Observing and understanding the Southeast Asian aerosol system by remote sensing: An initial review and analysis for the Seven Southeast Asian Studies (7SEAS) program

TL;DR: In this article, the authors focus on and repeatedly link back to the primary data source, satellite aerosol remote sensing and associated observability issues, and discuss aspects of SEA's physical, socio-economic and biological geography relevant to meteorology and observability problems associated with clouds and precipitation.

About: This article is published in Atmospheric Research.The article was published on 2013-03-01 and is currently open access. It has received 273 citations till now. The article focuses on the topics: Southeast asian & AERONET.

Summary

1. Introduction

• The 2007 IPCC Report on Impacts, Adaptation and Vulnerability lists Southeast Asia as one of the most vulnerable regions of the world to climate change (IPCC, 2007a, Table 10.11).
• Over the last several decades, the region extending from the Maritime Continent (MC) of Brunei, Indonesia, Malaysia, Singapore and Timor through the Indochina (IC) areas of peninsular SEA (e.g., Cambodia, Laos, Myanmar, Thailand, Vietnam) and the islands surrounding the South China Sea/East Sea (SCS/ES, including the Philippines and Taiwan) has seen significant economic and population growth.
• Air pollution originating from such sources as biomass burning, industry, mobile sources, biofuel, and domestic cooking, in conjunction with the domestic and international demand for agricultural products such as palmoil, sugar, and rice leading to deforestation and biomass burning, has further reduced air quality.
• 3) Satellite data records are temporally inconsistent, with sensors of varying efficacy entering and leaving the observing system; and 4) SEA has ubiquitous cloud cover which interferes with many relevant retrievals.

1.1. The 7SEAS program and its relation to this paper

• To help understand the SEA aerosol–environment system, the 7-Southeast Asian Studies (7SEAS) project was created.
• The cornerstone of 7SEAS is remote sensing through product analysis as well as data assimilation in models (Zhang et al., 2008).
• While a review paper on SEA could take a number of different approaches, such as a focus on emissions, particle chemistry, carbon pools, climate impacts, etc., here the authors wish to set the stage for the largely ground and network based 7SEAS activities.

1.2. Data and information used in this paper

• The authors should emphasize that this is a review paper with regional critical analysis.
• But in general, the authors use “standard” datasets that are easily available to the authors and potential users.
• In all cases, citations are given to papers describing in detail the data processing involved.
• The authors do point to localized studies to evaluate global product strengths and weaknesses.

2. Rationale: uncertainty in the SE Asian aerosol system

• Much of this manuscript on the study of aerosol lifecycles and impacts in SEA tries to bridge the gap between physical understanding and remote sensing.
• The authors note that nearly all of aerosol particles' impacts are interdisciplinary, both in science and in outcome.
• From 406 J.S. Reid et al. / Atmospheric Research 122 (2013) 403–468 satellite observations, scientists can only infer desired quantities from retrieved products by relating them to quantities that are measurable, such as spectral radiance.
• Remote sensing approaches systematically undersample environmental variability and can underestimate or miss significant events, which often have the largest impact signals.

2.1. Interdisciplinary nature of aerosol impacts in SEA

• In considering the role of aerosol particles in the SEA earth system, the authors can broadly assign impact categories such as environmental quality, climate, meteorology, etc.
• For SEA, notable studies include: Environmental quality and chemistry: SEA urban centers and rural receptors are known to suffer from poor regional air quality.
• It is hypothesized that absorption of solar radiation in the atmosphere coupled with the microphysical influence of aerosol particles on cloud droplets impact cloud formation and lifetime (Mcfarquahar and Wang, 2006; Dey et al., 2011) as well as precipitation (Barbel, 2007; Tosca et al., 2010).
• The carbon pools of SEA, especially in relation to peatland, are tremendous (Hergoualch and Verchot, 2011).

2.2. Nature of remote sensing science

• While in the Introduction the authors argued that atmospheric research, particularly in SEA, is heavily dependent on remote sensing, in most cases satellites do not directly provide what is needed to understand these relationships.
• If UV or polarization channels are applied, additional information on aerosol particle speciation, absorption or cloudy scene loadings can be derived.
• Higher resolution imagery improves these twomeasurements, at the expense of temporal coverage.
• Estimates of fire emissions 407J.S. Reid et al. / Atmospheric Research 122 (2013) 403–468 remain highly under-constrained, even if fire observations were completely accurate.
• That is, almost everything scientists need requires inference from under-determined retrievals.

2.3. Perceived variability in remote sensing data

• As the authors demonstrate, taken at face value, satellite products tend to have some qualitative agreement.
• At the same time, the dynamic range in Fig. 1 suggests that diversity between climatological averages from commonly used products often varies by a factor of 2 or more.
• Many individual smoke plumes are visible that are not detected by the fire hotspot algorithm.
• As an example, Système Pour l'Observation de la Terre‐SPOT imagery (©CentreNational d'Etudes Spatiales— CNES, 2005 and 2006) showing forest and agricultural burning on Sumatra is provided in Fig. 3(e) and (f) respectively.
• This review paper will not resolve the issue, but the authors hope that it will alert readers to those processes that interfere with observations, confuse the physical interpretation, and obstruct quantitative analysis of aerosol processes in SEA.

3. Fundamentals of Southeast Asia geography

• The physical review begins with a brief section on the physical, socio-economic, and biological geography of SEA.
• Population densities are high, with numerous mega cities in most SEA countries.
• The human encroachment on SEA forests, and in particular peatlands, is an area of high scientific interest.
• The continental portion of SEA includes Burma , Thailand, Laos, Cambodia, and Vietnam.
• First, the Philippines are commonly considered to be part of the MC, yet as the authors will show, climatologically are better grouped with IC.

3.2. Physical geography

• The topographic geography of SEA is complex including a series of high mountain ranges and large lowlands which influence their later interpretation of regionalmeteorology and aerosol lifecycle.
• The IC hosts a series of mountain ranges fanning out from the southeast corner of the Tibetan Plateau, separated by plateaus and river drainage basins.
• The Tenassem Hills run from the SE corner of Tibet roughly along the Burmese and Thai border, and down the Malay Peninsula with peaks up to 1km.
• This region has over 200 Holocene (active within the last 10,000years) volcanoes,with the vastmajority in Indonesia (145) and the Philippines (47) (Smithsonian Institute, Global Volcanism Network).
• As discussed in Section 6.5, not only is MC volcanic activity important for transport but it is also an important primary and precursor aerosol source, particularly for SO2.

3.3. Social geography

• The social geography of SEA is clearly demonstrated in Fig. 5(c), which logarithmically plots population density derived from the Gridded Population of the World, version 3 open dataset (http://sedac.ciesin.columbia.edu/gpw/).
• In IC in particular, the north–south mountain ranges clearly delineate population, cultural, and economic boundaries, which in turn influence the spatial distribution of emissions.
• Central Luzon/Manila in the Philippines, and the Hanoi–Haiphong corridor and Ho Chi Minh City in Vietnam also have ultra-high population densities, and total urban populations in the 30–35M range.
• The socio-economic disparity among the countries of SEA, as well as between urban and rural populationswithin each country, is extreme.
• Fuel for domestic cooking and heating use varies from electric, to propane, to various forms of biofuel (including wood, charcoal, dung, and rice stubble).

3.4. Biological geography

• The combination of physical and socio-economic geography of SEA also strongly relates to SEA's biological geography.
• Land use issues are critically important to climate change as awhole, and in regard to this review, to fire emissions.
• The drier areas of east Java, Bali and Timor contain natural dry tropical forest, with a more seasonal climate (Whitmore, 1984).
• Both of these products are based on MODIS data, but use different processing and ancillary data.

4. Southeast Asia meteorology and related satellite products

• Straddling the equator, Southeast Asia has a monsoonal climate largely dictated by the annual migration of the Inter-tropical Convergence Zone (ITCZ), with a dry and wet IC in boreal winter and summer respectively.
• This treatment is informative but ultimately reductive, glossing over the multitude of multi-scale meteorological features, ranging from interannual El Niño Southern Oscillation (ENSO), the Madden Julian Oscillation (MJO), tropical Kelvin, easterly and Rossby waves, tropical cyclones, the sea breeze, and isolated convection, to name a few.
• Cloud cover in SEA is nearly ubiquitous, with thin cirrus covering 80% of the sky.
• In a brief review of meteorological remote sensing in SEA, the authors show how SEA is not only one of the most challenging regions of the world for remote sensing, but displays large diversity in cloud and precipitation products as well.
• The human element (Vayda, 2006), including how anthropogenic activities such as biomass burning co-vary with meteorology, must also be considered.

4.1. Key monsoonal features

• There is no single review paper that covers all of the important aerosol relevant meteorology for all of SEA.
• Fig. 6(a) and (b) intend to demonstrate the overall nature of the SEA monsoon.
• What precipitation is visible in IC in Fig. 1(a) is largely from the transitionalmonth ofMay, although on very rare occasions cold fronts from the north enter IC.
• In the summer, the precipitation feature is associated with what local forecasters term the “West Sumatran Low”, a vorticity maximum formed by counter flowing monsoonal winds, with easterlies to the south, westerlies to the north in the vicinity of the Barisan Range of Sumatra (Reid et al., 2012).
• As one would expect, lightning over SEA is concentrated over land, with very specific hotspots, particularly around Sumatra and the Malay Peninsula in the MC (Sow et al., 2011), and southern Vietnam/Cambodia and Thailand in the IC.

4.2. Meteorological scales

• The two phase view of SEA meteorology in this section is enlightening, but one must consider that the monsoon is an evolving entity.
• It must be recognized that the meteorology of SEA has interconnecting influences among all of the above meteorological scales.
• Boreal summer precipitation over the MC is strongly negatively correlated with ENSO indices, and these dry periods can carry over all through the following spring (McBride et al., 2003), a fact punctuated by the massive accompanying fire event of 1997.
• Thus, from an aerosol point of view, the wet and dry phases of the MJO largely dictate the timing of significant smoke events in the MC (Reid et al., 2012), and the MJO was hypothesized to influence overall AOD (Tian et al., 2008).
• In the boreal summer the convection associated with MJO propagates eastward and northward, contributing to the Asian monsoon active and break cycles.

4.3. Clouds in SEA

• Themeteorology associated with the wet and dry monsoon can also be examined in the context of cloudproperties.
• The authors also note the high boundary layer cloud fraction over the region, typically with cloud top coverage peaking in the 500–1000m altitude bin over water, and 1km to 2km above ground altitude over land, corresponding well to the expected ~500m marine boundary layer (MBL) and ~2000m planetary boundary layer (PBL) heights.
• In context, the RGB images provided in Fig. 3 are considered to be some of the best days available over the 10year MODIS data record for surface and aerosol imagery.
• Most recently with Cloud–Aerosol Lidar and Infrared Pathfinder Satellite Observations—CALIPSO data, Virts and Wallace (2010) found ~20% thin cloud fraction for cirrus over 15km in height.

4.4. Variability in satellite cloud properties in SEA

• Already from Section 4.3, it is clear that satellite cloud data products are weighted to different parts of the cloud system, depending on the satellite instrument utilized.
• Difficulty emerges is in the transition from case study to climatology, anytime a large number of samples is required, and in the subsequent interpretation of such climatologies by a broader scientific community who may be less familiar with the details of its generation.
• In Fig. 9 the CALIPSO lidar, arguably the most sensitive instrument for thin cirrus detection, shows greater than 90% cloud cover over half of the SEA domain, much of which comes from high clouds.
• Note that while these plots show large differences amongst the datasets, these differences can be interpreted into meaningful information about cloud properties when one is familiar with the remote sensing techniques used to generate these datasets, a point that is nicely explored further in Marchand et al. (2010).
• Difficulties also arise in assigning a proper crystal habit during the retrieval.

4.5. Satellite precipitation products in SEA

• Given the strong coupling between precipitation, fire, emissions, and perhaps feedbacks between them, an understanding of the precipitation fields in SEA is critical.
• Most importantly, the joint NASA and Japanese Space Agency—JAXA Tropical Rainfall Measuring Mission (TRMM) continues (as of early-2012) to collect joint passive microwave (PMW) and Ku-band (14GHz) precipitation radar (PR) observations since late 1997.
• In recent years, a series of satellitebased and model-based precipitation validation experiments performed under the direction of the International Precipitation Working Group (IPWG) has shown that NWP models tended to underperform these multi-satellite based precipitation techniques (on a daily, 1/4-degree scale) in the tropics, with the opposite behavior for the mid-latitudes and cold seasons [Ebert et al., 2007].
• But, such diversity does demonstrate the relative amount of baseline uncertainty in these products.
• The utility of these products for quantitative uses is an outstanding issue for SEA meteorology in general.

5. Biomass burning

• Based on their knowledge of regional geography and meteorology, in this section the authors examine the nature of fire and smoke particle emissions in SEA.
• While the largest events are indeed associated with these El Niño peat burning regimes, fire activity and smoke emissions in the Maritime Continent are important in all years.
• Indochina, which actually has higher observed fire prevalence compared to the MC, has received much less attention; this is despite significant deforestation there as well.
• This comparison, which demonstrates diversity between products, does little to aid in actual verification.
• These products can capture broad spatial and temporal patterns of fire activity, but their quantitative use requires significant statistical adjustment.

5.1. The nature of biomass burning emissions in SEA

• Because of the publicity associated with Indonesian fire events, the public and scientists alike often associate SEA aerosol pollution with biomass burning; hence SEA aerosol research tends focuses on this aerosol source.
• While rapid regrowth of secondary forest at times prevents a clear “arc of deforestation” seen in other tropical regions, fire activity can nonetheless help indicate land change practices at work (e.g., Stolle and Lambin, 2003; Miettinen et al., 2010).
• Peat fires themselves are often thought of as being directly responsible for large carbon emissions.
• Carbon emissions directly from burningmay be small fraction of the total carbon released due to land use conversion.

5.2. Socio-economics of biomass burning in Southeast Asia

• Throughout SEA nearly all fires can be considered to be of anthropogenic origin, and fire is strongly connected to land use practices in the region.
• Historically, deforestation, agriculture and fire are often viewed as part of a shifting agricultural system such as swidden (i.e., slash and burn).
• The seasonal cycle of observed fires follows a very clear pattern that is loosely opposite of the winter and summermonsoonal trough (Giglio et al., 2006; Reid et al., 2012).
• El Niño also results in cooler waters around the MC, which may reduce maritime convection or latent heat flux; conversely during La Niña positive SST anomalies are located upstreamof enhanced convection over the MC (McBride et al., 2003).
• For Sumatra, Indonesia, Stolle et al. (2004) examined 8 active fire product (including mainstay MODIS, AVHRR, ATSR and OLS fire products), and found that two-third of the fires detected by one dataset were not detected by any other dataset, nor did any of the datasets detect fires in all of their test areas.

5.5. Biomass burning particle emissions

• As shown in the previous section, satellite fire counts from both polar (MODIS, AVHRR, ATSR/AATSR) and the geostationary sensors (WF_ABBA from MTSAT) are semi-quantitative at best in the IC, and likely qualitative in the MC, with extremely large discrepancies found in an intercomparison of satellite products in Indonesia (Hyer et al., 2013-this issue).
• Themagnitude of diversity in firemonitoring products can be gauged through a simple comparison of emissions products which make use of fire hot spots (such as FLAMBE, Reid et al., 2009) and burn scars (such as GFEDv3 van Der Werf et al., 2010).
• Qualitatively the two products agree in relative geography and modal month.
• Fig. 3(d) illustrates some limitations of active fire detection methods; the scene shows massive smoke production, but few detected hot spot fires.

6. Bulk properties of Southeast Asian aerosol particles

• The bulk chemical, microphysical, and optical properties of SEA aerosol particles are reviewed, emphasizing those needed for aerosol remote sensing.
• Compared to other parts of the world, there are very few direct in situ measurements of aerosol microphysical properties in SEA, with none to their knowledge within Indochina itself.
• Some information can be inferred from existing chemistry studies.
• Tohelp bring some consistency to comparisons across SEA, climatological averageAERONETdata are provided and compared to other isolated retrievals in the literature.

6.1. Nature of aerosol particle properties and their measurement in SEA

• To apply aerosol remote sensing to practical problems such as air quality, radiative forcing, or cloud impacts, the authors are interested in microphysical models which drive the relationship between mass, size, angular scattering, and absorption.
• To put it another way, the authors are interested in methods that link particle mass, number, speciation, thermodynamics, optical properties, or radiative flux which can be measured in situ, with what can be measured by satellite, such as spectral or angular radiance, lidar backscatter, or polarization.
• Relative to other parts of the world there are few peer reviewed papers on measurements of aerosol properties in SEA.
• Such environmental diversity is a truism for frequent travelers of the region.
• The realization that aerosol particles in the MC have a substantial semi-volatile fraction (He et al., 2010) may change previous thinking about the formulation of sampling strategies.

6.2. Urban and industrial aerosol environment

• Travelers in SEA will likely notice differences in emission sources between urban centers.
• Singapore has a modern vehicle fleet and air quality controls, but has significant petroleum refining activity.
• There is to their knowledge only one published paper on fundamental air quality parameters for Jakarta, the most populous urban area in SEA (Zou and Hooper, 1997).
• But there are significant differences for individual cities even between studies.

6.3. Biomass burning

• There are even fewer comprehensive papers with measurements of biomass burning in SEA than there are on urban pollution.
• While there have been fewmeasurements of optical properties, there have been a few lab measurements and extrapolations which give very different views of the biomass burning system.
• Most troubling are divergent estimates of ambient ionic mass fraction and hygroscopicity.
• A number of studies on aerosol composition for smoke in SEA have been reported, but most of these studies were focused on measuring specific aerosol components, and the specificmeasurements reported are not easy to apply to remote sensing problems.

6.4. Rural

• In addition to urban and biomass burning emissions the authors must consider rural emissions.
• In some cases such as on Java, these mosaic landscapes can have very high population densities.
• In Indonesia, which has rapid forest regrowth, residents are heavily dependent on wood fuels, leaving 73% of crop residue to burn in the field.
• There have been few published studies of aerosol chemistry and absorption in rural SE Asia.
• Optically derived black carbon mass fraction was still high however, ~15%.

6.5. Volcanic

• Given the recent improvements in space-based observation of atmospheric constituents, there has been a radical advance in monitoring volcanic emissions of such species as ash and SO2 (e.g., review papers bookended by Rose et al., 2000 to Thomas and Watson, 2010).
• Mt. Pinatubo on the island of Luzon, Philippines, for example, erupted in 1991 and emitted approximately 11km3 of tephra (ash, rock, etc.) and 10Mt of sulfur [Bluth et al., 1992], with effects felt worldwide.
• The magnitude of volcanic activity affects the emission behavior of volcanic aerosols and their subsequent local, regional and global effects.
• So, if size can be constrained, the authors may be able to infer some key optical properties from measurements.
• For venting of SO2 into a tropical marine environment, the authors would expect half lives on the order of 6h (Porter et al., 2002), although plumes aloft may have 10 times that rate (Carn et al., 2001).

6.6. AERONET and other sun–sky derived properties

• Inference of aerosol properties from sun–sky inversions is one of the few consistent methods practiced over the world.
• Cloud screening mechanisms are implemented in the AERONET level 2 products (Holben et al., 1998 and Smirnov et al., 2000).
• Given the lower overall AOD, the relative fine/coarse partition in the MC is more sensitive to cirrus contamination.
• Hong Kong shows high AODs (~0.85); this is not surprising given its location the middle of the highly populated Pearl River Delta.
• With particle size and ωo slightly higher than Chiang Mai in the source region, aerosol optical properties are consistent with known evolution (Reid et al., 2005b).

7. Long range aerosol transport patterns

• Because large aerosol events in SEA are almost always associated with biomass burning, the focus of the literature has been biomass burning transport.
• Once in the free troposphere, the smoke can be transported out over the Pacific Ocean and beyond.
• But even for the MC, most previous studies that addressed transport treated it as a supporting role rather than a core topic.
• Some transport information can be found in the review paper by Lawrence and Lelieveld (2010), but much of SEA is dealt with only on the periphery.
• Average profiles are the standard CALIOP products aggregated from 5km cloud free retrievals for 2007–2009.

7.2. Long-range aerosol transport in Indo-China

• Emissions ramp up in January–February, peak inMarch–April, and decay quickly in May with the onset of the summer monsoon.
• The shape of this plume aloft is very well captured in the OMI aerosol index climatology (Fig. 1(g)), which is enhanced by both the smoke's absorbing nature and altitude.
• As evident in the topographic maps of Fig. 5(b), the mountains between Laos and Vietnam do not extend all the way to the south.
• The land area of the IC is within the monsoonal trough and does not have a significant prevailing direction other than light westerly.

7.3. Long-range aerosol transport in the Maritime Continent

• For the MC there are few observations of aerosol vertical distribution.
• Similarly, the dominance of boundary layer aerosol makes meso-scale phenomenon important, including the sea breeze which is quite complex in the MC (e.g., Section 4).
• Smoke from the fire hotspots of central Sumatra and Borneo is largely influenced by the SCS/ES southwesterlies (Fig. 6(b)) and carried up into the Philippines for eventual scavenging into the summer monsoonal trough (Reid et al., 2012; Xian et al., 2013-this issue).
• While studies of the winter monsoon have focused on the northern half of the SCS/ES, there is no reason not to expect transport to the south.

7.4. Convective pumping and the summer monsoonal anti-cyclone

• While the vast majority of aerosol dynamics occurs in the lower troposphere, the nature of the mid and upper tropospheric aerosol environment has recently been receiving increased attention, notably through the SEAC4RS campaign.
• Indeed, Gonzi and Palmer (2010) found several cases during the 2006 El Niño event in the MC of both CO and aerosol particles being pumped to nearly 15km.
• The MLS CO product provided in Fig. 2(b) suggests this may be a common occurrence.
• The authors note here that while the pumping of primary particles is quite possible, under most circumstances the associated convection scavenges most particles.

8. The view of aerosol particles in Southeast Asia from space: diversity in common aerosol products

• In the final section of this review,the authors examine the diversity among common satellite based aerosol products in SEA.
• Ubiquitous cloud cover, shallow/ sediment loaded waters, complex land surfaces, evolving microphysics, and strong diurnal cycles conspire to create both random and systematic biases in all kinds of retrievals.
• This said, if the strengths and weaknesses of different sensors are considered, taken as a whole, a more consistent picture of SEA begins to emerge.
• Clearly, for aerosol problems in SEA, the selection of specific satellite products heavily depends on the application, and requires considerable forethought by investigators.
• MISR and OMI also have important roles to play.

8.1. Introduction to aerosol remote sensing in Southeast Asia

• In the previous sections the authors have provided an overview of the nature of the SEA environment, meteorology, aerosol properties and transport.
• Wewill present an overview of what the standard products from remote sensing can provide to understand the aerosol system, as well as key points on the magnitude of uncertainty.
• The authors review below is, in part, an attempt to clarify the key issues when utilizing these products at face value, pointing to examples of relevant literature for greater detail.

8.2. Dark target aerosol products

• Dark target retrievals (DTRs) are the most basic of retrievals (Fraser et al., 1984), and given their broad use and application to large swath sensors, they are the backbone of the global aerosol system.
• Conceptually, over ocean DTRs are the simplest retrievals, and hence it is easier to examine individual areas of bias for SEA.
• All of the products presented in Fig. 13, show the same basic patterns of AOD (except for Aqua Deep Blue because of albedo constraints in the MODIS Deep Blue product precluding retrieval over most of the area).
• High AODs exist in the Bay of Bengal and the coast of Asia.

8.2.1. Comparison of Terra and Aqua MODIS collection 5

• Fig. 15(a) and (b) shows the mean monsoonal differences between MODIS Terra and Aqua.
• This could very well be a result of radiance calibration for channels in the retrievals.
• Diurnal cycle of burning could systematically lead to higher AODs in the afternoon.

8.2.2. SeaWiFS Deep Blue and the lower boundary conditions

• This over-land bias between Terra and Aqua leads us to the next issue; the lower boundary condition.
• Shallow water where the ocean bottom contributes to the water leaving radiance, sediment loads from rivers, and the surf zone/shoals all make defining the lower boundary condition a challenge in all optical and IR wavelengths.
• Southeast Asia has thousands of islands, large delta regions (most notably theMekong and Sittang), and large shallows (such as the Java Sea and south of Vietnamwith depth ranges 20–60m).

8.2.3. Optical properties

• Optical retrievals do not constrain the variability of aerosol particle bulk optical properties.
• Many of these retrievals attempt to simultaneously constrain multiple atmospheric properties, such as spectral AOD and, over ocean, fine/coarse AOD partition and effective radius (variability in the land surface currently precludes such retrievals).
• There are two key difficulties in this process.
• Second, even if such a model exists, the radiances do not provide a unique solution to particle size distributions, chemistries and absorptions (i.e., solution degeneracy).
• This is exemplified in the regressions of Table 3, where the authors see regional differences in slope among satellite sensors, yet still strong correlations.

8.2.4. Cloud screening and impacts

• The use of DTRs is predicated on accurately isolating cloud-free parts of the scene, and cloud bias is an important issue.
• Similarly, Myhre et al., 2004, 2005a found little evidence for cloud bias in the NOAA AVHRR product (though there was bias in the AVHRR Global Aerosol Climatology Project—GACP product), suggesting stringent cloud clearing.
• In particular, the pedagogical study by Yang and Di Girolamo (2008) showed that the different sampling strategies commonly used in aerosol retrievals for mitigating these radiative perturbations can produce very different and potentially large biases within aerosol climatologies.

8.2.5. Comparison to AERONET

• The combination of lower boundary condition, variable aerosol particle microphysics, and cloud issues makes interpretation of individual retrievals in SEA difficult.
• A full AERONET comparison is outside this review.
• There is some limited direct verification data available for SEA for MODIS from Reid et al. (2009) and Hyer et al. (2011).
• While the regressions for Thailand behave reasonably well, the same cannot be said for regressions for other parts of SEA.
• Not only are thin cirrus clouds difficult to detect by passive satellite sensors, they can even corrupt well screened sun photometer verification data (Chew et al., 2011; Huang et al., 2011).

8.4. UV methods

• A set of products completely different from dark target and multi-angle retrievals comes from UV observations from the TOMS and OMI instruments.
• The OMI Aerosol Index (AI) is shown in Fig. 1(e) and (f). Fig. 1(e) and (f) shows that the AI very nicely highlights the IC smoke transect and smoke on Borneo.
• OMI AOD has a significant amount of variability, but it does, in some cases, pick up seasonal trends over water.
• Over land, in southern the product appears not to contain the signal of aerosol particles.

8.5. Space-based lidar

• Space-based lidars provide perhaps the most unique dataset for studying aerosol and cloud environments.
• Their ability to profile aerosol layers near the planet surface, even when attenuated by optically-thin cirrus layers in the upper troposphere such as are prevalent in SEA, adds a critical third dimension to studies.
• The last 5years has seen a revolution in 451J.S. Reid et al. / Atmospheric Research 122 (2013) 403–468 aerosol and cloud research with the launch and application of the NASA Cloud–Aerosol Lidar with Orthogonal Polarization , which has provided global datasets since the summer of 2006 (Winker et al., 2009).
• Consecutive orbital tracks on a single day near the equator are nominally ~2800km apart.
• Given the difficulty of obtaining accurate AOD measurements in SEA, this may not be a complete solution.

8.6. Radiative fluxes

• Sections 8.1–8.5 dealt with the monitoring of aerosol particles from space.
• Given the direct nature of aerosol particles' influence on the energy budget, a discussion of the measurement of radiative fluxes belongs here.
• InSEAenvironments, the separationof theaerosol signal from the ground and clouds is a difficult task.
• A careful assessment of clear sky reflected fluxes is needed since forcing is calculated as the difference in shortwave fluxes between clear and aerosol skies (Bellouin et al., 2008).

8.7. Gas products

• Finally, in this review, a few statements on gas products are appropriate.
• Gas retrievals benefit from better constraint of the optical properties of the target species relative to aerosol retrievals.
• They must generally separate multiple absorbing species in the spectral regions where they have sensitivity, and in some cases can experience interference from aerosol particles (the sensitivity to aerosol particles used to generate the TOMS and OMI AI may be considered a noise factor in the retrieval of ozone).
• NO2, SO2, CO, CO2, and a variety of trace products are now regularly produced from multiple platforms, and are powerful tools for understanding atmospheric composition (Logan et al., 2008).
• The interested reader should refer to the review paper by Wagner et al. (2008).

9. Discussion and conclusions: moving forward in the Southeast Asian aerosol system

• This review was conducted for the benefit of aerosol scientists wishing to begin research in the Southeast Asian (SEA) region, including the scientists in the 7 Southeast Asian Studies and Southeast Asia Composition, Cloud, Climate Coupling Regional Study (SEAC4RS).
• Compounding scientific problems are great observability difficulties posed by the SEA environment for both in situ and remote sensing measurements.
• To proceed past qualitative or semi-quantitative descriptions of processes to quantitative numbers, however, will require considerable effort in the community.
• All of these data streams need to be maintained and verified.
• In conclusion, SEA presents one of aerosol science's greatest challenges, where poor air quality and high climate change vulnerability meet with limited observability and spatial complexity.

Figures

Fig. 3.MODIS true-color images, depicting the complexity of the Southeast Asian biomass burning system. (a) Terra MODIS for 7 April 2002, with regional smoke and fire (red) transporting over a stratus deck in Vietnam; (b) Zoom of (a); (c) Massive Indonesian fire event on Borneo and Sumatra captured by Aqua MODIS Sept. 18, 2002; (d) Zoom of (c) with contrast enhancement; (e) SPOT 2 image of forest burning in Riau, Sumatra August 8, 2005 (RGB = NIR, red, green; that is vegetation is red, no vegetation/burn scar is green, smoke is white); (f) SPOT 4 Image of agricultural burning in south Sumatra, September 28, 2006 (RGB= SWIR, Near IR, red; that is fire is red, vegetation green, burn scar black, smoke bluish white ). (g), (h), (i) Photos of nontraditional burning: Invasive grass burn, stacked rice stubble burn and a brick kiln, respectively. Credits: (a)–(d) Adapted from NASA Visible Earth; (e) and (f) © CNES (2005 and 2006), acquired by Center for Remote Imaging, Sensing and Processing, National University of Singapore. (g)–(i) Robert Yokelson, University of Montana.

Fig. 17. Average 2006–2008 monsoonal (December–May and June–November) CERES (a) and (b) all sky and (c) and (d) clear sky fluxes.

Fig. 8. CALIPSO cirrus cloud statistics for 3×3 degree grids from August to October 2007, derived from the CALIPSO Level 2 5-km Cloud Profile product. (a) Total cloud probability. (b) Mean cloud top altitude for cloudy fields of view. (c) Mean cloud optical depth (COD) for non-attenuated retrievals. (d) Probability of the total cloud column exhibiting COD b0.3.

Fig. 12. Average 2007–2009 seasonal (see insets) CALIOP 532nm extinction profiles over Southeast Asia taken from the analysis by Campbell et al. (2013-this issue). Given in (a) is a map of the domain, (b) the Gulf of Tonkin; (c) the Bay of Bengal; (d) the southern South China Sea; (e) the Java Sea; (f) the biomass burning areas of Borneo; (g) an oceanic zone west of Sumatra; and (h) a zone on the southern border of the West Sumatran Low. Similar profiles for Singapore and Thailand can be found in Campbell et al. (2013-this issue).

Fig. 2. Same as Fig. 1 for seasonal OMI total column NO2 and MLS CO at 147 and 215hPa (~14 and 11km, respectively). MLS data courtesy of Ryan Fuller and Nathaniel Livesey, JPL.

Table 4 Published verification metrics from MODIS from Hyer et al. (2011) and MISR verification metrics from Kahn et al. (2010). “A”=AERONET.

Table 3 Regressions between aerosol optical depth (AOD) products of daily data for 2005–2007. Data are from daily averaged 2×2 degree latitude longitude boxes. “TM”=Terra MODIS. “AM”=Aqua MODIS.

Fig. 5.Maps of key geographic features of Southeast Asia. (a) NASA MODIS Blue Marble RBG product labeled with key countries, regions, and major urban centers with total urban area populations. (b) NASA Blue Marble Elevation map; (c) 2010 population Density (CIESIN, 2004); (d) MODIS 2001 Global Land Use (Friedl et al., 2002); (e) Regional subset of Miettinen MODIS composite. Note, middle/high montane forest and mosaic classes grouped here into single categories for simplicity.

Table 1 Southeast Asian regional reports of particulate matter and black carbon (BC) concentration.

Fig. 15. Average 2005–2007 monsoonal average (December–May and June–November) AOD differences between (a) and (b) Aqua and Terra MODIS Col 5, (b) and (c) Aqua MODIS and Deep Blue, (d) and (e) Aqua MODIS and SeaWiFS Deep Blue and (g) and (h) Aqua MODIS and MISR.

Fig. 9. Total, high, middle and low cloud fraction (as defined by ISCCP) for the (a) December–May and (b) June–November periods. Periods used in average are 2001–2009 for MISR and MODIS, 2000–2007 for ISCCP, and 2007–2008 for CALIPSO. This figure was drawn from the ongoing GEWEX cloud Assessment dataset.

Fig. 14. Mean 2005–2007 monsoonal averages (December–May and June–November) for OMAERO v3 from (a) and (b) OMI AOD, (c) and (d) OMI AAOD, and (e) and (f) single scattering albedo (ωo).

Fig. 16. Regression statistics for MISR versus MODIS, drawn from Shi et al. (2011b) (2005–2007).

Fig. 6. Key 2003–2009 meteorology variables divided by December to May winter and June to November summer monsoonal periods. (a) and (b) NOGAPS 30m winds and CMORPH precipitation rate (white b1mmday−1). (c) and (d) NOGAPS 700hPa winds and 500hPa vertical velocity (omega). (e) and (f) NOGAPS 200hPa winds and Tlightning flash frequency from LIS sensor on TRMM (updated form Christian et al., 2003a).

Fig. 4. Airborne photo with associated MPLnet normalized relative backscatter image of the cloud and aerosol environment at Singapore on 28 March 2012 at ~10Z. Lidar image is from 9 to 15Z, with a vertical coordinate from 0 to 18km. Annotated are a number of aerosol and cloud features demonstrating the complex vertical characteristics of the Southeast Asian environment.

Fig. 13.Mean 2005–2007monsoonal (December–May and June–November) AOD from (a) and (b) AVHRR PATMOS (1°×1°), (c) and (d) TerraMODIS Col 5 (0.5°×0.5°), (e) and (f) Terra MODIS Deep Blue(0.5°×0.5°), (g) and (h) SeaWiFS Deep Blue (0.5°×0.5°) and (h) and (i) Terra MISR (0.5°×0.5°). Included on (c) are the locations for the regressions in Table 1, and the time series of Fig. 15. 1. Thailand; 2) Gulf of Tonkin; 3) Southern Kalimantan; 4) Java Sea; 5) South China Sea; and 6) South of Sumatra.

Fig. 10. (a)–(f) Bi-monthly (see corresponding inset) MODIS fire prevalence for 2003–2009 based on Giglio et al. (2006) and Reid et al. (2012). (g) Fire count time series for Indochina and the Maritime Continent for 2003–2009 derived from Reid et al. (2012).

Fig. 18. Time series of AOD products for 2006: (a)Thailand (b) the Gulf of Tonkin; (c) Southern Kalimantan; (d) Java Sea. The OMI aerosol index (AI) for these four locations is given in (e). Locations correspond to the regressions in Table 2. Because of the 8day overpass schedules for MISR, data is presented as a series of bars.

Fig. 1. (a)–(f) Seasonal maximum, minimum and max–min difference of aerosol optical depths (AOD) composites fromMODIS Collection 5 (550nm), MODIS and SeaWiFS Deep Blue (550nm), and MISR (553nm) for each 0.5×0.5 degree point. Also shown (g) and (h) is the OMI 1×1 degree average Aerosol Index (AI). (a), (c), (e), and (g) December to May. (b), (d), (f), and (h) June to November.

Table 2 Basic optical depth, absorption, and size properties from AERONET over the region. Included is the aerosol optical depth (AOD; 500nm), fine mode optical depth (Fine-AOD), Angstrom exponent, precipitable water vapor (Water), single scattering albedo retrievals (ωo; 440nm for AODs>0.4), and retrieved effective radius (reff; μm).

Fig. 11. Comparison of PM2.5 emission properties (2003–2009) from the original FLAMBE and GFED3.1 products. Included are (a) and (c), emissions and (b) and (d) the modal month of emissions. (e) Ratio of FLAMBE to GFED3.1. (f) Difference in modal month, FLAMBE-GFED3.1.

Frequently asked questions

Q1. What are the contributions in "Observing and understanding the southeast asian aerosol system by remote sensing: an initial review and analysis for the seven southeast asian studies (7seas) program" ?

In this paper, the authors present an overview of what the standard products from remote sensing can provide to understand the aerosol system, as well as key points on the magnitude of uncertainty. 

Q2. What are the future works mentioned in the paper "Observing and understanding the southeast asian aerosol system by remote sensing: an initial review and analysis for the seven southeast asian studies (7seas) program" ?

Second, the complex meteorological context of any measurement must be considered in the analysis to minimize the possibility that any derived quantity is unduly influenced by confounding and sampling bias. The creation of such datasets can be achieved by several organizations, mining multiple international data streams. Once created, the data can be mined, cross checked and verified. Addressing SEA 's environmental issues will require cooperation from all levels of research, from “ boots on the ground ” to hand map changes on the land surface in local areas to large international programmatic efforts. 

Q3. What makes interpretation of individual retrievals in SEA difficult?

The combination of lower boundary condition, variable aerosol particle microphysics, and cloud issues makes interpretation of individual retrievals in SEA difficult. 

Q4. What are the main seasonal features of biomass burning in SEA?

Burning associatedwith agriculture, including rice, sugar cane, and pasture as well as deforestation burning, are common seasonal features throughout SEA. 

Q5. Why is Indonesia often highlighted in the media?

Because land cover conversion in Indonesia is often associated with fire activity and subsequently haze in the popular press and Indonesia is the largest country in the region, Indonesia is often highlighted in regard to peatland destruction in Sumatra and the Kalimantan provinces of Borneo. 

Q6. Why is the information content of satellite products anticorrelated to total area coverage?

Because of both fiscal and technological constraints, the information content of satellite products is typically anticorrelated to total area coverage. 

Q7. Why do the public and scientists often associate SEA aerosol pollution with biomass burning?

Because of the publicity associated with Indonesian fire events, the public and scientists alike often associate SEA aerosol pollution with biomass burning; hence SEA aerosol research tends focuses on this aerosol source. 

Q8. What is the benefit of multi-angle viewing instruments?

Multi-angle viewing instruments such as MISR have the benefit of additional view angles on the same location, thus allowing the land surface and atmospheric components of the at-sensor radiance to be more effectively isolated. 

Q9. What is the main reason for the presence of significant biomass burning in the MC?

the presence of significant biomass burning in the MC coupled with summer time convection and, at times severe storms (e.g., Fig. 6(f)), suggests that smoke can be convectively pumped to the upper troposphere. 

Q10. What is the complicated example of a IC smoke above a stratus deck?

With IC smoke above a stratus deck fed bypolluted air from China, the Crachin may be an especially complicated example of aerosol–stratus interaction in heavily polluted environments. 

Q11. Why is it difficult to compare field results to those from laboratory fires?

Because of the rapid evolution of biomass burning aerosol it is hard to meaningfully compare field results to those from laboratory fires.