Such information is needed to support decision-making which can place public benefits such as health costs alongside the potential private costs to farmers. Here we aim to inform efforts to mitigate the adverse impacts of crop residue burning by quantifying how small-scale and targeted changes could affect the air pollution and health risks of the entire Indian population. We use the Global Fire Emissions Database v4.1s to provide an estimate of emissions from open burning of crop residue, combined with district-level crop production data for India, to obtain a comprehensive map of the time and location of crop production and residue burning. We then use a regional atmospheric chemistry and transport model to perform inverse simulations, computing the sensitivity of a given population’s exposure to PM2.5 with respect to emissions in any location at any time, and date. In combination with an Integrated Exposure Response function3 and an India-specific Value of Statistical Life, we use this data to estimate which burning events, in what locations, and at what times are responsible for the greatest increases in population exposure, premature mortality, and monetary societal cost during the period 2003–2019.We use an adjoint modeling approach as our primary tool for air quality impact attribution. We first perform three sets of adjoint runs from which we obtain sensitivities that quantify the effect of emissions on population exposure for the whole Indian population . Each set of runs represents one typical rainfall condition for a year , “drought” , or “normal” year, but all three use the same population distribution.
We combine this sensitivity data with fire emissions for each year from 2003 to 2019 to estimate the impact of agricultural residue burning on India-wide population exposure. Modifying the emissions dataset allows us to quantify the benefit of different targeted mitigation strategies,hydroponic bucket while using each of the three sensitivity datasets allows us to quantify the role of meteorological variability. For each year from 2003 to 2019, one of the three sensitivity datasets is chosen based on monsoon rainfall record in India . Two additional sets of adjoint simulations are performed for the “normal” meteorological year, in which the cost function J is modified to include only either urban or highly populated areas. Comparison of results using these datasets to those using the full India-wide population allows us to evaluate the distribution of impacts between urban and rural areas. For additional context, we perform 23 pairs of forward simulations with the conventional GEOS-Chem Classic model. Each pair simulates the post-monsoon burning season with and without India agricultural emissions, and the 23 pairs collectively cover the period from 1997 to 2019. This allows us to evaluate the impacts of Indian residue burning on surrounding countries, and to gain a more complete understanding of the interaction between population growth, meteorological variability, and historical changes in fire emissions. See Methods for a detailed description.We define air quality impacts due to crop residue burning in terms of the premature deaths attributable to PM2.5 exposure and the associated monetized cost. Figure 2 shows how crop residue burning on each day contributed to the national mean PM2.5 exposure. From 2003 to 2019 we estimate that the annual mean population-weighted PM2.5 exposure due to burning activities in India averaged 6.7 μg m−3 .
Premonsoon and post-monsoon residue burning contribute 28% and 64% of this total, respectively. Geographically, more than 90% of the Indiawide fire-related exposure increase is due to agricultural fire emissions from the northwest states, with 64% from Punjab, 11% from Haryana, and 5.7% from Uttar Pradesh. It is consistent throughout 17 years that the pre-monsoon and post-monsoon fire seasons are responsible for 90% of PM2.5 exposure from residue burning, and that Punjab, Haryana, and Uttar Pradesh together contribute over two thirds of the nation-wide exposure burden . Demographically, urban and densely populated areas, defined by regions with a population density above 400 and 1,000 people per km2 , respectively, are exposed to 2.0 μg m−3 and 4.8 μg m−3 greater annual PM2.5 concentrations than the national average due to residue burning . This is supported by observations of elevated PM2.5 levels during burning seasons in large cities downwind of agricultural fires. By applying the IER and India-specific population distribution, we estimate 69,000 total premature mortalities on average across India resulting from ambient PM2.5 exposure due to crop residue burning . Our estimate for 2015 is consistent with Global Burden of Disease 2018 India Special Report’s estimate of premature mortality attributable to agricultural fires in India for the same year. Our estimate for the same period is 4.7–14% of premature mortality attributable to total ambient PM2.5 presented by earlier studies covering 2010-2019. This fraction is consistent with findings from GBD 2018, where 6.1% of premature deaths in India due to ambient PM2.5 are attributed to agricultural fires. The central estimates of early deaths as well as confidence intervals depend on the choice of relative risk function. While the IER the has been applied worldwide, including India, to quantify attributable deaths due to biomass burning episodes, we compare the IER-based results to those using other methods in the Methods and Supplemental Information.
Using a VSL adjusted for India, we estimate the annual, monetized cost of premature mortality due to crop residue burning as 23 billion USD. This is equivalent to 38% of the total health expenditure, or 7.8% of the gross value added from agricultural activity on average . These two ratios have increased from 29% to 40%, and from 6.1% to 9.2%, respectively between 2003 and 2019 . We find that for any year in this period, the three largest contributors to these impacts are consistently Punjab , Haryana , and Uttar Pradesh .Figure 3 shows the premature deaths per unit of emissions from burning against the emissions per unit of crop production for each district in Punjab and Haryana. The contribution to premature mortalities per crop produced for these two states is larger than for the rest of India combined . Variations between these districts are due not only to different crop production quantities but also different meteorology, population distributions, and agricultural practices. For the 43 districts in the two states, we use the two dimensions shown in Fig. 3 to define four categories . Punjab and Haryana, under the rice-wheat rotation system,stackable planters are collectively responsible for 60% of all rice and 30% of all wheat entering the central grain pool in India. This results in a large amount of residue being burned per unit of crop production, as these two crops produce relatively high quantities of residue per unit of product compared to crops such as oilseed and sugarcane . The contribution of crop residue burning in each district to total premature mortality in India is a product of three factors: first, premature deaths per unit of agricultural burning emissions ; second, agricultural burning emissions per unit of crop production ; and third, the total crop production . Districts in C1 rank high in the first two factors, partly because they grow coarse varieties of rice that generate more residue to be burned for the same amount of crop production32 and are mostly upwind of densely populated regions. As a result, they are on average responsible for 40% of the total air quality impacts in India due to burning, with Patiala and Sangrur alone contributing 20%. Compared to C1, districts in C2 and C3 have lower emissions per unit of crop production and fewer premature deaths per unit of emissions , respectively. The contribution to the total air quality impact of residue burning is therefore 11% and 14% for districts in C2 and C3, respectively. The remaining districts are minor contributors to the total impacts, among which Kaithal has the largest contribution due to its cultivation of coarse rice , lower sensitivity and lower overall crop production . All Haryana districts fall in C2 and C4, whereas districts in C1 and C3 are all from Punjab, distinguished by the first factor . Although both states are large rice producers, Haryana mainly grows basmati rice, the residue of which is used as animal feed. Punjab instead mainly grows non-basmati rice, the residue of which is not fed to livestock because of its high silica content and which is thus more often burned in-field.Figure 4 shows the percentage change in annual air quality impacts due to burning achieved by a 1% reduction in emissions from burning in each state or district during two burning seasons averaged over 2003–2019.
Specifically, during post-monsoon season, the efficacy of reducing burning is greatest in Punjab and Haryana, with a 0.57% and 0.065% reduction respectively in total India-wide impacts from crop residue burning per1% reduction in burning emissions. This means 380 and 45 averted premature deaths, valued at 130 and 15 million USD, respectively, averaged across 17 years. Considering only burning emissions from Punjab, two thirds of the achievable benefit comes from reductions in burning in Sangrur, Patiala, and Ludhiana, where a 1% reduction in emissions from post-monsoon residue burning would result in a 0.18%, 0.05%, and 0.068% reduction, respectively, in India-wide burning-related air quality impacts. One reason that farmers choose to burn agricultural residue rather than adopting mechanized alternatives is that there is a limited window of time, around two to three weeks, to manage crop residue between harvesting and sowing. A water table preservation policy instituted in Punjab and Haryana in 2009 to align rice irrigation with the summer monsoon further shortened this window, which may have further intensified burning activities as farmers seek to ensure timely sowing for the next planting season. This suggests that promoting burning earlier or later in the season may be unattractive or unachievable. It may instead be possible to burn earlier or later within a given day without incurring the same difficulties for farmers, but no study to date has quantified the potential benefits of such a change. We therefore quantify the air quality impacts of shifting burning within the day by applying different local time diurnal cycles of emissions . Figure 5 shows the changes in attributable air quality impacts resulting from shifts ranging from six hours earlier to one hour later. Averaged over the period 2003–2019, burning earlier by one to four hours could reduce the total air quality impacts of crop residue burning by 0.5–19%, while burning too early or burning later could instead increase the impacts by 2−30%. For any individual year in that period, burning two to three hours earlier in November reduces the total, annual, residue burning-related contribution to India-wide air quality impacts by 15–23% . If the target region is restricted to Punjab only, burning earlier by two hours in November yields an average 14% reduction in air quality impacts resulting from that region’s residue burning . This means 9600 averted early deaths annually, valued at 3.2 billion USD. This is greater than the sum from all other states if the same timing shift were applied. Although these improvements could be subject to the specific diurnal cycles we use, our assumptions of diurnal fire activity agree with previous findings that fires are typically set during early to mid-day and burn out by the evening , lasting 13–15 hours, and that fire activity generally peaks in the afternoon. Liu et al. 2020 collected survey data from households in India, finding that despite regional variations, 97% of burning activities happen between 10:00 and 23:00 LT, nearly 30% of which happen in late evening. While we do not have district-level information about hourly diurnal cycles of fire activities due to a lack of comprehensive in-field studies, our broad conclusion is that there may be significant benefits yielded by encouraging fires to be set earlier in the day rather than later . One contributing factor to these changes could be natural, diurnal changes in the depth of the planetary boundary layer .It is a key meteorological parameter in pollutant dispersion because a higher PBLH favors dispersion and reduces aerosol accumulation . However, the PBLH is also directly affected by aerosol loading meaning that the concentration of pollution can itself affect dispersion. The PBLH decreases with increasing aerosol concentration, which enhances atmospheric stability and in turn favors even higher pollutant concentrations––a positive feedback loop. However, the diurnal variations of PBLH do not change significantly on heavily polluted days compared to clean days.