Supplementary Fig. 6 also shows similar diurnal cycles of PBLH in polluted and less polluted areas. Therefore, despite local aerosol-PBL interactions, a shifted burning cycle on an hourly basis is not likely to significantly affect the broad patterns of PBLH and the timing effects of emissions. However, further investigation will be needed to evaluate how a shift in the diurnal pattern of burning might in turn modify the diurnal pattern of PBLH––and therefore pollution dispersion––compared to the effects observed to date for existing emission patterns. Other meteorological parameters including relative humidity , temperature, and wind speed also vary diurnally and may affect the benefits of burning earlier. Studies in India found higher pollutant concentrations with higher RH due to incomplete combustion, causing more secondary aerosol formation. Supplementary Fig. 6 shows that, on average, RH is roughly constant from 11:30 to 15:30 local time. Our proposed shift moves the peak burning time from 14:30 to 13:30, suggesting that the RH effect is not significant. In addition, lower temperature and lower wind speed could trap more aerosols within the PBL. This is consistent with our finding of decreased air quality benefits when the burning peak is too early or late . Changes in wind speed due to high aerosol loading should also be considered in future online studies of specific interventions. These findings suggest that aerosol-meteorology interactions are unlikely to eliminate the benefit of a shift in the diurnal pattern of burning. We recognize however that further studies involving sensitivity experiments will be needed to evaluate the degree to which these feed backs suppress or enhance the benefit.

In general,stackable flower pots for any intervention our approach is a first-order tool, and more modeling, experiments and observations would be warranted before advancing any policy recommendation. However, our broad explanation is that the aerosols emitted when meteorology favors dispersion are more diluted and less harmful as they travel downwind . This work focuses on how India-wide air quality impacts are affected by any conceivable change in fire emissions using an “adjoint” modeling approach. Rather than tracking the transport of pollution and the distribution of air quality impacts for a single scenario, the adjoint model tracks the sensitivity of exposure over a disperse population with respect to changes in emissions sources. This is essential for efficiently evaluating different cost-effective pollution abatement policies. We therefore do not calculate the local effects of shifting the timing of burning, e.g. the evolution of air pollution to downwind areas such as New Delhi.To provide context to the sensitivity-based results above, we use forward modeling to simulate the post-monsoon residue burning season over a 23-year period . We first quantify population exposure to BC and OC from all sources. For these simulations we include changes in residue burning emissions, population , and meteorology . We find that, while the daily-mean population exposure varies, the geographical distribution of exposure and meteorological context remain similar. From 1997 to 2019, the national average population-weighted PM2.5 exposure has increased from 54 μg m−3 to 75 μg m−3 , possibly due to increasing crop production . Northwest India is the most polluted region during post-monsoon burning season, and the daily-mean population-weighted PM2.5 exposure in these areas is consistently higher than the national average. For every year from 1997 to 2019, we estimate that Delhi’s air consistently had levels of PM2.5 exceeding 120 μg m−3 when averaged over the two month post-monsoon burning period.

We also perform 23 counterfactual simulations, covering the same period, in which emissions from residue burning in India are not included. While our sensitivity simulations show that our broad conclusions are consistent throughout 17 years of differing fire emissions and three different meteorological conditions , our extended set of forward simulations show that the average air quality impacts attributable to agricultural fires is 2.4% lower in drought years and 4.8% higher in flood years . For years with normal rainfall , close to the 17-year average, we estimate the annual fire-related premature deaths at 68,000 , valued at 22 billion USD. Specifically, crop residue burning in 2016 contributes to the largest enhancement of nationwide PM2.5 exposure , resulting in the most premature deaths 98,000 . This is consistent with the large number of agricultural fires observed by the Moderate Resolution Imaging Spectroradiometer instrument in that year, and the unprecedented enhancement in PM2.5 levels observed in Delhi. These results only include the post-monsoon burning season, and rely on monthly emissions data. Using our sensitivity data with daily emissions data and assigning each year from 2003 to 2019 an appropriate meteorological condition , we can quantify the impacts of each full year’s residue burning. From 2003 to 2019, along with an increase of food grain production from 210 MT to 300 MT , the monetized cost associated with burning increases from 7.2 to 44 billion USD . A point of interest is the impacts in 2018 and 2019. In 2018, a subsidy was introduced to encourage mechanization as an alternative to burning. Despite this policy change, we find air quality impacts attributable to crop residue burning to be similar between 2018 and 2019 at 86,000 premature deaths each year. To understand this, we rederived impacts for 2015-2019 using fixed meteorological data.

Under these conditions, mortalities in 2018 and 2019 are 4.2% and 11% lower than the previous three-year average, but 15% and 6.9% greater than for 2017, respectively . For just Punjab, Haryana, and Uttar Pradesh, the average exposure resulting from burning in 2018 and 2019 was 16% lower than the 2015–2017 average, but still 18% greater than the 2017 value alone . This does not necessarily imply that the initiative was unsuccessful. Prior work has suggested that overall post-monsoon residue burning was decreased, showing a 18% reduction in fire counts observed by the MODIS satellite in northwest India in 2019 compared to the previous year . The GFEDv4.1s dataset suggests that the total dry matter burned in Punjab, Haryana, and Uttar Pradesh in 2018 and 2019 was only 9 and 3% less than the 2015–2017 average , respectively––but this masks two compensating changes. Focusing on the post-monsoon period only, the total dry matter burned in the three states in 2018 and 2019 was 36% below the 2015–2017 average, and only 1% greater than the 2017 value. Dry matter burned during the pre-monsoon period, however, was 52% and 130% greater in 2018 and 2019 respectively than the 2015–2017 average . With fixed meteorology, pre-monsoon residue burning contributed 38% of the total premature mortality resulting from fire emissions in these three states in 2019, compared to 9% in 2015 . Although more data are needed to determine the significance of this trend, tower garden our results suggest that health benefits due to a reduction in emissions during the post-monsoon period in 2017-2019 have been offset by an increase in emissions during the pre-monsoon period and that the effectiveness of interventions should be considered across both seasons.In India, the total amount of crop residue generation has increased from 80 MT in 1950–1951 to 520 MT in 2017-2018 . In recent years, crop residue burning has contributed to levels of PM2.5 concentrations that are 15–45 times higher than the WHO safety guidelines in northern India. Our results suggest that this burning has a monetized annual cost of 23 billion USD averaged from 2003 to 2019, which has grown by a factor of six over the same period. Existing governmental efforts, including the National Policy for Management of Crop Residues, the National Green Tribunal Act, and the Straw Management System, are in place to reduce the practice of burning. However, in-field burning remains prevalent, especially in the states of Punjab and Haryana. First, we find that under similar meteorology and rice-wheat cropping system, the attributable air quality impacts from Punjab are six times as much as those from Haryana, partly because Haryana districts mainly grow basmati rice, which is more utilizable than nonbasmati rice cultivated in Punjab.

Therefore, in addition to diversifying rice with crops that produce less residue such as pulses and oilseeds , adopting rice varieties that are less burning intensive could also reduce the amount of burned residue and eventual attributable fire-related impacts. Second, current regulations, including bans and fines, are mainly implemented at national or state levels, which may overlook the possibilities that local-scale actions can bring significant benefit. Our study shows that a small number of administrative areas may be prioritized for interventions to effectively reduce the attributable impacts from fires , with burning in six districts in Punjab responsible for 40% of India-wide exposure to fire related PM2.5. Such information is helpful for spatially targeted decision-making. Third, we also find that burning earlier by a few hours within a day could avert up to 14% of the air quality impacts resulting from residue burning, over 90% of which is borne by Indians. Such temporally targeted interventions may therefore allow effective and potentially low-cost reductions in harm while local effects on pollution distribution can be further investigated. In addition, a combination of targeted decision-making and more permanent solutions, such as mechanization, may help to optimize resources and minimize disruption to farmers. While this study implies that significant societal benefits can potentially be achieved by small-scale actions, a comprehensive cost benefit analysis and consideration of extra incentives for farmers is needed for actionable planning and wide adoption of alternatives. Our hope is that work such as ours can provide quantitative data for near term measures to effectively reduce the harms of agricultural residue burning while more holistic solutions can be pursued.In the case of crop residue burning, most existing work focuses on air pollution at local and urban scales, highlighting the influence of agricultural fires on regional air quality. However, impacts of fire emissions are likely to extend over a much larger area due to dispersion and transport. Epidemiological studies show that any additional exposure to PM2.5 causes an increased mortality risk even when baseline exposure is very low. This means that small exposure increases over large regions should be considered equally with focused increases over smaller regions. Our study equitably evaluates the impacts on everyone who is subject to fire-related air pollution in India, not just those in a typical pollution hot-spot. In addition, while severe air pollution has been observed in cities such as Delhi, we acknowledge that crop residue burning is one of several factors in urban air pollution, rather than necessarily the dominant factor. For example, although agricultural emissions can contribute up to 50% of PM2.5 in Delhi during post-monsoon fire season, the dominant emission sources of year-round PM2.5 are vehicle, industrial and energy emissions. For population in suburbs or rural areas, crop residue burning is a greater year-round contributor in absolute or relative terms. By using an adjoint modeling approach, our study relates the eventual impacts across the country back to burning in each hour and individual district, and shows that crop residue burning can be controlled independently to achieve the greatest reduction of aggregate exposure at the lowest cost. The main contribution of this work is the quantification and disaggregation of air quality impacts across India due to agricultural emissions, and the identification of a potential new form of impact mitigation. However, our baseline estimates of attributable premature deaths in India also compare well with existing health impact assessment studies. GBD MAPS Working Group attributed 66,200 ambient PM2.5-related premature deaths to open burning of agricultural residue in 2015, which is similar to our calculations for 2015 . The total premature deaths attributable to ambient PM2.5 exposure in India range from 570,000 to 1,450,000 in previous studies covering years from 2010 and 2019, with which our estimate is consistent .Although we include an uncertainty analysis in the Supplementary Discussion , our study is subject to several other sources of uncertainty that we do not quantify. First, we focus on premature mortality risk changes resulting from changes in exposure to primary PM2.5 released from residue burning and do not quantify exposure to other species such as ozone. Second, the IER function used in this study assumes equal toxicity for all PM2.5 species and ignores differences in composition, which still requires further investigation3 . We also recognize that the IERs, as with other commonly used concentration response functions, were not developed specifically for India. However, without comprehensive epidemiological studies and available established models for India, the IER function is still a practical solution to represent our best at-present understanding of relative risks attributable to PM2.5 exposure .