precursor emissions including NH3. More work will be required to quantify the consequences of these emission changes on ozone fields and resulting health and vegetation impacts. Second, the paper limits the analysis to 2040, in order to explore near-term policy interventions without excessive speculation about long-term technological, socio-economic and climate trends. At the same time, the long-term transformational energy and agricultural policies that are assumed in the Clean Air scenario involve measures that will fully unfold only by 2050, so that the 2040 focus does not reveal their full potential in the long run. Additional work would be required to assess the impacts beyond 2040. Furthermore, any visionary analysis requires assumptions and methods that are associated with uncertainties. The following aspects have been identified as particularly relevant.
Over the last decades, emissions of key air pollutants have decoupled from economic growth at the global level, and increasingly also in the developing world. This trend break was to a large extent a result of policy decisions on pollution control and energy, in addition to structural changes in the economy and in consumption patterns. It is estimated that without dedicated pollution control policies, global SO2 emissions would have been more than twice as high as in 2015, and PM2.5 and NOx emissions twice as high. Given the lack of large-scale policy interventions, the evolution of agricultural NH3 emissions closely followed global population growth. The dominating role of policy interventions will prevail in the future. Instead of autonomous improvement of air quality connected with increasing economic wealth as hypothesized in the environmental Kuznets curve, this analysis reveals that future air quality will be mainly determined by policy decisions and their implementation. Timely implementation and full compliance with all air legislation that has been decided by 2018 is likely to decrease global anthropogenic SO2 emissions by 35% by 2040, and NOx and primary PM2.5 by about 10%. By contrast, for NH3 emissions a growth of one-quarter is estimated, with no peaking in sight. However, as illustrated by the Clean Air scenario, with adequate political will further policy interventions could cut global SO2 and PM2.5 by about 90%, NOx by 70% and NH3 by 60% below today's levels. Decisions in four policy areas will be critical: environmental policies focusing on pollution controls, energy and climate policies aimed at global temperature stabilization in line with the Paris Agreement, policies to transform the agricultural production system, and policies to shift human food consumption patterns towards largely plant-based diets such as the ‘Planetary Health Diet’ proposed by the EAT-Lancet Commission on Food, Planet and Health. However, emissions from biogenic sources (e.g. VOC) might further increase due to other factors, e.g. climate change. Obviously, the policy interventions considered in the Clean Air scenario would require fundamental transformations of today's practices in many sectors. These are visionary but considered likely to be technically achievable in the future. As they exceed current policy ambitions, their implementation would require strong political will. Political will could emerge from a solid understanding of the full range of benefits. Most importantly, the policies and measures of the Clean Air scenario would drastically improve air quality throughout the world. Population exposure to PM2.5 from anthropogenic sources would decline by about 75% relative to 2015, or by 80% compared to a future without additional policies. This would deliver substantial health benefits and avoid a large portion of the 3–9 million cases of premature deaths from ambient air pollution that are estimated for 2015. In addition, enhanced access to clean household fuels will drastically reduce the health impacts from indoor exposure to PM2.5, currently estimated to cause another 1.6–4 million cases of premature deaths annually [3,73,89]. These reductions in PM2.5 exposure from anthropogenic sources would allow substantial progress in moving towards the WHO air quality guideline value, which is currently set at 10 µg m−³. In large areas of the world, the Clean Air scenario would reduce anthropogenic PM2.5 concentrations below 5 µg m−³ and thereby bring total PM2.5 below the current WHO guideline value, which includes contributions from natural sources. However, violations will remain in areas where natural sources alone deliver already more than 10 µg m−³, as well as at small-scale hot spots which are not captured by the spatial resolution of this analysis. In addition to improved human health (SDG 3—Improve human health and wellbeing), particularly in urban areas (SDG 11—Sustainable cities and communities), the Clean Air scenario would deliver a host of co-benefits in other policy areas through several pathways.
Importantly, the Clean Air scenario includes a range of fundamental transformative changes (e.g. in the energy, agricultural and food systems) which are required for global sustainability and whose benefits occur globally and in the long term. The tangible local and near-term health benefits of the Clean Air scenario could enhance social acceptance and political support for such transformative policies. Despite the importance of a deep involvement of and interactions with numerous other policy areas, the analysis reinforces the continued relevance of environmental air quality policies. Ambitious action in other policy domains such energy, climate, food and agriculture will not resolve the air pollution problem on its own and—if not coordinated within a comprehensive multi-sectoral air quality management approach—could even lead to counterproductive impacts on air quality.
Data are accessible in the online version of IIASA's Greenhouse gas—Air pollution Interactions and Synergies (GAINS) model (http://gains.iiasa.ac.at). Data will be made available after acceptance of the paper. Key data are provided in the electronic supplementary material. M.A. coordinated work and drafted the article. G.K. conducted the atmospheric dispersion calculations and computed population exposure for the various scenarios. W.S. managed the database system and developed routines to compute emissions for the policy scenarios. Z.K. led the estimates of past and future PM and VOC emissions. W.W. specified the control measures for the agricultural and food policies. J.C. and P.R. conducted the calculations of emissions of the energy system, and imported IEA energy scenarios into the GAINS model. L.H.-I. estimated the impacts of the policy packages on future CH4 emissions. A.G.-S. was responsible for emission estimates from the waste sector. C.H. developed the gridded emission fields for all scenario and managed the data exchange with the Norwegian Meteorological Institute. P.P. focused on emissions from developing countries, especially from the non-conventional non-point sources. J.B.-K. specialized on current and future emissions from the transport sector. F.W. estimated the emission control potentials for the various policy packages. R.S. managed the GAINS database system. H.F. and A.N. conducted the full atmospheric dispersion calculations with the EMEP chemistry-transport model of the Norwegian Meteorological Institute, and transferred results to IIASA. L.C. and C.P. contributed to the development of the IEA energy scenarios, transferred data to GAINS and assisted in the interpretation of results. We declare we have no competing interests. IIASA's modelling work was funded by core funds of the International Institute for Applied Systems Analysis. In addition, the development of the air pollution scenarios has been supported by the European Union project ‘Action on Black Carbon in the Arctic’ (Commission Implementing Decision on the 2016 Annual Action programme for the Partnership Instrument). The contributions of the International Energy Agency (IEA) were funded through the IEA resources (this sentence needs to be refined before final publications). The contributions of the Norwegian Meteorological Institute (met.no) have been funded by the met.no. We are grateful to GLOBIOM modellers, specifically to Andre Deppermann, for providing results of their diet scenario model runs. FootnotesOne contribution of 17 to a discussion meeting issue ‘Air quality, past present and future’. Electronic supplementary material is available online at https://doi.org/10.6084/m9.figshare.c.5105954. © 2020 The Authors. Published by the Royal Society under the terms of the Creative Commons Attribution License http://creativecommons.org/licenses/by/4.0/, which permits unrestricted use, provided the original author and source are credited. References
Page 17
You have to enable JavaScript in your browser's settings in order to use the eReader. Or try downloading the content offline DOWNLOAD |
When did pollution become a problem
Postagens relacionadas
direito autoral © 2024 comojogaruno Inc.
