Introduction

The escalating concerns over deteriorating air quality in Delhi during October to December, accentuate substantial health risks for residents in the National Capital Region (NCR). The declining air quality is primarily attributed to recurring post-harvest stubble-burning practices in the states of Punjab and Haryana1,2,3,4. Every year in India, 22,289 gigagrams (Gg) of paddy stubble are generated, with 62.42% burned in open fields, primarily in Punjab and Haryana, which together contribute 48% of the country’s total straw production5. This large-scale burning releases particulate matter and co-pollutant gases, severely degrading air quality in the NCR and posing serious health risks. The severe smog episodes in Punjab, Haryana, and the NCR during October 2016 and November 2017 are examples of such impacts, leading to adverse health effects on the exposed population5. These post-harvest stubble-burning practices are further exacerbated by substantial local anthropogenic emissions and meteorological conditions that favour less vertical mixing and substantial accumulation of pollutants close to the surface6,7. These conditions facilitate the gradual spread of smoke across the entire Indo-Gangetic Plain (IGP). The synergistic effect of these factors culminates in a marked increase in PM2.5 in NCR by about 20–60% on a different temporal scale8,9,10,11 thereby necessitating urgent and comprehensive mitigation strategies to safeguard air quality and public health.

Recent satellite-based studies have highlighted a notable shift in stubble-burning practices in the northwestern states of India8,11,12,13,14. Prior to the 1970s (pre-Green revolution), Punjab and Haryana predominantly cultivated less water-intensive crops. However, by the mid-1990s, the adoption of high-yield rice varieties and subsidised electricity for groundwater extraction led to a shift towards water-intensive rice paddy farming. This transition resulted in severe groundwater depletion, with rates falling to −6.23 cm/year by the early 200015. To address fast groundwater depletion, the 2009 Punjab/Haryana Preservation of Subsoil Water Act (Groundwater Conservation Policy (GWCP))16,17 was enacted, mandating that rice transplantation be postponed until mid-to-late June to align with monsoon rains. Mandatory enforcement of this act for the farmers left a narrow window between kharif paddy harvest and winter crop planting. Additionally, penetration of mechanised harvesters resulted in increased stubble on the ground. Consequently, farmers began burning stubble extensively in early November to expedite its disposal, effectively shifting post-monsoon stubble-burning activity from October to November11,14,18. This shift coincides with seasonal transitions in northern India, characterized by cooling temperatures, increased atmospheric stability, and slower prevailing northwesterly winds, which collectively facilitate the transport and spread of smoke towards the densely populated NCR. Despite ongoing conservation efforts, irrigated agriculture has increased by over 10% since 2009, continuing to deplete groundwater depletion (though at a slower rate of −3.35 cm/year) and exacerbating stubble burning, which has significantly deteriorated air quality across northern India10. Additionally, fluctuations in groundwater levels due to GWCP might influence water quality, particularly affecting its suitability for drinking and contributing to dental fluorosis in affected regions19.

The increase in PM2.5 levels in the Delhi NCR region may be linked to a shift in stubble burning patterns, potentially contributing to the frequent poor to severe air quality episodes observed during the post-monsoon season14. However, a crucial factor often overlooked in this hypothesis is the strong aerosol-radiative feedback prevalent in this region due to significant anthropogenic aerosol loading during winter20. This feedback mechanism can alter atmospheric stability, reduce boundary layer height, and suppress vertical mixing, thereby enhancing near-surface PM2.5 accumulation and exacerbating pollution events. In this study, we rigorously test both hypotheses by conducting extensive model simulations using the high-resolution WRF-Chem (Weather Research and Forecasting model coupled with Chemistry) for the period from 2017 to 2021. Our objective is to quantify the combined impact of the shift in fire activity linked to ground water act and aerosol-radiation interactions on the frequency of poor and severe air quality episodes in the Delhi NCR region. This study demonstrates that a shift in peak fire activity, coupled with strong aerosol-radiative feedback, results in a threefold increase in severe air quality episodes in Delhi NCR. Around 40% of this increase can be attributed to the impact of aerosol-radiative feedback during the peak burning season. The results emphasize the risks of implementing policies in isolation, showing that focusing on one issue can unintentionally lead to problems elsewhere. This underscores the importance of adopting a comprehensive approach to environmental management, ensuring that solutions are well-rounded and don’t inadvertently create new challenges.

Results and discussion

Observed fire shift

Figure 1a illustrates the daily averaged fire counts associated with stubble burning in Punjab, based on MODIS active fire count data from NASA’s Fire Information for Resource Management System (FIRMS). It compares fire counts from before 2010, prior to the enforcement of GWCP, with those from 2010 to 2021, after the enforcement of the policy. The results reveal a significant temporal shift, with peak fire activity delayed by approximately 15 days— moving from around October 25 in the pre-2010 period to roughly November 10 in the post-2010 period. This shift underscores the influence of the GWCP on agricultural burning practices in Punjab and Haryana. The spatial distribution of mean surface PM2.5 concentrations for November 5, 2021 (Fig. 1b) highlights the influence of extensive fires in Punjab and Haryana on air quality in adjacent regions. The PM2.5 concentrations reach up to 600 µg/m³ in the affected areas. Delhi, situated downwind, experiences elevated pollutant levels due to a combination of north-westerly winds and the topographical characteristics of the Indo-Gangetic Plain (IGP), which acts as a conduit for transporting emissions from stubble burning in Punjab and Haryana. This enables the pollutant plume from the north-western regions to markedly intensify the already high local pollutant concentrations.

Fig. 1: Fire count trend over Punjab and Haryana and PM2.5 distribution over North India.
figure 1

a Average fire count over Punjab and Haryana during pre and post decade of 2010 as obtained from FIRMS NRT active fire data. b Spatial distribution of PM2.5 (μg/m3) at the surface over north India on 5th November 2021. The white arrow denotes the direction and magnitude of prevailing wind. Black lines indicate the state borders of India.

Full size image

Nexus between shift in stubble burning activity and synoptic meteorology: Impacts on Delhi NCR Air Quality

The analysis, based on data from five years of model experiments, revealed a notable decline in air quality. This deterioration is linked to changes in stubble burning patterns due to the GWCP, particularly during the peak burning period from October 29 to November 16 (Fig. 2a). Comparisons with five years of quality controlled ground-based observations from Delhi’s air quality monitoring network indicate that the model simulations with control scenarios reasonably captured the day-to-day variability and magnitude of PM2.5 levels, confirming the model’s accuracy in simulating realistic air quality conditions over Delhi.

Fig. 2: Daily variation of PM2.5, meteorological parameters, and wind patterns over Delhi.
figure 2

a 5year daily mean variation of PM2.5 (μg/m3) averaged over Delhi for comparison to in situ measurement and Prepon scenario. b Daily mean variation of wind speed (m/s) and surface temperature (K). c Daily mean variation of ventilation index (m2/s) and PBLH (m). d Frequency of occurrence of wind direction for each station of Delhi, each the coloured line denotes N (0–45), NE (46–90), E(91–135), SE(136–180), S(181–225), SW (226–270), W(271–315) and NW(316–360) directions of wind. The concentric circles denote the magnitude of the occurrence.

Full size image

The observed 24 h mean PM2.5 concentrations during the biomass burning season are significantly elevated, with levels 4–7 times above the National Ambient Air Quality Standards of 60 µg/m³ set by the CPCB in 2009. Specifically, average PM2.5 concentrations reach ~250 µg/m³, with peaks ranging from ~150 to 400 µg/m³. Model simulations for the post-2010 fire scenario are consistent with these observations, indicating average PM2.5 levels of ~230 µg/m³ and a range of ~160 to 350 µg/m³. In contrast, pre-2010 simulations predicted lower average PM2.5 concentrations of ~170 µg/m³, with values ranging from ~130 to 210 µg/m³. This shift highlights a ~60 µg/m³ increase in daily average PM2.5 levels due to changes in burning practices following policy adjustments, with temporal variations ranging from ~30 and 140 µg/m³ during the biomass burning season. Thus, the introduction of the GWCP has resulted in a ~36% increase in daily mean PM2.5 concentrations over Delhi.

This shift in stubble burning activity in Punjab and Haryana and increase in daily mean PM2.5 concentrations in Delhi NCR coincides with a period of stable meteorological conditions with low ventilation capacity. Figure 2b–d illustrates the seasonal variation in key meteorological parameters, including temperature, wind speed, boundary layer height, and ventilation coefficient, that govern the dispersion of pollutants from their sources during the transition from the monsoon to the post-monsoon season. As the season changes, there is a weakening of the southeasterly monsoon winds, leading to a dominance of northerly and northwesterly winds over the Indo-Gangetic Plain (IGP). Wind direction data at 10 m from 38 stations in Delhi, corresponding to ground-based observations from air quality monitoring stations (Fig. 2d), shows that westerly and northwesterly winds were the most prevalent during October and November. Given Delhi’s position downwind of Punjab and Haryana, the combination of northwesterly winds and the valley-like geography of the IGP facilitates the transport of pollutants from stubble burning in these states to the Delhi NCR region (Fig. 1b). The transition from the monsoon to the post-monsoon season is also marked by a steady decline in the planetary boundary layer height (PBLH), temperature, and ventilation index over the region. The data in Fig. 2c show a twofold decrease in mean PBLH (from approximately 950 m to around 420 m) and a sharp drop in mean temperature (~13 K) from early October to late November over Delhi NCR. This steady weakening of wind speed, combined with the drop in PBLH, reduces the ventilation index by a factor of three, significantly limiting vertical dispersion and promoting a slow, steady horizontal flow toward the northeastern IGP. As a result, air quality in northern India worsens, as the influx of pollutants from the northwest adds to the already high pollution levels in the Delhi NCR region, further deteriorating air quality.

Table 1 presents the daily mean Air Quality Index (AQI) values at 4 PM over a five-year period, categorized into different AQI levels, using both observed and simulated PM2.5 data from pre-2010 and post-2010 fire scenarios. The AQI categories—Severe (401–500), Very Poor (301–400), and Poor (201–300)—are calculated based on the standards and breakpoints defined by the Central Pollution Control Board (CPCB) of India21. During the biomass burning season, the daily Air Quality Index (AQI) at 4 PM shifted markedly from a range of 300–350 to 350–460 due to the implementation of the GWCP, resulting in a ~90-unit increase in the mean AQI. This shift elevated the air quality classification from “Very Poor” to “Severe.” To assess the impact of the GWCP on AQI fluctuations and their implications for Delhi’s population, we analysed hourly AQI data with and without the GWCP (Table 1). Our findings indicate that the GWCP led to a threefold increase in the frequency of “Severe” AQI conditions, with the number of hours in this category increasing by 275%.

Table 1 Overall frequency of AQI category occurrence estimated over Delhi from observations, Cntl, Prepon, and Radoff simulations of PM2.5
Full size table

Nexus between shift in stubble burning and synoptic meteorology and atmospheric feedback

Aerosol-radiation feedback mechanisms exert a profound influence on meteorological conditions and air quality20,21,22. We hypothesize that the pronounced rise in PM2.5 levels during peak stubble burning periods intensifies this feedback loop, fostering more stable atmospheric conditions and a shallower planetary boundary layer (PBL), which in turn leads to higher PM2.5 concentrations in Delhi. To test this hypothesis, five additional simulations were conducted without aerosol-radiation feedback and compared them with those that incorporated this feedback (Fig. 3a; Table 1; Supplementary Table S1). The results demonstrate that aerosol-radiation feedback significantly enhances PM2.5 accumulation near the surface, particularly in areas heavily impacted by biomass burning emissions. This feedback mechanism increases PM2.5 concentrations by approximately 100 µg/m³ over Punjab, Haryana (Fig. 3b), and downstream regions, including Delhi NCR, highlighting its considerable impact on regional air quality.

Fig. 3: Impact of aerosol-radiation interactions on PM2.5 and PBL.
figure 3

a Difference in daily mean concentration of PM2.5 (μg/m3) and corresponding PBL (m) obtained from Cntl and Radoff simulations. b Distribution of difference (ctl-radoff) in daily mean PM2.5 (μg/m3) due to aerosol radiation interaction feedback effect.

Full size image

The aerosol-radiation interactions reduce the mean PBL height by approximately 20–230 m, resulting in an increase in mean PM2.5 concentrations of up to 76 µg/m³ in Delhi NCR. Supplementary Table S1 outlines the range of PBL height reductions and the corresponding hourly PM2.5 mass concentrations across various air quality index (AQI) categories, comparing scenarios with and without aerosol-radiation feedback. This reduction in PBL height further exacerbates PM2.5 accumulation (Fig. 3a), demonstrating the positive feedback loop between aerosol-radiation interactions and PM2.5 levels. Notably, this feedback mechanism is more pronounced on days with higher PM2.5 pollution in Delhi NCR, underscoring its critical role during the stubble burning season, when PM2.5 levels are at their peak. During this period of elevated pollution, the average daily reduction in PBL height due to aerosol-radiation feedback in Delhi NCR was approximately 152 m, with a maximum decrease of up to 225 m. Consequently, this decrease in PBL height led to an average daily increase in PM2.5 concentrations of around 36 µg/m³, reaching up to 76 µg/m³. Remarkably, aerosol-radiation feedback alone caused significant increases in PM2.5 concentrations, reducing PBL height by up to 660 m in the “very poor” AQI category and up to 200 m in the “severe” AQI category over Delhi NCR. This rise in PM2.5 levels pushed more hours into higher health risk categories that would otherwise have fallen into lower AQI categories. The effect was particularly significant in the “severe” AQI category, where positive aerosol-radiation feedback increased the frequency of severe AQI hours by 92%. The preceding discussion highlights the critical role of aerosol–radiation feedback in reducing boundary layer height and exacerbating air quality deterioration. Further investigation into this feedback mechanism during winter months (December and January) in Delhi is particularly important, as foggy and hazy episodes frequently occur under conditions of high humidity and elevated aerosol liquid water content (ALWC). A previous study has shown that increased light scattering due to higher ALWC under humid conditions can suppress boundary layer height by approximately 27 m (15%) during morning hours in Delhi, thereby contributing to the degradation of air quality23. Recent investigations have also highlighted substantial health risks associated with crop residue burning in Punjab and Haryana. For example, Chakrabarti et al.24 quantified the health impact of Acute Respiratory Infections (ARI) attributable to pollutants from crop residue burning, revealing a threefold increase in ARI risk for populations residing in districts with high stubble burning intensity. Chakrabarti et al.24 further estimated that the complete cessation of stubble burning could mitigate the health burden by saving 149,000 years of life lost over a five-year period. In the present study, changes in PM2.5 pollution levels in Delhi NCR due to shift in stubble burning pattern, influenced by the GWCP, are also analysed in terms of the resulting increase in short-term excess health burden caused by acute exposure to elevated PM2.5 concentrations. The short-term excess health burden of 2122 individuals is estimated during October-November 2021, providing insights into the associated health burden (Supplementary Table S2). Due to GWCP-induced shifts in fire patterns, the estimated additional short-term exposure burden in Delhi NCR is 205 individuals (range: 171–240), resulting in additional 9978 years of life lost (range: 8323–11,682) (Supplementary Table S2). This burden is associated with the combined effects of changes in stubble burning practices and aerosol-radiation feedback. In summary, our findings indicate that the worsening of PM2.5 pollution in Delhi NCR, driven by shifts in stubble burning activity linked to GWCP, is further exacerbated by the positive feedback from aerosol-radiation interactions. This feedback mechanism contributes to a higher frequency of air quality falling into the “severe” AQI category, with substantial implications for public health.

Given the link between stubble burning and poor air quality with associated health impacts, the Government of India (GOI) has undertaken several initiatives. To mitigate open stubble burning, the GOI has outlawed the practice, imposed bans, and implemented other interventions. Under Section 144 of the Civil Procedure Code (CPC), paddy burning is prohibited25. The GOI has also recommended measures such as utilizing paddy straw for biogas generation and mixing crop residue pellets (about 10%) with coal for power generation25,26. Similarly, the Ministry of Agriculture of India recently developed the National Policy for Management of Crop Residue27. Likewise, the Commission for Air Quality Management (CAQM), a statutory body constituted by the GOI, has developed a Graded Response Action Plan (GRAP) for the NCR28,29. This plan imposes restrictions on pollution sources at different stages when the predicted Air Quality Index (AQI), as forecasted by an air quality early warning system29, exceeds 200. GRAP is implemented in four stages based on AQI levels and includes predefined temporary controls on emission sources, ranging from raising public awareness to banning diesel-operated vehicles in the city28,29,30.

Discussion

This study highlights the critical interplay between changes in agricultural burning practices and aerosol-radiative feedback mechanisms in exacerbating air quality issues in the Delhi NCR region. Shifts in the timing of stubble burning, driven by groundwater conservation policies, significantly worsen PM2.5 pollution, particularly during the post-monsoon season when atmospheric conditions favour the accumulation of pollutants. The aerosol-radiative feedback further amplifies this effect, leading to a pronounced reduction in planetary boundary layer height and a substantial increase in near-surface PM2.5 concentrations. This feedback mechanism is especially pronounced during peak pollution days, contributing to more frequent and severe air quality episodes. The study underscores the importance of considering the combined impact of agricultural practices and atmospheric feedbacks on air quality. Isolated policy measures targeting only one aspect of the problem, such as GWCP, may inadvertently intensify pollution due to the complex interactions between aerosols and atmospheric conditions. These findings suggest that a more integrated approach to environmental management is necessary to mitigate the public health risks associated with severe air quality degradation. The insights gained from this research provide a compelling case for adopting comprehensive strategies to mitigate air pollution in heavily polluted regions like Delhi NCR that address both the sources of pollution and the atmospheric processes that exacerbate their effects.

Methods

Fire count data

Moderate Resolution Imaging Spectroradiometer (MODIS) active fire data from NASA’s Fire Information for Resource Management System (FIRMS) has been utilized to analyse fire activity over Punjab and Haryana from 2001 to 2021. We analysed inter-annual variation of active fire count over Punjab and Haryana in two epochs, (i) before 2010, prior to the enforcement of Groundwater Conservation Policy (GWCP), and (ii) after 2010, after to the enforcement of GWCP. It revealed clear evidence of a temporal shift of fire count maxima occurrence, the peak fire activity observed in the first epoch (2001–2005) was on 16 October, the second epoch (2006–2010) was on 27 October, the third epoch (2011–2015) was on 31 October, and the fourth epoch (2016–2020) was on 5th November.

Air quality observations

The near real-time, quality-controlled PM2.5 observations for Delhi are sourced from 38 air quality monitoring stations within the Delhi NCR region, managed by the Central Pollution Control Board (CPCB), Delhi Pollution Control Committee (DPCC), and the Indian Institute of Tropical Meteorology (IITM). Data exceeding 1500 µg/m³ and those affected by instrument malfunctions are filtered out to ensure high-quality observations before being used for chemical data assimilation and forecast verification.

Experimental design

We employed the WRF-Chem model (version 3.9.1), centred at Delhi, with a horizontal grid resolution of 10 km. The model has 50 vertical levels, extending to a pressure level of 50 hPa, with vertical grid spacing that varies from approximately 200 m in the boundary layer to about 1200 m near the model’s upper boundary. Meteorological initial and boundary conditions are derived from the ERA5 reanalysis data, provided by the European Centre for Medium-Range Weather Forecasting (ECMWF) at a spatial resolution of 0.25° × 0.25°, with hourly updates. Chemical boundary conditions are taken from the MOZART-4 model, using a 10-year climatology with six-hour intervals. Biogenic emissions are dynamically calculated using the MEGAN model31, while anthropogenic emissions are integrated from the EDGAR-HTAP v2 inventory (0.1° × 0.1° resolution)32. Fire emissions at a 1 km × 1 km resolution are incorporated using the NCAR FINNv1.5 inventory33 with plume rise parameterization.

The simulations utilize the MOZART chemical mechanism for gas-phase chemistry, including 85 gas-phase species, 39 photolysis reactions, and 157 gas-phase reactions. Aerosol interactions are modelled with the MOSAIC scheme, which uses four-size bins and is integrated with MOZART chemistry34,35. To evaluate the effects of fires on Delhi’s air pollution, tagged tracers for CO from fire emissions (COfire) are also included, following previous studies but adapted for MOSAIC aerosols20. These tracers undergo all standard physical and chemical processes in the model simulations but do not influence the overall model dynamics. The simulation period covers September 24 to November 30 for 2017 to 2021 across three experimental setups, totalling 15 simulations. The control setup uses current-day fire emission data. A second setup prepones fire emissions in Punjab and Haryana by 15 days, while a third examines the effect of disabling aerosol-radiation interactions in the control simulations. Each simulation’s initial seven days are considered a spin-up period, with analysis focused on September 30th to November 30th.

A recent report of a ~12% increase in overall Aerosol Optical Depth (AOD) in Delhi during the post-2010 post-monsoon period overlooked the specific impact of increased stubble burning and rising emission sources compared to the pre-2010 decade12. Here, in this study, the effect of the Groundwater Conservation Policy (GWCP) on air quality in Delhi was assessed by simulating daily mean PM2.5 concentrations under two scenarios: one reflecting current fire activity with a delay of approximately 15 days (Cntrl_PM2.5) and another anticipating fire activity 15 days earlier to mirror pre-2010 conditions (Prepon_PM2.5).

Estimation of health impacts

In the present study, increase in short-term excess health burden and Years of Life lost (YLL) have been estimated in Delhi due to shift in post-harvest paddy stubble burning activity over Punjab and Haryana. Through model experiments namely control simulations and simulations with preponed fire emissions from September 24 to November 30 for the years 2017 to 2021, the mean impact of shift in stubble burning activity on PM2.5 pollution in Delhi is estimated. The health impacts due to acute PM2.5 exposure are estimated corresponding to both experiments; model simulations with control setup and with preponed fire emissions. Subsequently, the difference in these health impacts are estimated, reflecting the change in health burden attributed to the shift in stubble burning activity.

The exponential integrated exposure-response (IER) function is used for estimating relative risk (Eq. (1)) following previous studies36,37,38.

$${RRd}=\exp \{\gamma \times ({Cd}-{Cr})\times 0.1\}$$
(1)

Where RRd is the daily relative risk, Cd is 6 days cumulative average PM2.5 (µg/m3), Cr is the reference counterfactual concentration taken as 2.4 µg/m3 (Burnett et al., 2018) and γ is the effect estimate. In the present study, the effect estimates for the winter period and for different age groups are taken from a Delhi based study39. They reported γ for winter (0.55% [0.45–0.66%]) season and for different age groups (Age 5–44 y: 0.52% [0.33–0.71%]; Age 45–64 y: 0.36% [0.21–0.51%]; Age ≥65 y: 0.70% [0.57–0.84%]) utilising long-term mortality data of Delhi. The short term-excess health burden is estimated using following formulation (Eq. (2))

$${Md}=P\times Id\times ({RRd}-1)/{RRd}$$
(2)

Where Md is the excess health burden (Persons/day), P is the population count of Delhi (persons) and Id is the daily health burden rate. The daily health burden rate is estimated by dividing annual health burden rate40 by 365.25 following previous studies41,42. For each age group, the losses in life expectancy attributed to acute PM2.5 exposure are estimated in terms of YLL (Eq. (3))

$${YLLi}={Mi}\times {LEi}$$
(3)

Where YLLi is years of life lost in age group “i”, Mi is excess health burden in age group “i” and LEi is life expectancy of age group “i”.