lable at ScienceDirect

Environmental Pollution 195 (2014) 245e256

Contents lists avai

Environmental Pollution

journal homepage: www.elsevier .com/locate/envpol

Analysis of Southeast Asian pollution episode during June 2013 usingsatellite remote sensing datasets

Krishna Prasad Vadrevu*, Kristofer Lasko, Louis Giglio, Chris JusticeDepartment of Geographical Sciences, University of Maryland College Park, 4321 Hartwick Road, College Park, MD 20740, USA

a r t i c l e i n f o

Article history:Received 27 February 2014Received in revised form5 June 2014Accepted 9 June 2014Available online 30 July 2014

Keywords:IndonesiaHazeAerosolsCOSoutheast Asia

* Corresponding author.E-mail addresses: krisvkp@umd.edu, krisvkp@ya

com (K.P. Vadrevu).

http://dx.doi.org/10.1016/j.envpol.2014.06.0170269-7491/© 2014 Elsevier Ltd. All rights reserved.

a b s t r a c t

In this study, we assess the intense pollution episode of June 2013, in Riau province, Indonesia from landclearing. We relied on satellite retrievals of aerosols and Carbon monoxide (CO) due to lack of groundmeasurements. We used both the yearly and daily data for aerosol optical depth (AOD), fine modefraction (FMF), aerosol absorption optical depth (AAOD) and UV aerosol index (UVAI) for characterizingvariations. We found significant enhancement in aerosols and CO during the pollution episode.Compared to mean (2008e2012) June AOD of 0.40, FMF-0.39, AAOD-0.45, UVAI-1.77 and CO of 200 ppbv,June 2013 values reached 0.8, 0.573, 0.672, 1.77 and 978 ppbv respectively. Correlations of fire countswith AAOD and UVAI were stronger compared to AOD and FMF. Results from a trajectory model sug-gested transport of air masses from Indonesia towards Malaysia, Singapore and southern Thailand. Ourresults highlight satellite-based mapping and monitoring of pollution episodes in Southeast Asia.

© 2014 Elsevier Ltd. All rights reserved.

1. Introduction

Vegetation fires are an immense source of air pollution inseveral tropical countries. In particular, biomass burning inSoutheast Asia is extensive and an important source of trace gasesand aerosols (Folkins et al., 1997; Chan et al., 2006; Hyer et al.,2013). In the region, fire is used as a management tool forclearing land through slash and burn agriculture (Tomich et al.,1998), disposing of agricultural residues (Dobermann and Fairhurst,2002; FAO, 2013) and for clearing forests for developing palm oilplantations by private companies (Greenpeace, 2013; Miettinenet al., 2011). The fire season in this region coincides with the dryweather season and the southwest monsoon. Earlier researchersreported several biomass burning-related haze episodes from theregion during August to September 1982, September 1983,September 1987, August 1990, AugusteSeptember 1991,AugusteOctober 1994, AugusteOctober 1997 (Nichol, 1998;Radojevic, 2003; See et al., 2006). During the air pollution events,aerosols and pollutants can be transported long distances andpersist for weeks to months, impacting not only air quality but alsohuman health, biogeochemical cycles, atmospheric chemistry,weather and climate (Radojevic, 2003).

hoo.com, krivadrevu@gmail.

AmongAsian countries, Indonesia accounts for 15.7%of totalfires(Vadrevu and Justice, 2011). For Indonesia, the most recent, signifi-cant recorded haze episode occurred during 1997/1998 from thecombination of human factors and dry conditions during an El Ni~no-Southern Oscillation cycle. During that time, the haze persisted forweeks in Indonesia and in the neighboring countries of SingaporeandMalaysia (Murdiyarso et al., 2004). It is estimated thatmore than11 Mha of forest were burned (Siegert, 2001; Taylor, 2010). Close tothe vegetation fires at Palangkaraya on Kalimantan and Jambi onSumatra up to 4000 mg/m3 total particulate matter was measured,exceeding the Indonesian national ambient air quality standard by afactor of 15 (Heil andGoldammer, 2001). In Indonesia, peatland firesare common (Nara et al., 2011; Hyer et al., 2013). Peatlands occupy170,000e270,000 km2 (Heil and Goldammer, 2001) and it is esti-mated that they store 57,367Gt of carbon (Page et al., 2011). Thus,peatland fires release huge amounts of carbon into the atmosphere.Most of the peatland fires are attributed to anthropogenic factors,however, theyget aggravatedby the coincidence of dry season (Usupet al., 2004). During recent times, one of the major projects that ledto peatland degradation was the Mega Rice Project in central Kali-mantan, Indonesia. The project aimed to create large rice cultivationarea, however, it failed and resulted in the drainage of 1 Mha ofunmanaged peatlands. The drained peatlands are highly susceptibleto fires as they are relatively dry with a low water table, enablingcombustion of top peat layers (Miettinen et al., 2011). Fires from thepeatlands are hard to detect from space, as peat fires are typicallylow temperature smoldering fires involving both above-ground as

K.P. Vadrevu et al. / Environmental Pollution 195 (2014) 245e256246

well as below-ground organic matter (Langmann and Heil, 2004;Hyer and Chew, 2010).

During last year, an intense haze event occurred from June18theJuly 1st 2013, in Indonesia that spread to other neighboringcountries, including Brunei, Indonesia, Malaysia, Singapore andsouthern Thailand. Themain source for the pollutionwas attributedto fires in the Riau Province, Indonesia (Eastasia forum, 2013). In theprovince, several residents had to flee their homes due to respira-tory problems and a local airport had to be closed due to poorvisibility caused by haze (Eastasia forum, 2013). The smoke plumestraveled large distances and as a result, Malaysia had to declare anemergency in two southern regions after the country's air pollutionreached hazardous levels. As a result of the haze event, the pollu-tion index in the capital Kuala Lumpur reached 200 indicating “veryunhealthy” conditions and in Johar State, it reached greater than700 indicating “hazardous” conditions. Similarly, Singapore wasshrouded with haze as the pollution index reached 371 on the June20th, 2013. The haze also affected some provinces in southernThailand and parts of Brunei. Important reasons for the fires wereattributed to palm oil, pulp and paper companies which areinvolved in forest clearance and developing plantations(Greenpeace, 2013). It is inferred that the Indonesia-caused airpollution event reached the worst levels in nearly 16 years after themajor 1997/1998 haze episode.

Remote sensing data with its multi-temporal, multi-spectral,repetitive and synoptic coverage can provide useful informationon land use/cover and pollution monitoring. In particular, remotesensing of the troposphere can yield useful measurements onatmospheric composition which can range from regional toglobal scales, that are not easily obtained from in situ observingsystems (Monks and Bierle, 2011; Ichoku et al., 2013). However,the potential of remote sensing datasets in capturing air pollutionevents needs a thorough investigation, as pollutants can haveshort residence times. Further, most of the emission sources arelocated close to the ground, whereas, satellite detection is mostlybased on the absorption spectra of atmospheric constituents(Richter and Wagner, 2011) and pollutants may not be alwaysassociated with ground sources due to dispersion. Satellite ob-servations can be a powerful tool for monitoring of atmosphericpollution if they can be related to the underlying emissionsources.

In this study, we evaluate the potential of remote sensingdatasets in capturing the air pollution episodes (June, 2013) inSoutheast Asia. We use remote sensing data for capturing thechanges in atmospheric composition and degradation due to firesin the region.We focused on aerosol physical and optical propertiesas well as Carbon monoxide (CO) retrieved from MODIS, OMI andMOPITT data corresponding to the pollution event. We also usedmeteorological data and a Lagrangian particle dispersion model tocharacterize the influence of wind direction and long range trans-port of pollutants. In view of the growing incidence of forest firesand the resulting haze events in the region, there is a need toexplore remote sensing data for capturing synoptic pollutionevents. The results from such an exercise can provide valuable in-sights useful for pollution control measures.

2. Study area

The study area includes Indonesia, Singapore, Brunei andpeninsular Malaysia (Fig. 1). The region is home to 250 millionpeople and is experiencing increasing economic development. Theweather is hot and humid all year round, and the dry period is fromJune till October, during which most biomass burning take place.Mostly, the dry seasons in the study area are attributed to the ElNi~no-Southern Oscillation (ENSO) (Corlett, 2011), however, even

during this period, rainfall exceeds evaporation inmany parts of theregion in normal years (Meittinen et al., 2004).

3. Datasets and methodology

3.1. Active fires

For characterizing the vegetation fires in the region, we used daily active firedetections from the MODIS instruments onboard the Aqua and Terra satellites. Thetwo MODIS sun-synchronous, polar-orbiting satellites pass over the Equator atapproximately 10:30 a.m./p.m. (Terra) and 1:30 p.m./a.m. (Aqua) with a revisit timeof 1e2 days. The MODIS Advanced Processing System (MODAPS) processes theresulting data using the enhanced contextual fire detection algorithm (Giglio et al.,2003) combined into the Collection 5 Active Fire product. For this study, we analyzedthe yearly data (2003e2013) for Indonesia and also daily data for local provincesduring June, 2013 to characterize fires. The algorithm routinely detects fires of atleast 1000m2. However, under ideal observing conditions such as near nadir, little orno smoke, relatively homogenous land surface, flaming fires one tenth of that sizecan be detected. Further, the smallest size of a flaming fire that can be detected is50 m2 under pristine conditions (Giglio et al., 2013). In many cases a large fire eventconsists of several small active fires burning simultaneously (Morisette et al., 2005) .

3.2. Post-fire vegetation loss in Riau Province

For assessing the vegetation changes before and after fire, we used the MODISAqua vegetation indices 16-day L3 Global 250 m SIN Grid product version 5(MYD13Q1). Accordingly, we obtained scenes before and after the peak fire event onJune 2nd (t1) and June 18th (t2) respectively; time period one is a composite ofnormalized difference vegetation index (NDVI) with data from June 2nd till June17th while time period two has data ranging from June 18th till July 3rd. The NDVIranges from 0 to 1 with 0 indicating very low vegetation (or none) to 1 indicatingvery high vegetation vigor or greenness. For the northern region of Riau Provincewhere most fire hotspots were observed, the NDVI data for time period two wereobtained after June 25th i.e., the end of the fire episode (Fig. 2d) . A change wascomputed between the two time periods as.

NDVI Change ¼ Time periodð2Þ � Time periodð1Þ

Negative values are indicative of a decrease in NDVI which is indicative of adisturbance and degradation or loss of vegetation while positive values are attrib-uted to and associated with vegetation growth/regrowth increasing leaf area index,and other factors (Carlson and Ripley, 1997).

3.3. Aerosol optical depth (AOD) (a)

AOD is a measure of atmospheric extinction through a vertical column of at-mosphere as retrieved by satellites visible channel corresponding with Beer-eLambert’s law. The higher the AOD value, the more aerosols are within a column;thus lower visibility within that column. Biomass burning activities are expected toincrease AOD's compared to the background. The aerosol properties are derived bythe inversion of the MODIS-observed reflectance using pre-computed radiativetransfer look-up tables based on aerosol models (Remer et al., 2005). For this study,we specifically used the MODIS Collection 5.1 monthly product (MYD08_M3.005)for June, 2008e2013 and daily product (MYD08_D3.005) for June 2013 for charac-terizing the spatial and temporal variations in relation to fires.

3.4. Fine mode fraction (FMF)

Aerosol finemode fraction is the ratio of small mode optical depth (thickness) tothe total AOD. In the aerosol research, quantifying the particle size distribution iscritical to estimating the Earth's energy balance (Kleidman et al., 2005) and FMF isuseful for the same. MODIS provides the aerosol fine mode fraction at 550 nm andwe used the level 3 version 5.1 daily product (MOD08_D3.051) corresponding to thestudy area.

3.5. Aerosol absorbing optical depth (AAOD) at 500 nm

The aerosol absorption optical depth is measure of concentration of near-UVabsorbing aerosol particles such as smoke and mineral dust and is retrieved fromthe Ozone Monitoring Instrument (OMI). The OMI measures the solar light back-scattered by the earth's atmosphere and surface (Bucsela et al., 2008). The instru-ment consists of two spectrometers, one measuring the UV spectral range from 270to 365 nm in two sub-ranges (UV1: 270e314 nm, resolution: 0.42 nm, sampling:0.32 nm; UV2: 306e380 nm, resolution: 0.45 nm, sampling: 0.15 nm), the othermeasuring the UVevisible spectrum from 350 to 500 nm (resolution: 0.63 nm;sampling: 0.21 nm). We used the AAOD daily product at 500 nm (OMAERUVd.003).

3.6. UV Aerosol Index (UVAI)

is an index that detects the presence of UV-absorbing aerosols such as dust andsoot. UV Aerosol index (AI) is based on a spectral contrast method in a UV regionwhere the ozone absorption is very small. It is the difference between the

Fig. 1. Study area map showing Insular southeast Asia. The inset shows western part of Indonesia with different provinces.

K.P. Vadrevu et al. / Environmental Pollution 195 (2014) 245e256 247

observations and model calculations of absorbing and non-absorbing spectralradiance ratios. For OMI, AI is defined as

AI ¼ 100�log10ðI360=I331Þmeasured � log10ðI360=I331Þcalculated

�Positive values of Aerosol Index generally represent absorbing aerosols (dust

and smoke) while small or negative values represent non-absorbing aerosols (sul-fate, sea-salt) and clouds (Torres et al., 2007). We specifically used the daily product(OMTO3d.003) to infer UVAI variations for the study region.

3.7. Carbon monoxide

For characterizing the spatial variations and geographical extent of CO pollutionin Southeast Asia, we used theMOPITT CO datasets. MOPITT instrument, launched in

December 1999 onboard the NASA EOS Terra satellite, is a thermal nadir-viewing gascorrelation radiometer (Drummond, 1992; Deeter et al., 2004). It measures CO at aspatial resolution of 22 km � 22 km. CO profiles are retrieved from radiance mea-surements in the thermal infrared channel at 470 nm (Deeter et al., 2004) andMOPITT achieves a global coverage within 3 days. We used the Version 6 Level 3product which includes CO concentrations, averaging kernels and error co-variancesat 10 pressure levels. Of these columnar data, we used only surface data (~900 hPa)to infer fire-CO variations. Retrieved CO total columns are calculated by integratingthe retrieved mixing ratio profile and each retrieval 'level' corresponds to a uni-formly weighted layer immediately above that level. For example, the surface-levelretrieval product corresponds to the mean volume-mixing ratio over the layer be-tween the surface and 900 hPa. For CO vertical profiles, estimated errors are avail-able in the error field of the “retrieved CO Mixing ratio profile” and “retrieved CO

Fig. 2. Fire characteristics for Indonesia and Pelalawan district retrieved from MODIS data. a. Indonesia fires (2010e2013); b. Fire seasonality in Indonesia showing peak duringSeptember (2012); c. Fires in Pelalawan district, Riau province (2010e2013); d. Daily fires in Pelalawan district, June, 2013.

K.P. Vadrevu et al. / Environmental Pollution 195 (2014) 245e256248

Surface mixing ratio” variables of the MOPITT files. These values represent the cu-mulative error from smoothing error, model parameter error, forward model error,geophysical noise, and instrument noise (Deeter and MOPITT team, 2009) contrib-uting to ~20e30% uncertainty with 5% difference in the monthly mean CO.

In addition, trajectory analysis was performed using the NOAA Hybrid singleparticle Lagrangian integrated trajectory (HYSPLIT) model (Draxler and Hess, 1998)with six hourly archived meteorological data provided from the US National Centerfor Environmental Prediction (NCEP) Global Data Assimilation System (GDAS). Thetrajectory starting points were selected primarily from collocating the MODIS activefires during June, 2013 over the Riau Province, Indonesia.

We computed the mean absolute deviation (MAD) for the fire, aerosol and COdatasets to assess the spread of values from 2008 to 2013. The MAD is calculated as.

MAD ¼ 1=n�����x1 � x

�����þ x2 � x

�����þ ::::þ

����xn � x������

¼ 1=nXni¼1

����x1 � x�����

where, n is the number of samples, x1,2.n are the number of data values and x�is the

mean of values.To assess the deviation in fire counts, aerosol properties and CO during the

pollution episode (June, 2013) from the mean values (June, 2008e2012), we usedabsolute deviation (Di) as.

Di ¼ jx�mðXÞj

WhereDi is the absolute deviation, x is the data element (in our case June 2013 valuefor fire counts, aerosol properties and CO) and m(X) is the mean value of fire counts,aerosol and CO values from 2008 to 2012.

Further, to compare the June 2013 fire episode valuewith the previous 2012 yearvalue, we used percent difference (PD) as.

PD ¼ jE1 � E2j=1=2ðE1 þ E2Þ*100

Where E1 is the 2012 value, E2 is the 2013 value for aerosol properties and CO.In addition, we used Pearson-product moment correlations to test the rela-

tionship betweenMODIS fires and aerosol properties. The coefficient represented bythe symbol rx;y for the variables x and y is calculated by.

rx;y ¼ covðX;YÞ�sxsy ¼ EððX� EðXÞÞðY� EðYÞÞÞ�sxsy

Where E(.) denotes the expected value of the variable, cov(.) the covariance, sxsy thestandard deviations of two random variables X and Y respectively.

A two-tailed t-test was used to test the null hypothesis (H0) that there is nolinear relationship between the two data sets as.

t ¼����r

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiN � 2

�1� r2

q ����where, r is the correlation between two variables and N is the number of observa-tions. The t follows a t-distribution with N-2 degrees of freedom. The two-tailedsignificance level can be calculated as

p ¼ 2ð1� tðt;N � 2ÞÞAll of the coefficient calculations are based on pairs of data with a critical value

of r at the 99% level of significance.

4. Results

4.1. MODIS active fires

The total number of fire counts retrieved from the MODISdatasets for Indonesia from 2010 to 2013 are shown in Fig. 2(a) andfire seasonality in Fig. 2(b). For the entire Indonesia, 2012 had thehighest number of fires. To arrive at the fire seasonality forIndonesia, we averaged MODIS Aqua and Terra fire counts from2003 till 2012 for individual calendar months. Results indicate thatAugust till October is the peak fire season with the maximumnumber of fires occurring during September (Fig. 2b).

In this study, we specifically focused on the Pelalawan District ofRiau province, Sumatra, Indonesia, as it was the hotspot of fireevents and the major contributor to the June 2013 haze episode. Incontrast to entire Indonesia which had the highest number of firesduring 2012, the Pelalawan District and Riau Province showed the

Fig. 3. aec. Pre and post fire vegetation changes during June, 2013 in Riau province Indonesia. a. Pre-fire NDVI, June 2nd, 2013; b.Post-fire NDVI, June 18th, 2013; c. NDVI changeshowing highly negative values in the northeast Eastern provinces; The figure also shows MODIS active fires overlaid on vegetation change image.

K.P. Vadrevu et al. / Environmental Pollution 195 (2014) 245e256 249

K.P. Vadrevu et al. / Environmental Pollution 195 (2014) 245e256250

highest number of fires during 2013 (Fig. 2c). For Riau Province, themean number of fires from 2010 to 2012 were 6258 fire countscompared to 2013 value of 15,076, which were significantly higherthan previous years (figure not shown). Similarly, for the PelalawanDistrict, the mean fire counts (2010e2012) were 1160, whereas for

Fig. 4. aeh. Yearly (2008e2013) and monthly variations (JaneAugust, 2013) in aerosol propemode fraction (SMF); e-f. Aerosol Absorption Optical Depth (AAOD); g-h. UV Aerosol Index

2013 the 2949 fire counts were recorded. Daily fire counts for themonth of June, 2013 are shown in Fig. 2(d). Results clearly suggestthe peak fires occurred during June 15theJune 25, with the highestfire counts on June 19th (319), June 23 (135), June 21st (121), etc.The fires subsided after June 29th and were negligible (less than 5

rties for the Pelalawan district, Indonesia. aeb. Aerosol optical depth (AOD); ced Small(UVAI).

K.P. Vadrevu et al. / Environmental Pollution 195 (2014) 245e256 251

fire counts) during the first week of July 2013. The MADwas 984.17,absolute deviation during 2013 fire episode 984 fire counts, with a37.6 percent increase in fire counts during June 2013, as comparedto 2012.

4.2. Post-fire vegetation loss in Riau Province

Vegetation changes within the Riau Province, Indonesia areshown in Fig. 3(aed) . Highly negative values range from �0.25 to�0.99with the latter indicative of complete vegetation cover loss asthe maximum NDVI value is 1 and values ranging from 0.00 to 0.15are often bare ground areas (Wittich and Hansing, 1995). Wefocused on Riau Province, as it was a hotspot of fires during June18th�24th 2013 and additionally much of the haze, responsible forthe abnormally high aerosol and carbon monoxide values in theregion, originated here. In Fig. 3(c), it is clear that most of the highlynegative NDVI change areas are within Riau Province, and inparticular the northeastern portion. Parts of Riau Province with themost active fires and highly negative NDVI change are located 50miles or less from Malaysia, which subsequently reported veryhigh, unhealthy air quality index values.

The districts in Riau with the most highly negative NDVI changeand resulting vegetation loss in descending order are Bengkalis,Rokan Hilir (northern most district), Kota Dumai, Siak, and Pelala-wan. All of the above districts are located on the east coast ofSumatra and the majority of the vegetation loss is clustered in the

Fig. 5. aed. Aerosol characteristics during June, 2013 for southeast Asia with overlaid MOD(SMF) for June 19th, 2013; c. Aerosol Absorption Optical Depth (AAOD) during June 24th, 2

northeastern portion of Riau. Parts of Pelalawan and Bengkalis thathad highly negative NDVI change and corresponding MODIS activefire detections are only 80 miles or less from Singapore which re-ported astronomically high air quality index values.

Accordingly, of the highly negative NDVI change districts themost fires during June 2nd e 18th were found in Rokan Hilir,Bengkalis, Kota Dumai, Pelalawan, and Siak respectively. With thecombination of over 1200 fires occurring in these districts alone,and the active fires coinciding with the highly negative NDVI areas(Fig. 3d), the fires not only contributed significantly to high aerosollevels, but also to land cover disturbance and vegetation loss.

4.3. Yearly trends and spatial patterns

Trends in aerosol properties for Pelalawan District are shown inFig. 4(aeh) . AOD for June 2013 was significantly higher than theprevious years. For June, five year average (2008e2012) AOD wasless than 0.40 compared to the June 2013 value, which reached 0.80due to the fires. Spatial patterns in the AOD for June 19th, 2013 forthe entire region are shown in Fig. 5a. MODIS datasets captured asignificant increase in AOD, reaching 1.9 in the hotspot region ofnorthern Riau Province, Indonesia (Fig. 1 inset) and extending tothe other regions of Singapore, Malaysia and Brunei. MAD for AODwas found to be 0.15, absolute deviation to be 0.35 and percentdifference in AOD to be 9.78 which is significantly high. FMF washighest during June, 2013 (0.573) compared to the previous years'

IS fire data. a. Aerosol Optical Depth (AOD) for June 19th, 2013; b. Smal mode fraction013; d. UV Aerosol Index (UVAI) during June, 24th, 2013.

K.P. Vadrevu et al. / Environmental Pollution 195 (2014) 245e256252

(2008e2012) mean of 0.396, however, the June 2013 SMF valueswere lower than the August SMF value (0.683) (Fig. 4c,d). The MADfor SMF was 0.078, absolute deviation was 0.145 and percent dif-ference of 8.46 (compared to previous years). Similar to AOD, AAODvalue was high (0.672) during June 2013 compared to previousyears (2008e2012) mean AAOD of 0.45 (Fig. 4e,f). The MAD was0.016, absolute deviationwas 0.029 and percent differencewas 3.17.Spatial patterns in AAOD for June 19th, 2013 are shown in Fig. (5b)and the spatial patterns in AAOD closely matched the AOD(Fig. 3a).Similar to AOD and AAOD, UVAI had the highest valueduring 2013 (1.77) compared to previous years UVAI mean of 0.84(Fig. 4g,h). The MAD for UVAI was 0.257, absolute difference of0.781 and percent difference was 16.64. Spatial patterns in SMF and

Fig. 6. aec. MOPITT CO surface data at 900hPa. 6a. CO concentrations during June for differe

UVAI are shown in Fig. 5(c,d), which showed highest values duringthe fire episodes. Interestingly, all the aerosol data except for SMFi.e., AOD, AAOD, and UVAI were significantly higher during June2013 as compared to August, 2013, the latter being the typicalmonth of burning season in the region. The pattern clearly suggeststhat June 2013 burning is an unusual event that released hugeamounts of aerosols in the region compared to traditional burningmonths.

MOPITT CO columnar data at a surface level for the southeastAsia for June (2008e2012) is shown in Fig. 6a, months (Januar-yeAugust) in Fig. 6b and spatial patterns in COwith overlaidMODISactive fire data for June 24th, 2013 in Fig. 6(c). The mean CO valuesfor June for previous years (2008e2012) were below 200 ppbv,

nt years; b. Monthly trends during 2013; c. Spatial variations in CO on 24th June, 2014.

K.P. Vadrevu et al. / Environmental Pollution 195 (2014) 245e256 253

however it reached 978 ppbv during June, 2013 close to RiauProvince in Indonesia and Malaysia. The MAD for CO data was211.3 ppb, absolute deviation of 987.16 ppb and 29.61 percent dif-ference for 2013 June pollution episode compared to June 2012.

4.4. Daily patterns and correlations

Most of the aerosol properties for June 2013 followed similarpattern as the fires (Fig. 5aed). Fire counts coincided with AOD andthe correlation coefficient was found to be 0.36. Similarly, thecorrelation between fire counts and AAODwas 0.64, Fire counts andSMF 0.20 and Fire counts-UVAI-0.50. All correlations were signifi-cant (p < 0.01) using a two tailed test of significance.

5. Discussion

Aerosol and greenhouse gas concentrations in the atmospherecan vary in space and time, based on the underlying sources and avariety of factors including residence times and meteorologicalconditions (Lin et al., 2010). Increase in AOD coinciding with fireevents in Indonesia and Southeast Asia were reported earlier byseveral researchers (Kanniah et al., 2010; Rajab et al., 2011; Fengand Christopher, 2013). Similar to the previous studies, we notesignificant AOD increase during June fires which were not thetypical fires that occur during AugusteOctober.

Fig. 7. aed. Time series plots of daily fire counts with daily aerosol data during June 2013. a.Fire counts-Aerosol Absorption Optical Depth (AAOD); d. Fire counts-UV Aerosol Index (UV

Biomass burning plumes have carbonaceous aerosols withvarying radiative properties, both absorbing and non-absorbing.The non-absorbing nature is due to volatile organic compoundsemitted during burning, whereas absorbing nature of carbonaceousaerosols is mainly due to black carbon (Andreae and Gelencs�er,2006). Further, the absorbing properties of aerosols varythroughout the combustion process, based on flaming or smol-dering conditions. Previous studies have established that absorbingaerosols are mostly released during flaming conditions (Radkeet al., 1988) and that black carbon exhibits absorption throughoutthe entire solar spectrum due to a relatively spectrally constantimaginary index of refraction (Bergstrom et al., 2002). Also, blackcarbon aerosols are predominately fine mode in nature, producinghigh fine mode fraction values, whereas mechanically produceddust and sea-salt aerosols are larger (coarse mode) resulting inlower FMF values (Tanre et al., 2001; Kaufman et al., 2005;Badarinath et al., 2007). Thus, the biomass burning smokeplumes are supposed to have high FMF. In this study, although weobserved a significant increase in FMF during June 2013 fireepisode, those values were not higher than typical biomass burningmonth of August, 2013 values. However, FMF was high during June,compared to previous years (2008e2013) mean values. In general,AI shows positive values for UV absorbing particles such as desertdust, smoke and volcanic ash whereas negative values for non-absorbing particles such as sea-salt particles and sulfate aerosols.

Fire counts-Aerosol Optical Depth (AOD); b. Fire countsdSmall mode fraction (SMF); c.AI).

K.P. Vadrevu et al. / Environmental Pollution 195 (2014) 245e256254

Earlier, Torres et al. (1998), Hsu et al. (1999), and Ginoux and Torres(2003) all noted an important sensitivity of OMI to UV absorbingaerosols, as a function of height. Also, Torres et al. (2012) showedthat the magnitude of the positive AI depends on the AAOD andheight of the aerosol layers and that AI signal is meaningful above0.5 and AI signal is amplified when the absorbing layer lies abovethe clouds (Torres et al., 1998). Previously, Badarinath et al. (2009)and Guan et al. (2010) had shown increasing AI greater than 1.5 inbiomass burning plumes. Similar to these previous studies, in thisstudy, we found UVAI to be a good indicator of enhanced aerosolsfrom biomass burning, as the values were significantly high (1.77)during June, 2013 for the Pelalawan District. Further, most of thestudy region covering Indonesia and Singapore on June 24, 2013showed >2.0 UVAI suggesting intense aerosol load over the cloudlayer (Fig. 5d). Also, UVAI showed significant correlation (r ¼ 0.50)with fire counts (Fig. 7d) indicating fires as an important source ofabsorbing aerosols. A more direct measure of absorbed UV solarradiation at 388 nm is obtained in terms of the AAOD and weobserved a relatively higher correlation of fire counts with OMIAAOD (0.64) than the MODIS AOD (0.36). These correlation differ-ences might be attributed to smoke released from fires duringflaming and smoldering stages and impacting the cloud layers andsatellite viewing geometry. AAOD is less affected by sub-pixel cloudcontamination than AOD (Ahn et al., 2008) and thus might be areason for relatively good correlation strength observed in ourstudy. However, in-depth study might be needed to account forcloud contamination at a sub-pixel level. The AI value is positivewhen absorbing aerosols are present, whereas clouds yield a nearlyzero value of AI (Guan et al., 2010). AI varies with aerosol layerheight, optical depth and single scattering albedo and thatabsorbing aerosols can still be detected if mixed with clouds orlocated above clouds (Torres et al., 1998; Jeong and Hsu, 2008). Inaddition, as the surface albedo of land is much less in the UV range,

Fig. 8. a-b. Hysplit model showing air mass trajectories from fire locations in Riau ProvinceMalaysia and southern Thailand. The colors denote air parcels released at different heights

aerosol detection using UV is more robust than other wavelengths(De Graaf et al., 2005). These inferences from previous literatureand our current study support OMI AAOD and OMI UVAI asimportant indicators of biomass burning smoke plumes.

In the study region, forest clearing fires that occur during Augustand September are tightly linked to the monsoonal systems(Murphy, 2006). The southwest monsoon winds blow from Aprilthrough September while the northeast monsoon occurs fromNovember to February. From April through September, the eastcoast of Sumatra in Indonesia is sheltered from the southwestmonsoon, resulting in a dry season. It is therefore common for thelocals to clear land for farming during the dry season. However, thefire episode that occurred during June 2013 is an unusual fire event.In general, most of the winds during June are westerly, however,during June, 2013, the smoke plumes from the fires in Riau Prov-ince, Indonesia traveled towards southern Malaysia and Singaporepushing pollution to the highest levels. Results from the forwardtrajectories showed similar information from the Hysplit model(Fig. 8a,b) . We used the forward Hysplit runs as the sources ofpollution were known, i.e., fire hotspot locations in the Riau Prov-ince. Thus, the trajectory starting points for Hysplit were selectedprimarily from collocating the MODIS active fires during June, 2013over the Riau Province. We ran the Hysplit model forward in timeevery six hours to track the air parcels from the source. Tick marksand the numbers 00, 06, 12 and 18 h are used along each trajectoryto indicate the position of the air parcel over the six day period(Fig. 8a,b). Hysplit forward trajectory results from the source loca-tions showed the air mass at 500 and 1000 m altitudes movingfrom west-to-east towards Malaysia and Singapore on 17th and19th June till 23rd and 25th June, 2013 respectively. Coinciding withthe air masses, increased pollution were reported from Singaporeand Malaysia (Greenpeace, 2013). Our results highlight the use-fulness of satellite remote sensing data in capturing the pollution

, Indonesia. Air masses traveled towards east impacting pollutant levels in Singapore,(m) above ground level (AGL) (green-1100 m; blue-1000 m; red-900 m).

K.P. Vadrevu et al. / Environmental Pollution 195 (2014) 245e256 255

events. We advocate the integration of satellite remote sensing datafrom different sources for building innovate pollution monitoringsystems in the region to address haze impacts and health concerns.

6. Conclusions

A massive haze and pollution episode that occurred during June2013 in Southeast Asia has been analyzed using satellite remotesensing products. We integrated multiple satellite datasets to inferthe aerosol and CO characteristics. MODIS fire data were quiteuseful for inferring fire seasonality and relating to aerosol param-eters. We observed enhanced pollution levels in AOD, AAOD, SMF,UVAI as well as CO coinciding with the fire events. All these valueswere significantly higher than previous years (2008e2012). Ingeneral, most of the winds during June in the region are westerly,however, during June, 2013 the winds blew east, as a result thesmoke plumes from the fires in Riau Province, Indonesia traveledtowards southern Malaysia and Singapore causing intense pollu-tion. These results were confirmed from the Hysplit trajectorymodel. Satellite datasets were quite useful at inferring spatial aswell as temporal variations in fires, aerosols and CO.

References

Ahn, C., Torres, O., Bhartia, P.K., 2008. Comparison of ozone monitoring instrumentUV aerosol products with Aqua/Moderate Resolution Imaging Spectroradi-ometer and Multiangle Imaging Spectroradiometer observations in 2006.J. Geophys. Res. Atmos. 113.

Andreae, M.O., Gelencs�er, A., 2006. Black carbon or brown carbon? the nature oflight-absorbing carbonaceous aerosols. Atmos. Chem. Phys. 6, 3131e3148.

Badarinath, K.V.S., Kharol, S.K., Latha, K.M., Chand, T.R., Prasad, V.K., Jyothsna, A.N.,Samatha, K., 2007. Multiyear ground-based and satellite observations of aerosolproperties over a tropical urban area in India. Atmos. Sci. Lett. 8, 7e13.

Badarinath, K.V.S., Kharol, S.K., Sharma, A.R., Vadrevu, K.P., 2009. Analysis of aerosoland carbon monoxide characteristics over Arabian Sea during crop residueburning period in the Indo-Gangetic Plains using multi-satellite remote sensingdatasets. J. Atmos. Solar-Terrest. Phys. 71, 1267e1276.

Bergstrom, R.W., Russell, P.B., Hignett, P., 2002. Wavelength dependence of theabsorption of black carbon particles: predictions and results from the TARFOXexperiment and implications for the aerosol single scattering albedo. J. Atmos.Sci. 59.

Bucsela, E.J., Perring, A.E., Cohen, R.C., Boersma, K.F., Celarier, E.A., Gleason, J.F.,Wenig, M.O., Bertram, T.H., Wooldridge, P.J., Dirksen, R., Veefkind, J.P., 2008.Comparison of tropospheric NO2 from in situ aircraft measurements with nearreal- time and standard product data from OMI. J. Geophys. Res. 113.

Carlson, T.N., Ripley, A.A., 1997. On the relation between NDVI, fractional vegetationcover and leaf area index. Remote Sens. Environ. 62, 241e252.

Chan, C.Y., Wong, K.H., Li, Y.S., Chan, L.Y., Zheng, X.D., 2006. The effects of SoutheastAsia fire activities on tropospheric ozone, trace gases and aerosols at a remotesite over the Tibetan Plateau of Southwest China. Tellus B 58, 310e318.

Corlett, R.T., 2011. Impacts of warming on tropical lowland rainforests. Trends Ecol.Evol. 26, 606e613.

De Graaf, M., Stammes, P., Torres, O., Koelemeijer, R.B.A., 2005. Absorbing AerosolIndex: sensitivity analysis, application to GOME and comparison with TOMS.J. Geophys. Res. Atmos. (1984e2012) 110.

Deeter, M.N., Emmons, L.K., Edwards, D.P., Gille, J.C., Drummond, J.R., 2004. Verticalresolution and information content of CO profiles retrieved by MOPITT. Geo-phys. Res. Lett. 31.

Dobermann, A., Fairhurst, T.H., 2002. Rice straw management. Better Crops Int. 16,7e9.

Draxler, R.R., Hess, G.D., 1998. An overview of the HYSPLIT_4 modelling system fortrajectories. Aust. Meteorol. Mag. 47.

Drummond, J.R., 1992. Measurements of pollution in the Troposphere (MOPITT). In:Gille, J.C., Visconti, G. (Eds.), The Use of EOS for Studies of Atmospheric Physics.North-Holland, New York, pp. 77e101.

East Asia Forum, 2013. Indonesia's Fires: a Hazy Challenge for Southeast Asia.http://www.eastasiaforum.org/2013/07/08/indonesias-fires-a-hazy-challenge-for-southeast-asia/.

FAO, 2013. http://data.fao.org/dataset?entryId¼76a06e13-635c-47e5-a899-d66fd99eb3e6.

Feng, N., Christopher, S.A., 2013. Satellite and surface-based remote sensing ofSoutheast Asian aerosols and their radiative effects. Atmos. Res. 122, 544e554.

Folkins, I., Chatfield, R., Baumgardner, D., Proffitt, M., 1997. Biomass burning anddeep convection in southeastern Asia: results from ASHOE/MAESA. J. Geophys.Res. Atmos. (1984e2012) 102, 13291e13299.

Giglio, L., Descloitres, J., Justice, C.O., Kaufman, Y.J., 2003. An enhanced contextualfire detection algorithm for MODIS. Remote Sens. Environ. 87, 273e282.

Giglio, L., Randerson, J.T., Werf, G.R., 2013. Analysis of daily, monthly, and annualburned area using the fourth-generation global fire emissions database(GFED4). J. Geophys. Res. Biogeosci. 118, 317e328.

Ginoux, P., Torres, O., 2003. Empirical TOMS index for dust aerosol: applications tomodel validation and source characterization. J. Geophys. Res. Atmos.(1984e2012) 108.

Greenpeace, 2013. http://www.greenpeace.org/international/en/press/releases/Palm-oil-leading-cause-of-Indonesia-forest-destruction-RSPO-leaves-big-brands-exposed/.

Guan, H., Esswein, R., Lopez, J., Bergstrom, R., Warnock, A., Follette-Cook, M.,Iraci, L.T., 2010. A multi-decadal history of biomass burning plume heightsidentified using aerosol index measurements. Atmos. Chem. Phys. 10,6461e6469.

Heil, A., Goldammer, J., 2001. Smoke-haze pollution: a review of the 1997 episode inSoutheast Asia. Reg. Environ. Change 2, 24e37.

Hsu, N.C., Herman, J.R., Torres, O., Holben, B.N., Tanre, D., Eck, T.F., Lavenu, F., 1999.Comparisons of the TOMS aerosol index with Sun-photometer aerosol opticalthickness: results and applications. J. Geophys. Res. Atmos. (1984e2012) 104,6269e6279.

Hyer, E.J., Chew, B.N., 2010. Aerosol transport model evaluation of an extremesmoke episode in Southeast Asia. Atmos. Environ. 44, 1422e1427.

Hyer, E.J., Reid, J.S., Prins, E.M., Hoffman, J.P., Schmidt, C.C., Miettinen, J.I., Giglio, L.,2013. Patterns of fire activity over Indonesia and Malaysia from polar andgeostationary satellite observations. Atmos. Res. 122, 504e519.

Ichoku, C., Ellison, L., 2013. Global top-down smoke aerosol emissions estimationusing satellite fire radiative power measurements. Atmos. Chem. Phys. Discuss.13, 27327e27386.

Jeong, M.J., Hsu, N.C., 2008. Retrievals of aerosol single-scattering albedo andeffective aerosol layer height for biomass-burning smoke: synergy derived from“A-Train” sensors. Geophys. Res. Lett. 35.

Kanniah, K.D., Yaso, N., 2010. Preliminary analysis of the spatial and temporalpatterns of aerosols and their impact on climate in Malaysia using MODISsatellite data. ISPRS J. Photogram. Remote Sens. 38, 386e391.

Kaufman, Y.J., Koren, I., Remer, L.A., Rosenfeld, D., Rudich, Y., 2005. The effect ofsmoke, dust, and pollution aerosol on shallow cloud development over theAtlantic Ocean. Proc. Natl. Acad. Sci. U. S. A. 102, 11207e11212.

Kleidman, R.G., O'Neill, N.T., Remer, L.A., Kaufman, Y.J., Eck, T.F., Tanr�e, D.,Holben, B.N., 2005. Comparison of Moderate Resolution Imaging Spectroradi-ometer (MODIS) and aerosol robotic network (AERONET) remote-sensing re-trievals of aerosol fine mode fraction over ocean. J. Geophys. Res. 110.

Langmann, B., Heil, A., 2004. Release and dispersion of vegetation and peat fireemissions in the atmosphere over Indonesia 1997/1998. Atmos. Chem. Phys. 4,2145e2160.

Lin, J., Nielsen, C.P., Zhao, Y., Lei, Y., Liu, Y., McElroy, M.B., 2010. Recent changes inparticulate air pollution over China observed from space and the ground:effectiveness of emission control. Environ. Sci. Technol. 44, 7771e7776.

Miettinen, J., Shi, C., Liew, S.C., 2011. Influenceofpeatlandand landcoverdistributiononfire regimes in insular Southeast Asia. Reg. Environ. Change 11, 191e201.

Monks, P.S., Beirle, S., 2011. Applications of satellite observations of troposphericcomposition. In: Burrows, J.P., et al. (Eds.), The Remote Sensing of TroposphericComposition from Space, Physics of Earth and Space Environments. Springer-Verlag, Berlin, Heidelberg.

Morisette, J.T., Giglio, L., Csiszar, I., Justice, C.O., 2005. Validation of theMODIS activefireproduct over SouthernAfricawith ASTER data. Int. J. Remote Sens. 26, 4239e4264.

Murdiyarso, D., Lebel, L., Gintings, A.N., Tampubolon, S.M.H., Heil, A., Wasson, M.,2004. Policy responses to complex environmental problems: insights from ascienceepolicy activity on transboundary haze from vegetation fires in South-east Asia. Agric. Ecosyst. Environ. 104, 47e56.

Murphy, K., 2006. The ENSO-fire dynamic in insular Southeast Asia. Clim. Change74, 435e455.

Nara, H., Tanimoto, H., Nojiri, Y., Mukai, H., Machida, T., Tohjima, Y., 2011. Onboardmeasurement system of atmospheric carbon monoxide in the Pacific byvoluntary observing ships. Atmos. Meas. Tech. 4 (11), 2495e2507.

Nichol, J., 1998. Smoke haze in Southeast Asia: a predictable reoccurrence. Atmos.Environ. 32, 2715e2716.

Page, S.E., Rieley, J.O., Banks, C.J., 2011. Global and regional importance of thetropical peatland carbon pool. Glob. Change Biol. 17, 798e818.

Radke, L.F., Hegg, D.A., Lyons, J.H., Brock, C.A., Hobbs, P.V., Weiss, R., Rasmussen, R.,1988. Airborne measurements on smokes from biomass burning. AerosolsClim., 411e422.

Radojevic, M., 2003. Chemistry of forest fires and regional haze with emphasis onSoutheast Asia. Pure Appl. Geophys. 160, 157e187.

Rajab, J.M., Tan, K.C., Lim, H.S., Mat Jafri, M.Z., 2011. Investigation on the carbonmonoxide pollution over Peninsular Malaysia caused by Indonesia forest firesfrom AIRS daily measurement. Adv. Air Pollut., 115e136.

Remer, L.A., Kaufman, Y.J., Tanr�e, D., Mattoo, S., Chu, D.A., Vanderlei Martins, J.,Li, R.R., et al., 2005. The MODIS aerosol algorithm, products and validation.J. Atmos. Sci. 62, 947e973.

Richter, A., Wagner, T., 2011. The use of UV, Visible and Near IR solar back scatteredradiation to determine trace gases. In: Burrows, J.P., et al. (Eds.), The RemoteSensing of Tropospheric Composition from Space, Physics of Earth and SpaceEnvironments. Springer-Verlag, Berlin, Heidelberg.

See, S.W., Balasubramanian, R., Wang, W., 2006. A study of the physical, chemical,and optical properties of ambient aerosol particles in Southeast Asia duringhazy and nonhazy days. J. Geophys. Res. 111.

K.P. Vadrevu et al. / Environmental Pollution 195 (2014) 245e256256

Siegert, F., Ruecker, G., Hinrichs, A., Hoffmann, A.A., 2001. Increased damage fromfires in logged forests during droughts caused by El Nino. Nature 414,437e440.

Tanr�e, D., Kaufman, Y.J., Holben, B.E.A., Chatenet, B., Karnieli, A., Lavenu, F.,Smirnov, A., 2001. Climatology of dust aerosol size distribution and opticalproperties derived from remotely sensed data in the solar spectrum. J. Geophys.Res. Atmos. (1984e2012) 106, 18205e18217.

Taylor, D., 2010. Biomass burning, humans and climate change in Southeast Asia.Biodivers. Conserv. 19, 1025e1042.

Torres, O., Bhartia, P.K., Herman, J.R., Ahmad, Z., Gleason, J., 1998. Derivation ofaerosol properties from satellite measurements of backscattered ultravioletradiation: theoretical basis. J. Geophys. Res. Atmos. (1984e2012) 103,17099e17110.

Torres, O., Tanskanen, A., Veihelmann, B., Ahn, C., Braak, R., Bhartia, P.K., Levelt, P.,2007. Aerosols and surface UV products from ozone monitoring instrumentobservations: an overview. J. Geophys. Res. Atmos. (1984e2012) 112.

Torres, O., Jethva, H., Bhartia, P.K., 2012. Retrieval of aerosol optical depth aboveclouds from OMI observations: sensitivity analysis and case studies. J. Atmos.Sci. 69.

Usup, A., Hashimoto, Y., Takahashi, H., Hayasaka, H., 2004. Combustion and thermalcharacteristics of peat fire in tropical peatland in Central Kalimantan, Indonesia.Tropics 14, 1e19.

Vadrevu, K.P., Justice, C.O., 2011. Vegetation fires in the Asian region: satelliteobservational needs and priorities. Global Environ. Res. 15, 65e76.

Wittich, K.P., Hansing, O., 1995. Area-averaged vegetative cover fraction estimatedfrom satellite data. Int. J. Biometeorol. 38, 209e215.