Journal of Geophysical Research: Atmospheres

Aerosol optical depth and direct radiative forcing for INDOEX derived from AVHRR: Observations, January–March 1996–2000

Authors


Abstract

[1] Visible and near infrared reflectances from NOAA-14 Advanced Very High Resolution Radiometer (AVHRR) daytime passes are used to derive optical depths at 0.55 μm, an index of aerosol type, continental or marine, and the direct effect of the aerosol on the top of the atmosphere and surface solar radiative fluxes for the oceans in the Indian Ocean Experiment (INDOEX) region (30°S to 30°N and 50°–110°E) during the January–March 1996–2000 winter monsoons. Comparison of aerosol optical depth and radiative forcing in the Northern Hemisphere with those in the Southern Hemisphere suggests that the additional pollution sources augment the 0.55-μm optical depth by, on average, 0.1 in the Northern Hemisphere. As a result of the aerosol, the region of the Indian Ocean in the Northern Hemisphere loses about 1.6 Wm−2 in reflected sunlight and the ocean surface loses about 5 Wm−2 during the months of the winter monsoon. Aerosol burdens and the aerosol direct radiative forcing are a relatively constant feature of the Northern Hemisphere, although the southeastern Arabian Sea experienced considerably larger aerosol burdens during the February–March 1999 INDOEX Intensive Field Phase (IFP) than in other years. Frequency distributions of the optical depth for 1° × 1° latitude-longitude regions are well represented by gamma distribution functions. The day-to-day and year-to-year variability of the optical depth for such regions is correlated with the long-term average optical depth. Interannual variability of the monthly mean optical depths for such regions is found to be as large as the day-to-day variability. Such large variability suggests that long-term records of in situ observations will be required in order to assess the performance of models that generate climatologies of aerosol concentrations.

1. Introduction

[2] Anthropogenic aerosols affect the Earth's climate in two ways: they reflect and absorb solar radiation, thereby giving rise to what is referred to as the direct radiative forcing, and they increase the numbers and decrease the sizes of cloud droplets, thereby giving rise to the indirect radiative forcing [Charlson et al., 1992; Andreae, 1995; Charlson and Heintzenberg, 1995]. Both effects alter sunlight reflected and absorbed by the Earth-atmosphere system. These changes are expected to lead to changes in atmospheric and oceanic temperatures, and consequently, altered weather patterns. Knowledge of the spatial distribution of aerosols and their effect on the Earth's energy budget is key to reliable assessments of climate change. The variability of aerosol concentrations coupled with the variability of their physical and optical properties, makes direct measurements of aerosols and their properties on global scales impractical. Consequently, remote sensing of aerosols from satellites is essential to determining the temporal and spatial distribution of aerosols and estimating the aerosol direct radiative forcing of climate.

[3] The Indian Ocean Experiment (INDOEX) was an intensive field experiment designed in part to determine the radiative forcing due to anthropogenic aerosols on regional (Arabian Sea and Indian Ocean) and seasonal (winter monsoon) scales. During the winter monsoon months of January–March, the Arabian Sea and the Indian Ocean are an ideal laboratory for the study of aerosols and their effect on the Earth's radiation budget. Monsoon winds from Asia transport aerosol rich continental air from heavily populated areas of the Asian subcontinent to largely cloud-free ocean areas. Here, observations from the NOAA-14 Advanced Very High Resolution Radiometer (AVHRR) are used to determine the spatial and temporal distribution of aerosol optical depth and the effect of the aerosols on the radiation budget for the Indian Ocean for the period January–March 1996–2000. The results provide the context for the 15 February–26 March 1999 INDOEX Intensive Field Phase (IFP), during which extensive surface, shipboard, and aircraft observations were made of the aerosol.

[4] The method used to retrieve optical depths employed the optical properties of continental and marine aerosols in conjunction with the reflected sunlight at visible (0.64-μm, AVHRR Channel 1) and near infrared wavelengths (0.84-μm, AVHRR Channel 2) to determine the relative contributions of the two aerosol types and the optical depth of the mixture for each cloud-free pixel found to be suitable for retrieval [Coakley et al., 2002]. The aerosol models chosen for the retrieval scheme were the average continental aerosol and tropical marine aerosol models described by Hess et al. [1998]. The continental aerosol was chosen because it had relatively small particles and contained a substantial fraction of soot so that it absorbed sunlight. The marine aerosol by comparison had relatively large particles and was practically nonabsorbing. The 2-channel, 2-aerosol model retrieval scheme, unlike single-channel, single-model schemes [Stowe et al., 1997; Rajeev et al., 2000], was used to determine aerosol type and burden for each retrieval, so that in principle, variations in absorption due to aerosols could be allowed for in space and time. Consequently, the link between aerosol type and burden and the effects of the aerosol on the atmospheric and surface radiation budget could be explored. In the end, the retrieved mixing fractions of continental and marine aerosols proved to be reasonably constant for all regions and all times. In other words, the retrieval scheme in combination with the AVHRR reflectances at 0.64 and 0.84-μm were insufficient to distinguish between a small particle, absorbing aerosol, the continental aerosol, and a large particle, nonabsorbing aerosol, the marine aerosol. Nevertheless, when compared with surface measurements, optical depths retrieved using the 2-channel, 2-aerosol model scheme and those obtained using single-channel, single-model schemes, agreed equally well. Biases in visible and near infrared optical depths were typically within 0.05 and RMS differences were within 0.07. In addition, estimates of the top of the atmosphere aerosol direct radiative forcing under cloud-free conditions retrieved using the 2-channel, 2-model scheme and the single-channel, single-model schemes, fell within 40% of each other, regardless of the scheme used [Coakley et al., 2002]. The results presented here for the top of the atmosphere aerosol direct radiative forcing are likely to be as accurate as any estimate derived from retrieved aerosol optical depths. Results presented for the surface forcing remain rather uncertain. The results for the surface depend on the amount of sunlight absorbed by the aerosol, which is uncertain.

[5] Even with the seemingly large uncertainties, the spatial distributions of the optical depth and radiative fluxes derived for the Indian Ocean in the Northern and Southern Hemispheres suggest that pollution sources on the Asian subcontinent have a nearly permanent influence in the Northern Hemisphere during the months of the winter monsoon. In addition, the observations allow analysis of the day-to-day and year-to-year variations on regional scales. Because aerosols have short atmospheric residence times and are subject to relatively efficient and rapid removal processes, such as precipitation, their concentrations are best represented by asymmetric distribution functions, such as the gamma distribution function. Analysis of the variability in aerosol optical depths suggests that lengthy records of in situ measurements of aerosol concentrations are required in order to assess the performance of models that calculate climatological concentrations of aerosols [e.g., Chin et al., 1996; Roelofs et al., 1998; Adams et al., 1999; among others].

2. Analysis

[6] Daily NOAA-14 AVHRR 4-km Global Area Coverage (GAC) measurements [Kidwell, 1994] were collected for January–March 1996–2000. Analyses were performed for ascending daytime passes in the INDOEX region (30°S to 30°N and 50°–110°E). Prior to performing the aerosol retrievals, reflectances in Channel 1, at 0.64 μm, and Channel 2, at 0.84 μm, were calibrated using the extensive ice sheets of Greenland and Antarctica [Tahnk and Coakley, 2001a, 2001b].

[7] The methods used to retrieve aerosol optical depths and the aerosol direct radiative forcing were described by Coakley et al. [2002]. Briefly, each 4-km AVHRR GAC pixel was identified as being either cloud-free, partly cloud covered, or overcast by a single-layered, optically thick cloud system. For pixels identified as cloud-free ocean, outside the region of Sun glint, and viewed in the direction of backscattered sunlight, reflectances in Channels 1 and 2 were used in a 2-channel, 2-aerosol model retrieval scheme to determine the relative concentration of continental and marine aerosols for the pixel and the optical depth at a standard wavelength, 0.55 μm, for the mixture. The relative concentration was expressed as a mixing fraction: f = 1 for a mixture that was entirely continental, and f = 0 for a mixture that was entirely marine. The relative concentration and optical depth were used along with the date and latitude of the observation to calculate the cloud-free, direct radiative forcing of the aerosol mixture at the top of the atmosphere and surface.

[8] For each 1° × 1° latitude-longitude region, the aerosol properties and cloud-free radiative forcing were composited along with the numbers of cloud-free, partly cloudy, and overcast pixels as well as other cloud related indices to form daily, monthly, and seasonal composites for each 1° scale region. Estimates of the diurnally averaged radiative forcing were derived assuming that the aerosol mixture, optical depth, and cloud conditions encountered at the time of the NOAA-14 overpass were constant for the daylight hours.

[9] Most of the cloud cover encountered in INDOEX failed to fill the 4-km field of view of the AVHRR GAC data. For broken cloud systems, cloud properties are defined by the method used to detect the clouds and extract the properties. Different methods produce different results [Wielicki and Parker, 1992]. Consequently, the effect of clouds on the aerosol direct radiative forcing for INDOEX is rather uncertain. As discussed by Coakley et al. [2002], several approximations were used in order to estimate the uncertainty in the aerosol direct radiative forcing for average cloud conditions. Here, the simplest of the approximations was adopted: the radiative forcing was set to zero for all 1° × 1° latitude-longitude regions that contained upper-level clouds, regardless of the fractional coverage, and for regions that had low-level clouds, the forcing was set to zero for the portion of the region that was overcast by the low-level clouds. In all of the approximations, the radiative forcing was set to zero for all 1° × 1° latitude-longitude regions for which upper-level clouds completely covered the region. The region was taken to be overcast by upper-level clouds when the average 11-μm emission fell below IC = 75 mWm−2 sr−1 cm, equivalent to a brightness temperature of 275 K and to an opaque cloud at 4 km in a tropical atmosphere, and the 0.64-μm reflectance was greater than 0.44, an average reflectance found for pixels overcast by optically thick, layered clouds. Setting the forcing to zero for such regions mimics the attenuation of the effects of the aerosol on the top of the atmosphere and surface radiative fluxes when the aerosol layer lies beneath optically thick clouds. In an extreme approximation, referred to as “Average High,” for any region that contained upper-level clouds, but the clouds failed to completely cover the region, the top of the atmosphere and surface radiative forcing was set equal to the average of the forcing obtained when the region was either cloud-free or contained only low-level cloud. A region was identified as containing upper-level clouds if the 5th percentile of the 11-μm emission for the region was less than IC. Of the regions that met this condition, fewer than 15% had cloud-free pixels, indicating that such regions were heavily cloud covered. Setting the forcing for such regions equal to the average obtained under cloud-free conditions and conditions when only low-level clouds were present leads to an upper estimate of the forcing. Such an estimate does not account for the attenuation of the effect of the aerosols on the surface and top of the atmosphere solar radiative fluxes by the upper-level clouds. In another extreme approximation, the one used here and referred to as “Average-Zero, All Clouds,” the forcing was set to zero for all regions that contained upper-level clouds, regardless of cloud cover, and for portions of regions overcast by low-level clouds. For regions that had no upper-level clouds, the fraction of the region overcast by low-level clouds was estimated by taking the average fractional coverage for partly cloudy pixels to be 0.48. This value was consistent with comparisons of 0.64-μm reflectances for pixels overcast by low-level, layered clouds, and the reflectances for pixels that were only partly covered by these clouds. In the case of low-level clouds, setting the forcing to zero at the top of the atmosphere is consistent with the near cancellation of the forcing for the average continental and tropical marine aerosols when low-level clouds are imbedded in the haze [Coakley et al., 2002]. Because it absorbs sunlight, the average continental aerosol leads to a reduction in planetary albedo for regions overcast by low-level clouds that are imbedded in the haze. The tropical marine aerosol is nonabsorbing and, for a given optical depth, increases the albedo of regions overcast by low-level clouds by an amount that is comparable to the reduction caused by the average continental aerosol with equal optical depth. The near cancellation of the top of the atmosphere aerosol direct radiative forcing for overcast, low-level clouds arose in INDOEX because, as will be shown later, the mixing fractions of the two aerosols were approximately the same for all regions and all years. The relative difference between the extreme estimates of the top of the atmosphere forcing for average cloud conditions reached 45%.

[10] At the surface, the aerosol direct radiative forcing, is nonzero for all but the thickest of clouds. The surface forcing obtained for the “Average High” approximation is likely to overestimate the actual forcing because it does not account for attenuation of the effect of the aerosol by the upper-level clouds that fail to completely cover the 1° × 1° latitude-longitude analysis region. The “Average-Zero, All Clouds” approximation, on the other hand, will surely underestimate the forcing. Setting the forcing to zero for overcast regions will lead to estimates that are too small. Relative differences in the extreme estimates obtained using these approaches reached 70%. Here, the “Average-Zero, All Clouds” approach was adopted as it is the simplest to implement.

[11] For estimates of the aerosol optical depth, mixing fraction, and cloud-free radiative forcing, tests were performed to determine how many cloud-free pixels, suitable for retrievals, within a 1° × 1° latitude-longitude box were required to ensure that the average properties derived for the box represented the true mean for the region. For this test, sixty passes were randomly chosen from the 3 months of each of the 5 years so that each pass had at least one 1° square that contained at least 50 cloud-free pixels suitable for aerosol retrievals. For each pass, a set of 20 randomly selected optical depths were extracted from a single 1° square that contained more than 50 pixel-scale aerosol retrievals. As shown in Figure 1, RMS differences of the mean optical depth from the true mean for the square were calculated as a function of the number of observations included in the average. The true mean was obtained by averaging over all cloud-free pixels suitable for aerosol retrievals in the square. For all of the years studied, 12 observations within a square produced an average 0.55-μm optical depth that differed from the true mean with an RMS difference of 0.004. Consequently, a minimum of 12 cloud-free pixels suitable for aerosol retrievals within a 1° × 1° latitude-longitude region was imposed in order for the region to be included in the daily composite.

Figure 1.

RMS departure of the daily average aerosol optical depth from the true average optical depth for 1° × 1° latitude-longitude regions as a function of the number of 4-km cloud-free pixels used to obtain the average. RMS differences are plotted for years 1996 (plus), 1997 (asterisk), 1998 (diamond), 1999 (triangle), and 2000 (cross).

[12] As was done for the daily composites, tests were performed to determine the number of daily observations required for each 1° × 1° latitude-longitude region in order to obtain a representative monthly mean. Using daily composites for all satellite passes for each month of the 5-year period, RMS differences between the mean for a 1° square and its true mean were calculated as a function of the number of daily composites contributing to the monthly mean. Only regions that contained at least 20 daily composites were included in this test. Results for the Arabian Sea, Bay of Bengal, and Southern Hemisphere are shown in Figure 2. The regions were treated separately as they appeared to exhibit distinctly different aerosol regimes. In addition, a distinction was made between low and high optical depths within a region. As will be discussed later, for regions that have mean optical depths that are small, the standard deviations are also small. Consequently, the RMS differences for these regions showed little sensitivity to the number of daily estimates used to form a monthly mean. Regions with large mean optical depths, on the other hand, also had large standard deviations in optical depth. For such regions RMS differences between the mean and true mean generally decreased as the number of daily estimates used to compute the monthly mean increased. For all of the years studied, monthly means constructed from 5 daily estimates produced an optical depth that departed from the true mean with an RMS difference of 0.04. At least 5 daily composites were required in order for a 1° × 1° latitude-longitude region to be counted in the monthly composites.

Figure 2.

RMS departure of the monthly average aerosol optical depth from the true average for 1° × 1° latitude-longitude regions as a function of the number of daily values included in the average. RMS differences are plotted for low and high optical depths in the Arabian Sea and Bay of Bengal and low optical depths in the Southern Hemisphere. The standard deviation of the 1° × 1° latitude-longitude regional means is shown for each region and optical depth condition.

[13] Seasonal composites for each year and each 1° × 1° latitude-longitude region were then constructed from the monthly mean composites. The seasonal composites for each of the 5 years were combined to construct the 5-year composite and the year-to-year variations within the 5-year period.

3. Aerosol Optical Depths and Mixing Fractions

[14] Figure 3 shows the 5-year seasonal average optical depth at 0.55 μm and its year-to-year standard deviation for the INDOEX region. The boxes in the figure identify the regions referred to as the Arabian Sea, the Bay of Bengal, and the Indian Ocean in the Southern Hemisphere. The figure also shows 10° × 10° latitude-longitude subregions which will be discussed later. These subregions were chosen because the aerosol properties were expected to be reasonably homogeneous with regard to sources and they represented regions that were heavily polluted and regions that were relatively pristine. Table 1 gives the regional, January–March averages of the aerosol mixing fraction, optical depths, cloud conditions, and aerosol direct radiative forcing for each year and each region.

Figure 3.

5-year January–March average and year-to-year standard deviation of 0.55-μm optical depth. The boxes in the figures identify regions and 10° × 10° latitude-longitude subregions referenced in the analysis.

Table 1a. Composite Statistics (mean and Da-to-Day regional Variation) for Jannuary–March 1996 and the Specified Regions
ParameterArabian SeaBay of BengalNorthern HemisphereSouthern Hemisphere
Optical Depth0.23 (0.03)0.31 (0.09)0.25 (0.03)0.15 (0.03)
f0.70 (0.11)0.76 (0.12)0.72 (0.10)0.64 (0.09)
 
Frequency of Occurrence
Cloud-free pixels0.19 (0.08)0.10 (0.06)0.15 (0.06)0.09 (0.03)
Overcast pixels0.01 (0.01)0.04 (0.04)0.02 (0.02)0.03 (0.01)
Partly cloudy pixels0.79 (0.08)0.86 (0.06)0.82 (0.05)0.89 (0.03)
Regions containing upper level clouds0.22 (0.10)0.43 (0.14)0.32 (0.08)0.48 (0.11)
Regions overcast by upper level clouds0.04 (0.03)0.09 (0.07)0.06 (0.03)0.07 (0.04)
 
Radiative Forcing (Wm−2) Cloud-free
Top of the atmosphere−7.4 (1.0)−9.5 (2.3)−8.1 (1.0)−5.3 (0.9)
Surface−15.2 (2.4)−20.8 (5.9)−17.1 (2.4)−10.3 (1.4)
Atmosphere7.8 (1.6)11.4 (3.7)9.0 (1.6)5.0 (0.6)
 
Average
Top of the Atmosphere−3.5 (0.8)−2.7 (0.8)−3.2 (0.6)−1.6 (0.3)
Surface−7.3 (1.8)−6.0 (2.0)−6.7 (1.4)−3.0 (0.5)
Atmosphere3.8 (1.3)3.3 (1.4)3.5 (1.0)1.4 (0.4)
Table 1b. Composite Statistics (Mean and Day-to-Day Regional Variation) for January–March 1997 and the Specified Regions
ParameterArabian SeaBay of BengalNorthern HemisphereSouthern Hemisphere
Optical Depth0.27 (0.04)0.31 (0.06)0.29 (0.03)0.17 (0.04)
f0.73 (0.13)0.79 (0.12)0.75 (0.11)0.73 (0.10)
 
Frequency of Occurrence
Cloud-free pixels0.20 (0.07)0.12 (0.06)0.16 (0.05)0.08 (0.03)
Overcast pixels0.01 (0.01)0.03 (0.02)0.02 (0.01)0.04 (0.02)
Partly cloudy pixels0.79 (0.07)0.85 (0.06)0.82 (0.05)0.88 (0.03)
Regions containing upper level clouds0.22 (0.13)0.41 (0.13)0.31 (0.10)0.45 (0.12)
Regions overcast by upper level clouds0.04 (0.04)0.09 (0.06)0.06 (0.04)0.07 (0.04)
 
Radiative Forcing (Wm−2) Cloud-free
Top of the atmosphere−8.7 (1.3)−9.6 (1.6)−9.1 (1.0)−6.0 (1.4)
Surface−18.1 (2.3)−21.3 (4.0)−19.4 (2.1)−12.4 (2.4)
Atmosphere9.4 (1.5)11.8 (2.6)10.3 (1.5)6.4 (1.2)
 
Average
Top of the Atmosphere−4.2 (0.9)−3.1 (1.0)−3.7 (0.8)−1.9 (0.3)
Surface−8.9 (1.9)−7.0 (2.3)−7.9 (1.8)−3.7 (0.6)
Atmosphere4.7 (1.4)3.9 (1.6)4.2 (1.3)1.8 (0.4)
Table 1c. Composite Statistics (Mean and Day-to-Day Regional Variation) for January–March 1998 and the Specified Regions
ParameterArabian SeaBay of BengalNorthern HemisphereSouthern Hemisphere
Optical Depth0.23 (0.04)0.30 (0.06)0.26 (0.04)0.16 (0.03)
f0.67 (0.11)0.80 (0.10)0.72 (0.09)0.69 (0.08)
 
Frequency of Occurrence
Cloud-free pixels0.18 (0.08)0.13 (0.06)0.15 (0.06)0.08 (0.03)
Overcast pixels0.01 (0.01)0.03 (0.02)0.02 (0.01)0.02 (0.01)
Partly cloudy pixels0.81 (0.08)0.85 (0.06)0.83 (0.06)0.89 (0.02)
Regions containing upper level clouds0.24 (0.14)0.38 (0.12)0.31 (0.09)0.51 (0.12)
Regions overcast by upper level clouds0.04 (0.04)0.09 (0.04)0.06 (0.03)0.08 (0.04)
 
Radiative Forcing (Wm−2) Cloud-free
Top of the atmosphere−7.5 (1.2)−9.4 (1.7)−8.3 (1.0)−5.5 (1.1)
Surface−15.2 (2.3)−21.2 (4.2)−17.7 (2.2)−11.2 (2.0)
Atmosphere7.6 (1.5)11.9 (2.7)9.4 (1.4)5.6 (1.1)
 
Average
Top of the Atmosphere−3.5 (0.9)−3.2 (0.8)−3.3 (0.6)−1.6 (0.3)
Surface−7.0 (1.9)−7.2 (1.9)−7.1 (1.5)−3.1 (0.5)
Atmosphere3.5 (1.4)4.0 (1.4)3.8 (1.0)1.5 (0.4)
Table 1d. Composite Statistics (Mean and Day-to-Day Regional Variation) for January–March 1999 and the Specified Regions
ParameterArabian SeaBay of BengalNorthern HemisphereSouthern Hemisphere
Optical Depth0.24 (0.04)0.31 (0.06)0.27 (0.03)0.13 (0.02)
f0.53 (0.13)0.59 (0.17)0.55 (0.12)0.61 (0.06)
 
Frequency of Occurrence
Cloud-free pixels0.20 (0.07)0.10 (0.05)0.15 (0.06)0.08 (0.02)
Overcast pixels0.01 (0.01)0.04 (0.03)0.02 (0.02)0.04 (0.02)
Partly cloudy pixels0.79 (0.07)0.86 (0.06)0.83 (0.05)0.88 (0.02)
Regions containing upper level clouds0.23 (0.12)0.46 (0.10)0.35 (0.08)0.42 (0.11)
Regions overcast by upper level clouds0.04 (0.03)0.08 (0.05)0.06 (0.03)0.06 (0.03)
 
Radiative Forcing (Wm−2) Cloud-free
Top of the atmosphere−8.0 (1.3)−9.9 (1.7)−8.7 (1.0)−4.8 (0.9)
Surface−14.7 (2.7)−19.2 (3.5)−16.3 (2.1)−9.1 (1.4)
Atmosphere6.7 (1.7)9.3 (2.6)7.6 (1.6)4.4 (0.6)
 
Average
Top of the Atmosphere−3.8 (0.9)−2.7 (0.9)−3.2 (0.7)−1.7 (0.3)
Surface−7.1 (1.8)−5.3 (1.8)−6.2 (1.4)−3.1 (0.5)
Atmosphere3.3 (1.4)2.6 (1.4)3.0 (1.0)1.4 (0.4)
Table 1e. Composite Statistics (Mean and Day-to-Day Regional Variation) for January–March 2000 and the Specified Regions
ParameterArabian SeaBay of BengalNorthern HemisphereSouthern Hemisphere
Optical Depth0.24 (0.05)0.27 (0.07)0.25 (0.04)0.15 (0.04)
f0.41 (0.14)0.59 (0.14)0.47 (0.12)0.54 (0.10)
 
Frequency of Occurrence
Cloud-free pixels0.22 (0.08)0.09 (0.05)0.16 (0.06)0.07 (0.02)
Overcast pixels0.02 (0.01)0.04 (0.03)0.03 (0.01)0.03 (0.01)
Partly cloudy pixels0.75 (0.11)0.87 (0.11)0.81 (0.10)0.89 (0.10)
Regions containing upper level clouds0.21 (0.11)0.56 (0.13)0.37 (0.09)0.50 (0.10)
Regions overcast by upper level clouds0.03 (0.03)0.13 (0.08)0.07 (0.04)0.10 (0.04)
 
Radiative Forcing (Wm−2) Cloud-free
Top of the atmosphere−8.2 (1.5)−8.6 (2.1)−8.3 (1.4)−5.4 (1.2)
Surface−13.7 (2.6)−16.6 (4.2)−14.6 (2.5)−9.8 (1.8)
Atmosphere5.5 (1.6)8.1 (2.5)6.3 (1.4)4.4 (0.7)
 
Average
Top of the Atmosphere−4.2 (0.7)−1.7 (0.6)−2.9 (0.5)−1.4 (0.2)
Surface−7.0 (1.3)−3.5 (1.2)−5.1 (1.1)−2.7 (0.4)
Atmosphere2.8 (1.0)1.8 (0.9)2.2 (0.8)1.3 (0.3)

[15] Both the mean optical depth and its year-to-year standard deviation mark the regions affected by pollution haze and suggest the likely sources of the haze. As indicated by the optical depths in the Southern Hemisphere, pristine oceans appear to have low optical depths, rarely exceeding 0.2. Polluted regions have optical depths that range from 0.2 to 0.3 and higher, depending on their proximity to sources. On the seasonal scale, differences between optical depths in the Northern and Southern Hemispheres leads to the conclusion that the 0.55-μm optical depths may be elevated by about 0.1 due to the sources of haze pollution, with the increases being substantially larger, 0.1–0.3, depending, of course, on the distance to the pollution sources.

[16] Figure 4 shows monthly averages and daily standard deviations of regional means for the Arabian Sea, the Bay of Bengal, and the Indian Ocean in the Southern Hemisphere. There is little trend in the regional scale optical depths either with month or year. Climatologically, the elevated optical depths in the Northern Hemisphere appear to be a constant feature. For the 10° × 10° latitude-longitude subregions, differences between monthly mean optical depths are, of course, more apparent, as shown in Figure 5. During the February–March 1999 IFP the optical depths were among the largest obtained for the Arabian Sea and the Bay of Bengal during the 5-year period.

Figure 4.

Monthly mean 0.55-μm optical depth for January–March 1996–2000 for the specified regions. The bar represents the daily standard deviation about the mean for the region.

Figure 5.

Monthly mean 0.55-μm optical depths for January–March 1996–2000 for the specified 10° × 10° latitude-longitude subregions. The bar represents the standard deviation of the daily means for the subregion.

[17] The regional scale monthly means and daily standard deviations for the aerosol mixing fraction, f, are shown in Figure 6. Again, with the exception of the results for 1999 and 2000, there is no obvious trend in the mixing parameter. The radical change in the mixing fraction obtained for 2000 might be due either to the drift in orbit of the NOAA-14 satellite toward later equator crossing times, or to changes in the radiometric response of the AVHRR. Because of the later equator crossing time, the solar zenith angles were distinctly higher for 2000 than for the other years. With the higher solar zenith angles, retrievals must be made with lower levels of reflected sunlight, thereby amplifying any errors in calibration and the calculated radiances used in the retrievals. The calibrations for the Channel 1 and Channel 2 reflectances were checked using extensive ice sheets of Greenland and Antarctica. The radiometric response of the NOAA-14 AVHRR for 2000 had undergone a marked change from its behavior in previous years [Tahnk and Coakley, 2001b]. Once the altered response had been incorporated, retrieved optical depths were compared with those measured at KCO, and the comparisons were not noticeably different for 2000 than those obtained for the previous years. Unfortunately, as Coakley et al. [2002] noted, comparisons of retrieved and measured optical depths provide few constraints on the aerosol properties retrieved using AVHRR radiances. Consequently, the source of the trend in f, whether real or due to fundamental errors in the calculated radiances used in the retrievals, or to an error in the calibration of the AVHRR, remains elusive.

Figure 6.

Monthly means of the aerosol mixing fraction, f, for January–March 1996–2000 for the specified regions. The bar represents the standard deviation of the daily means for the region. f = 1 for the average continental aerosol and f = 0 for the tropical marine aerosol.

4. Day-to-Day and Year-to-Year Variability in Optical Depth

[18] The similarity in spatial patterns shown in Figure 3 for the optical depth and its year-to-year standard deviation suggests that the variability of the optical depth for a region is proportional to the long term average optical depth. Figure 7 shows the March 1996–2000 average optical depth and the standard deviation of the daily optical depth for the 1° × 1° latitude-longitude boxes that are part of the 10° × 10° latitude-longitude subregions identified in Figure 3. The figure shows linear, least squares fits of the daily standard deviation as a function of the monthly mean optical depth. The fits were obtained by dividing the monthly mean optical depths into quartiles and using the average standard deviation and mean optical depth for each quartile, as indicated by the cross, in the least squares analysis. As shown in the figure, the standard deviation of the daily optical depth was reasonably well predicted by the monthly mean optical depth. Such behavior is expected of species with short atmospheric residence times. Concentrations of particles build over a period until removed, primarily through washout by precipitation. On monthly time scales, regions with heavy aerosol burdens experience a much wider range of aerosol concentrations than do regions with low burdens.

Figure 7.

Average and day-to-day standard deviation of the 0.55-μm optical depth for March 1996–2000. Each point gives the average optical depth and the daily standard deviation for a 1° × 1° latitude-longitude box drawn from the specified 10° × 10° latitude-longitude subregion. The line represents a least squares fit to the averages for each quartile of the monthly mean optical depth.

[19] The relationship between the mean and standard deviation of the optical depth is captured by the gamma distribution function. The probability of finding the daily optical depth to be between τ and τ + dτ is given by

equation image

where equation image is the monthly average optical depth, Γ(ν) is the gamma function, and the standard deviation of the optical depth is given by

equation image

Figure 8 shows the gamma distribution functions obtained for the four subregions. Clearly, the distribution function provides a useful representation for the variability of the daily optical depths for the 1° × 1° latitude-longitude boxes. Note, the distributions in Figure 8 reflect variability of the optical depth in both time and space while those in Figure 7 reflect variability in time only.

Figure 8.

Frequencies of the daily 0.55-μm optical depths for 1° × 1° latitude-longitude boxes within the specified 10° × 10° latitude-longitude subregion. The observations are for March 1996–2000. The dashed line is the gamma distribution function fit to the observations.

[20] The relationship between monthly mean and the standard deviation of the daily optical depths also holds for the 5-year monthly means and the standard deviation of the monthly mean optical depths as shown in Figure 9. The figure gives the 5-year mean optical depth for each of the 1° × 1° latitude-longitude boxes of the subregion and the corresponding standard deviation of the 5 monthly means. As with Figure 7, the distributions reflect the variability of the optical depth in time only. Figure 10 shows the gamma distribution functions for the four subregions.

Figure 9.

Same as Figure 7 but averages of March for 5 years and year-to-year standard deviations.

Figure 10.

Same as Figure 8 but for the monthly means.

[21] The results in Figures 7 and 9 indicate that whether affected by pollution or pristine, for the 1° × 1° latitude-longitude scale, the standard deviations of the monthly means are practically identical to the standard deviations of the daily means. The relatively large ratio of the monthly to daily variability in optical depths results from the short atmospheric residence times of the particles coupled with the episodic character of the sources and sinks that allow burdens to rise to large values. The results in Figures 8 and 10, on the other hand, indicate that while for polluted regions the standard deviation associated with the spatial distribution is somewhat smaller for the monthly than for the daily means, the spatial structure for pristine regions is considerably smaller for the monthly means, as would be expected for species with short residence times.

[22] As noted in the Introduction, in recent years models have been developed to estimate concentrations of various aerosols. In assessing model performance, averages of station data for the chemical components of aerosols are compared with concentrations of the components produced by the models. For the 1° × 1° latitude-longitude scale, the monthly mean optical depths appear to be as variable as the daily means. For polluted regions, like the Arabian Sea and the Bay of Bengal, both the monthly means and the day-to-day and year-to-year variability in the 0.55-μm optical depths are >0.1 and thus substantially larger than errors estimated for the retrieval of optical depth, mean error ∼0.05 and RMS error ∼0.06. The day-to-day and year-to-year variability appears to be real. Consequently, short-term station data is of little use in establishing trends on either seasonal or regional scales. Records of many years appear to be required to establish such trends.

5. Aerosol Radiative Forcing

[23] Figure 11 shows the seasonal average aerosol direct radiative forcing at the top of the atmosphere, the surface, and in the atmosphere, for cloud-free conditions and Figure 12 shows the seasonal averages for average cloud conditions. As mentioned in section 2, the radiative forcing is the diurnally averaged estimate derived by taking the aerosol mixture, optical depth, and cloud conditions retrieved at the time of the NOAA-14 daytime overpass to be constant for the daylight hours on the day of the observations. As with the optical depth, perturbations to the top of the atmosphere solar radiative flux by the aerosol are clearly evident in the Northern Hemisphere when comparisons are made with the Southern Hemisphere. The perturbation to the radiation budget of the northern Indian Ocean due to the aerosol direct radiative forcing is −2 Wm−2 for most of the Northern Hemisphere, with −6 Wm−2 near pollution sources. The surface forcing is typically more than twice the top of the atmosphere forcing. The higher surface forcing is due to the absorption of sunlight by the aerosol. Note, because the 2-channel, 2-model retrieval scheme infers an absorbing aerosol for the Southern Hemisphere, and because in situ observations of the aerosols in the Southern Hemisphere indicate that aerosols there exhibit little, if any absorption [Clarke et al., 2002; Quinn et al., 2002], the atmospheric and surface forcing shown for the Southern Hemisphere are overestimated. For nonabsorbing aerosols, there is little atmospheric forcing, and the aerosol direct radiative forcing at the surface is only slightly larger in magnitude than that at the top of the atmosphere [Coakley et al., 2002].

Figure 11.

5-year seasonal average aerosol direct radiative forcing for top of the atmosphere, surface, and atmosphere for cloud-free conditions.

Figure 12.

Same as Figure 11 but for average cloud conditions.

[24] Clearly, the aerosol direct radiative forcing is related to the optical depth. The sensitivity of the forcing to optical depth is presented for the 10° × 10° latitude-longitude subregions in Figure 13 for cloud-free conditions and in Figure 14 for average cloud conditions. Each point in Figures 13 and 14 is contributed by each of the 1° × 1° latitude-longitude boxes for each of the five years included in the study. Linear, least squares fits constrained to pass through the origin were used to determine the sensitivity of the fluxes to the optical depth. For cloud-free conditions, the departures of the points from a strict, linear relationship with optical depth is the result of using the 2-channel, 2-aerosol model retrieval scheme. The departures are particularly noticeable for the surface and atmospheric forcing. Use of a single aerosol model as done, for example, by Ramanathan et al. [2001] would have led to little, if any, departure from a linear least squares fit line for the range of optical depths observed.

Figure 13.

Aerosol direct radiative forcing and optical depth for cloud-free conditions and the following 10° × 10° latitude-longitude subregions: (a)–(c) North Arabian Sea; (d)–(f) South Arabian Sea; (g)–(i) Bay of Bengal; and (j)–(l) Southern Hemisphere. The line represents a linear least squares fit passing through the origin.

Figure 14.

Same as Figure 13 but for average cloud conditions.

[25] Despite the use of the 2-channel, 2-model scheme, the sensitivity of the radiative forcing to optical depth is nearly identical for the four subregions. As indicated by the results in Table 1, the 0.64 and 0.84 μm reflectances did not place the absorbing average continental aerosol with its relatively small particles in the polluted regions and the nonabsorbing tropical marine aerosol with its relatively large particles in the pristine regions. Based on these retrievals, there is little difference between the aerosols and consequently little difference in the relationship between the top of the atmosphere direct radiative forcing for cloud-free conditions and the optical depths. For the four subregions under cloud-free conditions, the top of the atmosphere radiative forcing ranges from −30 to −33 Wm−2 per unit 0.55-μm optical depth; the surface forcing ranges from −64 to −71 Wm−2, and the atmospheric forcing ranges from 32 to 38 Wm−2. The sensitivity of the atmospheric forcing to optical depth in the Southern Hemisphere is expected to be close to zero. As has been noted already, the retrieval scheme used here failed to produce such a result. Satheesh and Ramanathan [2000] obtained −25 Wm−2 per unit 0.5-μm optical depth for the sensitivity of the top of the atmosphere cloud-free aerosol direct radiative forcing and −70 Wm−2 to −75 Wm−2 per unit 0.5-μm optical depth for the surface forcing. They based their estimates on surface measurements of the optical depth and surface radiative fluxes at KCO and top of the atmosphere radiative fluxes obtained by the Cloud and Earth's Radiant Energy System on the Tropical Rain Measurement Mission satellite. The results obtained here give larger magnitudes (25%) for the top of the atmosphere forcing and generally smaller magnitudes (10%) for the surface forcing.

[26] The magnitude of the aerosol direct radiative forcing under average cloud conditions is, of course, less than the magnitude for cloud-free conditions. The 10° × 10° latitude-longitude subregions of the Northern Hemisphere were remarkably devoid of cloud cover during the winter monsoon, and as a result, clouds reduced the sensitivity of the forcing by only about 15% in these subregions. Cloud cover was more extensive in the subregion of the Southern Hemisphere and the sensitivity of the top of the atmosphere forcing to optical depth was reduced by 40% there.

[27] Because of the nearly linear relationship between radiative forcing and optical depth, day-to-day and year-to-year variability in the forcing is given by the product of the sensitivity of the forcing to optical depth and the variability of the optical depth. For example, the year-to-year variability of the seasonal mean 0.55-μm optical depth in the Bay of Bengal is 0.07 (Figure 10), the year-to-year variability in the top of the atmosphere forcing is 0.07 × −27 Wm−2 = −2 Wm−2, −4 Wm−2 at the surface, and 2 Wm−2 in the atmosphere.

[28] Assessment of the errors in the top of the atmosphere and surface aerosol direct radiative forcing, entails the following benchmarks. First, as was noted earlier, the top of the atmosphere aerosol direct radiative forcing under cloud-free conditions, like those shown in Figure 11, appear to overestimate the forcing by 25% when compared with the empirical estimate made by Satheesh and Ramanathan [2000] for KCO. Second, tests revealed that the top of the atmosphere radiative forcing for cloud-free conditions was somewhat insensitive to the retrieval scheme and aerosol model used in the retrieval. Estimates based on one scheme were likely to fall within 40% of the estimates obtained with other schemes [Coakley et al., 2002]. Thus, the top of the atmosphere radiative forcing for cloud-free conditions is likely to be within 40% of the value presented in Figure 11. Likewise, the aerosol direct radiative forcing at the surface under cloud-free conditions appears to be within 10% of the empirical estimate made by Satheesh and Ramanathan [2000]. Unlike the top of the atmosphere forcing, however, the surface forcing is rather sensitive to the aerosol model used to obtain the forcing. Comparisons of the surface forcing under cloud-free conditions for the 2-channel, 2-model retrieval scheme and for the aerosol model based on in situ observations at KCO for the February–March 1998 INDOEX First Field Phase (FFP) [Satheesh et al., 1999] and used by Rajeev et al. [2000] give rise to relative differences in the forcing of as much as 70%. The surface forcing, like that shown in Figure 11 remains rather uncertain. Furthermore, as aerosols in the Southern Hemisphere are unlikely to absorb sunlight [Clarke et al., 2002; Quinn et al., 2002], the surface forcing shown in Figure 11 for the Southern Hemisphere is incorrect. For a nonabsorbing aerosol, the surface forcing is approximately 9% greater than the top of the atmosphere forcing [Coakley et al., 2002].

[29] Clouds reduce the effect of the aerosols at both the top of the atmosphere and the surface. Furthermore, the method used here to account for the clouds overestimates the reduction at both the top of the atmosphere and the surface. Consequently, the aerosol direct radiative forcing under average cloud conditions is likely to fall between the estimate for cloud-free conditions given in Figure 11 and that for average cloud conditions given in Figure 12. In some regions, cloud cover is sparse, and the two estimates converge. In such cases, the uncertainty in the direct radiative forcing approaches that for the cloud-free conditions, 40% for the top of the atmosphere forcing. The surface forcing remains rather uncertain as the amount of sunlight absorbed by the aerosol is uncertain. In addition, as Figures 13 and 14 suggest, the forcing under both cloud-free and average cloud conditions is proportional to the aerosol optical depth and thus the forcing under average cloud conditions is proportional to the forcing under cloud-free conditions. As indicated by the results in Table 1, however, the reduction due to clouds is sizable for the large-scale regions. Results obtained by Coakley et al. [2002] in which clouds had the least reduction on the direct aerosol forcing (“Average-High”) and in which the clouds had close to the largest reduction (“Average-Zero, All Clouds”) indicate that the relative difference in the constant of proportionality between the cloud-free and average cloud conditions reaches 45% for the top of the atmosphere forcing and 70% for the surface forcing. Taking the errors in the estimate of the aerosol direct radiative forcing for cloud-free conditions (40%) to be statistically independent of those due to the effects of clouds on the direct forcing (45%), leads to a potential error of 60% for the top of the atmosphere forcing under average cloud conditions. Again, because aerosols in the Northern Hemisphere absorb sunlight, the surface forcing in the Northern Hemisphere remains highly uncertain. As noted earlier, for the Southern Hemisphere, in situ observations suggest that the aerosol is nonabsorbing. Consequently, for the Southern Hemisphere, the forcing at the surface is only slightly larger than that at the top of the atmosphere with the corresponding accuracy for the top of the atmosphere forcing, 60%. Although relatively small compared with the Northern Hemisphere counterparts, the direct aerosol radiative forcing for atmosphere and surface for the Southern Hemisphere in Figures 11 and 12 is incorrect.

6. Summary

[30] Aerosol optical depths and the relative concentrations of aerosol type, continental or marine, were retrieved for the INDOEX region using 5 years of NOAA-14 observations covering the winter monsoons, January–March 1996–2000. The optical depths and relative concentrations coupled with the optical properties of the aerosol models used in the retrievals were employed to derive estimates of the diurnally averaged, cloud-free aerosol radiative forcing for 1° × 1° latitude-longitude boxes. The effect of clouds on the aerosol direct radiative forcing was estimated by setting the forcing to zero for all regions that contained upper-level cloud and for all portions of regions overcast by low-level clouds. For the monthly mean aerosol direct radiative forcing at the top of the atmosphere under average cloud conditions, the estimates are expected to be within 60% of those obtained using other retrieval schemes with other aerosol models and with alternative estimates for the effects of clouds on the aerosol direct radiative forcing. For the aerosol forcing at the surface, on the other hand, the estimates are uncertain, as the amount of sunlight absorbed by the aerosol is uncertain. The estimates given for the direct radiative forcing under cloud-free conditions at the surface are within 10% of those derived empirically by Satheesh and Ramanathan [2000] based on observations made at KCO. The relative differences between the estimates given here and those produced using the aerosol model employed by Rajeev et al. [2000] reach 70%. Comparable uncertainties arise when the effects of clouds are considered. In addition, for the Southern Hemisphere, in situ observations indicate that aerosols absorb little, if any, sunlight [Clarke et al., 2002; Quinn et al., 2002]. The values given here include substantial absorption. The aerosol direct radiative forcing at the surface in the Southern Hemisphere is likely to be slightly larger in magnitude than that given here for the top of the atmosphere forcing.

[31] For the Arabian Sea, Bay of Bengal, and the Indian Ocean in the Southern Hemisphere, the optical depths were remarkably constant over the five-year period (Figure 4). For subregions, such as the southeastern portion of the Arabian Sea and the subregion of the Bay of Bengal, the optical depths retrieved for the February–March 1999 IFP were substantially larger than those encountered during the February–March 1998 INDOEX FFP. Differences in optical depths between the polluted Northern Hemisphere and the relatively pristine Southern Hemisphere indicated that the 0.55-μm aerosol optical depth in the Northern Hemisphere was about 0.1 larger as a result of the pollution. In some regions, the seasonal mean aerosol optical depth climbed to 0.2–0.3, depending on the proximity of the pollution sources.

[32] Despite the use of a mixture of aerosols, the continental model with relatively small absorbing particles, and a maritime model with relatively large, nonabsorbing particles, and the reflected sunlight at visible (AVHRR Channel 1) and near infrared (AVHRR Channel 2) wavelengths to deduce the relative concentrations of the aerosol components and associated optical depths, the derived radiative forcing was found to be almost linearly related to the retrieved optical depth. The relative concentrations of the two aerosols were approximately the same for all regions, and with the exception of the later years (1999 and 2000), were the same for all periods. In later years the relative concentrations shifted toward more maritime conditions. While this shift may have been caused by errors in calibration or errors in the retrievals that were amplified by the shift in the NOAA-14 orbit to later equator crossing times, no evidence for such errors could be found in comparisons of the retrieved optical depths with optical depths measured by surface instruments. Because of the constant mixing fraction, the link between optical depth and the direct radiative forcing also remained constant. The results in Table 1 indicate that the difference in the top of that atmosphere aerosol direct radiative forcing, between the Northern and Southern Hemispheres is about 1.6 Wm−2, with the region of the Indian Ocean in the Northern Hemisphere losing energy due to the presence of the haze. Likewise, taking the aerosol in the Southern Hemisphere to be nonabsorbing, so that the surface forcing and the top of the atmosphere forcing are nearly the same, the difference in the aerosol direct radiative forcing at the surface for the two hemispheres is about 5 Wm−2 with the ocean in the Northern Hemisphere losing energy due to the presence of the haze.

[33] Because they have short atmospheric residence times, aerosols have concentrations which vary in accordance with their long-term average concentration. Large monthly mean optical depths arise because of the proximity of sources. Large optical depths lead to, on average, large day-to-day variability in the optical depths. Likewise, regions with large seasonal mean optical depths exhibit, on average, correspondingly large year-to-year fluctuations in the seasonal means. The variability in optical depths was well represented by a gamma distribution function. For 1° × 1° latitude-longitude regions the monthly mean aerosol optical depths exhibited year-to-year variability that was comparable to their day-to-day variability. Such large variability in monthly mean optical depths suggests that many years of in situ observations would be needed in order to construct a climatology in aerosol concentrations suitable for assessing the performance of models that calculate climatological concentrations.

Acknowledgments

[34] This work was supported in part by the National Science Foundation (NSF), ATM-9612886, and the Center for Clouds, Chemistry, and Climate at the Scripps Institute of Oceanography, a NSF Science and Technology Center.

Ancillary