A methodology to estimate source-specific aerosol radiative forcing

Journal of Aerosol Science 42 (2011) 305–320

Contents lists available at ScienceDirect

Journal of Aerosol Science journal homepage: www.elsevier.com/locate/jaerosci

A methodology to estimate source-specific aerosol radiative forcing Ramya Sunder Raman a,b,n,1, S. Ramachandran a, Sumita Kedia a a b

Space and Atmospheric Sciences Division, Physical Research Laboratory, Navrangpura, Ahmedabad 380 009, India Department of Earth and Environmental Sciences, Indian Institute of Science Education and Research Bhopal, Govindpura, Bhopal 462 023, India

a r t i c l e i n f o

abstract

Article history: Received 4 August 2010 Received in revised form 19 November 2010 Accepted 25 January 2011 Available online 3 February 2011

This paper presents a novel approach to estimate source-specific radiative forcing by combining source apportionment results for particulate matter mass with satellite (moderate resolution imaging spectroradiometer (MODIS)) derived aerosol optical depth (AOD). Positive matrix factorization (PMF) was applied to particulate matter (PM) mass and its chemical constituents measured during a winter intensive study (December 2004) at Hisar, Haryana, India. The model resolved four factors including carbonate rich dust, combustion rich aerosol, secondary sulfate/nitrate, and an unidentified factor likely to be emission from polymer industries. Carbonate rich dust was the highest contributor to the measured PM mass closely followed by combustion rich aerosol with their average contributions accounting for 34.0% and 33.6%, respectively. Model apportioned species concentrations corresponding to each factor were then used to estimate factor specific optical and radiative properties, and radiative transfer calculations were performed for the shortwave regime. During the study period, although carbonate rich dust and combustion rich aerosol mass contributions were comparable, carbonate rich dust contributed to only 22% of top of the atmosphere forcing while combustion rich aerosol contributed nearly 56%. Overall, the results suggested that the aerosol radiative forcing was primarily governed by the aerosol optical and radiative properties, while the mass concentrations played a secondary role. & 2011 Elsevier Ltd. All rights reserved.

Keywords: Source apportionment of ambient aerosols Positive matrix factorization (PMF) Potential source contribution function (PSCF) Conditional probability function (CPF) Aerosol optical depth (AOD) Radiative forcing

1. Introduction Prediction of climate and its variations in different time scales is a major area of research, globally. Atmospheric aerosols influence the earth’s climate in many important ways. Aerosols interact directly and indirectly with solar and infrared radiation in the atmosphere and give rise to radiative forcing. The direct and indirect aerosol radiative forcings (ARFs) remain a significant uncertainty for climate studies (Intergovernmental Panel for Climate Change (IPCC), 2007). Studies on specieswise determination of ARF from in-situ measurements of optical, physical and chemical properties of aerosols are quite rare. A recent study apportions assimilated monthly mean aerosol optical depth (AOD) values of seven species from a model simulation over Hanimaadhoo and Gosan (Adhikary et al., 2008). Another study during the Indian Ocean Experiment (INDOEX) simulated ARFs using a Monte Carlo radiation model for several species (Podgorny & Ramanathan, 2001). However, no studies on the source-specific aerosol extinctions and/or source apportioned ARFs exist over the tropics, and perhaps over the globe, to the best of our knowledge.

n Corresponding author at: Department of Earth and Environmental Sciences, Indian Institute of Science Education and Research Bhopal, Govindpura, Bhopal 462 023, India. Tel.: +91 755 4092 321; fax: + 91 755 4092 392. E-mail address: [email protected] (R. Sunder Raman). 1 A part of this work was done while the author was at PRL, Ahmedabad, India and the rest was completed at IISER Bhopal, India.

0021-8502/$ - see front matter & 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.jaerosci.2011.01.008

306

R. Sunder Raman et al. / Journal of Aerosol Science 42 (2011) 305–320

The sources of aerosols can be natural (e.g., sea salt, biogenic, volcanic, and dust) or manmade (e.g., combustion of fossil fuel from urban/industrial processes and biomass burning), and they differ on a regional basis leading to regional variations in the Earth’s radiation budget. Aerosols are abundant near source regions, however, they can impact global climate as their radiative influence gets transported due to changes in the mean atmospheric circulation patterns. The potential for aerosol forcing of climate can vary according to regional differences in aerosol columnar concentration, chemical composition and the age of air mass (Spencer, Holecek, Corrigan, Ramanathan, & Prather, 2008). The combined effects of aerosols produced by local sources and transported from long-range in different seasons may alter large-scale heating and induce changes in the atmospheric general circulation, thereby affecting the processes of generation of clouds and rainfall (Lau et al., 2008). Information on radiative effects of different kinds of aerosols is still not available in order to permit the examination of aerosol effects on hydrological cycle and monsoon climate variability. Studies focusing on the source-specific aerosol radiative forcing will be useful in examining the dynamical changes due to aerosols and their effect on climate. The Indo-Gangetic plain in Northern India is a densely populated region. Emissions from biomass burning, industries, power plants and vehicles are dominant contributors to ambient aerosols throughout the year. Consequently, aerosol optical depths are higher over this region throughout the year compared to other regions in India (Ramachandran & Cherian, 2008). Dense foggy conditions accompanied by reduced visibility occur during winter over this region when cold air from higher latitudes mixes with the local moist air. Low temperatures and frequent temperature inversions during winter enhance fog formation resulting in accumulation of pollutants near the surface. Thus, to document the levels and variations in aerosols and trace gases, a major land campaign was conducted over the Indo-Gangetic plain in December 2004 during which a variety of aerosol and trace gas related measurements were made at eight fixed stations including Delhi, Hisar, and Kanpur (Ganguly, Jayaraman, Rajesh, & Gadhavi, 2006; Ramachandran, Rengarajan, Jayaraman, Sarin, & Das, 2006; Rengarajan, Sarin, & Sudheer, 2007). In this study, aerosol optical depth (AOD), aerosol size distribution, mass concentrations, and aerosol chemical composition measured during a land campaign in the Indo-Gangetic plain over Hisar (29.11N, 75.71E), Haryana during December 2004 were used to perform source apportionment of mass, and derive sourcespecific aerosol extinction and radiative forcing.

2. Measurements, data analysis, and approach Hisar is a semi-urban location with a population of about 1.25 million and is an industrial town with several cotton and steel industries. Hisar is influenced by industrial emissions, traffic, and agricultural stubble burning to some extent. Details of the sampling site, prevailing meteorology during winter and measurement details are described in Ramachandran et al. (2006) and Rengarajan et al. (2007). Aerosol optical depths were measured using an indigenously built hand held sun photometer in the 0.4–1.02 mm wavelength range (Ramachandran et al., 2006). Ambient bulk aerosol samples (integrated over 8–10 h) drawn through pre-combusted (at 450 1C) Whatman quartz fiber filters using an Andersen high volume sampler (HVS) were measured by Rengarajan et al. (2007). These integrated filter samples were then used to determine the total aerosol mass and the chemical composition. Samples were analyzed for water soluble ionic species, carbonaceous aerosol and the results are reported in Rengarajan et al. (2007).

2.1. Mass and chemical composition 2.1.1. Data description and quality Concentrations of total suspended particles (TSP) mass and 12 chemical species reported in Tables A1 and B1 of Rengarajan et al. (2007) were used for source apportionment of aerosols over Hisar. The data set was first examined for mass closure (Malm, Sisler, Huffman, Eldred, & Cahill, 1994). In this study, mass closure was not achieved. An earlier source apportionment study (Larson et al., 2006) suggested that when mass closure is not achieved the ‘missing mass’ may be calculated and included as a model input. Thus, missing mass was calculated as the difference between TSP mass and the sum of ionic species, elemental carbon (EC), carbonate carbon (CC), and organic matter (OM). It must be noted that organic carbon (OC) was only measured and OM was estimated as 1.8 times OC. Since several mineral dust marker elements (Si, Fe, Al, Ti) were not measured in this study, it is suggested that the missing mass is likely to be mineral dust. It was desirable to retain ‘missing mass’ as a chemical species for source apportionment since it serves as a surrogate for mineral dust. Positive matrix factorization (PMF) was applied to the measured chemical species concentrations to identify the sources of TSP mass. The key statistics for all of the chemical species used as model inputs, and TSP mass are shown in Table 1. In this table, DL indicates detection limit and S/N indicates signal to noise ratio. Detection limits for all ions were reported in Rengarajan et al. (2007). The S/N ratio is useful in assessing the quality of data points and is discussed in detail by Paatero and Hopke (2003). Since all species had S/N ratios greater than 2, none of the data points were downweighted (Paatero & Hopke, 2003). Thus, a total of 13 species (including missing mass) and 31 samples (collected during December 2004) were retained for the analysis.

R. Sunder Raman et al. / Journal of Aerosol Science 42 (2011) 305–320

307

Table 1 Summary statistics of all of the chemical species used as inputs to PMF at Hisar. Concentration (mg m  3), N = 31 Species + NH4 +

Na K+ Mg2 + Ca2 + Cl  NO3  SO42 HCO3  CC OC EC Missing TSP

Mean

Geo mean

Maximum

Minimum

DL

S/N

6.05 0.75 2.19 0.34 3.55 0.59 12.84 12.73

4.34 0.70 2.04 0.33 3.35 0.34 10.84 11.25

15.20 1.97 3.81 0.54 6.74 2.58 24.40 25.50

0.27 0.22 0.75 0.13 1.26 0.04 2.40 3.85

0.0016 0.0170 0.0025 0.0021 0.0049 0.0160 0.0050 0.0098

14.1 13.0 14.1 13.8 14.1 13.8 14.1 14.1

5.36 0.98 29.01 3.62 63.32 169.42

4.07 0.89 27.23 3.44 58.44 162.05

15.20 1.77 58.90 8.58 148.32 295.00

0.16 0.17 14.50 2.13 21.70 67.00

0.0370 0.3100 0.3100 0.0100 – 0.0225

13.8 6.1 16.5 10.5 – 34.5

Fig. 1. (a) Comparison between MODIS (Terra, Aqua) derived AODs and in situ Sun photometer AODs measured at Hisar during December 2004. (b) Scatter plot of daily mean MODIS (Terra, Aqua mean) AODs against model estimated AODs for Hisar during December 2004.

2.1.2. Estimation of uncertainties for model input PMF requires individual data point uncertainties corresponding to the measured values as a model input. Since individual data point uncertainties were not reported, they were estimated by propagating the maximum measurement uncertainty (5% of the measured value for ions and OC, and 8% for CC and EC) and the flow uncertainty (5% of the sampled volume) for each species. For ‘missing mass’, the assigned uncertainty was 400% of the calculated value to prevent this variable from driving the analysis. As a final pre-processing step, measured concentrations and the corresponding uncertainties were replaced with appropriate values for input to PMF (Polissar, Hopke, Malm, & Sisler, 1998). 2.2. Aerosol optical depth The MODerate resolution Imaging Spectroradiometer (MODIS) is a remote sensor onboard the two Earth Observing System (EOS) Terra and Aqua satellites (Remer et al., 2008). In this study, Level 3 MODIS Collection V005 daily aerosol optical depth (AOD) at 11  11 grid from Terra and Aqua were utilized (Remer et al., 2008). MODIS Terra and Aqua satellites operate at an altitude of 705 km with Terra spacecraft crossing the equator at about 1030 local standard time, LST (ascending northward) while Aqua spacecraft crosses the equator at around 1330 LST (descending southward) (Remer et al., 2008). MODIS Terra and Aqua derived aerosol products over land and oceans are tested, validated, compared, and are being extensively used to investigate spatio-temporal variations in aerosol optical characteristics (e.g., Remer et al., 2008). The predicted retrieval uncertainty of MODIS derived AODs were found to be 7 (0.05 +0.15 AOD) over land (Remer et al., 2008). Over continental India MODIS Terra and Aqua derived AODs have been validated (e.g., Ramachandran, 2007). MODIS derived AODs were found to compare well with in-situ Aerosol Robotic Network (AERONET) sun photometer results over Kanpur (Ramachandran, 2007). Daily AODs obtained from Terra and Aqua for Hisar during 1–31 December 2004 were utilized in the study (Fig. 1), as MODIS derived AODs were available for all the days in December 2004 compared to only 10 days of in situ sun photometer measurements. Mean AODs were obtained and used on days when both MODIS Terra and Aqua were available; otherwise AOD from either Terra or Aqua were used. The ground based sun photometer measures the solar intensity at six different wavelength bands centered at 0.40, 0.50, 0.65, 0.75, 0.875, and 1.02 mm. The bandwidths of the filters used for the measurement in the present study were about 0.01 mm and the field of view of the sun photometer was about 41 (Ramachandran et al., 2006).

308

R. Sunder Raman et al. / Journal of Aerosol Science 42 (2011) 305–320

The sun photometer was periodically calibrated and Io values (the solar radiation intensities for zero air mass) were obtained using the Langley plot technique for all the wavelengths from the measurements made at Gurushikhar, Mount Abu (24.61N, 72.71E), a hill station and a relatively clean site located at a height of about 1.7 km above mean sea level. These Io values were used in Beer–Lambert’s law to derive the optical depths. The maximum uncertainty in the AODs retrieved using the sun photometer is found to be r15% (Kedia, Ramachandran, Kumar, & Sarin, 2010). The in situ AODs are found to compare well with MODIS retrieved AODs (Fig. 1a, correlation coefficient of 0.9). The high correlation between handheld sun photometer AODs measured locally over Hisar and MODIS AODs retrieved over a larger grid (11  11) suggests a uniform aerosol distribution over the study region. 3. Methodology 3.1. Application of PMF, CPF, and PSCF for source apportionment PMF is an advanced factor analysis method (Paatero & Tapper, 1994). The model is extensively described in literature and the methodology of applying it for source apportionment of airborne particles is well documented (e.g., Kim & Hopke, 2007; Larsen & Baker, 2003; Lee, Chan, & Paatero, 1999; Sunder Raman & Hopke, 2007). Briefly, the task of PMF is to determine the factor profiles and the corresponding contributions that account for the measured mass with the constraint that all of the elements of the factor profile and contribution matrices be non-negative. This task is achieved by minimizing an object function ‘Q’. Q can be defined as follows: " #2 P n X m X xij  pK ¼ 1 gik fkj Q¼ i¼1j¼1

sij

where xij is the jth species concentration measured in the ith sample, gik is the mass concentration from the kth factor contributing to the ith sample, fkj is the jth species mass fraction from the kth factor, sij is the estimated uncertainty in the jth species measured in the ith sample, and p is the number of independent factors. In this study, robust mode PMF2 implementation of PMF was used (Paatero, 1997). Determination of the number of factors is an important step in the application of PMF for source apportionment. The value of the object function ‘Q’ as a function of the number of factors chosen for PMF analysis were examined. Literature suggests that the number of factors that result in a value approximately equal to the theoretical ‘Q’ value can be used as a starting point, assuming that the input uncertainties for different species are well estimated. This method was used to choose the appropriate number of factors. In order to determine the appropriate number of factors, several other diagnostic methods suggested by Lee et al. (1999) were also used. In addition, scaled residuals ‘eij’, a PMF output was examined for all species included as model inputs. As suggested in literature (e.g., Kim & Hopke, 2007; Lee et al., 1999) model solutions that yielded symmetric distribution of all the residuals between  3 and +3 were accepted. Physical interpretability of the resolved sources was also considered in choosing an appropriate solution. A four factor solution was found appropriate in this study. Typically, the rotational ambiguity in PMF is controlled by setting suitable FPEAK values in the PMF program. The ‘Q’ value is determined as a function of FPEAK, and the region in which it reaches a minimum and is stable is examined. In this study, the ‘Q’ value appeared stable for FPEAK values between  0.2 and 0.2. The need for rotation, and suitable FPEAK values were also determined by examining the G-space plots, following a procedure suggested by Paatero, Hopke, Begum, and Biswas (2005). A FPEAK value of 0.2 was chosen. Finally, in order to ensure the stability of the solution several random starts were performed. Once a stable final solution was obtained, model resolved factor profiles and contributions were normalized using a multiple linear regression technique in which the model apportioned factor contributions were regressed against the TSP mass (Kim & Hopke, 2004). Conditional probability function (CPF) and Potential source contribution function (PSCF) are based on the ideas originally developed by Ashbaugh, Malm, and Sadeh (1985). The use of CPF or PSCF to determine likely geographical source locations depends on the scale of influence of the source under consideration. Local source impacts from various wind directions are analyzed using CPF. The mathematical details of CPF are described by Kim, Hopke, and Edgerton (2003). In this study, hourly wind speed and wind direction corresponding to the sampling interval were obtained, and calm winds ( o1 m s  1) were excluded from the analysis (Kim et al., 2003). CPF was calculated using a combination of factor contributions and meteorological data. The cut-off point was set at the upper 20th percentile contribution. Since hourly meteorological data were available, the same value of daily average factor contributions was assigned to each wind speed/ direction data during that day. For regional sources, PSCF is an appropriate technique to assess source regions, and its application to identification of airborne pollutant sources are discussed in literature (e.g., Sunder Raman, Ramachandran, & Rastogi, 2010). 5-day backward trajectories calculated using the hybrid single particle Lagrangian integrated trajectory (HYSPLIT) model with FNL meteorological data (Draxler & Rolph, 2003) and model resolved factor contributions were used to calculate the PSCF values. For each sample, back trajectories commencing every 2 h between 0900 and 1900 (local standard time, LST) were calculated. Based on previous study results, the trajectory arrival heights were set at 500 m above the ground level (Cheng et al., 1993). The cut-off point was set at the 50th percentile contribution and the weighting procedure discussed in

R. Sunder Raman et al. / Journal of Aerosol Science 42 (2011) 305–320

309

Han, Holsen, Hopke, and Yi (2005) was followed to account for cells with a small number of end points compared to the average end points per cell. The PSCF maps obtained for TSP are likely to be biased towards fine (PM2.5) and coarse PM (PM10) source regions, since the residence times of the ‘large’ particles in TSP is quite low, and they are unlikely to travel long distances. However, as discussed in the following sections the PSCF results (especially for sources that do contribute ‘small’ particles) are believed to provide useful information.

3.2. Reconstruction of spectral aerosol optical properties PMF resolved species (for each factor) were grouped into four categories for input to Optical Properties of Aerosols and Clouds (OPAC) model (Hess, Koepke, & Schult, 1998), including water insoluble species, water soluble species, soot, and dust. Assuming that on average 15% of the total organic carbon in each factor was water-soluble, water insoluble species mass was calculated as the sum of CC and 1.8 times water insoluble OC (to convert OC to organic matter (OM)). Water soluble species were calculated as the sum of all ionic species and 2.1 times water soluble OC (Turpin & Lim, 2001). Soot was equal to EC, while dust was the calculated ‘missing mass’. Chemical characterization of organics in ambient particles is not extensively documented for locations over India. Thus, OC to OM conversion factors in this study are approximations that were used due to the non-availability of site specific information. The principal input parameters required for calculating aerosol radiative forcing are AOD, single scattering albedo (SSA) and asymmetry parameter (g). Ambient relative humidity (RH) and state of mixing of the aerosol must also be taken into consideration while calculating the aerosol radiative forcing (ARF). Higher ambient RH can result in an increase in the size ¨ of hygroscopic aerosol particles (Hanel, 1976). In this study, however, model estimated total and component mass concentrations were compared with measured abundance of total and species mass concentrations obtained at 50% RH (Ramachandran et al., 2006; Rengarajan et al., 2007). An important parameter which introduces uncertainty in aerosol radiative forcing estimates is the mixing state of aerosols (IPCC, 2007). In OPAC the aerosol components are treated as externally mixed. Single particle aerosol analysis studies reveal that aerosols need not remain as external mixtures after long-range transport and can undergo transformation (e.g., Guazzotti, Coffee, & Prather, 2001; Murphy et al., 2006; Spencer et al., 2008). However, a recent study (Dey, Tripathi, & Mishra, 2008) over the Indo-Gangetic basin suggests that the most probable aerosol mixing state over this region during winter is external mixing. As such, since no information on the mixing state of aerosol during the study period is available for Hisar, it will be difficult to examine and quantify its influence on aerosol radiative forcing. It is expected though that the impact would be small on aerosol radiative forcing, as during the measurement period on most of the days the winds were calm over Hisar, and as aerosol transported from long distances over urban regions can only marginally affect the aerosol optical properties (e.g., Arimoto et al., 2006). MODIS AODs, aerosol size distribution, total mass and chemical species were used to retrieve AODs, SSA and g in the shortwave region between 0.25 and 4.0 mm (Ramachandran et al., 2006) using OPAC. The aerosol components that form the aerosol type were maintained as in Hess et al. (1998), but the number concentrations were changed to derive the AODs. These number concentrations may not be unique, but they were scrutinized based on the number mixing ratios, mass (mg m  3) and the single scattering albedo values. A closure between the measured and derived aerosol characteristics (total mass, mass of chemical species and 0.55 mm AODs) was achieved by iteration until all the following criteria were satisfied: (1) the root mean square (rms) difference between the measured and model AOD spectra was o0.03, thus, constraining the rms difference to within 0.10 AOD. (2) the OPAC estimated total mass concentrations at 50% RH were within 71s of the HVS measured TSP mass which was calculated at 50% RH (Rengarajan et al., 2007), and (3) OPAC estimated mass concentrations of water soluble, soot, sea salt and mineral dust were within 71s of the respective concentrations analyzed from HVS at 50% RH (Ramachandran et al., 2006). Aerosol radiative forcing due to different aerosol species is calculated using spectral aerosol optical depths of different aerosol species. The aerosol species are assumed to follow a constant vertical distribution with higher extinction near the surface, as the pollutants near surface dominate the columnar aerosol distribution during winter over northern India. This assumption was supported by the vertical profile measurements of aerosol extinction made in Delhi (28.61N, 77.11E), about 150 km from Hisar, during December 2004 (Ganguly et al., 2006). Aerosol extinction near surface was about 2–5 km  1 during December 2004, sharply decreased from ground level, and became negligible above 1 km. The distribution of aerosol particles with height is governed by scale height, which is found to exhibit spatial and temporal variation (e.g., Hess et al., 1998). During December 2004 as the near surface aerosol extinction was much higher and fell steeply as function of height in the lowest 1 km, a scale height of 2 km was used in this study while retrieving aerosol optical parameters. Note that the coefficient of determination (square of the correlation coefficient, r2) between MODIS and OPAC model estimated AODs (total) at 0.55 mm was 1.0 (Fig. 1b) underscoring the robustness of the hybrid approach.

3.3. Aerosol size distribution: number and effective radius The dominant size of aerosol particles that contributed to aerosol size distribution in Hisar can be determined from the total aerosol number concentration N (cm  3) and effective radius reff (mm). N and reff for an aerosol size distribution can be

310

R. Sunder Raman et al. / Journal of Aerosol Science 42 (2011) 305–320

determined by the following equations: N¼

Z

r2 r1

Z

reff

dnðrÞ dr dr

ð1Þ

r2

dnðrÞ dr dr ¼ Z r1r2 dnðrÞ dr r2 dr r1 r3

ð2Þ

where ‘r’ is the particle radius. The importance of N and reff stems from the fact that the variations in N and reff will indicate whether an aerosol size distribution consists of larger or smaller particles. A high value of N indicates a dominance of smaller particles in a size distribution while a relative increase in larger particle concentration gives rise to higher value of reff. Among the four aerosol components that can appropriately describe the aerosols over Hisar only the water soluble component is hygroscopic (Table 2). Water insoluble, EC and mineral dust components are hydrophobic. The mode radius (rm) of insoluble aerosols is above 0.1 mm, while for all others the mode radius is less than 0.1 mm. The mode radius of water soluble aerosols increases from 0.021 mm in the dry state to 0.035 mm at 90% RH. The aerosol extinction coefficient of the aerosol components (corresponding to 1 particle/cm3) show a large difference based on their sizes (Table 2). Extinction coefficient of water soluble species increases by an order of magnitude as RH increases from 0% to 90% (Table 2). EC has the lowest single scattering albedo (SSA) among the aerosol components. SSA increases for water soluble aerosols as RH increases owing to the increase in scattering coefficient. Mineral dust is more of a scatterer in the mid-visible when compared to insoluble aerosols (Table 2). The physical and optical properties (N, reff, extinction coefficient and SSA) as function of RH for pre-defined aerosol models given in Hess et al. (1998) which are relevant to the study area and comprising water soluble, insoluble and EC aerosols with widely varying number densities are given in Table 3. The chosen aerosol models are continental (clean, average and polluted) and urban. Continental clean aerosol model depicts remote continental locations with low anthropogenic influence. Continental average aerosol type represents continental areas influenced by man-made activities. Continental polluted aerosol model represents areas highly polluted by anthropogenic activities. Continental polluted aerosol model has 15,700 water soluble, 0.6 insoluble and 34,300 soot particles per cm3 (Hess et al., 1998). Continental polluted has more than twice the amount of water soluble aerosols and 4 times higher soot aerosol content when Table 2 Physical (mode radius, rm (mm)) and optical properties (extinction coefficient and single scattering albedo) of aerosol components used in the study from Hess et al. (1998). Aerosol extinction coefficient (km  1) and single scattering albedo (SSA) are given at 0.55 mm. Water soluble aerosol species is hygroscopic and its properties are given as function of relative humidity (RH (%)). Insoluble, soot and mineral dust are hydrophobic and their properties correspond to dry state (0% RH). Aerosol component

RH (%)

rm (mm)

Ext. coefficient (km  1)

SSA

Water soluble Water soluble Water soluble Insoluble Soot Mineral dust

0 50 90 0 0 0

0.021 0.026 0.035 0.471 0.012 0.070

3.91  10  6 6.37  10  6 1.37  10  5 8.49  10  3 5.54  10  7 7.14  10  5

0.96 0.98 0.99 0.73 0.21 0.97

Table 3 Physical and optical properties of continental and urban aerosol types as function of relative humidity (RH). Effective radius (reff) and total number (N) are given. Aerosol extinction coefficient and single scattering albedo (SSA) corresponding to 0.55 mm are also given. Aerosol type

RH (%)

reff (mm)

N (cm  3)

Ext. coefficient (km  1)

SSA

Continental Continental Continental Continental Continental Continental Continental Continental Continental Urban Urban Urban

0 50 90 0 50 90 0 50 90 0 50 90

0.055 0.058 0.063 0.051 0.053 0.058 0.046 0.049 0.055 0.043 0.046 0.049

2534 2481 2364 15,060 14,972 14,657 49,564 49,248 48,542 157,153 156,590 155,330

0.011 0.018 0.037 0.035 0.053 0.104 0.085 0.124 0.239 0.194 0.263 0.469

0.94 0.96 0.98 0.84 0.89 0.95 0.78 0.85 0.92 0.67 0.75 0.86

clean clean clean average average average polluted polluted polluted

R. Sunder Raman et al. / Journal of Aerosol Science 42 (2011) 305–320

311

compared to continental average aerosol model. Urban aerosol model is for urban areas which have strong pollution. Urban aerosol model has 28,000 water soluble particles, 1.5 insoluble particles and 130,000 soot particles per cm3. Total number and effective radius are determined for r1 = 0.01 mm and r2 = 10.0 mm following Eqs. (1) and (2). The lower and upper limits are chosen based on the fact that aerosols in this radius range contribute the maximum to aerosol extinction, mass, number and optical depth (d’Almeida, Koepke, & Shettle, 1991). Effective radius (reff) increases, while N decreases for the same aerosol model as RH increases. It is also clear that the effective radius decreases from clean continental aerosol to urban aerosol (Table 3), while total number increases by more than 2 orders of magnitude, thus, corroborating that with more number of fine mode aerosols reff will decrease while N will increase. Aerosol extinction coefficient varies by an order of magnitude from continental clean aerosol model to urban aerosol model (Table 3). Extinction coefficient and SSA increase as RH increases owing to the increase in the size of water soluble aerosols (Table 3). SSA of continental aerosol models increase from clean to polluted because of the increase in the EC particle concentration. The number of soot particles in urban aerosol is the highest among the continental models and hence the SSA is the lowest (Table 3). It should be noted that the pre-defined aerosol types may have additional aerosol components with varying number densities according to the actual location for which they are assumed to be valid (Hess et al., 1998). For example, in Hisar we found that in addition to insoluble, water soluble and EC aerosol species, mineral dust is also present. When such aerosol regimes are encountered users can define new and appropriate mixture of aerosol types in OPAC. The number concentrations of the above four aerosol components are modified to obtain best fits of AOD, extinction, and mass as described in Section 3.2. Day-to-day variability of physical and optical properties of aerosols retrieved over Hisar during 1–31 December 2004 are given in Table 4. Daily mean RH during December 2004 over Hisar varied from a low of 36 (6 December) to a high of 94 (22 December) (Table 4). 21–25 December were characterized by dense foggy conditions accompanied with a high RH (Ramachandran et al., 2006). Effective radius (reff) is found to be remarkably stable during December 2004 over Hisar with a mean value of 0.0666 mm indicating the dominance of sub-micron size aerosols. However, the total number (N) exhibits quite a bit of variation. N and aerosol extinction coefficient are found to be the lowest on 1 December (Table 4). N is maximum on 31 December ( 42,00,000 particles) while aerosol extinction is maximum on 22 December (0.975 km  1). On 22 December dense fog occurred which was evident from the highest RH and the lowest visibility measured on that day during the campaign (Ramachandran et al., 2006). SSA at 0.55 mm is found to be higher than 0.84 over Hisar during December 2004 with a mean value of 0.91. The features in physical and optical Table 4 Physical and optical properties of aerosols measured over Hisar during 1–31 December 2004. Effective radius (reff), total number (N), extinction coefficient and single scattering albedo (SSA) are given. Extinction coefficient and SSA correspond to 0.55 mm. Date (December 2004)

RH (%)

reff (mm)

N (cm  3)

Ext. coefficient (km  1)

SSA

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 Mean 71s

63 64 53 52 43 36 51 52 51 50 56 64 70 54 53 47 59 67 53 72 84 94 85 86 74 66 65 64 72 68 82 63 14

0.0674 0.0670 0.0665 0.0668 0.0665 0.0663 0.0666 0.0669 0.0665 0.0668 0.0661 0.0666 0.0669 0.0668 0.0665 0.0668 0.0665 0.0667 0.0648 0.0663 0.0669 0.0673 0.0663 0.0661 0.0660 0.0658 0.0662 0.0666 0.0665 0.0668 0.0677 0.0666 0.0005

59,877 93,832 76,322 128,607 93,118 95,885 127,662 121,077 150,454 134,267 153,877 177,167 135,907 160,234 149,816 144,650 186,574 182,829 94,146 194,323 81,051 114,724 89,003 117,655 103,531 106,359 96,391 142,031 189,596 144,265 222,812 131,227 39,682

0.114 0.227 0.222 0.316 0.274 0.309 0.372 0.289 0.416 0.347 0.518 0.484 0.380 0.391 0.436 0.379 0.609 0.569 0.471 0.721 0.339 0.975 0.490 0.702 0.440 0.468 0.372 0.466 0.616 0.416 0.427 0.437 0.168

0.85 0.88 0.90 0.88 0.90 0.91 0.90 0.88 0.89 0.89 0.91 0.90 0.90 0.88 0.90 0.89 0.91 0.91 0.94 0.92 0.93 0.97 0.95 0.95 0.93 0.93 0.92 0.91 0.91 0.90 0.85 0.91 0.03

312

R. Sunder Raman et al. / Journal of Aerosol Science 42 (2011) 305–320

Table 5 Percentage factor contributions to the measured TSP mass at Hisar during December 2004. Factor

Carbonate rich dust Combustion rich aerosol Secondary sulfate/nitrate Unidentified (polymer industries)

Percentage factor contribution to measured mass Average

10th percentile

90th percentile

34.0 33.6 27.1 5.3

13.7 14.2 4.8 0.6

58.8 51.1 58.7 12.9

properties of aerosols obtained over Hisar are different than those obtained for the known aerosol types (Table 3). For example, for the known aerosol models as RH increases reff, aerosol extinction and SSA increase while N decreases. Over Hisar such a relationship is not evident due to the presence of additional aerosol component (e.g., mineral dust). In addition, for known aerosol types the variations in physical and optical properties of aerosols are documented for the same range of RH consisting of the same aerosol components. This was not true over Hisar as RH varied on a day-to-day basis and the source contributions also varied (Fig. 3). On 19 December when calcium carbonate rich dust dominated the aerosol mass contribution (Fig. 3) reff decreased when compared to 18 and 20 December, while N decreased significantly—almost by factor of two (Table 4). In comparison, the mean SSA during Southern African Regional Science Initiative (SAFARI) 2000 in which physical and optical properties of aged regional aerosol haze rich in biomass burning aerosol were measured was 0.91 70.04 (Haywood et al., 2003). The daily SSA values obtained in Hisar were lower than the mean SAFARI SSA value almost half the month indicating the dominance of absorbing aerosols, while during the rest of the month the SSA values were Z0.91 (Table 5). SSA during the Tropospheric Aerosol Radiative Forcing Observational Experiment (TARFOX) (Russell, Hobbs, & Stowe, 1999) conducted on the US east coast to measure and analyze the aerosol properties and effects due to plumes of urban/ industrial haze as they move from the continent over the Atlantic Ocean was 0.95 70.03 (Hartley, Hobbs, Ross, Russell, & Livingston, 2000). SSA during the Chesapeake Lighthouse and Aircraft Measurements for Satellites (CLAMS) conducted to analyze the aerosol properties off the East coast of the United States during summer 2001 was measured to be 0.9670.03 (Reidmiller, Hobbs, & Kahn, 2006). In contrast the mean SSA for Hisar during the study period was 0.91 70.03. These comparisons illustrate that the aerosol extinction obtained over Hisar are higher and the SSA values are lower corroborating the abundance of both scattering and absorbing aerosols. 3.4. Factor specific radiative forcing estimates using SBDART The radiative transfer calculations are made using the Santa Barbara Discrete Ordinate Radiative Transfer (SBDART) (Ricchiazzi, Yang, Gautier, & Sowle, 1998) algorithm which is proven to be a useful tool to address issues related to Earthatmosphere radiation budget. SBDART computes plane-parallel radiative transfer in clear sky conditions within the Earth’s atmosphere and at the surface. Aerosol optical depths, single scattering albedos and asymmetry parameters determined following the procedure described in Section 3.2 are used as inputs in SBDART in the wavelength range of 0.25–4.0 mm for radiative forcing calculations. The shortwave aerosol radiative forcing (ARF) calculations are performed using 8 radiation streams at 1-hour interval for a range of solar zenith angles and 24-h averages are obtained. Aerosol radiative forcing (DF) at the top of the atmosphere (TOA) and surface (SFC) can be defined as the change between the net (down-welling (k) minus up-welling (m)) flux with and without aerosols as,

DFTOA,SFC ¼ FluxðnetÞwith aerosol TOA, SFC FluxðnetÞwithout aerosol TOA, SFC

ð3Þ

The difference between the radiative forcing at the top of the atmosphere (which is 100 km in this case) and the surface is defined as the atmospheric forcing (ATM) and can be written as,

DFATM ¼ DFTOA DFSFC

ð4Þ

DFATM represents the amount of energy trapped within the atmosphere due to the presence of aerosols. If DFATM is positive, the aerosols cause a net gain of radiative flux to the atmosphere leading to a heating (warming), while a negative DFATM indicates a net loss and thereby cooling. To perform aerosol radiative forcing calculations atmospheric profiles of temperature, pressure, columnar ozone, water vapor and surface reflectance characteristics are necessary in addition to aerosol properties. Standard tropical atmospheric profiles of temperature and pressure which is applicable up to 301N latitude are used (McClatchey, Fenn, Salby, Volz, & Garing, 1972). The columnar ozone and water vapor for tropical atmosphere are 253 Dobson Units and 4.12 cm respectively (Ricchiazzi et al., 1998). The shortwave clear sky aerosol radiative forcings did not exhibit any significant differences for varying water vapor and ozone column amounts in the tropical atmosphere (Ramachandran et al., 2006). Surface reflectance is a crucial parameter which can introduce errors in the aerosol radiative forcing estimates (Wielicki et al., 2005). Surface reflectance measured by MODIS onboard Terra and Aqua satellites (8-day, Level-3 Global 500 m ISIN Grid product, MOD09A1 (Terra) and MYD09A1 (Aqua)) at seven wavelength bands centered at 0.645, 0.859, 0.469, 0.555, 1.24, 1.64 and 2.13 mm are used. The surface reflectance data in

R. Sunder Raman et al. / Journal of Aerosol Science 42 (2011) 305–320

313

the above wavelength bands are utilized to reproduce the spectral dependence of surface albedo in the entire shortwave region using a combination of vegetation, sand and water surface types (e.g., Ramachandran et al., 2006). The relative standard error in radiative forcing values reported here, taking into account the input aerosol parameters, and flux estimates, is estimated to be 20% (Kedia et al., 2010).

4. Results and discussion 4.1. Source apportionment of PM mass PMF resolved four factors for the Hisar wintertime aerosol. They were carbonate rich dust, combustion rich aerosol, secondary sulfate/nitrate, and an unidentified factor likely to be polymer industry emissions. A comparison of the reconstructed TSP mass contributions from all the model resolved factors with measured TSP mass concentrations yielded a regression line with slope of 0.9170.05 and r2 = 0.92. Thus, the resolved factors effectively reproduce the measured values and account for most of the variability in TSP mass. The summary statistics of percentage factor contributions to the

Fig. 2. PMF resolved factor loadings of chemical species at Hisar during December 2004.

Fig. 3. PMF resolved factor contributions at Hisar during December 2004.

314

R. Sunder Raman et al. / Journal of Aerosol Science 42 (2011) 305–320

measured TSP mass at Hisar during the study period are shown in Table 5. Among the model resolved factors, carbonate rich dust was the highest contributor to the measured mass, while the unidentified factor was the lowest contributor. The model resolved factor mass loadings are shown in Fig. 2. The corresponding factor contributions are shown in Fig. 3. Throughout this paper all figures involving model resolved factors are organized in descending order of their mass contribution. In order to better understand the temporal behavior of factor contributions, the study period was divided

Fig. 4. CPF plots for PMF resolved factors at Hisar: (a) carbonate rich dust, (b) combustion rich aerosol, (c) secondary sulfate/nitrate, and (d) unknown (polymer industries).

Fig. 5. PSCF map for PMF resolved carbonate rich dust.

R. Sunder Raman et al. / Journal of Aerosol Science 42 (2011) 305–320

315

Fig. 6. PSCF map for PMF resolved secondary sulfate/nitrate.

into ‘clear’, ‘hazy’, and ‘foggy’ days based on the classification in a previous study over the same location during the same time period as this study (Ramachandran et al., 2006). Clear/hazy/foggy day average contributions for individual factors are shown in Supplemental Fig. S1. The vertical lines in this figure represent one standard deviation. The CPF plots for all of the factors are shown in Fig. 4, and the PSCF maps (Figs. 5 and 6) indicating the likely geographical location (relevant for regional factors) are discussed in the following sections corresponding to the identification of the model resolved factors. Factor 1 (carbonate rich dust): The chemical species profile for this factor was characterized by high loadings of HCO 3, ‘missing’, and moderate loadings of CC, along with some crustal elements (Fig. 2). This factor explained nearly 79% and 46% of HCO 3 and ‘missing’ variances, respectively. The apportionment of the missing mass to this factor re-affirms the suggestion that most of the unmeasured mass in this study is accounted for by mineral dust. High concentration peaks were observed between 12/17/04 and 12/20/04 and the highest concentration was on 12/19/04 (Fig. 3). An examination of the 5-day back trajectory plot on 12/19/04 for this factor revealed that the air mass spent a large interval of time in Rajasthan and Pakistan, accounting for high dust concentrations at the receptor site (Fig. S2). Overall, mean factor concentrations on clear and hazy days were higher than that during foggy days (Fig. S1), suggesting that mineral dust was not effectively transported to the receptor site during foggy conditions. The CPF analysis identifies north–north-east and east-south-east as the influential directions resulting in high concentrations of this factor (Fig. 4a), while the PSCF map for this factor identifies regions in central Rajasthan, and parts of north eastern Pakistan and southern Afghanistan as potential source locations (Fig. 5). At Hisar, this factor is predominantly regional and results obtained using CPF analysis (local surface wind data) is not particularly appropriate. Thus, it is suggested that airborne dust from Rajasthan and Pakistan are major contributors to this factor. Factor 2 (combustion rich aerosol): This factor was identified due to high loadings of OC and EC. This factor is likely to be a combination of vehicular emissions and biomass emissions (including those from agricultural stubble burning, and wood smoke). Additionally, a substantial fraction of water soluble K + (a wood smoke tracer) was also apportioned to this factor (Fig. 2). However, some CC and crustal elements were also apportioned to this factor, possibly due to the re-suspension of road dust by vehicular traffic. This factor explained 60% each of the variances of OC and EC, and 28% and 30% of the  variances of NO 3 and SO4 , respectively. The time series behavior of this factor is shown in Fig. 3. Fig. S1 shows an increasing trend in the absolute value of mean concentrations from clear to foggy days. Since, vehicular emissions (a component of this factor) are predominantly a local source, the observed trend can be explained by the decrease in boundary layer depths between clear and foggy days. Due to shallow boundary layer during winter the aerosols from fossil fuel and biomass emissions are higher near surface, and signature of vehicular emissions which are predominantly local is clearly seen in CPF plot. The CPF plot for this factor points to north, north-east and west, south-west (Fig. 4b). National highways 65 and 10 run to the north and east of the sampling site, and some local roads run along the west of the site. Thus, the plot indicates that vehicular traffic on these roads contributes to this factor. The PSCF map for this factor is shown in the supplemental material (Fig. S3), and is not particularly relevant since local sources operate on spatial scales that are much smaller than the resolution of PSCF analysis.

316

R. Sunder Raman et al. / Journal of Aerosol Science 42 (2011) 305–320 +

Factor 3 (secondary sulfate/nitrate): This factor was characterized by high loadings of NH4 , SO42 , and NO 3 (Fig. 2). This + factor accounted for 85% of the variance in NH4 concentrations and 50% and 46% of the variances in SO42 and NO 3, respectively. The time series plot is shown in Fig. 3, while Fig. S1 shows that the average foggy day mean concentration for this factor was higher than clear and hazy days. This factor is predominantly regional, and its PSCF map is shown in Fig. 6. This map identifies potential source locations in Rajasthan, regions in Uttar Pradesh, Haryana, Punjab, Chandigarh, and north eastern Pakistan. Thermal power plants in these regions include Anta in Rajasthan, Faridabad in Haryana, Dadri, Badrapur, and Auraiya in Uttar Pradesh. Further, agricultural and cattle farms located in Punjab, Haryana, and Uttar Pradesh are rich sources of gaseous NH3 emissions from fertilizer application and cattle excreta, respectively. Thus, PSCF + analysis identifies secondary sulfate/nitrate precursor NOx, SOx, and NH4 emission ‘hot spots’. Factor 4 (unidentified (polymer industries)): This factor was characterized by a high loading of Cl  (Fig. 2). It explained 73% of the variance in chloride and 21% of the variance in Na + . Further, it also explained 15% of OC variance. While this factor may well account for a small fraction of transported sea salt, it is not immediately clear as to what source this factor really represents. The time series shows several peaks between 12/22/04 and 12/29/04 (Fig. 3), and Fig. S1 shows that the absolute value of the mean concentration on clear days was higher than those on hazy and foggy days. However, it must be noted that the standard deviation for clear days is quite large (Fig. S1), and thus there is no statistical difference between clear, hazy, and foggy day means. The CPF plot indicates north east, north west, and west as the influential source directions (Fig. 4d). Several polymer industries are located to the north east of the sampling site, potentially accounting for chloride particles formed from solvents and other raw materials used during polymer manufacture and processing. As discussed earlier, roads are also located in the identified directions, suggesting the contribution of polymers from tyre abrasion particles. Thus, although this factor remains largely unidentified it is suggested that it may represent polymer industrial emissions. The PSCF map for this factor is shown in supplemental material (Fig. S4). The map identifies hot spots in northern Uttar Pradesh, yet again, sources of ambient chloride in this region are unclear. 4.2. Source apportionment of extinction co-efficients (bext) The contribution of different aerosol types to the total aerosol extinction of that type (bext_factor,type/bext_total,type) for each factor are shown in Fig. 7. The percentage contribution of aerosol extinction due to different aerosol species in each factor were found to exhibit day-to-day variability very similar to mass concentrations (Fig. S5). The carbonate rich dust aerosol extinction was dominated by dust and water soluble species, while combustion rich aerosols are dominated by water

Fig. 7. Chemical species bext profile for PMF resolved factors.

R. Sunder Raman et al. / Journal of Aerosol Science 42 (2011) 305–320

317

insoluble (potentially OC) and EC (Fig. 7). Mineral dust on average was found to contribute  45% to carbonate rich dust followed by 29% to combustion rich aerosol emissions. The dust in combustion rich aerosol factor was probably because vehicular emissions (a part of combustion rich aerosol) are likely to associated with re-suspended road dust resulting from traffic. Dust extinction contributed about 21% to secondary sulfate/nitrate factor. Secondary sulfate/nitrate is a regional source and dust was found owing to the similarity in their source regions (Figs. 5 and 6). In the combustion rich factor extinctions due to EC and water insoluble species during December 2004 was 460%. Water insoluble species extinction in combustion rich aerosol was high due to higher loading of OC. The mean water soluble aerosol extinctions were 33% and 36% in secondary sulfate/nitrate and carbonate rich dust factors, respectively confirming the regional nature of this factor. The water soluble aerosol which includes sulfate (Fig. 2) contributed nearly 25% to combustion rich aerosol factor extinctions. 4.3. Source apportionment of radiative forcing The overall ARF due to all of the factors at Hisar are shown in Table 6, while the variability of factor specific ARFs at the surface, atmosphere, and at the top of the atmosphere for Hisar are shown as box-and-whisker plots in Fig. 8. In this figure, boxes enclosed by the 25th and 75th percentile limits, the median (solid line) and mean (dotted line) within the box, whiskers indicating the 10th and 90th percentile limits, as well as 5th and 95th percentile outliers are shown. During the study period the overall ARF at the top of the atmosphere was 20 W m  2 while that at the surface was  22 W m  2 (Table 6). Thus, during the study period the ARF of the atmosphere was 42 W m  2 (warming) on average (Table 6). As far as factor-specific ARF was concerned, the combustion rich aerosol factor dominates the aerosol radiative forcing over Hisar, although the mass apportioned to combustion rich aerosol and carbonate rich dust are comparable (Table 5). This behavior emphasizes clearly that chemical nature and radiative properties determine the magnitude of aerosol radiative forcing rather than the absolute mass. The mean SSA values at 0.5 mm for December 2004 are 0.94 (carbonate rich dust), 0.80 (combustion rich aerosol), 0.93 (secondary sulfate/nitrate) and 0.90 (unidentified (polymer industries)), respectively. Table 6 Summary of the total aerosol radiative forcing at Hisar during December 2004. Region

Surface Atmosphere Top of the atmosphere

Radiative forcing (W m  2) Average

10th percentile

90th percentile

 22 42 20

 30 31 14

 15 58 28

Fig. 8. Box–whisker plots of PMF resolved factor specific radiative forcing. The upper and lower limits of the box denote the 25th and 75th percentile values and the enclosed central line denotes the 50th percentile. 10th and 90th percentile whiskers and 5th and 95th percentile outliers (dots) are also shown.

318

R. Sunder Raman et al. / Journal of Aerosol Science 42 (2011) 305–320

Table 7 Summary of factor-specific aerosol radiative forcing at Hisar during December 2004 (SFC= surface, ATM= atmosphere, TOA = top of the atmosphere). All units in (W m  2)

Days of highest magnitude of aerosol radiative forcing Minimum Maximum Mean

Carbonate rich dust SFC

ATM

10

2

 0.04  20.54  7.40

0.02 8.98 3.04

Combustion rich aerosol

TOA

SFC

ATM

20

16

25

 0.03  11.55  4.37

 0.26  24.78  9.29

0.21 21.11 7.71

TOA

Secondary sulfate/nitrate SFC

2

5

 0.05  3.67  1.58

 0.03  14.63  5.83

ATM 4 0.01 7.00 2.68

TOA

Unknown (polymer industries) SFC

9

0

 0.02  7.63  3.15

 0.02  6.38  1.48

ATM 0 0.01 3.80 0.88

TOA 0  0.01  2.57  0.61

Mean TOA forcings for the four factors were in the range of  1 to 4 W m  2 over Hisar during December 2004. Among the four factors, TOA forcing was most negative for carbonate rich dust ( 4.4 W m  2) and least negative for the unidentified factor (  0.6 W m  2) (Fig. 8). TOA forcing for combustion rich aerosol was  1.6 W m  2. In contrast, among all factors the maximum, the minimum and the mean SFC aerosol radiative forcings due to combustion rich aerosol were the most negative (highest cooling). Atmospheric forcing, the difference between TOA and SFC forcings (Eq. (4)), was most positive (highest warming) for combustion rich aerosol (Fig. 8). The mean ATM warming due to combustion rich dust aerosol over Hisar was 8 W m  2. ATM warming due to the unidentified factor was the lowest ( o1 W m  2), while ATM warming was about 3 W m  2 due to carbonate rich dust and secondary sulfate/nitrate species. Aerosol species such as sulfate and sea salt are more efficient scatterers (higher SSA) than elemental carbon and mineral dust which have lower SSA (strong absorbers). Scattering and absorbing aerosols produce a net cooling (loss of surface reaching solar flux) at the Earth’s surface, however, the magnitude of their radiative effects in the atmosphere as a function of altitude are distinct. For weakly absorbing aerosols with an SSA 1.0 in the entire shortwave region the TOA forcing is similar to SFC forcing; in contrast for absorbing aerosols (SSA r0.8) the SFC forcing is much higher (more negative) than TOA forcing, thus resulting in a higher (more positive) ATM warming. Therefore, the ratio of SFC to TOA forcing will be lower for scatterers while it will be higher for absorbers. In the present study mean 0.5 mm SSA of combustion rich aerosol was 0.80 while that of carbonate rich dust was higher (0.94). TOA forcing was less negative for combustion rich factor over Hisar than that of carbonate rich dust and secondary sulfate/nitrate factors. The mean SFC forcing was most negative for combustion rich aerosol which combined with a less negative TOA forcing resulted in the highest ATM warming (Fig. 8). TOA forcing was 2.8 times less for combustion rich aerosol when compared to carbonate rich dust. Strongly absorbing aerosols can absorb the radiation reflected upward from the lower layers of the atmosphere and the land surface; in such a scenario SFC forcing becomes more negative while the TOA becomes either less negative or positive, resulting in a higher atmospheric warming (e.g., IPCC, 2007; Podgorny & Ramanathan, 2001; Ramachandran et al., 2006) as seen in Hisar. SFC to TOA forcing ratio was found to be 1.7 (carbonate rich dust), 5.8 (combustion rich aerosol), 1.8 (secondary sulfate/nitrate), and 2.5 (unidentified). SFC to TOA forcing ratio was highest for aerosol species with lower SSA and vice-versa. Combustion rich aerosol factors dominated the SFC and ATM radiative forcing by contributing 39% and 54%, respectively, to the total forcing (Tables 6 and 7). The percentage contribution of carbonate rich dust was highest (45%) at the TOA followed by secondary sulfate/nitrate (33%). Secondary sulfate/nitrate contributed 24% of SFC and 19% of ATM forcings. Unidentified sources contributed 6% to the aerosol radiative forcing. A summary of factor specific aerosol radiative forcings classified on the basis of days of highest magnitude of forcing for Hisar during December 2004 reiterates the dominance of combustion rich aerosol (Table 7). Combustion rich aerosol factor dominated the ATM aerosol radiative forcing on 25 out of 31 days; it dominates the surface forcing on 16 days during December 2004. Carbonate rich dust dominated the TOA forcing on 20 days while secondary sulfate/nitrate dominates the TOA forcing on 9 days (Table 7). Both these factors dominated the TOA forcing owing to SSA values 40.9. Thus, the combustion rich aerosol not only dominated the atmospheric warming over Hisar during December 2004 but also did so on a maximum number of days.

5. Conclusions Employing a novel approach by combining source apportionment results of aerosols with aerosol optical properties and a radiative transfer algorithm source-specific aerosol radiative forcings have been estimated. Measurements of aerosol optical depths, aerosol size distribution, mass concentrations, and chemical composition were conducted over Hisar, a semi-urban region in the Indo-Gangetic plain during December 2004 (Ramachandran et al., 2006; Rengarajan et al., 2007). The measured aerosol total mass and chemical compositions are utilized to apportion the sources using positive matrix factorization (PMF), which is a factor analysis method. Model resolved factor contributions were used in conjunction with meteorological data to determine the potential geographical locations of local and regional source regions contributing to each factor using the conditional probability function (CPF) and potential source contribution function (PSCF), respectively. PMF resolved aerosol species concentrations corresponding to each factor were then used to determine source-specific

R. Sunder Raman et al. / Journal of Aerosol Science 42 (2011) 305–320

319

optical and radiative properties using an optical properties model, which were then employed in a radiative transfer model to estimate clear sky shortwave aerosol radiative forcing. The major findings of the study are as follows:

 PMF resolved four factors for the wintertime aerosols over Hisar, namely, carbonate rich dust, combustion rich aerosol, 







secondary sulfate/nitrate, and an unidentified factor likely to be aerosol emissions from polymer industries. On average, the dust factor contributed about 34% to the measured TSP mass while combustion rich aerosol contributed 33.6%. PSCF analysis revealed that likely source regions for carbonate rich dust were Rajasthan and Pakistan. Combustion rich aerosol was identified due to the high factor loadings of OC and EC, and it was suggested that this factor is likely to be a combination of vehicular and biomass emissions. CPF analysis suggested that vehicular traffic on the national highways adjacent to the measurement location contributed to combustion rich aerosol. Secondary sulfate/nitrate factor was characterized by high loadings of ammonia, sulfate and nitrates. Thermal power plants, agricultural and cattle farms in Punjab, Haryana, and Uttar Pradesh were found to be the potential source regions of this factor. The mean factor specific SSA over Hisar were found to be 0.94 (carbonate rich dust), 0.80 (combustion rich aerosol), 0.93 (secondary sulfate/nitrate), and 0.90 (unidentified (polymer industries)), respectively. Combustion rich aerosol was found to dominate the aerosol radiative forcing over Hisar, despite the fact that the mass apportioned to combustion rich aerosol and carbonate rich dust were comparable, thus, emphasizing that the chemical composition and radiative properties of aerosol determine the magnitude of aerosol radiative forcing rather than the absolute mass. TOA forcing was less negative for combustion rich aerosol when compared to carbonate rich dust, and secondary sulfate/nitrate, while the surface forcing due to combustion rich aerosol was the most negative. SFC to TOA forcing ratio was 6 for combustion rich aerosol which had lower SSA, while the forcing ratio was o2 for carbonate rich dust and secondary sulfate/nitrate. ATM warming due to combustion rich aerosol was  8 W m  2 and contributed more than 50% to the total ATM warming. When strongly absorbing aerosols are abundant over continental surfaces SFC forcing becomes more negative while the TOA forcing becomes less negative, thereby resulting in higher atmospheric warming as seen in Hisar. In addition, combustion rich aerosol dominated the ATM aerosol radiative forcing on maximum number of days (25 out of 31) in December 2004; while carbonate rich dust dominated the TOA forcing on 20 days owing to its higher SSA.

The present study, the first of its kind in the tropics, and perhaps over the globe, underscores the importance of combining the optical, physical and chemical properties of aerosols to determine source-specific aerosol extinction and radiative forcing. The results reveal that the aerosol radiative forcing is primarily governed by the aerosol optical (optical depths) and radiative properties (SSA), while the physical property (mass concentrations) plays a secondary role. In general, multiple sources, some of them regional in nature, may contribute a variety of aerosol species to the total aerosol mass over a particular region. Thus, globally, studies of this kind are needed in different regions to accurately determine the spatio-temporal variability in aerosol radiative forcing due to different aerosol sources.

Acknowledgments The chemical composition data used in this study was obtained from Tables A1 and B1 of Rengarajan et al. (2007). Wind speed and wind direction were downloaded from http://www.wunderground.com. Air parcel back trajectories were obtained using the HYSPLIT (version 4) model via http://www.arl.noaa.gov/ready/hysplit4.html. The aerosol optical depth data used in this study are downloaded using the NASA’s GES-DISC. Appendix A. Supporting information Supplementary data associated with this article can be found in the online version at doi:10.1016/j.jaerosci.2011.01.008. References Adhikary, B., Kulkarni, S., Dallura, A., Tang, Y., Chai, T. Leung, L. R., et al. (2008). A regional scale chemical transport modeling of Asian aerosols with data assimilation of AOD observations using optimal interpolation technique. Atmospheric Environment, 42, 8600–8615. Arimoto, R., Kim, Y. J., Kim, Y. P., Quinn, P. K., Bates, T. S. Anderson, T. L., et al. (2006). Characterization of Asian dust during ACE—Asia. Global and Planetary Change, 52, 23–56. Ashbaugh, L. L., Malm, W. C., & Sadeh, W. Z. (1985). A residence time probability analysis of sulfur concentrations at Ground Canyon National Park. Atmospheric Environment, 19, 1263–1270. Cheng, M. D., Hopke, P. K., Barrie, L., Rippe, L. A., Olson, M., & Landsberger, S. (1993). Qualitative determination of source regions of aerosol in Canadian high arctic. Environmental Science and Technology, 27, 2063–2071. d’Almeida, G. A., Koepke, P., & Shettle, E. P. (1991). Atmospheric aerosols: Global climatology and radiative characteristics. Virginia, USA: Deepak Publishing. Dey, S., Tripathi, S. N., & Mishra, S. K. (2008). Probable mixing state of aerosols in the Indo-Gangetic basin, northern India. Geophysical Research Letters, 35, L03808, doi:10.1029/2007GL032622. Draxler, R. R., & Rolph, G. D. (2003). HYSPLIT (hybrid single-particle Lagrangian integrated trajectory) model. NOAA Air Resources Laboratory, Silver Spring, MD, access via NOAA ARL READY website /http://www.arl.noaa.gov/ready/hysplit4.htmlS. Ganguly, D., Jayaraman, A., Rajesh, T. A., & Gadhavi, H. (2006). Wintertime aerosol properties during foggy and non-foggy days over urban center Delhi and their implications to Shortwave radiative forcing. Journal of Geophysical Research10.1029/2005JD007029.

320

R. Sunder Raman et al. / Journal of Aerosol Science 42 (2011) 305–320

Guazzotti, S. A., Coffee, K. R., & Prather, K. A. (2001). Continuous measurements of size resolved chemistry during INDOEX-Intensive Field Phase 99. Journal of Geophysical Research, 106, 28,607–28,627. Han, Y.-J., Holsen, T. M., Hopke, P. K., & Yi, S.-M. (2005). Comparison between back trajectory based modeling and Lagrangian backward dispersion modeling for locating sources of reactive gaseous mercury. Environmental Science and Technology, 39, 1715–1723. ¨ Hanel, G. (1976). The properties of atmospheric aerosol particles as functions of the relative humidity at thermodynamic equilibrium with the surrounding moist air. Advances in Geophysics, 19, 73–188. Hartley, W. S., Hobbs, P. V., Ross, J. L., Russell, P. B., & Livingston, J.M (2000). Properties of aerosols aloft relevant to direct radiative forcing off the midAtlantic coast of the United States. Journal of Geophysical Research, 105, 9859–9885. Haywood, J. M., Osborne, S. R., Francis, P. N., Keil, A., Formenti, P. Andreae, M. O., et al. (2003). The mean physical and optical properties of regional haze dominated by biomass burning aerosol produced from the C-130 aircraft during SAFARI 2000. Journal of Geophysical Research, 108, 8473, doi:10.1029/2002JD002226. Hess, M., Koepke, P., & Schult, I. (1998). Optical properties of aerosols and clouds: The software package OPAC. Bulletin of the American Meteorological Society, 79, 831–844. Intergovernmental Panel on Climate Change (IPCC), I. (2007). Summary for policymakers. In S. Solomon, D. Qin, M. Manning, Z. Chen, M. Marquis, K. B. Averyt, M. Tignor, & H. L. Miller (Eds.), Climate change 2007: The physical science basis. Contribution of working group I to the fourth assessment report of the intergovernmental panel on climate change. Cambridge, UK, New York, NY, USA: Cambridge University Press. Kedia, S., Ramachandran, S., Kumar, A., & Sarin, M. M. (2010). Spatiotemporal gradients in aerosol radiative forcing and heating rate over Bay of Bengal and Arabian Sea derived based on optical, physical and chemical properties. Journal of Geophysical Research, 115, D07205, doi:10.1029/2009JD013136. Kim, E., & Hopke, P. K. (2004). Source apportionment of fine particles at Washington, DC utilizing temperature resolved carbon fractions. Journal of Air and Waste Management Association, 54, 773–785. Kim, E., & Hopke, P. K. (2007). Source identifications of airborne fine particles using positive matrix factorization and US Environmental Protection Agency positive matrix factorization. Journal of Air and Waste Management Association, 57, 811–819. Kim, E., Hopke, P. K., & Edgerton, E. S. (2003). Source identification of Atlanta aerosol by positive matrix factorization. Journal of Air and Waste Management Association, 53, 731–739. Larsen, R. K., & Baker, J. E. (2003). Source apportionment of polycyclic aromatic hydrocarbons in the urban atmosphere: A comparison of three methods. Environmental Science and Technology, 37, 1873–1881. Larson, T. V., Covert, D. S., Kim, E., Elleman, R., Schreuder, A. B., & Lumley, T. (2006). Combining size distribution and chemical species measurements into a multivariate receptor model of PM2.5. Journal of Geophysical Research, 111, D10S09, doi:10.1029/2005JD006285. Lau, K. M., Tsay, S. C., Hsu, C., Chin, M., Ramanathan, V. Wu, G. X., et al. (2008). The joint aerosol-Monsoon experiment: A new challenge for Monsoon climate research. Bulletin of the American Meteorological Society, 89, 369–383. Lee, E., Chan, C. K., & Paatero, P. (1999). Application of positive matrix factorization in source apportionment of particulate pollutants in Hong Kong. Atmospheric Environment, 33, 3201–3212. Malm, W. C., Sisler, J. F., Huffman, D., Eldred, R. A., & Cahill, T. A. (1994). Spatial and seasonal trends in particle concentration and optical extinction in the United States. Journal of Geophysical Research, 99, 1347–1370. McClatchey, R. A., Fenn, R. W., Salby, J. E. A., Volz, F. E., & Garing, J. S. (1972). Optical properties of the atmosphere (3rd ed.). Environmetal research paper 411. Cambridge, MA: AFCRL. Murphy, D. M., Cziczo, D. J., Froyd, K. D., Hudson, P. K., Matthew, B. M. Middlebrook, A. M., et al. (2006). Single-particle mass spectrometry of tropospheric aerosol particles. Journal of Geophysical Research, 111, D23S32, doi:10.1029/2006JD007340. Paatero, P. (1997). Least squares formulation of robust non-negative factor analysis. Chemometrics and Intelligent Lab Systems, 37, 15–35. Paatero, P., & Hopke, P. K. (2003). Discarding and downweighting high noise variable in factor analysis models. Analytica Chimica Acta, 490, 277–289. Paatero, P., Hopke, P. K., Begum, B. A., & Biswas, S. K. (2005). A graphical diagnostic method for assessing the rotation in factor analytical models of atmospheric pollution. Atmospheric Environment, 39, 193–201. Paatero, P., & Tapper, U. (1994). Positive matrix factorization: A non-negative factor model with optimal utilization of error estimates of data values. Environmetrics, 5, 111–126. Podgorny, I. A., & Ramanathan, V. (2001). Modeling study of the direct effect of aerosols over the tropical Indian Ocean. Journal of Geophysical Research, 106, 24,097–24,105. Polissar, A. V., Hopke, P. K., Malm, W. C., & Sisler, J. F. (1998). Atmospheric aerosol over Alaska: 2. Elemental composition and sources. Journal of Geophysical Research, 103, 19045–19057. Ramachandran, S. (2007). Aerosol optical depth and fine mode fraction variations deduced from Moderate Resolution Imaging Spectroradiometer (MODIS) over four urban areas in India. Journal of Geophysical Research, 112, D16207, doi:10.1029/2007JD008500. Ramachandran, S., & Cherian, R. (2008). Regional and seasonal variations in aerosol optical characteristics and their frequency distributions over India during 2001–2005. Journal of Geophysical Research, 113, D08207, doi:10.1029/2007JD008560. Ramachandran, S., Rengarajan, R., Jayaraman, A., Sarin, M. M., & Das, S. K. (2006). Aerosol radiative forcing during clear, hazy, and foggy conditions over a continental polluted location in north India. Journal of Geophysical Research, 111, D20214, doi:10.1029/2006JD007142. Reidmiller, D. R., Hobbs, P. V., & Kahn, R. (2006). Aerosol optical properties and particle size distributions on the east coast of the United States derived from airborne in situ and remote sensing measurements. Journal of Atmospheric Science, 63, 785–814. Remer, R., et al. (2008). Global aerosol climatology from MODIS satellite sensors. Journal of Geophysical Research. D14S0710.1029/2007JD009661. Rengarajan, R., Sarin, M. M., & Sudheer, A. K. (2007). Carbonaceous and inorganic species in atmospheric aerosols during wintertime over urban and highaltitude sites in North India. Journal of Geophysical Research, 112, D21307, doi:10.1029/2006JD008150. Ricchiazzi, P., Yang, S., Gautier, C., & Sowle, D. (1998). SBDART: A research and teaching software for plane-parallel radiative transfer in the Earth’s atmosphere. Bulletin of the American Meteorological Society, 79, 2101–2114. Russell, P. B., Hobbs, P. V., & Stowe, L. L. (1999). Aerosol properties and radiative effect in the United States East Coast haze plume: An overview of the Troposphere Aerosol Radiative Forcing Experiment (TARFOX). Journal of Geophysical Research, 104, 2213–2222. Spencer, M. T., Holecek, J. C., Corrigan, C. E., Ramanathan, V., & Prather, K. (2008). Size-resolved chemical composition of aerosol particles during a monsoonal transition period over the Indian Ocean. Journal of Geophysical Research, 113, D13605, doi:10.1029/2007JD008657. Sunder Raman, R., & Hopke, P. K. (2007). Source apportionment of fine particles utilizing partially speciated carbonaceous aerosol data at two rural locations in New York State. Atmospheric Environment, 41, 7923–7939. Sunder Raman, R., Ramachandran, S., & Rastogi, N. (2010). Source identification of ambient aerosols over an urban region in western India. Journal of Environmental Monitoring, 12, 1330–1340, doi:10.1039/b925511g. Turpin, B. J., & Lim, H. J. (2001). Species contributions to PM2.5 mass concentrations: Revisiting common assumptions for estimating organic mass. Aerosol Science and Technology, 35, 602–610. Wielicki, B. A., Wong, T., Loeb, N., Minnis, P., Priestley, K., & Kandel, R. (2005). Changes in Earth’s albedo measured by satellite. Science, 308, 825.