High-resolution vertical profiles of clouds and aerosols were obtained from the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite (CALIPSO) (Winker et al.,2010; Winker et al., 2007). The CALIPSO payload, polarization-sensitive backscatter lidar, known as the Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP), operates at 532 and 1,064 nm wavelengths. The CALIPSO satellite has been observing the vertical distribution of aerosols since June 2006. The primary instrument on CALIPSO is CALIOP, a nadir-viewing dual-wavelength (532 and 1,064 nm) dual polarization at 532 nm, elastic back-scatter lidar (Hunt et al., 2009). Level 2 algorithms detect features, assign type classification for aerosols, and retrieve extinction coefficients from the attenuated backscattered signals. The extinction algorithm retrieves vertical profiles of extinction, reported separately for aerosols and clouds. Aerosol extinction is not reported within clouds because the lidar signals are dominated by cloud scattering and so atmospheric features are classified as either aerosols or clouds and the retrieved extinction is reported for only one or the other. The aerosol profile product combines the profiles retrieved within aerosol layers to report vertical profiles of extinction coefficients at 5 km horizontal resolution. The vertical resolution is 60 m from 0.5 to 20.2 km and 180 m above 20.2 km. For the 5 km along CALIPSO track horizontal averaging, 15 consecutive level 1 B profiles are used (Tackett et al., 2018). The lidar ratios and their uncertainties for several of the aerosol subtypes have been revised in version 4 (Kim et al., 2018). It is said that the reductions in the relative uncertainties associated with the improved lidar ratios will reduce the relative uncertainties in the retrieved extinction coefficients and optical depths. These improved lidar ratios in V4 are a better representative of actual conditions than in previous data releases (Young et al., 2018). Here, in the present study, we used the level 2 version 4.10/4.20 CALIPSO aerosol profile (APro) data for the period 2010–2020 (https://asdc.larc.nasa.gov/project/CALIPSO). The details of the lidar ratio selection algorithm are well documented by Kim et al. (2018). We used the extinction profiles of aerosol and the optical depth for dust and elevated smoke aerosols at 532 nm. These extinction profiles were used to calculate the aerosol optical depth profiles at pre-defined altitudes including higher numbers in the altitude-range 1–10 km. We gridded these profiles at a 1× 1 degree resolution.

The aerosol optical depth data obtained from the Moderate Resolution Imaging Spectroradiometer (MODIS) were also analyzed to understand aerosol variations over India. The MODIS sensor measures radiances at 36 bands, ranging from visible to infrared and varying spatial resolutions. Here, we used daily AOD at 550 nm from the MODIS (MOD08_D3) collection 6.1 level 3, combined dark target and deep blue, from 2010 to 2020 (Hsu et al., 2013; Levy et al., 2013; Wei et al., 2019a; Wei et al., 2019b) (https://giovanni.gsfc.nasa.gov/giovanni). The past studies show that CALIPSO AOD is biased towards lower values as compared to MODIS (Kim et al., 2013). Kittaka et al. (2011) found that the biases between them vary with season and are higher over land than ocean. The observed biases may due to various reasons, e.g., MODIS has a higher frequency than CALIPSO (Ma et al., 2013).

We also analyzed Absorbing Aerosol Index (AAI) from the Ozone Monitoring Instrument (OMI) launched in 2004 (Torres et al., 2007). The level 3 data for the period 2010–2020 were obtained from http://giovanni.gsfc.nasa.gov/giovanni/. MODIS fire data (https://firms.modaps.eosdis.nasa.gov/active_fire/) were used to plot the location of fires during March–April 2020. The fire data with a confidence level above 80 were used to locate the fires.

We also used the Copernicus Atmosphere Monitoring Service (CAMS) near-real-time observations of the location and intensity of active fires to estimate the emissions of pollutants. The Copernicus Atmosphere Monitoring Services (CAMS) uses wildfire as a general term to describe active vegetation fires detectable by the satellite. This also includes forest, grassland and peat fires, and open burning of agricultural waste (https://atmosphere.copernicus.eu/fire-monitoring). We referred to it as fires since during the lockdown period fires were mostly from agricultural burning. CAMS estimates are based on the Global Fire Assimilation System (GFAS). We also used the CAMS-derived fluxes of black carbon, organic carbon, total carbon, sulphate, ammonia, and particulate matter that have a diameter of less than 2.5 μm (PM2.5) for the period 2010–2020 (https://apps.ecmwf.int/datasets/data/cams-gfas/).

It should be noted that the horizontal resolution of all data sets used in this study (CALIPSO, MODIS, OMI, CAMS) is 1× 1 degree and for the period 2010–2020. We show changes in AOD, dust optical depth, and elevated smoke optical depth during the lockdown period in comparison to climatology (2010–2019). To check whether these changes are significantly different than climatology, we apply two-sided Student’s t-test (Zimmerman, 1987; Walpole and Raymond, 2006). If the p-value is less than 0.05 (95% significance level), then we reject the null hypothesis and conclude that the differences are significant.

The direct aerosol radiative forcing and heating rate are assessed through the one-dimensional parallel plane Fu-Liou-Gu (FLG) Radiative Transfer (RT) model (Fu and Liou, 1992; Fu and Liou, 1993; Gu et al., 2003; Gu et al., 2011; Lolli et al., 2019). The FLG RT code is initialized with the lidar vertically resolved CALIPSO optical depth profiles corresponding to the different aerosol species, which were matched to the Optical Properties of Aerosol and Clouds (OPAC) (D’Almeida et al., 1991; Tegen and Lacis, 1996; Hess et al., 1998). Catalog-based physical and optical models are embedded in the code (Gu et al., 2011). The number of levels of the RT model is adjusted to match CALIOP lidar resolution, i.e., the RT model and the vertically resolved lidar optical depth observation will have the same spatial resolution. The total aerosol radiative forcing is computed adding all contributions (in terms of optical depth) from the different aerosol species identified by the CALIPSO classification algorithm in the considered region and matched with the corresponding FLG aerosol species (Tosca et al., 2017).

From CALIOP data, among the 18 aerosol types parameterized within the FLG RT model from the OPAC catalog, we considered only the main two types of interest that match the CALIPSO classification: transported dust (CALIPSO: “dust”) and black carbon (CALIPSO: “smoke”). To compute the radiative forcing at the top of the atmosphere and at the surface, the FLG RT model, which also accounts for aerosol hygroscopicity, solves the radiative fluxes at each level for 18 spectral bands (12 short-wave, 6 long-wave, Fu and Liou, 1992; Fu and Liou, 1993). Nevertheless, an important source of error (potential) is represented by Version 4 CALIOP aerosol types parameterization into FLG RT. The CALIPSO classification algorithm is not able to distinguish between local urban pollution and an advected smoke aerosol layer from distant sources that descend below 2.5 km. However, it is reasonable to suppose that those episodes are infrequent, but they can still occasionally introduce a bias in the analysis. For each annually averaged lidar extinction profile used as input in the FLG code, the aerosol direct radiative forcing (DRF) at the bottom of the atmosphere (surface) and top of the atmosphere (TOA) and the vertically resolved heating rate (HR) are computed. These estimates are obtained by subtracting the net radiative flux when the aerosols are present in the atmosphere from the net radiative flux obtained during pristine conditions, as shown in the following equation: DRF , HR = FLG TotalSky − FLG Pristine

The other data that are needed as input to the FLG model (i.e., the temperature, the atmospheric thermodynamic variable profiles, the ozone concentration, and the mixing ratio) are obtained from the tropical standard atmosphere (USS976). The FLG radiative transfer model also needs the Solar Zenith Angle (SZA) for the computation. We use SZA for each box at noon local time of the 15th day of that month. Following the approach used in Landi et al. (2021), we applied a constant (wavelength-independent) albedo value of 0.12 for urban environments, 0.15 for vegetated areas, and 0.37 for desertic regions (obtained integrating the hemispherical directional reflectance, Strahler et al., 1999) while the infrared surface emissivity is set to a constant value of 0.98.

A significant reduction in AOD (∼40%) over Western and Northern India during the lockdown period is evident from MODIS observations (Figure 1A). A similar reduction (a drop of ∼45% in AOD with respect to climatology 2010–2019) is also showed by CALIPSO measurements over this region during the same period (Figure 1B). Recently, other studies also reported aerosol reduction over the parts of the Indian region during the lockdown period, e.g., Pathakoti et al. (2021) showed a decrease in AOD over the Indo-Gangetic plain by ∼24% (climatology 2015–2019) using MODIS data. While a study by Sanap (2021) showed a reduction of aerosols (∼16–27%) over north India (climatology 2000–2020). Mishra and Rathore (2021) also reported an overall decrease in AOD by 60% (in comparison to 2019) over the Indian landmass.

MODIS AOD (Figure 1A) shows enhancement (5–30%) over central India but it is faintly seen in CALIPSO (Figure 1B). It may be due to limited CALIPSO data (every 16 days overpass at the same location) during the lockdown period (Winker et al., 2007). The AOD enhancement over central India may be associated with aerosols emitted from fires (see discussions in Section 3.2). Interestingly, enhancement in AOD is also seen over the Bay of Bengal (12%) and parts of the North Arabian Sea (∼22°N) (5%) in MODIS data (Figure 1A). The atmospheric circulation (wind at 850 hPa) indicates that aerosol loading over the Bay of Bengal is associated with transport from India and Myanmar regions. During spring, anthropogenic and dust aerosols are transported from the Indo-Gangetic Plain and Myanmar region to the Bay of Bengal (Nair et al., 2016). In agreement with our results, past studies show evidence of anthropogenic (Satheesh et al., 2001; Kumar et al., 2014; Nair et al., 2016) and dust (Lakshmi et al., 2017) aerosol loading over the Bay of Bengal during the spring season. Trajectory analysis-based studies also show that anthropogenic aerosols over the Bay of Bengal are associated with transport from the Indian region (Nair et al., 2016). The aerosol enhancement over the Arabian Sea is due to transport from West Asia (Lau and Kim 2006). During spring, westerly winds transport dust from West Asia to the Arabian Sea (Vinoj et al., 2014; Fadnavis et al., 2017b). However, in spring 2020 dust transport from West Asia was suppressed (Fadnavis et al., 2021). A small enhancement (5%) in AOD over the Arabian Sea may be due to the transport of small amounts of dust and biomass-burning aerosols from Saudi Arabia (Figure 1A) (Discussed in Section 3.2). The enhancement in AOD over the Bay of Bengal (85°E–95°E, 0°N–10°N) is smaller in CALIPSO measurements than MODIS (Figure 1B) and no enhancement is seen over the Arabian Sea. It may be due to limited CALIPSO data as mentioned above.

We understand the influence of dust transport from West Asia on the Indian region which occurs in spring (Lau and Kim 2006), here we show the changes in dust during the lockdown period using CALIPSO measurements of dust optical depth and the OMI aerosol index (OMI-AAI). The OMI-AAI over the deep inland area indicates dust aerosols (Brooks et al., 2019). The CALIPSO dust optical depth and OMI-AAI shows a reduction of ∼30% over the Indian region (Figures 2A,B). The Arabian Sea and Bay of Bengal regions show a widespread decrease (∼20%) with pockets of a small enhancement in dust optical depth (∼14%) and OMI-AAI (∼20%) (Figures 2A,B). It shows that during the lockdown period, in spring 2020, transport of dust from West Asia was lower than the climatology. The model simulations for COVID-19 anthropogenic emission changes also showed a reduction of dust aerosols over the Arabian Sea in the spring of 2020. These simulations showed that the anthropogenic emission reductions had induced changes in atmospheric circulation that inhibited the transport of dust from West Asia to the Tibetan Plateau during spring 2020 (Fadnavis et al., 2021).

Further, we show anomalies in the vertical distribution of elevated smoke aerosols from the CALIPSO measurements during the lockdown period in Figure 2C. The elevated smoke is a name of a CALIPSO product for smoke layers with tops higher than the 2.5 km above the planetary boundary layer (McGrath-Spangler and Denning 2013; Kim et al., 2018). It shows positive anomalies over different parts of India that may be due to the presence of local fires as well as the long-range transport of biomass-burning aerosols. A striking feature seen in Figure 2C is a large enhancement (an increase of 0.08–0.1 with respect to climatology) in elevated smoke optical depth over 1) the central peninsular (72°E–81°E, 10°N–25°N), 2) North-East-India-Myanmar region (93°E–100°E, 15°N–25°N), and a part of IGP (81°E–88°E, 26°N–31°N. Also, a high amount of smoke AOD is seen over eastern parts of central India (15°N–24°N; 77°E–82°E). Other parts of India also show a small enhancement in smoke optical depth that is substantially less than over the central peninsula, eastern-central India, and the North-East-India-Myanmar region (positive anomalies 0.02–0.03) above two regions. Smoke aerosols may be associated with the fire events hence we show anomalies of MODIS fire during the lockdown period (Figure 2D). Figure 2D shows negative anomalies over North India (blue color) and positive anomalies over southern India, the Northeast-India-Myanmar region, and over central India (indicated by boxes in Figure 2D). The regions of dense fire are collocated with higher amounts of smoke optical depth (Figure 2C). The contribution to elevated smoke aerosol optical depth over central India is due to these local fires as well as the transport of smoke aerosols emitted from the surrounding regions. High amounts of smoke anomalies over the eastern parts of central India are due to transport (see circulation in Figure 2B) and not directly emitted by the fires, since the fire anomalies are negative over eastern parts of central India. Higher amounts of smoke aerosols (positive anomalies 0.04–0.06) are seen over the Arabian Sea. The wind vectors indicate this enhancement is due to transport from central India (Figure 2C).

In this section, we show anomalies in the fire fluxes of black carbon, organic carbon, ammonia, sulfur dioxide, particulate matter (PM2.5), and total carbon aerosols during the lockdown period (Figures 3A–F). Regions of positive anomalies (>1.5) are referred to as emission hotspots. The emission hotspots for organic carbon, black carbon, ammonia, sulphur dioxide fluxes (leads to the formation of sulfate aerosols), and particulate matter (PM 2.5) are collocated with the regions of dense fires during the lockdown (central peninsular and the North-East-India-Myanmar region). The positive anomalies of fire fluxes over southern India (Figure 3) are collocated with MODIS fire location (Figure 2D). This confirms that the increase in anomalous aerosol loading over 1) the central peninsular and 2) North-East-India-Myanmar region is due to fire emissions. Figures 1–3 show that, although the contribution of anthropogenic aerosols had reduced, the smoke aerosols over the central peninsular and North-East-India-Myanmar region caused an enhancement in AOD over these regions (Figures 1A,B, 2C,D).

Carbonaceous aerosols are key components of smoke that absorb solar radiation producing local atmospheric heating (Galanter et al., 2000; Zhang et al., 2020), while they produce a cooling effect on the climate via inhibiting solar radiation from reaching the surface (Shawki et al., 2018; Fadnavis et al., 2019b). During the lockdown period, enhanced smoke aerosols may have affected atmospheric heating. Here, we deliberate on heating rates and radiative forcing estimated from elevated smoke optical depth/profiles averaged for the lockdown period.

Figure 4A shows the spatial distribution of elevated smoke optical depth from CALIPSO at 532 nm during the lockdown period. It shows a high amount of elevated smoke optical depth (Figure 2C), over 1) the central-peninsular region and 2) Northeast-India-Myanmar region, and low smoke optical depth in 3) the IGP (depicted in Figure 4A by boxes). The high values of elevated smoke optical depth are also seen over eastern parts of central India due to transport (discussed in Section 3.2). Further we show vertical distribution of elevated smoke optical depth at the above three regions. There is a significant enhancement in elevated smoke aerosol optical depth (0.002 ± 0.001 to 0.0065 ± 0.002) in the lower troposphere (2–5 km) over the central-peninsular region and (0.008 ± 0.001 to 0.025 ± 0.003) over the Northeast-India-Myanmar region. The enhancement over the IGP is significantly less than the above-mentioned two regions due to comparatively fewer fires over the IGP. Interestingly, all three regions show an elevated layer of smoke aerosols between altitudes 2–5 km. The elevated layers of smoke aerosols corroborate well with the regions of high aerosol optical depths observed over the central-peninsular and Northeast-India-Myanmar regions (Figure 3). Our analysis shows that the mean contribution due to elevated smoke aerosols to altitudes ranging from 2 to 5 km is ∼57% of the total columnar optical depth over the central peninsular and ∼68% over the Northeast-India-Myanmar region. While elevated smoke aerosols over the IGP contributed ∼18% to columnar optical depth the during lockdown period. Sarangi et al. (2016) reported the mean contribution of the aerosol layer from a 1.5–5.5 km altitude as ∼51–60% to the total columnar aerosol optical depth for the years 2009–2011 during May–June over Kanpur.

Figure 4C shows profiles of heating rate estimated from CALIPSO observations of elevated smoke aerosols over 1) the central peninsular, 2) Northeast-India-Myanmar region, and 3) IGP. Heating rate profiles over the Northeast-India-Myanmar region show a higher amount of heating than over the central-peninsular region of India. It is quite evident that high amounts of elevated smoke aerosols during lockdown at altitudes between 2 and 5 km over the central peninsular have produced significant heating at 1.6 K/day ±0.5 K/day at those altitudes and ∼6.5 K/day ±0.5 K/day heating over the Northeast-India-Myanmar region. The IGP region shows comparatively less heating at ∼0.5 K/day ±0.3 K/day at those altitudes due to smaller amounts of smoke aerosols. Strong warming is seen locally in the altitudes corresponding to higher amounts of elevated smoke optical depth. Also, a large increase in the heating rates is noticed below the peak in the smoke aerosol profile. For example, a peak in heating rates over the central peninsular is seen at 3 km while smoke aerosols show a peak at 3.8 km. Similarly, over the Northeast-India-Myanmar region, heating rates are maximum at 2.2 km while the aerosol profile has a peak at 2.8 km. This indicates that the peak in smoke aerosols and heating rates occurs at different altitudes. This may be due to aerosol heating occurring within a layer of atmosphere that retains and changes pressure values (Tripathi et al., 2007). Past studies showed enhancement in carbonaceous aerosols and increases in the heating rates by 0.08 K/day in the lower troposphere over India (Fadnavis et al., 2017a). The annual mean atmospheric heating rate due to the BC aerosols was 0.86 K/day over the Guwahati region during 2014. Pani et al. (2018) estimated atmospheric heating of ∼1.4–3.6 K/day due to biomass-burning aerosols in the dry season over a station in southeast Asia.

Aerosol radiative forcing is defined as the net radiative change by aerosols present in the Earth system. Aerosols significantly impact the regional climate and this phenomenon has been largely studied (IPCC, 2014; Vinoj et al., 2014; Fadnavis et al., 2019a). Decoupling the elevated smoke aerosol optical depth from the atmospheric column, we estimate the biomass burning radiative forcing at the surface and top of the atmosphere (TOA) over 1) the central peninsular, 2) Northeast-India-Myanmar region, and 3) Indo-Gangetic Plain averaged for the lockdown period (Figure 5). The estimated radiative forcing at the TOA shows warming of ∼4.8 W/m² over the central-peninsular region, ∼15 W/m² over the Northeast-India-Myanmar region, and ∼1 W/m² over the IGP. The lower-tropospheric warming caused by elevated smoke aerosols has an implication on atmospheric circulation and cloud cover (Fadnavis et al., 2017b; Fadnavis et al., 2019b).

The estimated surface radiative forcing due to smoke aerosols over the central-peninsular region is ∼−38 W/m² (±15 W/m²), ∼−152 W/m² (±50 W/m²) over the North-India-Myanmar region, and ∼−8 W/m² (±9 W/m²) over the IGP. The amount of radiative forcing at the TOA (positive) and surface (negative) over the three regions is proportional to the amount of smoke aerosol, e.g., a large amount of smoke aerosols in the North-India-Myanmar region has imposed higher radiative impacts there than in the central peninsula and IGP (higher amount of surface cooling and warming at the TOA). All the three regions show warming in the atmosphere (in-atmospheric radiative forcing, TOA-surface), central India: ∼73 W/m² (±40 W/m²), North-India-Myanmar region: ∼167 W/m² (±50 W/m²), and IGP: ∼9 W/m² (±9 W/m²).

A previous study also showed positive radiative forcing due to black carbon aerosols at the top of the atmosphere (∼5 W/m² over Bangalore, ∼9.5 W/m² over Guwahati), negative radiative forcing at the surface (∼−23 W/m² over Bangalore and ∼−21.1 W/m² over Guwahati), and in-atmospheric warming (∼27 W/m² ± 9 W/m² over the Indian region) (Babu et al., 2002; Tiwari et al., 2016; Nair et al., 2017).

Importantly, these studies show that smoke/carbonaceous aerosols produce positive radiative forcing at the top of the atmosphere, negative radiative forcing at the surface, and in-atmospheric warming is agreement with our results. The atmospheric heating generated by smoke aerosols has implications on atmospheric circulation and cloud cover while the surface cooling might have effects on the precipitation changes (Lohmann and Feichter, 2005; Ward et al., 2012).