We describe a radiation experiment performed at the Kaashidhoo Climate Observatory (KCO), Republic of Maldives, during the Indian Ocean Experiment (INDOEX) field campaign in February through March 1999. Both the total solar broadband (0.3 to 3.81 μm) and visible spectral narrowband (seven spectral channels in the region from 0.4 to 0.7 μm) quantities were measured for the total (hemispherical), direct solar, and diffuse (including forward scattering) components of the radiation field. The aerosol optical depth at 500 nm, obtained from one of the narrrowband spectral channels, ranged from approximately 0.2 to 0.7 during the period encompassing the intensive field phase (IFP) of INDOEX. The diurnally averaged atmospheric forcing efficiencies determined for the total solar broadband and visible spectrum are −72.2 ± 5.5 W m−2 and −38.5 ± 4.0 W m−2, respectively. Model simulations driven by in situ measurements and realistic aerosol optical properties infer a single scattering albedo of 0.874 ± 0.028 at 500 nm.
 Aerosols have a major impact in determining the radiative properties of the atmosphere as well as affecting its total energy budget. The size, composition, magnitude, distribution, and other properties of aerosols all contribute to radiative structure of the atmosphere. The introduction of a large quantity of highly absorptive particles into the environment could potentially alter the climate through atmospheric heating or cooling, sea surface temperature changes and other effects that have been shown to impact climate on the planetary and local scales [Graham, 1995; Valero et al., 1997]. In addition, aerosol radiative forcing and the interaction between aerosol particles and solar radiation are major sources of uncertainty [Hansen et al., 1998] in modeling climate change in global circulation models.
 The Indian Ocean Experiment (INDOEX) [Ramanathan et al., 1996, available at http://www-indoex.ucsd.edu] was a multiagency international field experiment developed to assess the impact of aerosols on the environment. The intensive field phase (IFP) of INDOEX took place during February and March 1999 in a region of the Indian Ocean that is greatly affected by the presence of anthropogenic aerosols. The prevailing surface and lower atmosphere winds that exist during the monsoon season serve to transport pollutant particles generated on the Indian subcontinent to the entire Northern Indian Ocean. Atmospheric radiation, in situ sampling, meteorological, and other measurements made from a variety of platforms including land surface, ship, balloon, aircraft, and satellite platforms served to make up a very comprehensive data set available for addressing the climatic impact of the aerosols in this region.
 In this study, we use surface radiation measurements to quantify the effect of aerosols (natural and anthropogenic) in the region. This is accomplished by relating the variation in the net surface flux to the magnitude of the total columnar aerosol optical depth. By studying this relationship between the fluctuation of the net surface flux and aerosol loading, the overall solar radiative effect relative to the ideal aerosol-free environment, the surface forcing, can be estimated. We also compare our radiation measurements to model simulations to assess the various contributions from different aerosol types on the total measured optical depth and infer a mean single scattering albedo for the INDOEX IFP.
2. Radiometric Measurements
 Radiometric measurements from the Radiation Measurement System (RAMS) [Valero et al., 1997; Bush et al., 1999] took place during the INDOEX IFP from 12 February to 28 March 1999 at Kaashidhoo Climate Observatory (KCO). The tiny island of Kaashidhoo (4.965 N, 73.466 E) was selected as the observatory site due to its remoteness from major landmasses (approximately 700 km from the Indian subcontinent) as well as other neighboring islands (approximately 20 km from the closest Maldivian island). The specific location of the observatory on the island was selected such that the prevailing winds came from over the ocean, thus avoiding any local pollution (e.g., burning) generated by the few inhabitants of Kaashidhoo.
 RAMS consisted of two broadband (0.3 to 3.81 μm) radiometers - a Total Solar Broadband Radiometer (TSBR) and a Direct Solar Broadband Radiometer (DSBR) - and one narrowband Total-Direct-Diffuse Radiometer (TDDR). The TDDR covered the visible spectrum with 7 channels: one at 500 nm with a 10 nm band pass and six 50 nm band pass detectors spanning the 400 to 700 nm spectral region. The TSBR had a hemispherical field-of-view (FOV) and measured the solar irradiance. The TDDR also had a hemispherical FOV, but with the inclusion of a scanning shadow band (approximately 1 min period), permitted measurement of the total, direct, diffuse, and forward scattering components of the narrowband fluxes. The DSBR was mounted on a solar tracker and had a FOV with an opening angle of 5.7 degrees centered on the solar disk. Table 1 summarizes the instrument types, band passes, FOV, and absolute measurement uncertainties.
Table 1. Summary of RAMS Instrumentation and Absolute Uncertainties
0.3 to 3.81 μm
Hemispherical (cosine response)
0.3 to 3.81 μm
5.7° Full angle tracking the Sun
1) 0.495–0.505 μm
Hemispherical (cosine response)
2) 0.400–0.450 μm
3) 0.450–0.500 μm
Shadow band transits through the FOV once every approximately 45 s
4) 0.500–0.550 μm
5) 0.550–0.600 μm
6) 0.600–0.650 μm
7) 0.650–0.700 μm
 All instruments were fabricated and calibrated “in-house” [Valero et al., 1982; Valero and Ackerman, 1989]. The absolute radiometric response was determined for the broadband radiometers by referencing the signal to that of an absolute cavity radiometer, both instruments looking at the Sun with the same field of view under cloud-free conditions. The absolute TDDR response was measured in an optics laboratory using a National Institute of Standards and Technology (NIST) standard lamp. Finally, a complete zenith/azimuth angular response and dark signal characterization was accomplished. The absolute calibration accuracies are summarized below for each instrument type.
 The combination of the two broadband instruments gives the direct measurement of the total downwelling and direct components of the solar flux as well as the ability to infer the diffuse component. This diffuse component is calculated by differencing the TSBR measurement and the downwelling component of the DSBR measurement (as determined from the known position of the Sun). In this process, corrections due to slight deviations of the TSBR's angular response relative to that of an ideal cosine response are made, thus minimizing any systematic errors. These corrections are typically less than a few percent of the measured direct flux. The direct measurement of the DSBR contains both the unattenuated solar photons as well as those scattered into the narrow 5.7°FOV of the instrument. A technique for estimating this forward scattered component and correcting for the direct signal is described later. The absolute accuracy of the RAMS broadband instruments has been determined to be good to about 1% in previous surface and airborne radiation measurement campaigns [Bush and Valero, 1999].
 The shadow band analysis of the TDDR measurements results in the determination of the various components of the radiation field: when the shadow band is out of the FOV, the total flux is measured; the direct component is determined when the TDDR is directly shadowed by the band; the curvature near the “dip” region corresponds to the solar aureola and gives an estimate of the forward scattering; the diffuse portion is the difference of the total and direct terms and includes the forward scattering term [Valero and Ackerman, 1989]. Absolute accuracies are typically about 3% for the irradiance [Valero et al., 1997]. An example of a single shadow band scan is given in Figure 1 for a low aerosol period (8 March 1999; τ500 = 0.328) as well as a high aerosol period (27 March 1999; τ500 = 0.523). In each case the solar zenith angle is approximately 10.2°. The direct component, BN (N is the day in March), is calculated as the difference between the minimum of the dip, and the flux value just before the solar disk is block by the shadow band. Using this method, the forward scattered photons are not included in the representation of the direct (unscattered) flux, thus leaving the optical depth calculations free from this contamination. The forward scattered component itself, CN, can be estimated by subtracting the flux value just before the disk blocks the Sun with the flux when the shadow band is completely out of the field of view. Combining this term with the flux value at the minimum of the dip, AN, gives the entire diffuse component.
Table 2 summarizes the terms required to determine the various flux components from a single TDDR shadow band cycle. Note that it is important that the data rate of the TDDR is sufficient enough that the minimum of the dip as well as the points just before and after the Sun is blocked are adequately resolved. The nominal data rate of RAMS at 5 Hz was determined to be sufficient to satisfy these conditions. Also, errors in separating the various components as described above are minimized when the total flux is constant over the time period of one TDDR shadow band cycle (approximately 1 min). Since the aerosol optical depth measurements are restricted to clear sky conditions, these criteria are essentially met for all valid times.
Table 2. Terms Required in Calculating the 500–550 nm Total, Direct, Diffuse, and Forward Scattered Components From a TDDR Scana
 To demonstrate the variability of the flux components with respect to different aerosol loadings, Figure 1 shows sample profiles of a single TDDR dip for both low and high aerosol conditions. Both scans occur during predominantly clear sky conditions and a solar zenith angle of approximately 10.2°. On 8 March, the low aerosol day, the total flux is larger than that on 27 March, the high aerosol day. This is largely due to the fact that the direct component on the low aerosol day, B8, is much greater than that on the more polluted day, B27. The enhanced diffuse signal (A27 + C27 with respect to A8 + C8) typifies the additional scattering associated with the more abundant aerosol conditions. Furthermore, the decrease in the total signal during the greater optical depth day is a direct indication of the atmospheric forcing efficiency of the aerosol.
 The visible (400–700 nm) spectrum is determined via the measurements made by the multichannel TDDR instrument. The six channels, each with 50 nm band passes, spanning the 400 to 700 nm spectral range is summed to estimate the energy in the visible spectrum. Just as with the individual channel measurements, the total, direct, diffuse, and forward scattering components of this signal are also determined.
2.4. Aerosol Optical Depths
 The direct term of the TDDR 500 nm channel is used in calculating the optical depth at this wavelength. Attenuation of the line-of-sight (LOS) solar irradiance is determined relative to top-of-atmosphere (TOA) values and optical depths are calculated using the Lambert-Beer law. The optical depths are then converted from LOS to vertical values after correcting for atmospheric path lengths. The aerosol optical depth is then determined after subtracting the Rayleigh scattering contribution and any absorption components from minor atmospheric constituents (e.g., ozone). These nonaerosol components are estimated with the MODTRAN [Anderson et al., 1995] atmospheric model with no aerosol loading while using an in situ radiosonde to define the atmospheric profile. At KCO during the INDOEX IFP, the 500 nm aerosol optical depths were typically greater than 0.2 and occasionally increased as large as 0.7. Figure 2 shows a histogram of the 500 nm aerosol optical depths during the IFP after being screened for clear sky conditions.
 TDDR data was screened to identify clear sky periods during the IFP. Data was analyzed in 10-min intervals and the variability of the signals and optical depth were examined in these periods. A measurement was tagged as clear if the deviation in both the total and direct 500 nm flux measurements was less than 1% relative to a linear fit of the data in a 10 min period centered at measurement time. By comparing to the fit of the 10-min data, natural variation of the fluxes due to changes in the Sun's elevation angle is not restricted.
 Two other instruments were operated at KCO during the IFP that measured aerosol optical depths. The CIMEL [Holben et al., 1998] Sun photometer measured aerosol optical depths at 340, 380, 440, 500, 670, 870, 940, and 1020 nm. The MICROTOPS (S. K. Satheesh, unpublished data, 1999) hand held Sun photometer provided values at 380, 440, 500, 670, and 870 nm. For comparison purposes, the aerosol optical depths at 500 nm, common to both of the above instruments as well as the RAMS TDDR, are shown in Figure 3. The data points displayed in this figure correspond to times that are tagged as clear using the technique described earlier. The regression fits show no apparent biases between any of the data sets and only a few outliers exist for the two-month measurement period. The optical depths referred to hereon in this paper are taken from the RAMS 500 nm TDDR measurements.
 Atmospheric forcing is a commonly used parameter used in quantifying the effect that aerosols have on the energy budget of the atmosphere. Expressed in different forms (e.g., instantaneous, diurnally averaged, etc.), it represents the overall change in the net radiative behavior of the atmosphere resulting from the presence of various sources. It is simply calculated as the difference between the net (downwelling minus upwelling) radiative flux for the given atmospheric conditions and the same quantity in a pristine (aerosol-free) environment:
In determining the surface forcing, the net flux is estimated as
where F is the irradiance (F0 in the pristine case), α is the surface albedo, and the factor μ, the cosine of the solar zenith angle, is used to normalize the values for an overhead Sun. Since the upwelling flux at the surface was not measured at KCO, we estimate the broadband surface albedo (about 3% over open ocean) from the low-altitude legs of the research flights that took place concurrently during the INDOEX IFP (A. Bucholtz, unpublished data, 1999). The resulting surface forcing then becomes:
This equation is valid for the appropriate spectral region of interest. In this work, we will be mainly concerned with the broadband forcing, ΔFBB, and the visible forcing, ΔFVIS.
 The pristine atmosphere used as a reference in determining the atmospheric forcing is calculated using an atmospheric radiation and transmission model, MODTRAN [Anderson et al., 1995], free of aerosols and clouds. The profile of the major atmospheric constituents is taken from in situ balloon soundings. This model calculation represents the basis for estimating the cloud and aerosol/pollution free environment. For the observations on 27 March at KCO, a relatively polluted day, the surface flux for the pristine case is shown in Figure 4 for both the broadband and visible spectral regions. In this essentially cloud-free case, the difference between the pristine case (dashed) and measurement (solid) is the effect of the aerosols present in the environment. In comparison, the pristine atmosphere simulation and flux measurements for a less polluted day, 8 March, are presented in Figure 5. The difference between the measurement and aerosol-free environment is clearly less in the later case. In the next section, we quantify this difference in relation to the magnitude of the aerosol present.
 The broadband and visible forcings are calculated using equation (3) spanning the time period of the KCO measurements. These quantities are screened for clear-sky conditions (as described above) to eliminate the effects and contributions of clouds. Figure 6 depicts the instantaneous forcing for all clear sky conditions at KCO during INDOEX after being correlated with the 500 nm aerosol optical depth measurement. The scatterplot shows a definite dependence of both the broadband and visible forcing in relation to the magnitude of the optical depth. The average forcing per unit optical depth, better known as the forcing efficiency, is determined for both the broadband spectral regions:
The spread in the data points in Figure 6 is partially explained by the fact that the cases used do not represent clear skies entirely. There may be some clouds out of the field of view of the direct beam measurements that can cause some variability in the total downwelling fluxes. Also, when normalizing the fluxes to overhead Sun conditions, this approximation will tend to be less accurate for increasing solar zenith angles and also for increasing diffuse signals. Furthermore, the approximately 0.03 absolute accuracy in the measured optical depth will also contribute to the spread in the forcing versus optical depth relationship.
4. Direct Determination of Net Flux Forcing Versus τ Approach
 Another method of calculating the forcing efficiency, independent of any model simulations of the pristine atmospheric conditions, consists of relating the net flux at the surface to the optical depth. Figure 7 is a scatterplot of the net flux (normalized to overhead Sun) as given in equation (2) as a function of the 500 nm optical depth. The slope of the linear fit of the data points represents the change in the net flux per optical depth and is the forcing efficiency. Here, the τ500 = 0 intercept is surface net flux during aerosol-free conditions. The resulting fits of the broadband and visible net fluxes are
The instantaneous noontime forcing efficiencies determined using the two methods described above are comparable to one another. Meywerk and Ramanathan  measure a slightly smaller forcing efficiency in the 400–700 nm spectral region for noontime as −84 W m−2 during the first field phase of INDOEX in 1998. Jayaraman et al.  measure the same quantity during a pre-INDOEX cruise in January and February 1996 to be −156 ± 15 W m−2. The instantaneous visible forcing efficiency of −112.9 ± 7.6 W m−2 in this work is consistent with other measurements made in this region of the Indian Ocean [Valero et al., 1999].
5. Diurnally Averaged Forcing
 For the surface measurements at KCO in which a significant portion of a day is cloud free, it is possible to integrate the atmospheric forcing and determine a diurnal average. For example, the effects of clouds on the flux can be removed by fitting a curve to the clear sky portions during the day. The forcing is then calculated using the pristine atmosphere model simulation and then integrated over the entire day to obtain the total forcing for that day. For the sample high aerosol day, 27 March (see Figure 4), and low aerosol day, 8 March (see Figure 5), the average 500 nm aerosol optical depths are 0.575 and 0.334, respectively. For the broadband spectrum, the diurnally averaged forcing (total forcing divided by 24 hours) is −43.3 and −23.5 W m−2 resulting in forcing efficiencies (forcing per unit optical depth) of −75.3 and −70.4 W m−2. Similarly for the visible spectrum, the 27 March and 8 March forcings are −23.6 and −12.9 W m−2 giving efficiencies of −41.0 and −38.6 W m−2.
 For all the predominantly clear days at KCO during INDOEX, Figure 8 shows the diurnally averaged forcing in relation to the 500 nm aerosol optical depth average for the day. The mean diurnally averaged forcing efficiencies at KCO during the INDOEX IFP are −72.2 ± 5.5 W m−2 and −38.5 ± 4.0 W m−2 for the broadband and visible spectrum, respectively. This broadband forcing efficiency is consistent with the aircraft observations reported by Valero et al.  and with the measurements made by Satheesh and Ramanathan  at KCO during 1998 and 1999 who estimate the diurnally averaged broadband forcing to be from −70 to −75 W m−2.
6. Model Validation
 All model calculations were made using the MODTRAN [Anderson et al., 1995] atmospheric transmission model version 3.5 using the DISORT [Stamnes et al., 1988] radiative transfer code to simulate the radiative fluxes coincident with the measurements at KCO. Balloon sondes launched every 6 hours from KCO were used to parameterize the atmospheric environment. Furthermore, aerosol optical depths measured by the TDDR defined the amount of aerosol loading for the given simulation. Both the total downwelling irradiance and the unattenuated direct solar flux are calculated during the simulations. The diffuse component is inferred by subtracting the direct component multiplied by the cosine of the solar zenith angle from the irradiance. In all model calculations described below, simulations only occurred for periods that were identified as being essentially cloud-free.
 Previous sensitivity studies of this model have been completed for cloud-free conditions with various aerosol loadings - similar to conditions described in the work [Valero and Bush, 1999]. Table 3 summarizes these uncertainties for each of the total, direct, and diffuse components of the atmospheric radiation. The model uncertainties have roughly the same percentage for the broadband and visible components of the solar spectrum. In absolute terms (estimated from solar noon values), the uncertainty in the diffuse term is about twice that of the total or direct components.
Table 3. Estimated Uncertainties in Cloud-Free MODTRAN Calculationsa
Broadband Uncertainty (Peak, W m−2)
Visible Uncertainty (Peak, W m−2)
From Valero and Bush . Estimates of the absolute uncertainties are determined for solar noon conditions.
6.1. Flux Comparisons
 Direct comparisons can be made between the simulated irradiances and the measurements made by the TSBR. However, the calculated and measured direct components may not be explicitly compared until the DSBR signal is corrected for photons that are forward scattered into the 5.7° FOV. The model calculations do not consider photons singly or multiply scattered into the FOV contained in the line-of-sight to the Sun in the “direct” term. When there is a significant amount of aerosol present and photons are more effectively scattered as well as the case when the size of the scattering particle is large, a nonnegligible amount (up to ∼5%) of the DSBR signal may be from single or multiple scattered photons. Estimates of the magnitude of the broadband forward scattered component are made using the analysis of the TDDR shadowband measurements (see Figure 1 and Table 2). The ratio between the forward and total visible flux determined using the TDDR measurements is used in approximating the forward scattered component in the total solar broadband region from the measured irradiance. An example of this is given below.
 Initial model calculations were completed using a sulfate aerosol entirely [d'Almeida et al., 1991], consistent with a predominantly marine environment. In situ samples and previous studies indicate that the actual aerosol composition also had a significant contribution from a more absorptive soot component [Satheesh et al., 1999]. The optical parameters used in characterizing these two aerosols, sulfate and soot, in the MODTRAN simulations were take from tabulated of the extinction and absorption coefficients as well as the asymmetry parameter and are summarized in Figure 9 [d'Almeida et al., 1991]. Further simulations were completed partitioning the aerosol optical depth that was driving the simulations between these two aerosol types. Figure 10 depicts the results of these simulations for the clear sky portions of 27 March. The solid lines indicate the model calculations for the irradiance for various percentages of the sulfate/soot contribution to the measured optical depth: 100:0, 90:10, 80:20, 70:30, and 60:40%. The dashed and dotted lines represent the direct and diffuse components respectively. The fluxes derived from measurements at these times are plotted with diamonds (total irradiance), triangles (direct), squares (diffuse), and X's (forward scattering). The relatively slight variation in the direct flux calculations compared to that in the total and diffuse components is a result of the model run being constrained by the measured aerosol optical depth. The larger variations in the total and diffuse components are a direct consequence of the differing optical properties of the sulfate and soot aerosols: the sulfate being more scattering and the soot being more absorptive.
 The irradiance measurements best match the simulations when the contributions to the aerosol optical depth are approximately 81.0% from sulfate and 19.0% from soot. The uncertainty in these estimates is roughly 2.7%. In all model simulations, the direct flux is below the DSBR measurements. This is consistent with the forward scattering contamination mentioned above that is associated with the large particle size of the scattering aerosol. When the measured direct signal is reduced by the forward scattering estimate (about 5% of its absolute magnitude) and similarly the diffuse component is increased by the same amount, the model comparisons of each of the three radiation components are all roughly consistent with the same sulfate/soot partitioning (Figure 11). The best fit partitioning for the diffuse component implies an 80.4% sulfate and 19.6% soot contribution with an absolute error of 1.7%. After the forward scattering is accounted for, the measured and simulated direct flux components compare with experimental and modeling uncertainties.
 The same procedure used above for the broadband spectral region can be applied to the visible region. Figure 12 gives TDDR measurements as well as the calculated visible spectral fluxes for the same model simulations for 27 March as above. Here, the direct and diffuse fluxes are free from the effects of forward scattering. This comparison gives a sulfate contribution to the overall optical depth as 84.5% ± 1.9% using the total visible fluxes and 83.9% ± 1.8% using the diffuse components. These values are slightly larger than the same term determined with the broadband comparisons but are within the uncertainties of the simulations.
6.2. Estimate of Single Scattering Albedo
 The single scattering albedo, ω0, at 500 nm corresponding to the 81.0% sulfate and 19.0% soot mixing is 0.865 ± 0.029 and that for the 84.5% sulfate and 15.5% soot mixing is 0.890 ± 0.020. This parameter is calculated using the defined mixing ratios (in optical depth space) and the aerosol optical parameters given in Figure 9. These values are consistent with other measurements and estimates of the single scattering albedo in this region during the northeast monsoon. Satheesh and Ramanathan  used measurements at KCO during the 1998 and 1999 monsoon periods to determine ω0 in the range of 0.87 to 0.90.
 Measurement and model comparisons were also made for the lower aerosol day, 8 March. Here, the best fit sulfate and soot mixing ratio of the aerosol optical depth was 83.3% for sulfate and 16.7% for soot with an error of 1.9%. These values are statistically consistent with the values inferred from the 27 March data. The corresponding ω0 of 0.881 ± 0.020 also matches the 27 March values. Combining the 8 March and 27 March measurements, we estimate ω0 to be roughly 0.874 ± 0.028 in the Indian Ocean region during the INDOEX campaign.
 The RAMS measurements at KCO are consistent with other findings [Jayaraman et al., 1998; Meywerk and Ramanathan, 1999; Valero et al., 1999; Satheesh and Ramanathan, 2000] that the aerosols present in the Indian Ocean region during the northeast monsoon are numerous in magnitude (τ500 is as large as 0.7) and highly absorptive in nature (ω0 = 0.874 ± 0.028). In this work, we used realistic aerosol optical properties [d'Almeida et al., 1991] as inputs to clear sky model simulations that are based on in situ measurements and other observations [Satheesh et al., 1999] of the aerosol composition in the region. The difficulty in determining the overall wavelength dependence of the aerosols via in situ measurements necessary to adequately characterize them in the entire broadband spectral region lead us to use of tabulated extinction (kext), absorption, (kabs), and asymmetry parameter (g) as model inputs.
 The best comparisons between the model and measurements occur when the contribution to the total aerosol optical depth at 500 nm is approximately 83% from a highly scattering sulfate-type aerosol and about 17% from a very absorptive soot-like particle. The similarity in the contribution from sulfate and soot, determined from the measurements on a relatively low aerosol day (8 March) and high aerosol day (27 March), indicates that the aerosol composition remains somewhat unchanged whereas the fluctuation in the total aerosol optical depth is predominantly due to the variation in the magnitude of aerosol present. With this parameterization of the sulfate and soot contributions, all three of the measured radiative components – total, direct, and diffuse – match the simulations both the broadband and visible spectral regions within the uncertainties of each data set. Because of the large particle size, highly scattering nature of the sulfate aerosol, and large optical depths, the degree of forward scattering of solar photons into the narrow field of view of the direct beam measurement is significant and needs to be accounted for in the data interpretation. Only when properly accounting for forward scattering do each of the individual components match the model.
 The surface forcing associated with this aerosol present during INDOEX may be significant from the climate point of view. Diurnally averaged forcing efficiencies of −72.2 ± 5.5 W m−2 and −38.5 ± 4.0 W m−2, for the broadband and visible spectral regions, indicate that the aerosol has a major impact on the environment. This corresponds to a daily reduction in the radiative heating at the surface typically between −15 W m−2 and −40 W m−2 during a period covering a few months in the monsoon season. Furthermore, the TOA forcing is estimated using NASA's Cloud and Earth's Radiant Energy System (CERES) satellite [Weilicki et al., 1996] to be from −4 W m−2 and −10 W m−2 [Satheesh and Ramanathan, 2000]. This implies an increase in the total atmospheric radiative heating of roughly 10 W m−2 to 30 W m−2.
 We greatly acknowledge the government of the Republic of Maldives and the inhabitants of Kaashidhoo in their assistance during INDOEX and at KCO. Funding for operating RAMS at KCO came from NASA through contract NAG1-1259 and grant NAS1-98141 with the Langley Research Center and from the National Science Foundation grant NSF-ATM-9612890. Measurements and instrument operations were accomplished by efforts of Anthony Bucholtz, Shelly Pope, and Sabrina Simpson at the Atmospheric Research Laboratory, Scripps Institution of Oceanography. We are also grateful for assistance at KCO from Jurgen Lobert, William Conant, and Eric Wilcox. We also acknowledge the efforts of Hung Nguyen who managed the facility at KCO and coordinated many of the operations at that site.