Radiative Effect of Springtime Biomass-Burning Aerosols over Northern Indochina during 7-SEAS / BASELInE 2013 Campaign

The direct aerosol radiative effects of biomass-burning (BB) aerosols over northern Indochina were estimated by using aerosol properties (physical, chemical, and optical) along with the vertical profile measurements from ground-based measurements with integration of an optical and a radiative transfer model during the Seven South East Asian Studies/ Biomass-Burning Aerosols & Stratocumulus Environment: Lifecycles & Interactions Experiment (7-SEAS/BASELInE) conducted in spring 2013. Cluster analysis of backward trajectories showed the air masses arriving at mountainous background site (Doi Ang Khang; 19.93°N, 99.05°E, 1536 m above mean sea level) in northern Indochina, mainly from near-source inland BB activities and being confined in the planetary boundary layer. The PM10 and black carbon (BC) mass were 87 ± 28 and 7 ± 2 μg m, respectively. The aerosol optical depth (AOD500) was found to be 0.26–1.13 (0.71 ± 0.24). Finer (fine mode fraction ≈0.95, angstrom-exponent at 440–870 nm ≈1.77) and significantly absorbing aerosols (single­scattering albedo ≈0.89, asymmetry-parameter ≈0.67, and absorption AOD ≈0.1 at 440 nm) dominated over this region. BB aerosols (water soluble and BC) were the main contributor to the aerosol radiative forcing (ARF), while others (water insoluble, sea salt and mineral dust) were negligible mainly due to their low extinction efficiency. BC contributed only 6% to the surface aerosol mass but its contribution to AOD was 12% (2 times higher). The overall mean ARF was – 8.0 and –31.4 W m at top-of-atmosphere (TOA) and at the surface (SFC), respectively. Likely, ARF due to BC was +10.7 and –18.1 W m at TOA and SFC, respectively. BC imposed the heating rate of +1.4 K d within the atmosphere and highlighting its pivotal role in modifying the radiation budget. We propose that to upgrade our knowledge on BB aerosol radiative effects in BB source region, the long-term and extensive field measurements are needed.


INTRODUCTION
Biomass-burning (BB) aerosols are the significant contributor to the global aerosol loading and radiation budgets (Crutzen and Andreae, 1990;Penner et al., 1992).The emissions (i.e., gas and particle) from tropical BB influence the physicochemical and optical properties of the atmosphere (Andreae et al., 1988;Crutzen and Andreae, 1990; Andreae phase, environmental conditions, and atmospheric aging (Vakkari et al., 2014 and references therein).BB aerosols are mainly generated by agricultural burning in tropical regions (Andreae and Merlet, 2001); Africa, the Amazon basin and southern Asia are major source regions (Johnson et al., 2008).
The radiative effects of BB particles have remained poorly quantified since their optical and cloud activating properties are very diverse (Vakkari et al., 2014).The annual global mean radiative forcing at the top-of-atmosphere (TOA) was estimated as -0.2 W m -2 (Intergovernmental Panel on Climate Change (IPCC), 2001) for BB aerosols and with a large uncertainty.Model simulated direct radiative forcing of BB aerosols was found to be +0.03 ± 0.12 W m -2 globally (IPCC, 2007).IPCC (2014) reported the radiative forcing of BB aerosols as -0.20 (range; -0.60 to -0.07) W m -2 in the Second and Third Assessment Reports, +0.03 (-0.09 to +0.15) W m -2 in the Fourth Assessment Report, and 0.0 (-0.20 to +0.20) W m -2 in the Fifth Assessment Report.There were a few campaigns conducted to investigate the characteristics of regional BB emitted aerosols and their climatic effect, such as Southern African Regional Science Initiative (SAFARI) (Andreae et al., 1996), Smoke, Clouds and Radiation-Brazil (SCAR-B) (Kaufman et al., 1998), Indian Ocean Experiment (INDOEX) (Ramanathan et al., 2001), Transport and Chemical Evolution over the Pacific (TRACE-P) (Jacob et al., 2003), and Biomass Burning Aerosols in Southeast Asia: Smoke Impact Assessment (BASE-ASIA) (Tsay et al., 2013).Recently, Pani et al. (2016a) reported the enhancement of surface cooling and free troposphere warming by the transported Indochina BB plumes over the northern South China Sea during the Seven South East Asian Studies (7-SEAS)/Dongsha Experiment.
The 7-SEAS/BASELInE (Biomass-Burning Aerosols & Stratocumulus Environment: Lifecycles & Interactions Experiment) was conducted in spring 2013-2015 over northern Southeast Asia to further explore numerous key atmospheric processes and impacts of BB on radiation budgets of surface/atmosphere (Lin et al., 2013).BB activities in the form of wildland forest fires and agricultural burning are very pronounced in the northern Indochina regions mainly during dry/spring season from late February to mid-April (i.e., Gautam et al., 2013;Hsiao et al., 2016).Aerosol concentrations during this season are typically at peak associated with BB activity and contribute significantly to the regional emissions (Carmichael et al., 2003;Janjai et al., 2009;Streets et al., 2009).The impact of BB on air quality over Southeast Asia was evaluated within the framework of BASE-ASIA campaign by using chemical transport modelling (Fu et al., 2012;Huang et al., 2013).Moreover, some recent studies also emphasized the BB emissions from Indochina region and explained their impact on regional air quality, atmospheric radiation, hydrological cycle, and climate (e.g., Lin et al., 2013;Tsay et al., 2013;Lin et al., 2014;Tsay et al., 2016) by using in-situ measurements (e.g., Chuang et al., 2013b;Wang et al., 2015;Chuang et al., 2016;Hsiao et al., 2016;Khamkaew et al., 2016;Lee et al., 2016a;Pantina et al., 2016;Sayer et al., 2016), satellite observations (e.g., Hsu et al., 2003;Huang et al., 2011;Campbell et al., 2013;Lee et al., 2016b), remote sensing (Gautam et al., 2013;Reid et al., 2013), and modeling studies (Fu et al., 2012;Loftus et al., 2016) within the 7-SEAS framework.Wang et al. (2015) investigated the presence of strongly absorbing (single-scattering albedo ≈0.88 and absorption angstrom exponent ≈1.5) carbonaceous aerosols within the planetary boundary layer (PBL) over northern Indochina in 2014 spring.
The main goal of this work is to assess and quantify the shortwave direct radiative effects over the northern Indochina due to excess loading of near-source BB aerosols during 7-SEAS/BASELInE 2013 campaign (March 1-April 15).Contributions from variety of source regions and aerosol compositions to the total radiative forcing were carefully examined by matching up observational and modeling results.This study provides an better understating on linkage between aerosol surface mass concentration, optical depth and radiative effects of different aerosol components, and can be utilized effectively further in the context of regional and global climate studies.

Site Description
The study location, Doi Ang Khang meteorology Station [DAK; 19.93°N, 99.05°E, 1536 m above mean sea level (a.m.s.l)], is a high mountain vegetation field in northern Indochina region.The site is near Myanmar border and about 125 km north of Chiang Mai City.It is situated on top of the mountain surrounded by forest and some agricultural fields.The mean synoptic surface winds during spring (March-April) over the Chiang Mai Province, Thailand, where DAK is located, were northwest/west/southwesterly.

Source Region Identification: Back Trajectory Analysis
Five-day back trajectories ending at ground level of DAK at 06 UTC daily study period were calculated using HYSPLIT (Hybrid Single-Particle Lagrangian Integrated Trajectory, Version 4.9) model developed by NOAA Air Resources Laboratory (Draxler and Rolph, 2003) to allocate them in to groups of similar source origin, transport pathways and path length.The meteorological data was obtained from the NCEP Global Data Assimilation System (GDAS) with one-degree resolution and used to initialize HYSPLIT.Three clusters (i.e., Clusters 1-3) were identified over DAK.Individual trajectories representing each cluster are shown in Fig. 1.Cluster 1 (No. of days, N: 15) accounts for 48% of total trajectories and represents the path, originating from Bay of Bengal (BoB) and mostly locally near the source.Trajectories in Cluster 2 (N: 13) accounts for 43%, mainly originated from East-coast India and BoB.Cluster 3 (N: 3) accounts for only 9% of total trajectories and represents the longest path, originating from semi-arid regions over northwest India (NWI) and Indo-Gangetic Plain (IGP) and passing through the coastal areas of Indian subcontinent.Cluster 1 consists of trajectory confined below 2 km, while Cluster 2 and 3 exhibited greater variability with respect to origin heights.All the clusters were mainly influenced by the near-source BB activity.

Aerosol Measurements
Measurements of elemental carbon (EC) concentration and total suspended particulate with sizes ≤ 10 µm (PM 10 ) were being carried out at the study location during 01 March-8 April 2013.PM 10 was collected by using Teflon filters (R2PJ047 TefloTM, PALL Life Sciences, Inc., Ann Arbor, MI, USA) and quartz fiber filters (Tissuquartz 2500 QAT-UP, 20 cm × 25 cm, Pall Corporation, USA) for 24 hours (08:00 a.m. to 08:00 a.m.local time).The collected filters were refrigerated below 4°C before shipping and transported to a laboratory in Taiwan for filter weighting and subsequent analysis.Further water soluble ions (Na + , NH 4 + , K + , Mg 2+ , Ca 2+ , Cl -, NO 3 -, and SO 4 2- ) were analyzed from the Teflon filter sample by an ion chromatograph (DX-120 for cations and DX-1000 for anions, Dionex Co., Inc., Sunnyvale, CA, USA).The quartz fiber filters were analyzed for EC and organic carbon (OC) fractions by using a DRI Model 2001A OC/EC Carbon Analyzer (Atmoslytic Inc., Calabasas, CA, USA) by the thermal optical reflectance method.More details of the sampling procedure and analytical methods were presented in Chuang et al. (2013a, b).Black carbon (BC) is often used interchangeably with EC (Watson et al., 2005) and we further used BC throughout in this manuscript.
The aerosol optical properties i.e., aerosol optical depth (AOD), single scattering albedo (SSA), asymmetry parameter (AP), and absorption aerosol optical depth (AAOD) were being obtained from AERONET (AErosol RObotic NETwork) Version 2 Level 2 direct-sun and inversion products.The estimated uncertainty in AOD was 0.01-0.02and was primarily due to the calibration (Eck et al., 1999).Direct-sun data are cloud screened and quality assured based on Smirnov et al. (2000), while AERONET inversion products were quality assured based on Dubovik et al. (2000) and Holben et al. (2006).The uncertainty in the retrieved inversion product was estimated to be ± 0.03 (Holben et al., 2006).

Methodology and Criteria of the Study
An aerosol optical model (Optical Properties of Aerosols and Clouds (OPAC 3.1) optical model (Hess et al., 1998)) was adopted to calculate aerosol properties by using data from in-situ aerosol measurements as input.The OPAC model can easily incorporate with user defined aerosol mixtures, aerosol vertical profile, wavelength, and relative humidity (RH).The different aerosol components (water insoluble, water soluble, BC, sea salt and mineral dust) were used in the OPAC model based on aerosol source regions and transport pathways over the northern Indochina.Aerosol components used in this study were assumed to be spherical and externally mixed (Koepke et al., 1997, Hess et al., 1998).Water soluble and sea salt aerosols are predominant scatterers, water insoluble and mineral dust aerosols are absorbers, and BC is the single largest absorber of solar radiation.Atmospheric boundary layer height (the daytime values) obtained from Micro pulse lidar aerosol extinction profiles were used.Detail about the Micro pulse lidar was available in Wang et al. (2015).The OPAC works at eight different RH conditions in the range of 0-99% (Hess et al., 1998).The measured daily mean value of RH was 50% during the study period and thus used in this study.
Fig. 2 shows the flow diagram describing the methodology for the estimation of aerosol optical properties (AOD, SSA, and AP) in the entire short wave range (0.25-0.4 µm) and direct aerosol radiative effects using measured aerosol parameters (i.e., surface PM 10 and BC mass concentrations) in the OPAC optical model and a radiative transfer model.The number densities were obtained from the respective mass concentration of individual aerosol components.The number density of different aerosol components (i.e., water insoluble, water soluble, BC, sea salt, and mineral dust) were varied in an iterative process until all the following criteria were satisfied: (1) PM 10 aerosol surface mass concentration was the same as obtained from observations; (2) BC mass concentration in the OPAC was the same (within ±1) as that of the chemically measured BC mass; (3) the root-meansquare error between the measured and modeled AOD spectra was < 0.001; (4) model-derived AAOD (sum of AOD due to water insoluble, BC and mineral dust) was matched with AAOD obtained from observations; (5) model-derived SSA was matched with SSA obtained from observations (6) model-derived AP was matched with AP obtained from observations.Furthermore, number density of each aerosol components were used in the OPAC model separately to derive the AOD corresponding to each of the component and their relative contribution to total AOD were also estimated.
By adopting the above methodology, the estimated aerosol optical properties (i.e., AOD, AAOD, SSA, and AP) can be found well matched with the observations as shown in Table 1.Upon the data process, we ensured the constituency between the theoretical and observational results which can further improve the accuracy in radiative transfer calculations.

Aerosol Radiative Effect Estimation
Radiative transfer estimations in the shortwave (0.25-4 µm; 38 bands) range were made by using the SBDART (Santa Barbara DISORT Atmospheric Radiative Transfer) radiative transfer model (Ricchiazzi et al., 1998).The radiative transfer equations in this model are numerically integrated with Discreet Ordinate Radiative Transfer (DISORT) radiative transfer module (Stamnes et al., 1988).This approach provides a numerically stable algorithm to solve the equations for plane-parallel and non-isothermal radiative transfer in a vertically inhomogeneous atmosphere.
The algorithm computes the radiation fluxes for each wave length region in a line-by-line mode with a fixed spectral resolution of 0.005 µm in the shortwave region (Ricchiazzi et al., 1998).The radiative transfer calculations were performed using 65 atmospheric layers and 8 radiation streams (8 zenith angles and 8 azimuthal modes) to estimate the upward and downward fluxes at the surface and TOA for 24 hour period (at the interval of 1 hour), with and without aerosols.
The fluxes estimated by SBDART model were found to be within 2% of direct and diffuse irradiance measurements (Michalsky et al., 2006).This methodology has been widely used in solving the radiative transfer equations and available in literature (e.g., Satheesh, 2002Satheesh, , 2010;;Ramachandran et al., 2012;Pani, 2013;Srivastava and Ramachandran, 2013).Aerosol radiative forcing (ARF, W m -2 ) at the surface (ARF SFC ) and at the TOA (ARF TOA ) was calculated as the change in the net (difference between downward and upward) fluxes with and without aerosol conditions as ARF SFC = (NetFlux) with aerosol at SFC -(NetFlux) without aerosol at SFC (1) ARF TOA = (NetFlux) with aerosol at TOA -(NetFlux) without aerosol at TOA (2) ARF ATM , the amount of energy trapped by the aerosols present in the atmosphere, was estimated as the difference between the ARF TOA and ARF SFC .
Spectral values of aerosol optical properties (AOD, SSA, and AP) at wavelengths of 0.25-4 µm obtained from OPAC optical model for each cluster measurement were incorporated in SBDART to estimate ARF in clear sky.ARF corresponding to each aerosol components for each cluster were also calculated by taking the difference between ARF Total (Eqs.( 1) and ( 2)) and the ARF in absence of that particular component.In addition to aerosol optical properties, the atmospheric profiles of temperature, pressure, columnar ozone, water vapor and surface reflectance were also required in SBDART model to compute ARF.The standard tropical atmospheric profiles of temperature and pressure, as described in Ricchiazzi et al. (1998), were used in this study.The tropical climate is characterized by the average water vapor (4.1 g cm -2 ) and the average ozone level (253 Dobson Unit).SBDART model also includes the default values for the trace gases CO 2 (360 parts per million; ppm), CH 4 (1.74 ppm), and N 2 O (0.32 ppm).Surface reflectance is a crucial factor in determining the strength of ARF.Gautam et al. (2013) reported the composite mean map of CERES (Clouds and the Earth's Radiant Energy System) derived broadband surface reflectance during February-April over Indochina region in the 0.12-0.16range indicating low surface albedo over the largely vegetated regions.SBDART uses the default value of surface reflectance as 0.127 at 555 nm for vegetation (Manual of Remote Sensing, 1975) surface type.Moderate Resolution Imaging Spectroradiometer (MODIS) derived land surface products (Moody et al., 2005) like black-sky albedo (direct reflectance) and white-sky albedo (bi-hemispherical reflectance) were used to model the surface reflectance (e.g., Satheesh et al., 2010;Campbell et al., 2016) characteristics required for the radiative transfer calculation.MODIS surface reflectance data were obtained at seven different wavelengths centered at 469, 555, 645, 858.5, 1240, 1640, and 2130 nm and used to reproduce the surface reflectance in the entire short wavelength spectrum using a combination of water, sand, and vegetation surface types (e.g., Ramachandran et al., 2012;Pani, 2013;Verma et al., 2013).The spectral mean surface reflectance over DAK was obtained as 0.123.The combination of water (3%), sand (0%), and vegetation (97%) was made in appropriate proportions such that the resultant spectrum matched the MODIS derived surface reflectance.
The aerosol radiative efficiency (ARE), change in ARF per unit AOD at 500 nm, was also estimated.The ARF ATM in the shortwave region gets converted into heat and the heating rate at each layer (∆P) is given by where (∆T/∂t) is the heating rate (Kelvin per day; K d -1 ), g is the acceleration due to gravity (9.8 m s -1 ), C p is the specific heat capacity of air at constant pressure (1006 J kg -1 K -1 ), and P is the atmospheric pressure difference (Liou, 1980).∆P was considered as 200 hPa (equal to the pressure difference between ground and 3.5 km) in this study.

Aerosol Vertical Profiles
Aerosol vertical profiles from lidar observations were analyzed to determine the aerosol vertical distribution over DAK and used in this study.The mean aerosol extinction profile for the entire study period is shown in Fig. 3.The height of aerosol vertical distribution reached 5 km a.m.s.l was attributed to PBL dynamic over mountain area and transported aerosol plume (Wang et al., 2015).The peak of the mean aerosol extinction (0.37 km -1 ) was observed between 2.5 km to 3.5 km a.m.s.l indicating the presence of lofted aerosol layer.

Measured Properties of Aerosol: Day-to-Day Aerosol Variability
The PM 10 and BC mass concentrations over DAK were ranged between 25-135 µg m -3 (mean ± one standard deviation; 87 ± 28 µg m -3 ) and 5-10 µg m -3 (7 ± 2 µg m -3 ), respectively during the entire study period.The measured PM 10 and BC mass concentrations were strongly correlated (R 2 = 0.8) over DAK and suggested the common source origins.Fig. 4 shows the daily variations in various aerosol optical properties obtained from the AERONET observations.The columnar integrated AOD represents the extinction of incoming solar radiation by aerosols and its magnitude is directly proportional to the loading of aerosols in the total atmospheric column.The columnar AOD 500 (Fig. 4(a)) was found to be from 0.26 to 1.13 (0.71 ± 0.24) during the study period.
The spectral dependence of the AOD was used to compute the Angstrom exponent (AE).AE is also a qualitative indicator for the size of aerosol particles (Pani and Verma, 2014).The variations in the magnitude of AE 440/870 were shown in Fig. 4(b), which was found to vary from 1.64 to 1.95 (with a mean 1.77 ± 0.07).The results suggested the relative dominance of smaller aerosols over the region, mainly from the BB emissions.Fig. 4(c) shows daily average fine mode fraction (FMF), observed at 500 nm wavelength during the period of study.Predominant contribution from fine mode aerosol was clearly observed and the FMF 500 ranged between 0.88-0.98(with a mean 0.95 ± 0.02).The  The SSA is a very common and crucial measure of the relative contribution of absorption to total extinction and also a key parameter in climatic assessment of effects of aerosols (Jacobson, 2000;Dubovik et al., 2002).It is the ratio of scattering coefficient to the extinction coefficient (the sum of scattering and absorption coefficient).SSA 440 (Fig. 4(d)) varied in between 0.85-0.94(with a mean 0.89 ± 0.02) over DAK and suggested the dominance of BB aerosols.Wang et al. (2015) reported the presence of strongly absorbing (SSA 440 ≈0.88) smoke particles over northern Indochina in 2014 spring.It was also observed that SSA decreases with wavelength (SSA 440 ≈0.89) and SSA 1020 ≈0.84).Higher SSA at shorter wavelength indicates the dominance of absorbing aerosols mainly attributed to nearsource BB emissions.Long-range transport of dust/pollutants from NWI/IGP to the DAK may also play a role here.The SSA 440 over DAK was very close to the value observed in Senanga, Zambia (0.86), Mwinilunga, Zambia (0.88), Skukuza, South Africa (0.88), Etosha Pan, Namibia (0.90), Inhaca Island, Mozambique (0.88) and Bethlehem, South Africa (0.90) during SAFARI 2000 dry season (August-September) campaign in southern Africa (Eck et al., 2003).The SCAR-B experiment in Brazil measured SSA as low as 0.6 (at 550 nm wavelength) for fresh smoke (Reid et al., 1998) and increasing up to 0.91 as the smoke aged.Measurements of Canadian smoke during the Boreal Ecosystem-Atmosphere Study (BOREAS) found SSA in range from 0.70 for fresh smoke to 0.98 for aged smoke (Miller and O'Neill, 1997;Li and Kou, 1998).
The AP is the measure of angular distribution of light scattered by the aerosol particles and gives information on aerosol size and scattering properties (Andrews et al., 2006).It is another key parameter which regulates aerosol radiative forcing.However ARF is less sensitive to changes in AP when compared to AOD and SSA (Srivastava and Ramachandran, 2013).The daily variations in AP 440 (Fig. 4(e)) was opposite to the variations in AE (with R 2 = 0.55).AP 440 varied in between 0.63-0.70(with a mean 0.67 ± 0.02) over DAK.AAOD, a measure of the column aerosol light absorption, is expressed as AOD × (1 -SSA).AAOD 440 (Fig. 4(f)) was varied in between 0.05-0.21(with a mean 0.10 ± 0.04) during the entire study period and showed a moderate absorption over the study site.The highest AAOD (0.21) was observed on March 22 (SSA ≈ 0.86) and the lowest AAOD (0.05) was on April 4 (SSA ≈ 0.94).
Compared with results from Sayer et al. (2014) and Wang et al. (2015), higher values and similar trend in AE 440/870 was seen for BB aerosols over Amazon (≈1.9-2.0)basin, Savanna (≈1.9-2.0)forest, and Australia (≈1.9) followed by Indochina (≈1.7-1.8) and Boreal forests (≈1.4-1.7) of North America and Russia.The SSA440 (as shown in Table 2) for smoke varied notably (ranging from 0.86 to 0.95) over distinct BB regions in the world.SSA 440 was marked the lowest over Indochina (0.86-0.91) followed by Savanna (0.87-0.9),Australia (0.88), Amazon (0.89-0.92), and Boreal (0.94-0.95).The boreal forest demonstrated the weakest aerosol absorption.Indochina smoke exhibited the strongest absorption (lowest SSA value) among Savanna, Amazon, northern Australia and Boreal smoke.Moreover a little stronger absorption (SSA 440 ≈ 0.86) was observed over the Thai-Laos valley (Maesoon and Luang Namtha) than that of other Indochina (Mukdahan, Sonla, and DAK) regions (SSA 440 of 0.89-0.91 or ~0.9).Aerosol absorption (SSA 440 ≈ 0.89, AP 440 ≈ 0.67 and AE 440/870 ≈ 1.7) was observed closely similar during spring of 2014 and 2013 over DAK.As stated in the comparison Table 2, we can conclude that the optical properties of BB aerosols over northern Indochina are distinct from those over other BB regions by taking everything into account.

Cluster Variability
Cluster-wise AOD 500 showed a strong variation with the trend similar to PM 10 mass concentrations (Table 1) over the study region.The AOD (PM 10 ) was 0.64 (73 µg m -3 ), 0.76 (98 µg m -3 ), and 0.81 (104 µg m -3 ) for Clusters 1, 2, and 3 respectively.The highest value AOD (0.81) was observed for Cluster 3, with the highest SSA 440 (0.9) due to the relatively more influence of sea salt from BoB and mineral dust from NWI/IGP region to the study site.AAOD 440 also showed a distinct cluster variation and the trend was similar to the mean BC mass.The AAOD 440 (BC) was 0.08 (4.4 µg m -3 ), 0.1 (5.5 µg m -3 ), and 0.09 (5.3 µg m -3 ) for Clusters 1, 2, and 3 respectively mainly due to the presence of absorbing carbonaceous aerosols over DAK.The highest AAOD 440 was recorded for Cluster 2 with the highest BC (5.5 µg m -3 ) and AE 440/870 (1.79) values suggesting the presence of more fresh smoke particles in Cluster 2.

Modeled Properties of Aerosol: Relative Contribution of Different Components to Total Mass
Magnitude of net aerosol radiative forcing depends on both absolute mass concentration of individual components like BC and its mass fraction in total aerosol loading (Petzold et al., 1997).Estimated cluster mean relative contribution of different aerosol components to total aerosol mass (Figs.5(a), 5(b) and 5(c)) concentration reveals a larger variability in water soluble (77-83%), mineral dust (0-5%) components than that in water insoluble (4-7%), sea salt (4-7%), and BC (5-6%).The highest contribution of water soluble (which is optically scattering in nature and made up of various kinds of sulfates, nitrates, organics, etc.) aerosols to the total aerosol mass concentration was estimated and mainly attributed to frequent BB activities in spring.BB aerosols grow by coagulation, condensation and other gas-to-particle  exchanges (Jhonson et al, 2008) with time.Aerosol ageing increases the scattering coefficient and single scattering albedo (Reid et al., 1998;Abel et al., 2003).It is worthy to be noted that, since an exact match was obtained between the observations and the model estimated optical properties, so any impact due to internal mixes (core-shell or homogeneous mixing) of aerosols on the aerosol optical properties and radiative estimations were non-significant (Ramachandran et al., 2012).Water soluble contribution was the highest (83%) for Cluster 1 mainly due to the influence of local BB aerosols, followed by for clusters 2 (77%) and 3 (79%) due to the contribution of sea salt and mineral dust aerosols.Water insoluble (mainly suspended soil/road dust particles with a certain amount of organics matter) aerosols contributed the highest (7%) for Cluster 1. BC aerosols contributed the highest (6%) to the total aerosol mass for Cluster 1, mainly because of local and near-source BB activities.Sea salt aerosols contributed the lowest (4%) in Cluster 1, while relatively higher (7%) for Clusters 2 and 3, as aerosols were coming over BoB to the study site.No dust contribution was estimated for Cluster 1 but 5% for both the Clusters 2 and 3 as the influence of aerosols from NWI and IGP region of Indian subcontinent.SSA was relatively lower (0.89) for Clusters 1 and 2 owing to the dominance of fine mode absorbing aerosols.SSA was the highest (0.9) for Cluster 3 due to the increase in the abundance of sea salt aerosols which are more scattering in nature.

Relative Contribution of Different Components to Total AOD
AOD 500 due to different aerosol components (i.e., water insoluble, water soluble, BC, sea salt and mineral dust) for Clusters 1-3 were estimated in the OPAC model and presented in Table 1.Furthermore their relative contributions to total AOD (Figs. 4(d), 4(e), 4(f)) were also estimated.Cluster mean aerosol component AOD exhibits a low variability compared to their mass concentrations.The AOD due to water soluble aerosols (component AOD: 0.55-0.68;relative contribution to the total AOD: 83-84%) was predominant over DAK, followed by that due to BC (0.08-0.09; 11-13%), sea salt (0.02-0.03; 2-4%), mineral dust (0.004-0.005; 0.55-0.58%),and water insoluble (0.002-0.003; 0.3-0.5%)aerosols during the study period.Water soluble aerosols and BC contributed as high as 97% to total AOD for Cluster 1, followed by Clusters 2 and 3 (95%).Sea salt contribution to AOD was maximum 4% for Clusters 2 and 3 and only 2% for Cluster 1.The contribution of mineral dust was 5% to the total aerosol mass for Clusters 2 and 3, but its contribution to the total AOD was very little (0.6%).Negligible contribution of water insoluble component to total AOD was observed.The BC only contributed 5-6% to the aerosol mass loading, but its contribution to the total AOD was 11-13% (≈2 times higher).This showed the importance of BC aerosol mass fraction on the optical and radiative effect over near source BB region.Though the BC aerosol was present only in very small fractions, their effect on short wave radiation was significant.

Clear-Sky Shortwave Direct Aerosol Radiative Effects
The magnitude of estimated shortwave total ARF (ARF TOA , ARF SFC , and ARF ATM ) estimated from aerosol optical properties for each clusters is given in Table 3.The cluster variation in ARF over DAK was not so significant because of the similar contribution from the water soluble and BC aerosols.ARF TOA and ARF SFC for Clusters 1-3 were ranged between -7.1 to -9.1 and -28.3 to -33.5 W m -2 , respectively.Variations in ARF were mainly due to the variations in AOD and SSA (Ramachandran et al., 2012;Pani et al. 2016b).Cluster 3 showed the highest ARF (ARF TOA of -9.1 W m -2 and ARF SFC of -33.5 W m -2 ) due to the highest AOD 500 (0.81) and SSA 440 (0.9) as well.The model simulated monthly mean ARF TOA of BB aerosols in the southern African region during SAFARI-2000 was found to be -4.3W m -2 (Myhre et al. 2003).The mean ARF TOA during the BB season (August and September) for the 10-yr (2000 to 2009) was -5.6 ± 1.7 W m -2 over Amazonia (Sena et al., 2013).The overall mean ARF TOA value (-8.0 W m -2 ; see Table 3) over Indochina was 1.4 times higher than that over Amazonia.Table 4 shows that the mean daily ARF TOA results over the Amazon Basin were compatible with this present study.ARF ATM (+24.7 W m -2 ) was the highest for the Cluster 2 due to the highest BC (5.5 µg m -3 ) and AAOD 440 (0.1) as well.
The ARF (ARF TOA , ARF SFC , and ARF ATM ) values due to individual aerosol components for all clusters are shown in Fig. 6.ARF SFC values were negatives for each component.ARF SFC was the highest for water soluble (-17.5 to -21.2 W m -2 ) followed by BC (-17.6 to -18.7 W m -2 ), sea salt (-0.3 to -0.7 W m -2 ), water insoluble (-0.2 to -0.3 W m -2 ), and mineral dust (-0.2 to -0.3 W m -2 ).Absorbing aerosols (water insoluble and BC) only showed positive values at TOA, whereas scattering aerosols (water soluble and sea salt) showed negative values.BC aerosols showed warming of +10.2 to +11.1 W m -2 at TOA and water soluble (-9.6 to -11.5 W m -2 ) balanced the effect.ARF ATM was +27.8 to +29.8 W m -2 due to BC aerosols along with negligible contribution from water insoluble (+0.3 to +0.4 W m -2 ) and mineral dust (+0.2 to +0.3 W m -2 ).BB aerosols (mainly water soluble and BC) were the main contributor to the ARF values, while others were negligible mainly due to their low extinction efficiency.
ARE is a useful parameter for quantifying and comparing aerosol radiative effects at different places under different aerosol conditions (Satheesh and Ramanathan, 2000).The mean values of total ARE TOA (ARE SFC ) were -11.1 (-44.2),-10.4 (-42.8), and -11.2 (-41.3)W m -2 for Clusters 1, 2, and 3, respectively (Table 3).ARE TOA values were -15.7 ± 2.4 and -9.3 ± 1.7 W m -2 over forest and savannah-like vegetation, respectively over Amazonia (Sena et al., 2013).were much compatible with our results.Also it is very interesting to note that ARE TOA (-10.5 W m -2 ) over Amazon in 2005 (Procopio et al., 2004) was nearly close with our result (-10.9W m -2 ) over Indochina.Furthermore, value of the cluster mean atmospheric heating rate (estimated by Eq. ( 3)) was ranged between 1-1.2 and 1.4-1.5 K d -1 due to total and BC aerosols (Table 3), respectively.This present study showed that radiative impacts of aerosols were more dependent on AOD and SSA rather than their surface mass concentrations.A sensitivity analysis was also carried out to understand the uncertainty in the ARF estimated by the SBDART model.The uncertainty in ARF was calculated by constantly running the model by varying the value of AOD and SSA while keeping all other input parameters unchanged.It was estimated that an error of 0.02 in AOD can result the uncertainty of 2-4% and 5-8% in ARF TOA and ARF SFC , respectively.Likely, an error of 0.01 in SSA can result the uncertainty of 10-15% and 10% in ARF TOA and ARF SFC , respectively.However, the uncertainty value reported in this study is consistent with some earlier reported literature (e.g., Ramachandran et al., 2012;Srivastava and Ramachandran, 2012).Also our results revealed that the ARF is mainly governed by the BB aerosols, which contributed up to ≈50% to the ARF SFC and significantly contributing to solar dimming.The contribution of BC aerosols towards ARF ATM was up to ≈75% during the BB season over Indochina.The large negative ARF SFC and positive ARF ATM estimated in this study can have a significant influence on atmospheric stability and regional climate.Our measurements were carried out over mountain top and there is valley with scattering agriculture.Moreover, northern Indochina is a complex terrain and has a high prevalence of forest-conversion fires.So the radiative impact will be larger when we will consider more topography effects, and it should be an appealing topic for future study.

CONCLUSIONS
In this study, we evaluated aerosol radiative impacts of BB aerosols over near-source BB emission region in northern Indochina by using ground-based physical, chemical, and optical data obtained from the 7-SEAS/BASELInE in 2013 as well as by using aerosol optical and radiative transfer models.We analyzed data from in-situ measurements, and trajectory simulations, to explore the relationships between the origins and transport of atmospheric pollutants over northern Indochina.Cluster analysis of back trajectories revealed the mainly contribution from near-source origins for Cluster 1 followed by Cluster 2 and Cluster 3 (influence of mineral dust from Indian subcontinent or NWI/IGP).The mean PM 10 and EC mass concentrations were ranged between 25-135 µg m -3 (87 ± 28 µg m -3 ) and 5-10 µg m -3 (7 ± 2 µg m -3 ), respectively.Cluster variations in PM 10 (BC) mass was 73 ± 26 µg m -3 (4.4 ± 1.5 µg m -3 ), 98 ± 25 µg m -3 (5.5 ± 1.5 µg m -3 ), and 104 ± 19 µg m -3 (5.3 ± 1.5 µg m -3 ), for Cluster 1, 2, and 3, respectively.
BB aerosols (water soluble and BC) were found mainly predominate in both surface mass concentration and the in columnar burden over the northern Indochina in spring.Water soluble aerosols contributed as high as 80% and 83% to surface mass and AOD 500 , respectively.BC aerosols contributed only 6% to surface aerosol mass, but its contribution to the total AOD was 12% (2 times higher).ARF variations in Clusters 1-3 was mainly due to the variations in AOD and SSA.The overall mean ARF TOA and ARF SFC were -8.0 and -31.4 W m -2 , respectively.The mean ARF TOA and ARF SFC were +10.7 and -18.1 W m -2 , respectively due to BC aerosols with the heating rate of 1.4 K d -1 .The contribution of BC aerosols towards ARF ATM was up to 75% over Indochina region and draw special attention to its pivotal role in modifying the radiation budget and regional climate.Thus the results gain further interest to understand the large spatial and temporal variations in BB aerosol loading, their sources and radiative impacts over this region.

Fig. 1 .
Fig. 1.The location of DAK in northern Indochina.Also shown are the gridded fire counts during March-April 2013 from the MODIS instrument aboard the Terra satellite (http://disc.sci.gsfc.nasa.gov/neespi/data-holdings/mod14cm1.shtml).The 5-day back trajectories ending at ground level of DAK for dates in Clusters 1 to 3 (a-c).

Fig. 2 .
Fig. 2. Flow diagram of the methodology adopted for the estimation of aerosol radiative forcing using SBDART model, in conjunction with the aerosol optical properties estimated from the OPAC in the shortwave (0.25-4 µm) range.

Fig. 4 .
Fig. 4. Day-to-day variation in observed aerosol optical properties (a) AOD at 500 nm, (b) AE at 440 and 870 nm, (c) FMF at 500 nm, (d) SSA at 440 nm, (e) AP at 440 nm, and (f) AAOD at 440 nm over DAK during the entire study period.

Fig. 6 .
Fig. 6.ARF due to aerosol components (water insoluble, water soluble, BC, sea salt and mineral dust) (a) at TOA, (b) at atmosphere, and (c) at surface, for Clusters 1, 2, and 3 over DAK.The numbers over each bar represents the value of forcing in W m -2 .

Table 1 .
Comparisons of cluster-wise mean aerosol parameters obtained from observations and estimated from model over DAK.

Table 2 .
Aerosol optical properties from worldwide BB source regions.

Table 4
also shows the mean daily ARE TOA results over the Amazon Basin.The slight deviations in ARE TOA over Amazon and Indochina are mainly due to the difference in BB aerosol loading over two different geographical regions.The mean daily ARE TOA values obtained for Cerrado regions

Table 3 .
Summary of radiative effects over DAK.

Table 4 .
Comparison of BB aerosols ARF from this study with literatures.