Unveiling the effects of post-monsoon agricultural biomass burning on aerosols, clouds, and radiation in Northwest India

The post-monsoon agricultural biomass burning activities in Northwest India have been recognized as a significant socio-environmental problem in recent years, primarily due to their severe impacts on air quality degradation across a wide area, including the capital New Delhi. Although these biomass burning activities have been extensively studied from an air quality perspective, their potential impacts on the climate system, particularly through their influences on cloud and radiation fields, have been largely overlooked. In this study, we aim to address this research gap by analyzing fire, meteorological parameters, aerosol, cloud, and radiation data spanning nearly two decades (2002–2021), obtained from the Moderate Resolution Imaging Spectroradiometer (MODIS) aboard the Aqua satellite, Modern-Era Retrospective Analysis Research and Applications, Version 2 (MERRA-2), and the Fifth Generation of European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis (ERA5). Our analysis reveals a notable increase in agricultural biomass burning intensity in Northwest India over the past two decades, contributing significantly to air quality degradation. Our analysis further indicates a delay in peak burning time (day of the year) and a shortening of the period of intense burning, reflecting changes in farming practices and agricultural biomass burning in Northwest India over the past two decades. These agricultural biomass burning activities substantially elevate total and light-absorbing aerosols, thereby affecting cloud properties and altering the radiation budget. The intensification of these burning activities can cause an increase in cloud droplet size and a decrease in cloud optical thickness, suggesting an enhancement of the cloud droplet collision-coalescence process during the period of intense burning. Similarly, the intensification of burning activities leads to increased cooling effects at the surface and top-of-the-atmosphere across shortwave and longwave spectral ranges, while inducing a heating effect within the atmosphere. These findings highlight the potential impacts of agricultural biomass burning activities on the regional climate system and hydrological cycle, emphasizing the need for more detailed studies in the future.

Burning agricultural biomass left after harvest in the field is a prevalent practice in India and other developing countries to quickly clear and prepare fields for the next planting season. India, with a population of nearly 1.4 billion and the second-largest agro-based economy in the world, produces an average of 500 million tons (Mt) of crop residue annually. Of this, 92 Mt (approximately 18%) is burned each year, a significantly larger portion than the total production of agricultural waste in neighboring countries (Bhuvaneshwari et al. 2019). Notably, about two-thirds of the total food grains are produced in Northwest (NW) India, including Punjab and Haryana states, as a result of the multi-decade implementation of the Green Revolution movement that began in the mid-1960s (Sarkar and Das 2014; Jethva et al. 2019; Kaur et al. 2022). The substantial grain production in NW India leads to proportionate amounts of residue generation, which, under the traditional open-field burning practice, results in extensive burning activities in the region, multiple times higher than in other parts of India (Jain et al. 2014). With increasing agricultural activities and food production, the use of mechanized harvesting systems has also become popular (Singh and Kaskaoutis 2014). This system typically leaves several inches long root-bunded stacks and crop residue on the ground, which are subsequently burned in the open fields by farmers. NW India experiences two harvesting seasons: April to May (pre-monsoon) and October to November (post-monsoon), resulting in corresponding agricultural biomass burning activities. The burning activities are significantly more intense in the post-monsoon season than in the pre-monsoon season (Jethva et al. 2018; Sawlani et al. 2019). During the post-monsoon season, the boundary layer height in NW India is generally shallower due to lower air temperatures, leading to the trapping and accumulation of high concentrations of air pollutants near the surface. These pollutants, transported by north-westerly winds, impact not only the local source region but also downwind regions, extending throughout the Indo-Gangetic Plain (IGP) region(Kaskaoutis et al. 2014; Singh and Kaskaoutis 2014; Jethva et al. 2018; Singh et al. 2023) causing severe air quality degradation and consequent health impacts. For example, New Delhi, the capital city located in a downwind region, frequently ranks among the most polluted cities in the world during the post-monsoon to winter season, with post-monsoon agricultural biomass burning activities in NW India identified as a major contributing factor (Jethva et al. 2019; Singh et al. 2023). The IGP region reportedly has the highest premature mortality compared to other parts of India due to very poor air quality (Ghude et al. 2016), particularly during the winter season, partially attributed to such burning activities in NW India.

As a result, post-monsoon agricultural biomass burning activities have emerged as a significant socio-environmental problem in India, attracting attention from the media, local and central government authorities, policymakers, as well as agricultural and atmospheric scientific communities. Consequently, several studies have focused on investigating the physical, chemical, and optical properties of aerosols, as well as their long-term trends, in relation to biomass burning activities in NW India (Kaskaoutis et al. 2014; Rastogi et al. 2014; Singh and Kaskaoutis 2014; Sarkar et al. 2018; Jethva et al. 2019; Mhawish et al. 2022). However, it is essential to note that, beyond degrading air quality and affecting human health, aerosols can also influence the atmospheric heat budget, cloud properties, and hydrological cycle through aerosol–radiation and aerosol–cloud interactions (Twomey. 1977; Khatri et al. 2009, 2023; Bi et al. 2014; Fan et al. 2016). Despite extensive research on the connections between post-monsoon agricultural biomass burning activities and air quality degradation in NW India and surrounding regions, there is a lack of studies examining how these agricultural biomass burning activities can impact cloud properties and radiation budget perturbations via the emission of strong light-absorbing aerosols. Recognizing this gap, our efforts presented in this paper represent a step forward in developing an understanding of the link between agricultural biomass burning activities in NW India and their impacts on aerosol, cloud, and radiation fields, by analyzing data of nearly two decades obtained from multiple sources.

Focusing on the post-monsoon burning activities in NW India, this study analyzed data from September to November for the years 2002–2021. We selected a region of high agricultural burning activities falling within a boundary of longitude: 74°E–77°E and latitude: 29°N–32°N (Fig. 1), which covers most parts of Punjab and Haryana, and small parts of Rajasthan and Pakistan on the other side of the international border. The data used in this study are described below.

2.1 Fire data

This study used fire count data obtained from observations made by the Moderate Resolution Imaging Spectroradiometer (MODIS) aboard the Aqua satellite. The active fire in MODIS observation is detected by comparing the signal in the thermal infrared bands to the background conditions, such as surface temperature and sunlight reflection, after rejecting typical false alarm sources like sun glint and unmasked coastlines. Detailed information about the detection process is available in the MODIS Fire User Guide (https://modis-fire.umd.edu/files/MODIS_C6_Fire_User_Guide_A.pdf). The spatial resolution of fire data used in this study is 1-km ×ばつ 1-km. Additionally, only data with a confidence level value greater than 30% were considered in this study. A confidence level greater than 30% corresponds to the confidence classes of ‘nominal’ and ‘high,’ ensuring fewer false detections of fire pixels.

2.2 Vegetation index data

Normalized Difference Vegetation Index (NDVI) data derived from MODIS observations aboard the Aqua satellite were used. NDVI can be derived from atmospherically corrected surface reflectance measurements at 0.67 μm (Red) and 0.86 μm (near-infrared, NIR) wavelengths using the formula:

NDVI values range from −1 to 1, with higher positive values indicating greater vegetation density, and lower values indicating less vegetation. NDVI data from MODIS observations are provided at 16-day intervals and at various spatial resolutions. Detailed information about MODIS NDVI data is available in the document: ‘MODIS Vegetation Index (VI) Product Collection 6 Users Guide,’ accessible at https://vip.arizona.edu/documents/MODIS/MODIS_VI_UsersGuide_June_2015_C6.pdf. In this study, Level 3 NDVI data of a monthly temporal variation and a spatial resolution of 0.05° ×ばつ 0.05° were used.

2.3 Aerosol data

Aerosol data—aerosol optical thickness (AOT) and aerosol single scattering albedo (SSA) at a wavelength of 0.55 μm—obtained from satellite observations and reanalysis were used. Level 3 AOT data from MODIS observations, which have a spatial resolution of 1° ×ばつ 1° and are derived from Level 2 MODIS aerosol products ((Hsu et al. 2013; Remer et al. 2020), were used. On the other hand, SSA data from the Modern-Era Retrospective Analysis Research and Applications, Version 2 (MERRA-2) were used. MERRA-2 aerosol products are assimilated from the Goddard Earth Observing System-5 (GEOS-5) atmospheric general circulation model, coupled with the Goddard Chemistry Aerosol Radiation and Transport (GOCART) model (Chin et al. 2002), by incorporating aerosol observations from various platforms, such as MODIS, Multiangle Imaging SpectroRadiometer (MISR), Aerosol Robotic Network (AERONET), and Advanced Very-High-Resolution Radiometer (AVHRR) (Khatri and Hayasaka 2021). MERRA-2 data have a temporal resolution of 1 h and a spatial resolution of 0.5° (latitude) ×ばつ 0.625° (longitude). This study used hourly data closer to the Aqua overpass time and corresponding to 1° ×ばつ 1° grid of the MODIS Level 3 product. Further, absorption aerosol optical thickness (AAOT) was calculated by multiplying AOT with (1-SSA). It is important to note that MERRA-2 aerosol products rely on data from satellite observations and AERONET ground-based measurements, both of which face constraints such as coverage gaps due to cloud cover, bright surface reflectance, and the sparse spatial distribution of AERONET stations. Additionally, MERRA-2 aerosol products are subject to uncertainties in emission inventories. They further employ a relatively coarse spatial resolution along with assumptions in aerosol composition, size, and vertical distribution. Such factors can cause difficulties in capturing fine-scale, local, or rapidly fluctuating aerosol concentrations (Buchard et al. 2017; Randles et al. 2017). However, validation studies indicate that MERRA-2 aerosol products align reasonably well with satellite, aircraft, and ground-based observations, except during episodic events (Buchard et al. 2017). This consistency highlights the practical utility of MERRA-2 aerosol data in aerosol climatology studies, especially in regions where high-quality surface observations are limited.

Along with aforementioned satellite and reanalysis data, we further used ground-truth aerosol data. Aerosol observations from surface, such as those from AERONET (Holben et al. 1998) and SKYNET (Nakajima et al. 2020), over our study area (longitude: 74°E–77°E, latitude: 29°N–32°N) are sparse, and only a single AERONET site located in Lahore (31.47°N, 74.26°E) falls within the boundaries of our study area. Hence, AERONET observation data from the Lahore site offered an opportunity to validate satellite and reanalysis data used in this study. For this purpose, we used Level 2.0 (Version 3.0) quality-controlled AOT at 0.50 μm obtained from direct sun observations (Giles et al. 2019) and SSA at 0.44 μm obtained from almucantar measurements (Sinyuk et al. 2020) for September to November from 2007 to 2021, as AERONET data for Lahore site are available since the end of December 2006.

2.4 Meteorological data

Air temperature and specific humidity data of atmospheric layers (pressure level: 1000 hPa to 0.01 hPa, vertical levels: 137), air temperature and dew point temperature at 2 m above ground level (AGL), u- and v- components of wind at 10 m AGL, and total precipitation were obtained from the fifth-generation European Centre for Medium-Range Weather Forecasts (ECMWF) reanalysis, known as ERA5 (Hersbach et al. 2020). These data have a spatial resolution of 0.25° ×ばつ 0.25° and a temporal resolution of 1 h. The MetPy package (May et al. 2022) was employed to calculate wind speed and wind direction from u- and v- wind components; precipitable water content (PWC) from vertical profiles of air temperature, atmospheric pressure, and specific humidity; and relative humidity (RH) at 2 m AGL from air temperature and dewpoint temperature at 2 m AGL. Similar to the MERRA-2 data mentioned earlier, hourly data of ERA5 closer to the Aqua overpass time and corresponding to 1° ×ばつ 1° grid of the MODIS Level 3 product were used in this study.

As AERONET also provides PWC data estimated from direct sun observations (Giles et al. 2019), we further used Level 2.0 (Version 3) quality-controlled PWC data of Lahore site to validate PWC values of ERA5.

2.5 Cloud data

MODIS (Aqua) Level 3 water cloud products—Cloud Optical Thickness (COT), Cloud Particle Effective Radius (CER), and Cloud Liquid Water Path (LWP)—were used. These products have a spatial resolution of 1° ×ばつ 1° and are generated from MODIS Level 2 water cloud products of 1-km ×ばつ 1-km spatial resolution (Platnick et al. 2017). For CER, data corresponding to wavelengths of 1.6 μm, 2.1 μm, and 3.7 μm were used. It is important to note that cloud penetration is greater for shorter wavelength and less for longer wavelength. Consequently, CER (3.7 μm) predominantly provides information about upper cloud layers, whereas CER (2.1 μm) and CER (1.6 μm) provide insights into cloud layers deeper than those corresponding to 3.7 μm. COT, CER (2.1 μm), and LWP data corresponding to single-layered clouds were used. However, CER values for 1.6 μm and 3.7 μm wavelengths are not specifically provided for single-layer clouds. As a result, the CER values for these wavelengths (1.6 and 3.7 μm) used in this study may not exclusively represent single-layer clouds.

2.6 Radiation data

The radiation data used in this study are from ‘CER_SYN1deg-Day_Terra-Aqua-MODIS_Edition4A’ product (https://asdc.larc.nasa.gov/project/CERES/CER_SYN1deg-Day_Terra-Aqua-MODIS_Edition4A). This product is generated by combining top-of-atmosphere (TOA) radiative fluxes observed by Clouds and the Earth’s Radiant Energy System (CERES), cloud and aerosol properties observed by MODIS, and cloud properties and broadband fluxes observed by geostationary satellites (Loeb et al. 2013; Doelling et al. 2016). In this study, daily mean values of upwelling and downwelling shortwave (0.3 –5.0 μm) and longwave (5 –200 μm) fluxes at the surface and TOA for 1° ×ばつ 1° spatial resolution for aerosol-laden and aerosol-free atmospheres were used.

Before delving into the primary focus of this study, which is to examine the impacts of agricultural biomass burning activities on aerosol and cloud properties and radiation budget, it is essential to provide a brief insight of the temporal variation characteristics of agricultural biomass burning practices in NW India. Previous studies (e.g., Jethva et al. 2019) have demonstrated that fire count (FC) values in NW India exhibit characteristics that can be effectively described using a Gaussian distribution. To facilitate clarity and discern different levels of burning activity, MODIS observed FC data were smoothed using a 7-day running mean window and then fitted into a Gaussian distribution. Figure 2 illustrates observed FC (FC_raw), smoothed FC (FC_smoothed), and the Gaussian fitting of FC_smoothed for each year’s data. Additionally, the figure presents the median radius (μ) and the standard deviation (σ) of the distribution. In this context, μ corresponds to the day number of the year (DNY) at which the peak value of FC_smoothed occurs in the Gaussian distribution, while σ represents a time window (day number) containing approximately 68% of FC_smoothed. Figure 3 displays the yearly variations of Gaussian distribution parameters, total FC (FC_total), and the NDVI of the pre-harvest month (September). As illustrated in Figs. 2 and 3, an increasing trend of μ suggests a gradual shift in agricultural biomass burning activities toward the end of the year. This shift in agricultural biomass burning activities is largely attributed to agricultural intensification and groundwater conservation policies. It is because the agricultural intensification initiated during the Green Revolution, with the introduction of the high-yield rice–wheat cropping system, has transformed this region into a major agricultural hub (Swaminathan 2006). Over recent decades, crop production has continued to increase (Jethva et al. 2019), driven by the expansion of irrigation infrastructure, the intensive use of chemical fertilizers, and the widespread application of pesticides. Farmers have increasingly adopted high-yielding rice varieties with extended growing seasons, which push the rice harvest later and compresses the timeframe before wheat planting (Joshi 2018). Additionally, early rice planting, before mid-June, often leads to excessive irrigation, resulting in significant groundwater depletion during the hot summer months when evaporation rates are high. To conserve groundwater, the Punjab Preservation of Subsoil Water Act (2009) prohibits rice planting before mid-June, thereby aligning the growing period more closely with the monsoon season to reduce evaporation losses during peak summer months (Tripathi et al. 2016). This act also delays rice harvesting, narrowing the time available before wheat planting. This tight schedule compels farmers to burn rice stubble to clear fields quickly. Broadly speaking, FC_total exhibits an increasing trend over the past two decades, indicating a heightened intensification of agricultural burning activities in recent years. Conversely, σ demonstrates a decreasing trend implying that, despite the increasing trend in overall burning activities, farmers are taking relatively shorter time periods to clear their fields in recent years compared to the past. Further, the temporal trend of NDVI closely aligns with that of FC_total, indicating its potential utility as a proxy for both total crop production and post-harvest agricultural biomass burning activities (Jethva et al. 2019). This finding underscores the significance of leveraging remote sensing data, such as NDVI, for predicting agricultural burning activities and developing air pollution mitigation strategies preemptively.

For statistical analysis of long-term data, we classified agricultural biomass burning periods of each year into three groups: strong burning period (DNYs: μ-σ to μ + σ), moderate burning period (DNYs: μ−2σ to μ-σ−1 and μ + σ + 1 to μ + 2σ), and weak burning period (DNYs: μ−3σ to μ−2σ−1 and μ + 2σ + 1 to μ + 3σ). These classifications were based on μ and σ values from the Gaussian distribution of each year, as illustrated in Fig. 4. Accounting for possible differences in meteorological conditions during different burning periods, data analyses were performed for each type of burning period on either side of μ. For clarity, the negative and positive values of σ denote the burning periods occurring before and after μ, respectively.

Further, to investigate the effects of aerosols on the radiation field, aerosol radiative forcing (ARF) values at the surface and TOA and the forcing within the atmosphere were calculated as:

where subscripts sfc, toa, al, af, and wv denote surface, TOA, aerosol-laden, aerosol-free, and spectral range (e.g., shortwave or longwave), respectively.

5.1 Meteorological conditions

Figure 5 depicts box plots showing the mean, median, 5th, 25th, 75th, and 95th percentile values for different meteorological parameters for different burning periods: strong, moderate, and weak. In Fig. 5(a), the near-surface air temperature decreases as sigma increases, indicating temperatures during weak, moderate, and strong periods before μ being higher than those after μ. This suggests the decrease in inversion heights during the later burning periods after μ compared to those before μ. In Fig. 5(b), RH near the surface is relatively high during the early weak burning period before μ, but it shows a slight increasing trend in subsequent burning periods. As the capacity of air to hold water vapor decreases with the decrease in temperature, RH can increase with the decrease in temperature. Additionally, temperature inversions associated with colder temperatures can trap moisture in the lower atmosphere, further increasing RH near the surface. In Fig. 5(c), the precipitation is generally low across all burning periods, particularly after μ, primarily due to the reduced water-holding capacity of colder air, which limits the availability of water vapor for cloud formation and precipitation. Figure 5(d) and Fig. 5(e) reveals that wind speed does not vary significantly across the different burning periods, with prevailing winds generally westerly. This suggests that pollutants from burning activities in NW India can be easily transported toward Delhi and the IGP region. Moreover, consistent with the decrease in air temperature, as shown in Fig. 5(a), PWC shows a decreasing trend from the early weak burning phase to the dissipating weak burning phase in Fig. 5(f). This trend is primarily attributed to the reduced capacity of cold air to retain water vapor as temperature decreases.

5.2 Aerosol properties

To assess the quality of satellite and reanalysis data used in this study, daily mean values of quality-controlled Level 2.0 (Version 3) AERONET data for AOT at 0.50 μm, SSA at 0.44 μm, and PWC from the Lahore site were compared with corresponding data from MODIS (AOT at 0.55 μm), MERRA-2 (SSA at 0.55 μm), and ERA5 (PWC) for a 1° ×ばつ 1° spatial resolution that included the Lahore site. The comparisons showed the correlation coefficients of 0.87 for AOT, 0.51 for SSA, and 0.97 for PWC, alongside absolute mean bias error (MBE) values of 0.05, 0.00, and 0.6 cm, respectively. These results suggest that the satellite and reanalysis datasets used in this study are reasonably consistent with ground-based observations. The relatively weaker correlation for SSA may be attributed to differences in spatial and temporal resolutions, wavelengths, model uncertainties in reanalysis data, and challenges associated with SSA retrieval via surface-based remote sensing (Khatri et al. 2016). Further, Fig. 6 presents monthly box plots comparing AERONET data with MODIS, MERRA-2, and ERA5 data for AOT, SSA, and PWC, respectively. While a few quantitative differences between AERONET and satellite/reanalysis datasets in Fig. 6 can be attributed to spatial and temporal resolution discrepancies and other factors mentioned above, the figure demonstrates that these datasets successfully capture temporal trends consistent with AERONET observations. This further underscores the reliability of the satellite and reanalysis datasets employed in this study.

Figure 7 illustrates variations in FC, AOT, AAOT, and SSA during different burning periods: strong, moderate, and weak. The median, percentile, and whisker values shown in the box plots in Fig. 7 differ significantly among the different groups at the 95% confidence level (p value < 0.05), as indicated by the Kruskal–Wallis test. During periods of weak burning activity (sigma values between −3 to −2 and 2 to 3), the mean FC was approximately 0; however, the mean AOT (AAOT) remained higher than 0.5 (0.05). This suggests the presence of high aerosol loading, with a considerable amount of light-absorbing aerosols in the study region, even in the absence of substantial agricultural biomass burning activities. Notably, this mean AOT value is comparable to mean AOT values observed in highly polluted urban and industrial cities of IGP, such as Kanpur and Ballia (Khatri et al. 2021). With increasing burning activities, FC increases, leading to increase both AOT and AAOT. During periods of strong burning activity, mean FC reached approximately 600, resulting in mean AOT (AAOT) up to approximately 1.0 (0.1). The 1-sigma value ranged from approximately 7 to 14 days in different years (see Figs. 2 or 3). Therefore, Fig. 7 suggests that agricultural biomass burning activities can increase AOT (or AAOT) significantly within a short time period of less than 2 weeks. Moreover, the 95th percentile values of AOT and AAOT reaching up to approximately 2.5 and 0.25, respectively, during strong (intense) burning periods highlight the very strong impacts of agricultural biomass burning activities on air quality degradation in NW India. Aerosols emitted from such intense burning activities can transport over long distances, as mentioned above. An intensive field campaign conducted during October and November of 2022, involving Punjab, Haryana, and Delhi; by utilizing 29 Compact and Useful PM_2.5 instruments with gas sensors (CUPI-Gs) (Singh et al. 2023) revealed an increase in PM_2.5 from less than 60 μg m⁻³ to an extremely high value of 500 μg m⁻³ due to intense agricultural biomass burning activities. This suggests that extremely high AOT levels are reasonable during periods of intense agricultural biomass burning activities. Despite remarkable differences in PWC values for the same type of burning period falling before and after μ, as shown in Fig. 5(f), Fig. 7(b) and 7(c) show the decrease in both AOT and AAOT values for moderate burning period followed by weak burning period for both negative and positive values of sigma. This suggests the secondary role of meteorological factors in AOT and AAOT modifications and thereby emphasizing the dominant effect of agricultural biomass activities on air quality degradation. Unlike AOT and AAOT, SSA variations corresponding to different burning periods shown in Fig. 7(d) are consistent with PWC variations of different burning periods shown in Fig. 5(f). Since SSA is a metric describing the relative abundance of light-absorbing aerosols, increased AOT (or AAOT) does not necessarily decrease SSA unless the scattering aerosol optical thickness (SAOT) remains the same. Although increased PWC can increase aerosol size to enhance AOT (Khatri and Ishizaka 2007), Fig. 7 suggests that AOTs were largely driven by emissions from agricultural biomass burning activities rather than the change of PWC. It is also important to note that an increase in PWC can increase SSA by enlarging aerosol size and reducing refractive index, thereby increasing SAOT more promptly than AAOT (Khatri and Ishizaka 2007). Tanada et al. (2023) also demonstrated the strong influence of atmospheric water vapor on SSA values over biomass burning regions of Brazil, Angola, Australia, California, Siberia, and Southeast Asia. Due to such strong effect of PWC on SSA, both PWC and SSA exhibited similar variations in Fig. 5(f) and Fig. 7(d).

5.3 Cloud properties

Figure 8 illustrates box plots for the values of COT, CER (2.1 μm), and the difference in CER between 3.7 μm and 1.6 μm for different periods of varying biomass burning intensities. The median, percentile, and whisker values shown in the box plots in Fig. 8 differ significantly among the different groups at the 95% confidence level (p value < 0.05), as indicated by the Kruskal–Wallis test. This figure reveals a strong influence of agricultural biomass burning activities on the modification of cloud microphysical properties. As illustrated in Fig. 8, CER (2.1 μm) increased, but COT decreased during strong burning periods (sigma values between −1 and 1), which corresponded to higher values of FC, AOT, and AAOT (see Fig. 7). Conversely, CER (2.1 μm) decreased, while COT increased for weak biomass burning periods (sigma values between −3 to −2 and 2 to 3), which corresponded to relatively lower values of FC, AOT, and AAOT (see Fig. 7). High concentrations of light-absorbing aerosols, such as those emitted from biomass burning, can have complex impacts on cloud properties through both the aerosol indirect effect (Twomey. 1977; Albrecht 1989; Nakajima et al. 2001) and the aerosol semi-direct effect (Koch and Del Genio 2010; Jacobson 2012). In highly polluted regions, increased aerosols can provide abundant cloud condensation nuclei (CCN), resulting in a larger number of smaller cloud droplets. This aerosol indirect effect resulting the decrease in CER can suppress cloud droplet collision-coalescence, thereby increasing cloud lifetime and cloud albedo. Additionally, the decrease in CER can lead to an increase in COT due to the increase in total cross-sectional area of small-sized cloud droplet population (Khatri et al. 2023). Thus, such aerosol indirect effect can be ruled out as the primary factor for the increase in CER (2.1 μm) and the decrease in COT, as noted in Fig. 8, during strong biomass burning period. The next mechanism for aerosol–cloud interaction is the aerosol semi-direct effect, where light-absorbing aerosols modify cloud properties either by embedding within cloud layers or by altering surrounding atmospheric conditions. When embedded within clouds, light-absorbing aerosols enhance local droplet warming by absorbing solar radiation (Stier et al. 2007). This localized heating can promote evaporation from warmer droplets, redistributing water vapor to colder neighboring droplets, which enhances growth and fosters a droplet size distribution that favors collision-coalescence (Abade et al. 2018; Chandrakar et al. 2021). Additionally, localized heating can generate small-scale turbulence within the cloud, increasing droplet interactions and enhancing coalescence potential. Even without direct embedding within cloud, light-absorbing aerosols influence cloud droplet growth process by modifying atmospheric conditions around the cloud. Increased concentrations of light-absorbing aerosols reduce downwelling solar radiation (Ningombam et al. 2015), leading to surface cooling (Khatri et al. 2009) and weakening of the surface sensible heat flux (Yu et al. 2002). These changes inhibit lower-atmosphere convection while simultaneously heating the upper boundary layers (Ding et al. 2016; Wang et al. 2018). This phenomenon reduces the vertical transport of water vapor, creating competition among cloud droplets for available vapor and favoring the evaporation of smaller droplets, which can enhance collision-coalescence process. In addition to these aerosol effects, other factors, such as temperature variations, PWC, and precipitation efficiency, can have influence on cloud properties modifications. However, given the absence of significant meteorological anomalies during intense burning period, as indicated in Fig. 5, these factors likely played a secondary role in the present context. Therefore, we suggest that the observed decrease in COT coupled with an increase in CER (2.1 μm) during intense biomass burning period was primarily attributable to strong light-absorbing aerosols, enhancing cloud-droplet growth via collision-coalescence process. This result is further supported in Fig. 8(c) and Fig. 9, as discussed below. Figure 8(c) shows a relatively larger negative CER difference between the longer (3.7 μm) and the shorter (1.6 μm) wavelengths during the period of strong biomass burning compared to other periods. This suggests the presence of smaller cloud droplets in the upper cloud layers compared to those in the middle and/or lower cloud layers during strong biomass burning period, as opposed to moderate and weak burning periods. It is important to note that the terminal velocities of cloud droplets can increase with the increase in cloud droplet size, facilitating the accumulation of larger cloud droplets in the middle and/or lower cloud layers compared to the upper cloud layers, suggesting that clouds with the dominance of collision-coalescence process can result larger cloud droplets in the middle and/or lower cloud layers than in the upper cloud layers. Overall speaking, the three panels of Fig. 8 collectively suggest a very important role of agricultural biomass burning activities in modifying cloud properties by inducing the anti-Twomey effect.

In the literature, the Twomey and the anti-Twomey effects are typically discussed in terms of fixed LWP (e.g., Nakajima et al. 2001). The insufficient data samples made it difficult to study these effects separately for each type of burning period by dividing data into several LWP bins. However, to demonstrate the significant role of AAOT on the anti-Twomey effect, we aggregated data from the entire study period and then divided them into LWP bins with a 10 g/m² interval. Furthermore, to mitigate the possible influences of meteorological factors, we employed a statistical approach outlined by Khatri et al. (2022). In brief, a multiple linear regression method was applied, treating cloud properties (COT or CER) as dependent variables and AOT (or AAOT) and meteorological parameters as independent variables. We formulated their relationship as follows:

where CLD represents the cloud property (COT or CER), X represents AOT (or AAOT), and \(M_{i}\) represents meteorological factors (lower tropospheric stability LTS, i.e., potential temperature difference between 700 and 1000 hPa, and PWC). a and b_i are constant terms. \(CLD^{\prime}\), \(X^{\prime}\), and \(M_{i}^{\prime}\) were calculated as follows:

where \(\overline{CLD},\overline{X}\), and \(\overline{M_{i}}\) represent the mean values of CLD, X, and \({M_{i}}\), respectively. Similarly, \({\sigma }_{CLD}, {\sigma }_{X}, \text{and }{\sigma }_{M_{i}}\) represent the standard deviation values of CLD, X and \({M_{i}}\), respectively. Finally, we calculated a constant term a, which provides information about ∂CLD’/∂X’, a parameter describing the variation of cloud property with the variation of aerosol property.

Figure 9(a) displays the values of ∂CER ́/∂AOT ́ and ∂CER ́/∂AAOT ́, and Fig. 9(b) displays the values of ∂COT ́/∂AOT ́ and ∂COT ́/∂AAOT ́ for different LWP bins. These values are statistically significant at 95% confidence interval. Statistically insignificant values at 95% confidence interval are not shown in Fig. 9(a) and 9(b). Further, Fig. 9(c) illustrates the count of data samples for each LWP bin. Figure 9(a) reveals positive values of ∂CER ́/∂AOT ́ and ∂CER ́/∂AAOT ́ for all LWP bins. Additionally, except for very thin clouds with LWP < 10 g/m², the values of ∂COT ́/∂AOT ́ and ∂COT ́/∂AAOT ́ are negative for all LWP bins, suggesting that aerosols prevalent in NW India during the post-monsoon season are capable of inducing the anti-Twomey effect. Almost for all LWP bins, ∂CER ́/∂AAOT ́ is higher than ∂CER ́/∂AOT ́. Moreover, in general, ∂COT ́/∂AAOT ́ is lower than ∂COT ́/∂AOT ́ for all LWP bins, indicating that an increase in AAOT can strengthen the anti-Twomey effect in comparison to the increase in AOT. Considering the higher values of AAOT during the strong burning period followed by the intermediate and weak burning periods, the anti-Twomey effect is suggested to be the strongest during the strong burning period, with a gradual weakening in the intermediate and weak burning periods. The decline in ∂CER ́/∂AOT ́ and ∂CER ́/∂AAOT ́ values at larger LWP bins may be attributed to fewer data samples available for analyses. Further, Fig. 9 suggests that the strength of the anti-Twomey effect can increase with the increase in LWP, as thicker clouds provide more opportunities for cloud droplet collision and coalescence compared to thinner clouds. Similar results were reported for aerosol–cloud interactions over highly polluted urban areas of the IGP (Khatri et al. 2022).

5.4 Radiation budget

Light-absorbing aerosols play a very important role in trapping solar (shortwave) radiation within the atmosphere (Khatri et al. 2009). This warms the atmosphere and influences the emission of longwave radiation (Nasrtdinov et al. 2018). Consequently, agricultural biomass burning activities play a crucial role in modulating radiation budgets in both the shortwave and longwave spectral ranges.

Figure 10a-c displays ARF corresponding to the shortwave spectral range at the surface, TOA, and in the atmosphere for different burning periods. Similarly, Fig. 10d-f displays results for the longwave spectral range. Further, Fig. 10g-i illustrates total (shortwave + longwave) ARF at the surface, TOA, and in the atmosphere for different burning periods. The median, percentile, and whisker values shown in the box plots in Fig. 10 differ significantly among the different groups at the 95% confidence level (p value < 0.05), as indicated by the Kruskal–Wallis test. Figure 10a-c shows relatively stronger cooling effects both at the surface and TOA, but a relatively stronger heating effect within the atmosphere for the shortwave spectral range during the period of strong burning. Compared to the mean values for the weak burning period shown in Fig. 10a-c, the cooling effects at the surface and TOA and the heating effect within the atmosphere increased by ~ 30% or more during the period of strong burning for the shortwave spectrum. Such results arise from the fact that light-absorbing aerosols can reduce both downwelling shortwave irradiance at the surface and upwelling shortwave irradiance at the TOA by trapping solar radiation within the atmosphere. Simultaneously, the atmosphere absorbs this trapped energy and re-emits a portion of it as longwave radiation. Thus, a warmer atmosphere releases more longwave irradiance, and vice versa. Figure 10d-f depicts positive values of ARF at the surface and TOA (heating effects) and negative values in the atmosphere (cooling effect) for the longwave spectral range, suggesting that the effects of aerosols in the shortwave spectral range are opposed by those in the longwave spectral range. Despite such opposing effects in the shortwave and longwave spectral ranges, the total (shortwave + longwave) ARF values, as shown in Fig. 10g-i, suggest that light-absorbing aerosols emitted in significant amounts due to strong burning activities are effective in cooling the surface and TOA, but heating the atmosphere across the entire spectral range. Hence, agricultural biomass burning can have considerable effects not only on aerosol and cloud properties but also on radiation budget perturbation. Such effects of agricultural biomass burning can play important roles in modulating atmospheric dynamics (Wang et al. 2013), as well as climate system and hydrological cycle at a regional scale. Hence, the impacts of agricultural biomass burning on climate system and hydrological cycle require more detailed studies in the future.

This study used a suite of satellite and reanalysis data spanning nearly two decades (2002–2021) to examine the temporal characteristics of post-monsoon (September–November) agricultural biomass burning activities in NW India and their effects on atmospheric changes, including aerosol and water cloud properties, as well as the radiation budget in the shortwave and longwave spectral ranges. Our analyses reveal distinct temporal patterns in agricultural biomass burning activities in NW India, including increasing trends in the burning peak occurring day number of the year, burning intensity, and pre-harvest NDVI, but a decreasing trend in the duration of the burning period (1-sigma value of a Gaussian distribution). The good alignment between pre-harvest NDVI and burning intensity suggests the potential utility of pre-harvest NDVI information for estimating burning intensity in advance, aiding policymaking for air pollution mitigation strategies.

Our statistical analyses, based on detecting different strengths of burning intensity using Gaussian distributions, indicate that intense burning activities can significantly increase the concentrations of total aerosols and light-absorbing aerosols over a relatively short period of less than 2 weeks, overriding the influence of meteorological factors. However, PWC is found to have a stronger influence on SSA variation. This study further suggests that the increase in light-absorbing aerosols from agricultural biomass burning activities can contribute to the enhancement of the cloud droplet collision-coalescence process, leading to larger cloud droplets and a decrease in cloud optical thickness.

Furthermore, the increase in light-absorbing aerosols from agricultural biomass burning activities is found to have significant impacts on the radiation budgets across both shortwave and longwave spectral ranges, leading to strong cooling effects at the surface and TOA, but a heating effect within the atmosphere for the shortwave spectrum, and opposing effects for the longwave spectrum. Despite these opposing effects between the shortwave and longwave spectral ranges, the overall effect across both spectral ranges results in cooling at the surface and TOA, and heating within the atmosphere.

Overall, our analyses provide evidence that the influences of agricultural biomass burning activities on cloud and radiation fields are substantial, highlighting the need for future studies to better understand their potential links with the regional climate system and water budget.

Data pertaining to aerosol optical thickness and cloud properties, fire, and vegetation observed from space are from MODIS observations. Aerosol single scattering albedo data are from MERRA-2. Shortwave and longwave radiation data stem from observations conducted by CERES, MODIS, and geostationary satellites. These data are freely accessible and downloadable from https://search.earthdata.nasa.gov/search. Meteorological data (air temperature, specific humidity data, dew point temperature, u- and v- components of wind and total precipitation) are from ERA5, which are freely accessible and downloadable from https://cds.climate.copernicus.eu/datasets/reanalysis-era5-single-levels?tab=download. Further, surface-observed data of aerosols and precipitable water content for Lahore site are from AERONET, which are freely accessible and downloadable from https://aeronet.gsfc.nasa.gov.

AAOT:

Absorbing Aerosol Optical Thickness

AERONET:

AErosol RObotic NETwork

AOT:

Aerosol Optical Thickness

ARF:

Aerosol Radiative Forcing

AVHRR:

Advanced Very High Resolution Radiometer

CER:

Cloud Effective Radius

CERES:

Clouds and the Earth’s Radiant Energy System

ECMWF:

European Centre for Medium-Range Weather Forecasts

ERA5:

ECMWF ReAnalysis 5th Generation

GOCART:

Goddard Chemistry Aerosol Radiation and Transport

GEOS-5:

Goddard Earth Observing System Model Version 5

IGP:

Indo-Gangetic Plain

MERRA-2:

Modern-Era Retrospective analysis for Research and Applications, Version 2

MODIS:

Moderate Resolution Imaging Spectroradiometer

NDVI:

Normalized Difference Vegetation Index

NW:

Northwest

PWC:

Precipitable Water Content

SKYNET:

Skyradiometer Network

SSA:

Single Scattering Albedo

TOA:

Top of Atmosphere

The authors acknowledge the use of MODIS and CERES data, provided by NASA’s Earth Observing System Data and Information System (EOSDIS); MERRA-2 data, provided by NASA’s Global Modeling and Assimilation Office (GMAO); ERA5 data, provided by the European Centre for Medium-Range Weather Forecasts (ECMWF); and AERONET data from the Lahore site, provided by the principal investigator and co-investigators.

This study was supported by Research Institute for Humanity and Nature (RIHN: a constituent member of NIHU) Project No. 14200133 (Aakash) and JSPS KAKENHI Grant Number 24K07129.

Authors and Affiliations

1. Department of Science and Engineering for Sustainable Innovation, Faculty of Science and Engineering, Soka University, Hachioji-Shi, Tokyo, Japan

   Pradeep Khatri

2. Center for Atmospheric and Oceanic Studies, Tohoku University, Sendai, Japan

   Tadahiro Hayasaka

3. Research Institute for Humanity and Nature, Kyoto, Japan

   Prabir K. Patra & Sachiko Hayashida

4. State Key Laboratory of Remote Sensing Science, Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing, 100101, China

   Husi Letu

5. GESTAR-II, Morgan State University, Baltimore, MD, USA

   Hiren Jethva

6. National Aeronautics and Space Administration, Goddard Space Flight Center, Greenbelt, USA

   Hiren Jethva

Contributions

PK conducted the data analysis and drafted the manuscript. TH, HJ, and PP reviewed and corrected the manuscript. PP and SH secured research funding and contributed to the design of the manuscript and interpretation of results. HL provided support in data analysis and result interpretation. All authors have read and approved the manuscript.

Corresponding author

Correspondence to Pradeep Khatri.

Competing interests

The authors declare that they have no competing interest.

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions