Wintertime radiative effects of black carbon (BC) over Indo-Gangetic Plain as modelled with new BC emission inventories in CHIMERE

To reduce the uncertainty in the black carbon (BC) induced climatic impacts from the global and regional aerosolclimate model simulations, it is a foremost requirement to improve the prediction of modelled BC distribution. And that specifically, over the regions where the atmosphere is loaded with a large amount of BC, e.g., the Indo-Gangetic plain (IGP) in the Indian subcontinent. Here we examine the wintertime radiative perturbation due to BC with an efficiently modelled 5 BC distribution over the IGP in a high-resolution (0.1◦×0.1◦) chemical transport model, CHIMERE, implementing new BC emission inventories. The model efficiency in simulating the observed BC distribution was assessed executing five simulations: Constrained and bottomup (Smog, Cmip, Edgar, Pku) implementing respectively, the recently estimated India-based constrained BC emission and the latest bottom-up BC emissions (India-based: Smog-India, and global: Coupled Model Intercomparison Project phase 6 (CMIP6), Emission Database for Global Atmospheric Research-V4 (EDGAR-V4) and Peking 10 University BC Inventory (PKU)). A low estimated value of the normalised mean bias (NMB) and root mean square error (RMSE) from Constrained estimated BC concentration (NMB: < 17%) and aerosol optical depth due to BC (BC-AOD) (NMB: 11%) indicated that simulation with constrained BC emissions in CHIMERE could simulate the distribution of BC pollution over the IGP more efficiently than with the bottom-up. The large BC pollution covering the IGP region comprised of wintertime all-day (daytime) mean BC concentration and BC-AOD from theConstrained, respectively, in the range 14–25 (6– 15 8) μg m−3 and 0.04–0.08, with a strong correlation between the variance in BC emission and simulated BC mass concentration or BC-AOD. Five main hotspot locations were identified in and around Delhi (northern-IGP), Prayagraj/Allahabad-Varanasi (central-IGP), Patna-Palamu (upper/ lower mideastern-IGP), and Kolkata (eastern-IGP). The wintertime radiative perturbation due to BC aerosols from theConstrained included a wide-spread enhancement in atmospheric radiative warming by 2-3 times and a reduction in surface cooling by 10%-20%, with net warming at the top of atmosphere (TOA) of 10-15 W m−2, compared 20 to the atmosphere without BC, for which, a net cooling at the TOA was, although, exhibited. These perturbations were spotted 1 https://doi.org/10.5194/acp-2020-511 Preprint. Discussion started: 25 September 2020 c © Author(s) 2020. CC BY 4.0 License.


Introduction
Black carbon (BC) is released into the atmosphere from the incomplete combustion of carbon-based fuels (Bond et al., 2013; 25 Verma et al., 2013;Sadavarte and Venkataraman, 2014). It is one of the constituents of concern among the atmospheric aerosol pollutants because of its profound impact on climate through an imbalance of the Earth's radiation budget, besides, degradation of air quality and adverse effects on human health as well (Qian et al., 2011;Wang et al., 2014a;Fan et al., 2015;Zhang et al., 2015;Janssen et al., 2011Janssen et al., , 2012. Among aerosol constituents, BC aerosols are considered as the strongest absorber of visible solar radiation and, thereby, a contributor to tropospheric warming (Ramanathan and Carmichael, 2008;Gustafsson and Ramanathan, 2016). However, the magnitude of tropospheric radiative warming due to BC aerosols is highly uncertain and is classified with a medium to low-level understanding in the Inter-governmental Panel on Climate Change-Fifth Assessment Report (IPCC-AR5) (Myhre et al., 2013a, b;Wang et al., 2016;Boucher et al., 2016;Permadi et al., 2018a;Paulot et al., 2018;Dong et al., 2019). The direct radiative forcing (DRF) of BC averaged over the globe is estimated in the range 0.2-1 W m −2 (Myhre et al., 2013b;Bond et al., 2013;Gustafsson and Ramanathan, 2016). These estimates from global climate models used 35 in the latest assessment by the IPCC is noted to be about 2-times lower than the observation-based estimates from satellite and ground-based Aerosol Robotic Network (AERONET) observations (0.7-0.9 W m −2 ) (Chung et al., 2012;Myhre et al., 2013b;Gustafsson and Ramanathan, 2016;Stocker et al., 2013). The DRF of BC is inferred to be furthermore uncertain (e.g., −0.06 W m −2 to +0.22 W m −2 ) when estimated for BC rich sources comprising of BC emitted with different composition of short-lived co-emissions of species, e.g., sulphate and organic carbon (Bond et al., 2013). 40 Though the consensus is still to be achieved in BC DRF, nevertheless, the global atmospheric absorption attributable to BC was found to be too low in models and had to be enhanced by a factor of three to converge with observation-based estimates (Bond et al., 2013). The systematic underestimation of BC aerosol absorption by the global climate model predictions relative to atmospheric observations as noticed specifically over south Asia and east Asia (Chung et al., 2012;Gustafsson and Ramanathan, 2016) is also in compliance with studies evaluating atmospheric BC concentration between model and observations. 45 For example, recent evaluations of BC concentration from global and regional aerosol models over south Asia showed that the simulated BC concentration, though, exhibited a consistent correlation with, but was significantly lower (by a factor of about 2 to 11) than the measured concentration (Kumar et al., 2018;Verma et al., 2017;Kumar et al., 2015;Pan et al., 2015;Sanap et al., 2014;Moorthy et al., 2013;Nair et al., 2012). The factor of model underestimation was further noticed to be large specifically during wintertime over the Indo-Gangetic Plain (IGP) when the atmosphere is observed to be laden with a large 50 BC burden.
To assess BC aerosol absorption accurately and reduce the uncertainty in the BC DRF as estimated from global and regional aerosol-climate models, it is, therefore, a foremost requirement to improve the prediction of atmospheric BC estimates in models. And that specifically, over the regions where the atmosphere is loaded with a large amount of BC, e.g., the Indo-the large underestimation of BC concentration over the India mainland would primarily be due to BC emission dataset, instead of the model configurations.
The simulated atmospheric BC burden with atmospheric chemical transport models is related to the BC emission strength as input and simulated atmospheric residence time of BC . While the atmospheric residence time of BC aerosols is independent of the emission strength, it is an indication of model-specific treatments of transport and aerosol 70 processes affecting the simulated BC burden. The uncertainty in the mean model residence time for BC based on evaluation in sixteen global aerosol models, has been estimated as 33% , which is, though, noted to be much lower than the discrepancy found between the simulated BC and observation. Due to the inclusion of various complex physical-chemical atmospheric and aerosol processes in these models, in conjunction with the inherent uncertainty in inputs to the model (e.g., aerosol emissions and their properties), a systematic approach is required to improve the prediction of BC aerosols in the 75 models.
In this study, we examine the wintertime radiative effects of BC over the IGP evaluating the efficacy of simulated atmospheric BC burden in a high resolution (0.1 • ×0.1 • ) chemical transport model, CHIMERE, during winter when a large BC burden is observed. This is done executing multiple BC transport simulations with CHIMERE, implementing new BC emission inventories, which included the recently estimated India-based constrained BC emissions and the latest bottom-up 80 BC emissions (India-based: Speciated Multi-pOllutant Generator (Smog-India), and global: Coupled Model Intercomparison Project phase 6 (CMIP6), Emission Database for Global Atmospheric Research-V4 (EDGAR-V4) and Peking University BC Inventory (PKU)). A short description of the five BC emission datasets is provided in Section 2.1. The bottom-up BC emissions applied in the present study are being widely used in regional and global climate models in the assessment of spatial and temporal distribution of aerosol burden and aerosol-climate interactions (Eyring et al., 2016;Zhou et al., 2020;David et al., 2018;85 Lamarque et al., 2010;Meng et al., 2018;Wang et al., 2016), including (e.g. CMIP6) to support the IPCC climate assessment report (Myhre et al., 2013a). Henceforth, it is necessary to evaluate the performance of the new BC emissions (bottom-up and constrained), with a state-of-the-art chemical transport model, towards their adequacy to represent the BC distribution and thereby, the climatic impacts, over the IGP in the Indian subcontinent. The model efficiency in simulating the observed BC 3 https://doi.org/10.5194/acp-2020-511 Preprint. Discussion started: 25 September 2020 c Author(s) 2020. CC BY 4.0 License. distribution, including the spatial and temporal trend, is, thus, examined with the estimated BC concentration from five simu-90 lations subjected to the same aerosol physical and chemical processes with CHIMERE. Further, aerosol optical depth due to BC (BC-AOD) and its fractional contribution to total AOD, including the wintertime radiative perturbation due to BC aerosols is also examined.
The specific objectives of this study are, therefore, to (i) characterise the model efficiency from five simulations through a detailed validation and statistical analysis of simulated BC concentration with respect to ground-based measurements at stations 95 over the IGP, and identify the regional hotspots, (ii) utilise the multi-simulations to quantify the degree of variance in estimated BC concentration attributed to emissions corresponding to areas types (e.g., megacity, urban, semi-urban, low-polluted) and temporal distribution (e.g., daytime and evening hours), (iii) evaluate the spatial features of BC-AOD from five simulations, and analyse the association between simulated BC concentration and BC-AOD with BC emissions source strength, and (iv) examine the spatial distribution of wintertime radiative perturbation due to BC aerosols over the IGP and that compared with 100 the atmosphere considered without BC aerosols. and Forecasting (WRF-V3.7) model as a meteorological driver in offline mode, meaning that the meteorology is pre-calculated with WRF then read in CHIMERE. In case of our study, this configuration has an interest since we are performing emission scenarios. Having calculated the meteorology one time, we are sure that the differences between the simulation are due and only due to emission scenarions and not to possibly chaotic retroactions due to an online coulping between meteorology and aerosols. Simulations are carried out at a horizontal grid resolution of 0.

The CHIMERE chemical transport model
CHIMERE is a regional chemical transport model designed to model ten number of gaseous species and aerosols. For chemistry, the gaseous mechanism MELCHIOR2 is used (Derognat et al., 2003). The calculation of aerosols is as described in 115 Bessagnet et al. (2004) with ten bins, with a mean mass median distribution ranging from 0.039 to 40 µm and for primary particulate matter (black carbon BC, organic carbon, OC, and PPM the remaining part of primary emissions), sulphate, nitrate, ammonium, sea salt, and water. Secondary organic aerosols are formed following Bessagnet et al. (2009). Chemical concentration fields are calculated with a time-step of few minutes (using an adaptive time-step sensitive to the mean wind speed). For radiation and photolysis, the online FastJX model is used (Wild et al., 2000). The horizontal transport is calculated with the 120 4 https://doi.org/10.5194/acp-2020-511 Preprint. Discussion started: 25 September 2020 c Author(s) 2020. CC BY 4.0 License.
VanLeer scheme (van Leer, 1979) and vertical using an upwind scheme with mass conservation Menut et al. (2013). Note that additional information is provided in Table 1 (bottom). Boundary layer height is diagnosed using the Troen and Mahrt (1986) scheme, and deep convection fluxes are calculated using the Tiedtke (1989) scheme. Gaseous and aerosol species can be dry or wet deposited, and fluxes are computed using the Wesely (1989); Zhang et al. (2001) parameterizations. Initial and boundary conditions are estimated using global model monthly climatology calculated with the Laboratoire de Météorologie Dynamique 125 General Circulation Model coupled with Interaction with Chemistry and Aerosols (LMDz-INCA) (Szopa et al., 2009). The domain grid has twenty vertical levels in σ-pressure coordinates ranging from the surface (997 hPa) to 300 hPa. The horizontal grid has a resolution of 0.1 • ×0.1 • . Finally, the meteorological forcing is provided by the WRF regional meteorological model, described in the next section.

The WRF meteorological model 130
The WRF model is a state-of-the-art numerical weather forecast and atmospheric simulation system designed for both research and operational applications. The initial and boundary meteorological conditions for WRF simulation are obtained from Global Forecast System (GFS) National Center for Environmental Prediction -FINAL operational global analysis data (NCEP-FNL, http://rda.ucar.edu/datasets/ds083.2/) at a spatial resolution of 1 • × 1 • . Meteorological fields are simulated in WRF at the temporal resolution of one-hour with the horizontal resolution same as that for CHIMERE simulation. The meteorological 135 boundary conditions are updated every six hours. The optimized schemes applied in WRF simulation are as follows: Lin scheme for cloud microphysics (Lin et al., 1983), Grell 3D ensemble scheme for subgrid convection (Grell and Devenyi, 2002), Yonsei university (YSU) scheme for boundary layer (Hong et al., 2006), Rapid Radiative Transfer Model (RRTM) for radiation transfer (Mlawer et al., 1997), MM5 Monin-Obukhov scheme for surface layer and Noah LSM for land-surface model (Chen and Dudhia, 2001). Circulation Model (LMDZT-GCM)) with the observed BC by combining forward and receptor modelling approaches (Kumar et al., 2018;Verma et al., 2017). BC emission inventory based on bottom-up approach is generally compiled using information on activity data and generalised emission factors (see the references for bottom-up emissions,  (Pandey et al., 2014;Sadavarte and Venkataraman, 2014).
The CMIP6 BC emission used in the model simulations of CMIP6 is a combination of regional and global emission inventories and re-gridded as per EDGAR-V4 (Eyring et al., 2016). In the present study, global BC emission inventories utilised, viz. Besides BC emission, emission of aerosol species such as OC, SO 2 , primary particulate matter (PPM) are also implemented in CHIMERE. This implementation is done to perform atmospheric aerosol transport simulation for atmosphere with abundant 160 aerosol species (including BC), and that for atmosphere without BC. These simulations are required to calculate the radiative perturbations due to BC aerosol (refer to Section 2.3).
The spatial distribution of mean and percentage standard deviation (δ as represented in Equation 4) of BC emission flux from five BC emission inventories over the study domain is presented in Figures 1a and 1b, respectively. The mean BC emission flux is considerably high (450-1000 kg km −2 yr −1 ) over most of the IGP, with this being the highest (>2500 kg km −2 165 yr −1 ) over the megacities (Kolkata and Delhi). The divergence in BC emission flux is about 50%-75% over most of the IGP with this being relatively lower over the eastern and upper mideastern IGP. The divergence is large in and around megacities (100%-125%), and is noted to be specifically large (150%-200%) over the rural location in the lower mideastern IGP (in and around Palamu, refer to Figure 3e for details of location). Uncertainties in activity data and emission factors have been inferred leading to uncertainty in bottom-up BC inventories of about greater than 200% over India and Asia (Bond et al., 2004;Streets 170 et al., 2003;Lu et al., 2011). One of the drawbacks of the bottom-up approach is its inability to take into account possible unknown or missing emission sources. Bottom-up BC emissions are thus found to be often lower than the actual (Rypdal et al., 2005;Johnson et al., 2011;Zhang et al., 2005;Reid et al., 2009). Bottom-up BC emission over India includes a large missing source of BC emitted over India (Venkataraman et al., 2006). Hence, the divergence in emission data (refer Figure   1b) using five emission datasets (observationally-constrained and bottom-up BC emissions) is indicative of inadequacy in BC 175 emission source strength suggesting specific improvement required in bottom-up BC emission tabulation over the IGP and that at specific locations where the divergence is typically noted to be large.

Observational data for model evaluation and model sensitivity analysis
The spatial distribution of WRF simulated surface temperature over the IGP is compared with the available gridded distribution of observed temperature from Climatic Research Unit (CRU) (Morice et al., 2012). The observed temperature from CRU at 180 a horizontal resolution of 0.5 • ×0.5 • is re-gridded to the same resolution (0.1 • ×0.1 • ) as that from WRF and the bias in simulated temperature for each grid-cell is calculated using equation 1. The temporal trend of WRF simulated hourly mean of meteorological parameters (temperature, relative humidity) is also evaluated with that of observed from available measurements at stations over the IGP (Table 2). The monthly mean of simulated PBLH averaged is compared with that of measured available for stations at Delhi (mean of hourly PBLH during 1000-1600 LT), Kharagpur (during 1000-1100 LT and 1400-1500 LT),

185
Ranchi (at 1430 LT) and Nainital (during 0500-1000 LT) corresponding to the overlapping time hours from measurements ( Figure 2l in Section 3.1).
To compare simulated BC surface concentration with observations, measured BC surface concentration is obtained at stations over the IGP from available studies (refer to Table 2 and references therein). The selected stations correspond to area types identified as megacity (Delhi and Kolkata), urban (Agra, Kanpur, Prayagraj (or Allahabad) and Varanasi), semi-urban 190 (Kharagpur, Ranchi, and Bhubaneshwar) and low polluted (Nainital). Measurement data used in the present study are reported with an uncertainty (due to instrument artifacts, etc.) of about 10%-30% for PBLH (Seidel et al., 2010;Srivastava et al., 2010), 2%-3% for meteorological parameters, and 5%-20% for measured BC concentration measured (refer to Table 2 for details and references therein). It is to be noted that observational data used for BC surface concentration, belong to measurement during different years at stations over the IGP. Taking into account that the reported inter-annual variability (Safai et al., 2014;195 Surendran et al., 2013;Bisht et al., 2015;Ram et al., 2010b;Kanawade et al., 2014;Pani and Verma, 2014) of atmospheric BC concentration (5%-10%) is within the uncertainty range for measurements and also is much lower than the discrepancy between simulated and observed BC as reported in previous studies (refer to Section 1). The comparison between model and measurements at widespread geographical locations and area types as presented in this study is, therefore, justifiable and is primarily required for evaluating the model performance towards enhancing the statistical analysis.

200
The model and measured BC concentration are compared corresponding to daytime (1000-1600 LT) and all-day (24-hourly) winter monthly mean values. This comparison is made because measured BC concentrations are found exhibiting a strong diurnal variability, with a relatively lower value during daytime hours than that during the late evening to early morning hours attributed to prevailing wintertime meteorological conditions (Verma et al., 2013;Pani and Verma, 2014). Also, the daytime mean BC concentration exhibits a low hourly variability and corresponds to the well-mixed layer of atmosphere (Verma et al.,205 2013; Pani and Verma, 2014). Hence, the lower value of the daytime mean from the model than from observations is primarily attributable to a low emission strength. Evaluation of model estimates for both daytime and all-day mean, thus, provides a systematic hypothetical approach to identify the model discrepancy, if primarily due to emissions or that due to model processes attributed to meteorology (which is an input to the various aerosol processes that govern the atmospheric residence time of aerosols). This approach is further strengthened, implementing BC emissions from five new BC emission inventory 210 databases and simulating BC transport subjected to the same aerosol physical and chemical processes with CHIMERE.
Bias in simulated estimates (X modelled ) from simulations at stations mentioned above for hourly, all-day and daytime hours is estimated with respect to observed data (X obs ) with the equation as follows, where, X = BC concentration, temperature, relative humidity, wind speed and PBLH.

215
Statistical analyses are carried out corresponding to daytime and all-day winter monthly mean to evaluate the normalised mean bias (NMB, equation 2) and root mean square error (RMSE, equation 3) from the simulated results for N (=10 in this study) number of stations. We also evaluate the percentage deviation (δ) in simulated BC concentration attributed to BC emission, estimated as the variability about the mean of BC concentration from five simulations (refer to equation 4).
AOD from the simulations to estimate the sensitivity of simulated BC concentration or BC-AOD towards the variation in emission magnitude.
where σ is the standard deviation for the mean from five simulations (e.g., BC emissions, all-day, daytime mean of BC concentration, etc.).  Simulation for radiative flux with WRF-solar are performed for each of the three cases, as mentioned above, using respective simulated optical properties as input for each case. Shortwave (SW) radiative flux (at 550 nm) for clear sky condition is 250 estimated at the top (TOA) and bottom (SUR) layer of the atmosphere for atmosphere with BC and that without BC. This is done by subtracting the respective flux at TOA and SUR due to wAero from the flux due to wBC and BCaero, respectively.
The radiative perturbations represented as the direct radiative effects (DRE) due to BC aerosols at TOA (DRE T OA (BC)) and at SUR (DRE SU R (BC)), which are calculated by taking difference between the radiative flux from BCaero and that from wBC at the respective layers of the atmosphere (equation 5 and 6). The DRE at the atmosphere (ATM) due to BC is estimated by 255 subtracting the flux at the SUR from that estimated at TOA (equation 7).
3 Results and discussions High load of BC aerosols over the IGP as obtained (discussed later) in the present study is inferred due to confinement of pollution near the surface within the shallow boundary layer height in winter due to low vertical mixing and weak dispersion 270 of atmospheric pollutants, thereby, stagnant weather under the prevailing meteorological conditions, viz. low temperature and weak wind speed, the downdraft of the air mass, and a narrow PBLH (as presented above). Besides, the Himalayan mountains northward, further, inhibits the dispersion of aerosol pollutants and favours their confinement over the IGP. This inference is also in corroboration with the observational studies at stations over the IGP (e.g. Nair et al., 2007Nair et al., , 2012Pani and Verma, 2014;Verma et al., 2014;Vaishya et al., 2017;Rana et al., 2019). Further, the IGP also comprises of the highest population 275 density, and thereby the associated wintertime increased anthropogenic activities as perceived over the IGP, specifically from the combustion of biofuel, e.g., fuelwood and crop-waste for residential cooking and heating (Venkataraman et al., 2005;Verma et al., 2013;Sahu et al., 2015;Rana et al., 2019).
We compare the spatial distribution of monthly mean temperature from WRF simulations (Figure 2e) with that from gridded ground-based observations from CRU (Figure 2f). The bias in modelled temperature is found within ±5% over most of the 280 IGP ( Figure 2g) but is noticed to be slightly large (about ±10 to ±25%) over a few grids of the north-eastern, western and southern IGP.
A comparative study of the hourly distribution of winter monthly mean of the simulated surface temperature and relative humidity (RH), with the corresponding observed value from available measurements at Kharagpur (semi-urban) and Kolkata  (Figures 2h and 2j) is found to be comparing well with that from observations during daytime hours (1000-1600 LT) for both the stations; but is, however, seen to be underestimated (bias: −45% to −58%) during mid-night to early morning hours (0000-0500 LT). The WRF simulated meteorology is input to various aerosol processes that govern the atmospheric Figure 2. Spatial distribution of WRF simulated (a-c) winter monthly mean of (a) horizontal wind field (note the color scale is for wind speed in m s −1 and the arrows indicate the direction of the mean field), (b) vertical wind speed at 1000 hPa, (c) planetary boundary layer height (PBLH in m), and (d) topography (m above sea level, m asl); (e-g) spatial distribution of winter monthly mean of surface temperature from (e) WRF simulations, (f) observations from CRU, (g) percentage bias in winter monthly mean temperature from WRF; (h-k) validation of hourly distribution of winter monthly mean of (h,j) surface temperature, (i,k) relative humidity from WRF simulations with observations at stations (Kharagpur, KGP; Kolkata, KOL); (l) comparison between measured and simulated winter monthly mean of PBLH during day hours at stations under study. The error bars present the standard deviation (σ) in measured PBLH. Refer to Table 2  of simulated surface temperature than the observed during mid-night to early morning hours would lead to a decreased mixing of pollutants enhancing their accumulation in the atmosphere during these hours (as also evinced in the diurnal distribution of simulated BC concentration, refer to Section 3.2, Figure 4).
The WRF simulated RH at both stations (Figure 2i and 2k) is in good agreement with measurements (bias: −5% to +35%) with the mean RH during late evening to early morning hours (2000-0500 LT) being 2-times higher than that during daytime.

295
A comparison of winter monthly mean of PBLH during daytime hours (as described in Section 2.2) from WRF simulation with that available from observations at Delhi, Kharagpur, Ranchi, and Nainital is also presented (Figure 2l). The standard deviations (1σ) in measured values are within 10%-16% for Delhi, Kharagpur, and Ranchi and about 49% at Nainital. The simulated PBLH is close enough to measurements (bias estimated within ±10%) at all stations. Although, at Nainital the simulated bias is large (−28%), though, is within the range of uncertainty in observations as mentioned in Section 2.2.

300
Thus, overall, the meteorological parameters, evaluated as winter monthly mean are simulated consistently well with the WRF. A better temporally resolved meteorological boundary condition in WRF, aided with data assimilation (also discussed in Section 3.2), is believed would potentially lead to more adequately simulating the observed magnitude of diurnal distribution of meteorological parameters and reduce the discrepancy, specifically in simulated temperature during mid-night to early morning hours (as seen in the present study).     (Table 3 (top)) attributed to emissions is, specifically, the lowest for the low polluted location (e.g., Nainital) and is, generally, within 40% for all other locations under study. The δ for the megacity is noted as being, typically, amplified (51%-56%) during the late evening to early morning hours than that during daytime hours (36%-43%) compared to other locations under study; thereby suggesting that under the similar meteorological condition and with the same aerosol processes in the model the deviation in simulated BC concentration attributed to emissions increases from daytime (with well-mixed 330 atmospheric layer) to wintertime late evening hours (time of pollution confinement).
On comparing the temporal distribution of simulated BC concentration (presented only from the Constrained) with that of measured, it is seen that the pattern of simulated diurnal variability (shown for selected stations, refer to Figure 4) is consistent with that of measured. The diurnal variability comprising of BC concentration being relatively higher, by a factor of 2 to 5 in Constrained, during the late evening to early morning hours (2000-0500 LT) than that during daytime hours (1000-1600 LT) 335 at stations (except Nainital). Notably, this factor is equivalent to that obtained from bottomup simulations and also to that from observations (Surendran et al., 2013;Pani and Verma, 2014;Ram and Sarin, 2010;Nair et al., 2012;Dumka et al., 2010;Lipi and Kumar, 2014). The diurnal variability in BC surface concentration is mainly associated with the atmospheric mixing depth depending upon the stability characteristics of atmospheric layer linked with meteorology (Stull, 2012;Verma et al., 2013;Govardhan et al., 2015Govardhan et al., , 2019. It is worth noting that the specific feature observed in the temporal trend of BC concentration, comprising of peaked BC concentration during late afternoon hours (1500-1800 LT) at high altitude location, Nainital, unlike the temporal trend observed at plain locations (e.g., Kolkata, Kharagpur), conforms with measurements. This specific feature, as inferred from available studies (Dumka et al., 2010;Stull, 2012) is attributed to the deepening of atmospheric mixing depth during the late afternoon hours which flushes out pollutants, including BC to the high altitude locations from the valley (Dumka et al., 2010;Stull, 2012).

345
The bias in the simulated hourly distribution of winter monthly mean of BC concentration (refer to Figure 4) with respect to observation is, however, noted to be larger by 40%-60% during mid-night to early morning hours (0000-0500 LT) than that during daytime hours. A larger bias is attributable to that the simulated aerosol processes in CHIMERE are influenced by the simulated diurnal meteorology from the WRF. It is to be noted that the WRF simulated hourly mean temperature is found being 45%-60% lower than the observed, specifically during 0000-0500 LT (as mentioned in Section 3.1), which has implications on 350 the diurnal distribution of BC concentration. A better temporally resolved meteorological boundary condition in WRF (compared to 6-hourly from NCEP in the present study), aided with data assimilation at a fine temporal resolution (e.g., 1-hourly) using diurnal meteorological observations for India-based stations would potentially lead to simulate the observed magnitude of diurnal distribution of meteorological parameters more accurately, and, thereby reducing the bias in simulated diurnal BC distribution. The application of data assimilation in WRF using diurnal meteorological observations is under progress. Be-355 sides, it is also required to improve the representation of the factors for hourly disaggregation of the total emission of pollutants in CHIMERE  during late evening hours (1800-2200 LT), specifically for megacity (e.g., Kolkata) and urban location (e.g., Agra). This improvement is suggested taking into account the enhanced local hourly traffic emissions at these locations, hence a better representation of the factors. The results of the diurnal BC distribution including improved representation of local emissions (specifically for megacity and urban locations) in CHIMERE forced by assimilated diurnal 360 meteorological data will be presented in a future study.
We also provide an animation showing a representation of transport of BC concentration over the IGP as a supplement (please see BC-animation-1 in supplementary material). This animation shows the hourly monthly mean of surface BC concentration to highlight the diurnal cycle and its visualisation shows the diurnal evolution of the BC plume over the IGP. The BC surface plume is observed to be shrinking during daytime hours (1000 LT-1600 LT) and swelling-up during late evening till morning 365 hours (1800 LT-0600 LT) when it is visualised spreading towards the south (central India) and north (Himalayan side) and also from the upper/northern IGP towards the lower/eastern IGP. The diurnal feature of surface BC plume distribution thereby appears exhibiting the pollution breathing pattern by the IGP region.
Further, to statistically evaluate the simulated BC concentration from each of the five simulations with respect to observa-375 tions, we define the performance of the simulation considering the best, moderate, and poor efficiency based on their relative frequency to maintain the percentage bias in all-day (daytime) mean simulated BC concentration as about, respectively, ≤±25%, >±25% to ±50% and >±50% (refer to Table 3 (bottom)) corresponding to the observation data points under study.
This consideration leads to identify Constrained estimates delivering the best performance (percentage bias ≤±25%) among all simulations for most of the times, i.e., for 100% (100%) of the total data points corresponding to measured value at stations 380 under study. Estimates from P ku exhibit the best performance for about 50% (50%) of the total stations. These from  18 https://doi.org/10.5194/acp-2020-511 Preprint. Discussion started: 25 September 2020 c Author(s) 2020. CC BY 4.0 License.

Simulated wintertime BC-AOD with new BC emissions: Correlation analysis of variance
The spatial distribution of the monthly mean of AOD due to BC (BC-AOD) at 550 nm from simulations are presented in Figures 5a to 5e. The spatial pattern of BC-AOD distribution showing a large value over the IGP is consistent with the features of observed AOD from satellite retrievals (e.g., Verma et al., 2014). The value of BC-AOD distribution across the IGP from Constrained (0.04-0.1) is found to agree well with that from a recent study (0.05-0.1) -based on a designed constrained aerosol simulation approach inferred being delivering a good agreement between model estimates and observations of atmospheric aerosol species (Kumar et al., 2018;Santra et al., 2019). The BC-AOD from Constrained is also found to be matching consistently well (NMB: 11%) with absorption AOD (AAOD) from AERONET based observations at stations over the IGP and BC-AOD estimated at Kolkata from the configured aerosol model using in-situ ground-based observations for Kolkata, (Verma et al., 2013)) (refer to Figure 5f). Estimated BC-AOD from simulations-P ku, Smog, Cmip and Edgar, is lower in 400 magnitude by, respectively, 15%-30%, 30%-50%, 40%-60% and 50%-70% than the Constrained over most of the IGP.
The percentage BC-AOD fraction and BC mass fraction from Constrained (Figures 5g-h), are estimated by taking the ratio of BC-AOD to total AOD and that of BC concentration to the total submicronic aerosol concentration, respectively. The total AOD and submicronic aerosol concentration required for estimating fractional distribution are obtained from a previous study (as mentioned above), based on the designed constrained aerosol simulation approach (Kumar et al., 2018). The BC-

405
AOD fraction and BC mass fraction are about 10%-16% and 6%-10%, respectively, over most of the IGP. The estimated BC mass fraction in the present study is also seen to be in corroboration with values reported from wintertime measurements over the Indian region, e.g., noted as being 12% (wintertime average) of the total submicronic aerosol concentration over Kolkata, 4%-15% of the total aerosol concentration over Delhi and Kanpur, 3%-7% of PM 2.5 over Varanasi and Anantpur, including that over Kaashidhoo climate observatory in Maldives (Verma et al., 2013;Kumar et al., 2017;Tripathi et al., 2005;Ganguly 410 et al., 2006;Reddy et al., 2012;Satheesh et al., 1999). The location of hotspots for BC mass fraction (value > 16%), BC-AOD fraction (12%-16%), including that for BC-AOD (value > 0.08) is seen to overlap with that identified for BC surface concentration ( Figure 3e). It is also seen that the percentage fraction of BC-AOD, in general, is about twice larger than the BC mass fraction, indicating that even a low BC concentration in aerosol mass has the potential to contribute significantly to attenuation of solar radiation and thereby influence the regional radiation balance (which is examined in the next section).

415
To gain insight into the degree of association of the simulated BC burden with the BC emission strength, we utilise the five simulations to evaluate the correlation coefficient between the variation in emission strength and that in simulated BC-AOD for BC-AOD is moderate, but, however, is still stronger than BC concentration. Thereby indicating the potential influence to BC-AOD over the region from high rise BC emissions (corroborated by prevalence of open biomass burning emissions, Venkataraman et al. (2006)) and the elevated transport of BC aerosols as also inferred in a previous study (Verma et al., 2008).
3.4 Wintertime radiative perturbations due to BC aerosols: comparison with atmosphere eliminating BC Further, the wintertime SW radiative perturbation due to BC aerosols over the IGP (Figures 6a-c) is evaluated corresponding 430 to the layers of atmosphere (SUR, ATM, and TOA, refer to Section 2.3). We also compare the radiative perturbation due to BC with that estimated considering the atmosphere eliminating or without BC aerosols (Figures 6d-f) to evaluate the magnitude of radiative perturbation in the presence of BC aerosols. The positive value of radiative effect signifies warming due to BC aerosols and is vice versa for the negative value of the radiative effect. There is a reduction in the wintertime radiative flux 20 https://doi.org/10.5194/acp-2020-511 Preprint. Discussion started: 25 September 2020 c Author(s) 2020. CC BY 4.0 License.
due to BC at the SUR by −20 to −40 W m −2 (Figure 6a). The radiative warming (Figure 6c) due to BC aerosols at the ATM 435 (+30 to +50 W m −2 ) is estimated to be about 50%-70% larger than the cooling due to BC at the SUR. The magnitude of SUR cooling effect as noted due to BC aerosols is, however, found to be 10%-20% lower than that estimated considering the atmosphere eliminating BC aerosols (Figure 6d and 6g). Moreover, the magnitude of ATM radiative warming due to BC is seen to be larger by 2-3 times compared to the atmosphere without BC aerosols (Figure 6f and 6h). The radiative effect at the TOA due to BC aerosols (Figure 6b) is positive and, thereby, indicates a net radiative warming effect (+10 to +17 W m −2 ) over the 440 IGP during winter. In contrast, a cooling effect at TOA (−10 to −20 W m −2 ) is exhibited considering the atmosphere without BC aerosols (Figure 6e). It is also seen that the patch with the most substantial value (>15 W m −2 ) of the net radiative forcing due to BC is observed in and around megacities and is extended to the eastern coast. A comparison of the radiative effect due to BC from Constrained estimates with that from Smog estimates shows that bottom-up BC emissions (e.g., Smog-India) lead to a relatively lower wintertime radiative warming at ATM and TOA by 30%-50% than the constrained emissions over 445 most of the IGP and that by more than 80% over northern IGP (in and around Delhi). The comparison between the bottomup and the Constrained estimates, thus, indicates the potential underestimation of wintertime radiative perturbation due to BC aerosols over the IGP attributable to the low BC emission strength in the bottom-up BC emission database.
The uncertainty in estimated wintertime radiative perturbations in the present study is inferred to be within 40%. This estimation is based on taking into account NMB in simulated BC concentration (as presented in Section 3.2) and the model 450 variability (33%) in estimated DRF of BC based on the evaluation of twenty global aerosol models .

Conclusion
In the present study, wintertime radiative perturbation due to black carbon (BC) aerosols were examined over the Indo-Gangetic plain (IGP) evaluating the efficacy of the fine grid resolved (0. The meteorological forcing to CHIMERE as provided by the WRF regional meteorological model showed that the winter monthly mean of temperature, RH, and PBLH, as estimated from WRF simulations, resembled well (bias <±25%) the observations. However, a better temporally resolved meteorological boundary condition in WRF, aided with data assimilation at a fine temporal resolution (e.g., 1-hourly) using diurnal meteorological observations for India-based stations, would lead to simulate the observed magnitude of diurnal distribution of meteorological parameters more accurately, thereby reducing the 465 bias in simulated diurnal BC distribution, specifically, during mid-night to early morning hours (0000 LT to 0500 LT). Analysis of multi-simulations implementing new BC emission inventories in CHIMERE indicated the percentage deviation in simulated BC concentration attributed to emissions is, specifically, the lowest (20%-25%) for the low polluted location 475 (e.g., Nainital), with a, notably, amplified value (51%-56%) during the late evening to early morning hours (during the time of pollution confinement due to meteorology) for the megacities. A strong positive correlation between the variance in emissions and simulated BC mass concentration and BC optical depth from the five simulations was noticed over most of the IGP region where a large BC pollution load is obtained; thereby manifesting the sensitivity of simulated BC concentration and optical depth towards the change in input emission strength.

480
A strong association between modelled and measured monthly mean BC concentration for stations under study corresponding to each of the five simulations was noticed. The simulated spatial and temporal pattern of BC surface concentration was consistent with observations. Nevertheless, the efficacy to simulate the magnitude of observed wintertime BC distribution was found to be moderate to poor for bottomup estimates. Estimates from the Constrained could simulate the observed all-day (daytime) winter monthly mean of BC concentration with the lowest percentage bias ( ≤±25%) among five simulations for 485 each of the data points under study. An overall comparison of the Constrained and bottomup estimates with measurements, indicated the low BC emission strength as the primary reason for the underestimation of BC concentration from the bottomup.
Analysis of radiative perturbations due to BC aerosols showed that wintertime BC aerosol over the IGP enhances the atmospheric warming by 2-3 times more, and, reduces the surface cooling by 10%-20% lesser than considering the atmosphere eliminating BC aerosols. The BC induced net warming effect at the top of the atmosphere (TOA) from the Constrained was 490 estimated as 10-15 W m −2 over most of the IGP, in contrast to a net cooling at the TOA considering the atmosphere without BC. The radiative perturbation was spotted being spatially the largest in and around megacities (Kolkata and Delhi) and extended to the eastern coast. These were assessed to be about 30%-50% lower from the bottomup than the Constrained over most of the IGP.
The present study showed that an adequate BC emission strength and a meteorological forcing in a state-of-the-art chemical 495 transport model at a fine grid resolution led to successfully simulate the wintertime BC distribution (surface concentration and BC-AOD) over the IGP, unlike previous studies (as mentioned in Section 1). We believe this distribution provides a reasonable understanding of wintertime radiative perturbations due to BC aerosols with an identification of their hotspots over the IGP.
The wintertime radiative perturbation due to BC aerosols as simulated in the present study is further being utilized to evaluate the potential response on temperature, air quality, and regional climate over the IGP, outcome from these evaluations will be 500 presented in a future study. The present study is also further extended to evaluate the inter-seasonal BC distribution and asso-22 https://doi.org/10.5194/acp-2020-511 Preprint. Discussion started: 25 September 2020 c Author(s) 2020. CC BY 4.0 License. ciated radiative impacts over the Indian subcontinent with their implications on the southwest monsoon rainfall.
Data availability. The data in this study are available from the corresponding author upon request (shubha@iitkgp.ac.in).

505
Author contributions. SG conducted the BC transport simulations and radiative transfer simulations, evaluation, and validation of the model estimates, including the statistical analyses, and participated with SV in synthesizing and analyzing the results. SV planned and coordinated the study. SG and SV wrote the paper. JK and LM contributed to the writing and analysis of results. LM also advised for the technicality of the CHIMERE model configuration.

510
Competing interests. The authors declare that they have no conflict of interest.