Information on radiative fluxes that reach the ground is needed in numerous areas of climate research. On a global scale, such information is obtainable only from satellites. Top of the atmosphere satellite observations during clear and cloudy sky conditions have been found useful for inferring, among others, information on aerosol optical depth (AOD) and cloud optical depth (COD). These are important elements for estimating surface radiative fluxes. Satellite retrievals of AOD are based on the assumption that the aerosols are of a specific type. In certain climatic regions, dust aerosols and those from biomass burning are injected into the atmosphere simultaneously and are distributed distinctly in the vertical. In this study, it is demonstrated that in such scenarios inaccurate assumption on the vertical distribution of the aerosols can lead to errors in AOD estimates from satellites and, subsequently, in the inferred shortwave (SW) radiative fluxes that reach the surface. The magnitude of such errors can be as high as 80 W m−2. The frequency of these mixed situations has not as yet been documented, and it is possible that on global scale the impact is small. Using available records of observed SW radiative fluxes, their reduction (“dimming”) at numerous locations has been reported. Some relate this reduction to aerosol effects. The findings of this study have implications for the ability to assess such “dimming” from satellites in areas where the vertical structure of aerosol distribution needs to be accounted for.
 The radiation balance of the Earth's atmosphere can be altered by atmospheric constituents such as aerosols. Their influence on this balance is referred to as aerosol radiative forcing (ARF). In the Intergovernmental Panel on Climate Change (IPCC)  report, it is stated that radiative forcing is usually quantified as the “rate of energy change per unit area of the globe as measured at the top of the atmosphere” and is given in units of W m−2. Positive radiative forcing leads to an increase of the Earth-atmosphere energy leading to warming. Negative radiative forcing will decrease the energy, leading to cooling. ARFs can be direct (as the scattering and absorption of radiation) or indirect (as the mechanism by which aerosols modify the microphysical and radiative properties, amount, and lifetime of clouds). The combined effects of the direct and indirect ARF have resulted in net cooling in the global climate system [IPCC, 2007]. More recently the semidirect effect of aerosols has drawn attention, namely, that absorbing aerosols can heat the atmosphere, may increase cloud droplet evaporation, increase low-level atmospheric stability, and lead to a positive semidirect radiative forcing [Ackerman et al., 2000]. The present study addresses only the direct effects of anthropogenic aerosols, primarily from biomass burning and from mineral dust, at the surface, top of the atmosphere (TOA), and within the atmosphere.
 The effect of scattering by aerosols is a net negative direct forcing; partially absorbing aerosols may exert a negative TOA direct ARF over dark surfaces and a positive TOA ARF over bright surfaces [Hansen et al., 1997]. For scattering aerosols, the surface forcing is about the same as the forcing at the TOA. However, for partially absorbing aerosols, the surface forcing may be larger than the one at the TOA [Ramanathan et al., 2001; Pandithurai et al., 2004; Mallet et al., 2008]. The following information is needed for estimating direct ARF: aerosol optical properties (such as the optical depth), the single scattering albedo, extinction coefficient, the scattering phase function (which varies as a function of wavelength), and relative humidity. If indeed, as speculated, aerosols are the cause of “global dimming” [Stanhill and Moreshet, 1992], it is crucial to obtain accurate estimates of ARF at the surface.
 In this study, we use observations from the sub-Sahel for estimating ARF at the surface to illustrate the limitations of estimates of this parameter from satellite observations, when information on the vertical structure of the aerosols is unavailable. The data used includes surface observations of SW radiative fluxes and AOD, radiative fluxes derived from satellite observations, and unique observations of aerosol vertical profiles made in support of the Radiative Atmospheric Divergence Using ARM Mobile Facility Gerb and Amma Stations (RADAGAST) [Slingo et al., 2006; Miller and Slingo, 2007]. The vertical profiles were obtained in 2006 during a campaign in Niamey, Niger with the participation of the ARM Program and the Facility for Airborne Atmospheric Measurement (FAAM) of the U. K. Met Office. The aim was to improve the understanding of the radiative balance of the atmosphere by combining simultaneous measurements of radiative fluxes at the surface and at the top of the atmosphere and information on the vertical structure and properties of aerosols [Johnson et al., 2008a, 2008b; Osborne et al., 2008]. In the framework of the RADAGAST activity (Dust and Biomass Experiment, DABEX), numerous flights were conducted during January and February 2006; two were close to a ground observing station located in Ilorin, Nigeria. The data used in this study are described in section 2; the site background is presented in section 3; estimates of aerosol radiative forcing are discussed in section 4; experiments with variable vertical distribution of aerosols are described in section 5; findings from observed vertical profiles over the site are described in section 6; and implications for satellite retrievals of surface radiative fluxes are presented in section 7.
1.2. The Site
 During the dry season in West Africa, large amounts of Saharan mineral dust and aerosols from sub-Sahelian biomass burning cause one of the strongest, widespread, and persistent aerosol signals over the globe [Holben et al., 1998]. Each year, between November and March, a large amount of dust is transported from the Chad basin passing over Nigeria to the Gulf of Guinea [Kalu, 1979]. The agent for this dust transport is the dry “harmattan” wind, creating conditions conducive to forest fires. The dust plumes can have a thickness of up to 3 km and reduce the visibility to less than 1 km. This results in reduction of surface solar radiation by a magnitude that is not well documented.
 To better quantify these effects, continuous surface monitoring of aerosols, solar irradiance, and environmental parameters has been carried out at Ilorin, Nigeria (08°19′N, 04°20′E) [Pinker et al., 2001, 2006; Holben et al., 2001; Smirnov et al., 2002]. In this study, simultaneous measurements of aerosol optical properties, downward SW fluxes, and radiative transfer models are used for selected clear days during the dust-biomass burning season for estimating aerosol radiative effects. Unique aircraft observation made during DABEX on the vertical structure of aerosols and other gaseous species in the vicinity of the ground observing site were utilized in this study. The experiment took place during the dry season of January and February 2006, which was designated as the Special Observing Period SOP-0 of the African Monsoon Multidisciplinary Analysis (AMMA) Project [Lebel et al., 2003]. The flights were also part of the RADAGAST Experiment [Slingo et al., 2008]. These combined observations helped to illuminate the importance of information on the vertical distribution of aerosol properties, when satellite observations at the top of the atmosphere are used for inferring surface radiative fluxes.
2. Data Used in Study
2.1. Ground Observation
 Ilorin Nigeria (8°32′N 4°34′E) is the capital of Kwara State, situated in southwestern Nigeria within the savannah zone. During the dry season when the ITCZ (Intertropical Convergence Zone) shifts to the Southern Hemisphere, wind blown dust and biomass burning aerosols from “slash and burn” agricultural practices dominate this region. A CIMEL Sun-sky radiometer has been set up on the campus of University of Ilorin (as a member instrument participating in Aerosol Robotic Network (AERONET) since 1997 [Pinker et al., 2001; Pandithurai et al., 2001]) for monitoring these elevated aerosol events. The observations are centrally processed at NASA Goddard Space Flight Center (GSFC) [Holben et al., 1998], and the derived aerosol optical properties can be publicly accessed from http://aeronet.gsfc.nasa.gov. The instrument is an automatic Sun-tracking sky radiometer capable of measuring both Sun and sky radiance at eight spectral channels. The filters are centered at wavelengths 340, 380, 440, 500, 670, 870, 940, and 1020 nm, located in a filter wheel, which is driven by a stepper motor. The observations are transmitted almost in real time via the European Meteosat (D. Tanre, private communication, 1997). The instrument is periodically calibrated at the central calibration facility at GSFC against an instrument whose calibration is traceable to Maona Loa. Detailed information on measurement protocol, radiometric precision, calibration procedures, and processing methods are described by Holben et al. .
 Observations at the site include key components of the radiation budget and environmental parameters such as temperature, humidity, and wind speed. Results from these ground observations have been reported by Pandihurai et al. , Pinker et al. , Holben et al. , Smirnov et al. , and Pinker et al. . In this study, we used information on global SW radiation during the following three dry seasons (December 1998 to February 1999, December 1999 to February 2000, December 2000 to February 2001). Table 1 lists instruments that were used at Ilorin to measure SW radiation. Data used in this study were obtained with the following instruments: PSP 17883F3 and PSP 17675F3. For achieving the highest possible accuracy and referencing to international standards, the calibration of the instruments during the measurement period has been performed at recognized calibration centers. These include the Eppley Laboratory in Newport, Rhode Island, the Radiation Center at Davos, Switzerland, and the NOAA Climate Monitoring and Diagnostics Laboratory (CMDL) at Boulder. The entire calibration history for instrument 17883F3 is presented in Figure 1, which shows the gradual degradation of this instrument. The calibration constants for the period used in this study are based on daily values interpolated between 7.58 (15 September 1997, ACI) and 7.33 (14 September 2001, Boulder). The calibration history for instrument 17675F3 is presented in Figure 2, which reveals a large degradation in 1999, attributed to paint deterioration of the sensor. The instrument was sent to Eppley Laboratory for repainting and recalibration, after which it was redeployed in Ilorin and the data acquired were initially processed with the new calibration. When the instrument was due for recalibration, it was sent to CMDL (E. Dutton, NOAA Earth System Research Laboratory private communications, 2001–2010). The difference between this calibration and the one done after repainting seemed too large. Another indication that the Eppley calibration was too high was the discontinuity in the entire data record of SW fluxes when this calibration was used. Since during the period 1978–1994 the instrument behaved “well,” a linear fit was applied to the calibration points during this period and a degradation rate was derived. It is reasonable to assume that the instrument maintained the same characteristics after the repainting. Assuming that the 2004 calibration is correct, we have used the derived degradation rate of this instrument from the earlier record and went back in time from 2004 to derive a “calculated” calibration value of 8.55 (the black dot in Figure 2) for the period starting after the repainting. The processing of the data was redone using daily interpolated calibration values between 8.55 and 8.32. Using this calibration and adding the reprocessed data to the entire record since the beginning of the observations at the site, a homogeneous time series was obtained (not achievable with the calibration value after repainting) (Figure 3).
Table 1. Instruments Used to Measure SW Radiation at the Ilorin Site
Period and Instrument
1 Sep 1992 to 31 Aug 1994, Eppley 17675F3
1 Jun 1995 to 30 Apr 1998, Eppley 19227F3
1 Nov 1998 to 31 Aug 1999, Eppley 17883F3
Period 1: 1 Sep 1999 to 12 Jul 2004, Eppley 17675F3
Period 2: 28 Jul 2004 to 12 Jul 2005, Eppley 17883F3
2.2. Airborne Observations
 The ARM Mobile Facility (AMF) includes a wide range of passive and active instruments for measuring radiative fluxes at the surface and within the atmosphere. The AMF was deployed at the Niamey Airport for 2006 as part of the RADAGAST project [Slingo et al., 2006], in cooperation with the AMMA experiment [Lebel et al., 2003]. The FAAM facility provided flight support for the experiment. It was jointly funded by the U. K. Met Office and the U. K. Natural Environmental Research Council (NERC). Detailed overview of the Dust and Biomass Burning Experiment (DABEX) and the African Monsoon Multidisciplinary Analysis (AMMA) Special Observing Period-0 can be found in the study of Haywood et al.  with a focus on aerosol and radiation measurements. The aircraft measurements of biomass burning aerosol over West Africa during DABEX are detailed upon in the study of Johnson et al. [2008a, 2008b, 2009] who investigated the properties of the biomass burning aerosols using data from the U. K. FAAM aircraft. Physical and optical properties of mineral dust aerosols during DABEX are discussed in the study of Osborne et al. , who identified a mineral dust layer below 1–2 km (sourced from the north) and an overlying biomass burning (BB) layer (sourced from anthropogenic fires to the south) on all days of the experiment. Aerosol direct radiative forcing over northern Benin during the AMMA dry season experiment (Special Observation Period-0) has been estimated by Mallet et al. .
3. Background Information on the Ilorin Site
 Clear-sky solar radiation reaching the surface at Ilorin decreases with the increase of atmospheric aerosol as illustrated in Figure 4. The variations of daily averaged surface downward SW global flux and aerosol optical depth at 0.5 μm are plotted for three dry seasons (1998–2001), when both measurements are available. A negative linear correlation of −0.476 is present between the surface flux and aerosol optical depth, which cannot be rejected at 1% test level.
Table 2a lists the dry season monthly mean aerosol optical thickness at 0.50 μm, which ranges from 0.55 to 1.61 with a multiyear average value of 0.96. Such high concentrations of atmospheric aerosols along with the predominance of clear-sky conditions during the dry season make Ilorin an extremely valuable observational site for assessing direct radiative effects of aerosols. Statistical information on aerosol properties during the dry seasons for 1998–2006 is presented in Table 2b.
Table 2a. Dry Season Monthly Mean Aerosol Optical Thickness at 0.50 μm as Observed at the Ilorin Site
Table 2b. Statistical information on aerosol properties and total precipitable water over Ilorin, Nigeria during the dry seasons (Dec–Feb) during 1998–2006
Aerosol Optical Depth at 0.5 μm
Aerosol Angström Exponent (0.44–0.87 μm)
Aerosol Single Scattering Albedo
Aerosol Asymmetry Parameter
Aerosol Effective Radius [μm]
Precipitable Water [cm]
4. Estimates of Aerosol Radiative Forcing: Simulations
4.1. Daily Time Scale
 For assessing reduction in surface SW fluxes due to ambient aerosols, the SBDART model [Ricchiazzi et al., 1998] was used to simulate aerosol-free conditions over a vegetative surface. Model input on atmospheric profiles has been taken from the National Centers for Environmental Prediction (NCEP)-DOE Reanalysis 2 data [Kanamitsu et al., 2002] and daily column O3 amount from TOMS retrievals [Herman et al., 1997] (http://toms.gsfc.nasa.gov/ozone/ozone_v8.html). Broadband (0.3–0.7 and 0.7–5.0 μm) surface albedo was calculated from collection 4 MOD43C2 16 day L3 climate modeling grid MODIS/Terra BRDF/Albedo parameters [Schaaf et al., 2002], and AERONET-derived column water vapor was used to scale the NCEP relative humidity profiles to better represent the condition at Ilorin. Sixty-five days were identified as cloud free based on the examination of 1 min SW radiative fluxes. Daily averaged depletion of surface SW flux is calculated as the difference between pyranometer-measured and model-simulated irradiance and is displayed in Figure 5 as functions of AOD and Ångström exponent (α). At Ilorin, the reduction (“dimming”) of daily surface irradiance due to ambient aerosols ranges from 50 to 140 W m−2 (about 14%–40% of solar radiation), and 1 unit of AOD at 0.5 μm corresponds to a decrease of surface irradiance of about 90 W m−2 (about 24% of the solar radiation). Cases with large Ångström exponents, most likely associated with biomass burning with strong aerosol absorption, tend to have a larger effect than dust-dominated conditions (smaller α) for the same optical depths. In Niamey, Haywood et al.  reported a reduction in the midday solar irradiance of over 25%; about 50% of the reduction has been estimated to be due to smoke from biomass burning and 50% due to mineral dust. Mallet et al.  studied the direct radiative forcing over Benin (a RADAGAST supersite) during the AMMA SOP-0 and found that aerosols reduced the radiation reaching the ground by about 61.5 W m−2 by reflecting back to space but predominantly by absorption of the solar radiation within the atmosphere (about +41.1 W m−2). They estimate the mean heating rate at the surface and within the elevated biomass burning layer to be enhanced by 1.50 and 1.90 K d−1, respectively.
4.2. Instantaneous Time Scale
 SBDART simulations of 330 instantaneous cases at Ilorin were performed at times when AERONET detailed aerosol information from almucantar retrievals was available; it included spectral values of aerosol optical depth, single scattering albedo, and asymmetry parameter. SBDART default aerosol vertical profile (exponentially decreasing with scale height varying between 1.05 and 1.51 km) was used in the simulations. Figure 6 shows the comparison between ground measurements and SBDART simulation results for global, direct, and diffuse surface SW fluxes. The model simulated total surface irradiance has a positive bias of 22.24 W m−2 (7.8%), which can be largely attributed to the diffuse component. Figure 6c shows the overestimation of diffuse flux from the forward computations using observed values of atmospheric composition.
5. Experiments With Assumed Aerosol Vertical Distributions
 The disagreement found in Figure 6 can be due to the quality of the ground observations or lack of sufficient information on aerosols. The simultaneous occurrence of dust outbreaks and biomass burning complicates the situation. The RADAGAST activity at Niamey aimed at improving the understanding of the radiative balance of the atmosphere by combining simultaneous measurements (1) of radiation at the surface, (2) of radiation at the top of the atmosphere; and (3) of the vertical structure of the atmosphere. Probing of the vertical distribution and properties of aerosols was a major component of the RADAGAST effort [Johnson et al., 2008a, 2008b; Osborne et al., 2008]. Aircraft measurements and lidar observations indicated that aerosols from biomass burning tended to be located above the dust layer [Slingo et al., 2006]. Several simulations have been performed to better understand the effect of vertical distribution of the aerosols on the surface SW radiative fluxes and how this can impact the retrieval of these fluxes from satellite observations. Details on the assumptions that were made regarding the vertical structure are provided in Table 3. Additional assumptions include the following: total column τ0.55 μm = 0.6, solar zenith angle = 60°. For experiment 3, it has been assumed that aerosols from biomass burning had τ0.55 μm = 0.3 and were within 3–6 km; dust aerosols have τ0.55 μm = 0.3 and were within 0–3 km. For experiment 4, it has been assumed that dust aerosol have τ0.55 μm = 0.3 and were within 3–6 km; aerosols from biomass burning had τ0.55 μm = 0.3 and were within 0–3 km.
Table 3. Tests Designed to Investigate the Effect of Vertical Distribution of Aerosol on Surface Shortwave Radiationa
In the simulation, aerosol is assumed to be homogeneously distributed within its layer.
Within 0–3 km with aerosol optical depth as 0.6
Within 0–3 km with aerosol optical depth as 0.6
Smoke over dust
Smoke within 3–6 km with aerosol optical depth as 0.3; dust within 0–3 km with aerosol optical depth as 0.3
Dust over smoke
Dust within 3–6 km with aerosol optical depth as 0.3; smoke within 0–3 km with aerosol optical depth as 0.3
 The results from these experiments indicated that the assumptions on the optical properties of the aerosols (dust or biomass burning) has an impact on the derived SW fluxes (Table 4). The location of the dust layer with respect to the biomass burning aerosol layer had an impact on the TOA SW flux, but a lesser one on the surface flux (Figure 7), namely, the accuracy of surface fluxes inferred from satellite observations at the TOA (more details in section 7) without prior information on their vertical location will be affected.
Table 4. Results From Experiments Listed in Table 3
Surface Downward Total Flux (W m−2)
Surface Downward Direct Flux (W m−2)
Surface Downward Diffuse Flux (W m−2)
TOA Reflected Flux (W m−2)
6. Profiles From RADAGAST
 The ambient conditions during the DABEX flights (under the RADAGAST project) for the days when the aircraft reached Ilorin were described by Haywood et al.  and can also be found at http://www.faam.ac.uk. Most of the DABEX flights passed over the main AMF site Banizoumbou. On 17 January 2006, DABEX Flight B157 continued southeast toward Ilorin. As reported for this day “curious very thin midlevel clouds above the highest aerosol layer were noticed.” On 24 January 2006, DABEX Flight B162 also continued southeast to Ilorin. As reported, it was mainly clear with some clouds appearing around midday. The aircraft measurements taken during these two flights are presented in Figures 8 and 9. The wavelength dependence of the scattering coefficient in Figure 8 for 17 January indicates that fine-mode particles dominate. The peak in scattering is at about 2 km, and the aircraft profiles of CO also indicate a peak at that level, which can be attributed to biomass effects. On 24 January (Figure 9), there are two peaks. The stronger and higher one can be attributed to biomass burning as indicative from the wavelength dependence of the scattering coefficient. The lower peak has a much weaker spectral dependence and could be due to dust. Since there is some enhancement in the CO distribution, possibly, biomass aerosols were also present. The spectral variation of aerosol optical depth at Ilorin on 17 and 24 January 2006 as derived from the CIMEL instrument in the AERONET framework is presented in Figures 10 and 11. As the plots show, the AOD has little variability during the day. Johnson et al. [2008a, 2008b], who investigated the properties of biomass burning aerosols over the primary sites in West Africa with data from the U. K. FAAM aircraft during (DABEX), report that aged biomass burning aerosols were widespread across the region, often at altitudes up to 4 km. Fresh biomass burning aerosols were observed at low altitudes by flying through smoke plumes from agricultural fires. The airborne instruments measured aerosol size distributions, optical properties, and vertical distributions. Single scattering albedo varied from 0.73 to 0.93 (at 0.55 μm) in aerosol layers dominated by biomass burning. Much of this variation was attributed to the variable proportion of mineral dust and aerosols from biomass burning. They estimated the single scattering albedo of aged biomass burning aerosol to be around 0.81 with an instrumental uncertainty of ±0.05. External mixing and possibly internal mixing between the biomass burning aerosol and mineral dust present an additional source of uncertainty in this estimate. The size distributions of biomass burning aerosols were dominated by particles with radii smaller than 0.35 μm. A 20% increase of mean radius was observed when contrasting fresh and aged biomass burning aerosols, accompanied by changes in the shape of the size distribution. These changes suggested growth by coagulation and condensation. Extinction coefficients, asymmetry parameters, and Angstrom exponents were calculated from Mie theory, using the lognormal fits to the measured size distributions and assumed refractive indices.
7. Implication of Findings for Satellite Retrievals
7.1. Background on Satellite Issues
 For assessing the existence of large-scale global “dimming” or “brightening,” information on radiative fluxes at such scales is needed. Many attempts have been made to use satellite observations at regional or global scale to derive such information. Using the best available ground truth (such as the Baseline Surface Radiation Network (BSRN) [Ohmura et al., 1998] or the Surface Radiation Network (SURFRAD) [Augustine et al., 2005]), the satellite retrievals have been evaluated and limits on the errors of these estimates are now known. Since most of the ground observing stations are in areas of relatively average climatic conditions, the statistics presented on the accuracy of the satellite retrievals are not representative of unique environmental conditions. These include polar regions, high-elevation regions (Tibet Plateau), heavy dust locations, heavy biomass burning locations (Amazon), and dust and biomass burning mixtures (sub-Sahel).
 At present, lack of information on all the relevant optical properties of aerosols poses a challenge for estimating radiative fluxes and radiative forcing at large scales. Aerosol characterization in most satellite-based retrievals is rudimentary. The objective of this section is to illustrate how improved aerosols information impacts the estimation of surface radiative fluxes from satellite observations at several locations of interest. It will also demonstrate that when dust and biomass burning aerosols are known to be present, there is room for additional improvements. The tool used is the University of Maryland Shortwave Radiation Budget (UMD/SRB) scheme [Liu and Pinker, 2008] based on two different treatments of aerosols.
 In version 1 of the UMD/SRB inference scheme, four aerosol models are applied (land, oceans, deserts, and polar regions) with temporally and spatially invariant climatological optical depths. This rudimentary treatment has been updated in the version 2 of the model. Specifically, global gridded (2.5 × 2) monthly mean values of the extensive property (τ0.55 μm) and of the intensive parameters (single scattering albedo, asymmetry parameter, and normalized extinction coefficients averaged in five spectral intervals 0.2–0.4, 0.4–0.5, 0.5–0.6, 0.6–0.7, and 0.7–4.0 μm) have been derived by merging information from Moderate Resolution Imaging Spectroradiometer (MODIS) retrievals, Global Ozone Chemistry Aerosol Radiation and Transport (GOCART) model simulations [Chin et al., 2002], and Aerosol Robotic Network (AERONET) [Holben et al., 2001] measurements for the year of 2001 [Liu et al., 2005, 2008]. Estimates of aerosol optical depth at 0.55 μm is achieved by combining the accurate point measurements from AERONET and the spatial and temporal variation patterns from model and satellite data. Single scattering albedo is calculated by extrapolating the GOCART monochromatic value at 0.55 μm to the entire solar spectrum using the spectral variation deduced from AERONET almucantar retrievals. Global scale asymmetry parameter is estimated from MODIS Ångström exponent based on an empirical relationship derived from AERONET data. Normalized extinction coefficient is computed from MODIS Ångström exponent. It should be noted that further improvement of aerosol characterization would be feasible with progress in satellite retrievals, model simulations, and ground measurements.
7.2. Implementation of Satellite Inference Schemes
 Both aerosol treatments (versions 1 and 2) were implemented with the ISCCP D1 satellite data [Rossow and Schiffer, 1991]. Results for Ilorin and several additional BSRN sites, known to have high dust loading, are presented in Table 5. It is evident that both aerosol treatments produce fluxes that are in better agreement with ground observations during clear-sky-dominated cases (cloud fraction <20%) during year 2001. The other locations where high-quality BSRN observations are available (such as the Solar Village, Saudi Arabia, Alice Springs, Australia, Desert Rock, United States, Tamanrasset, Algeria) indicate that version 2 with improved aerosol treatment yields better agreement with ground observations than over Ilorin (Table 5) (except for Solar Village).
Table 5. Comparison of SRB Model Retrieval of Daily Surface Downward Shortwave Flux With Ground Measurements at Selected BSRN Stations During Clear-Sky Dominated Conditions (Cloud Fraction Less Than 20%)a
Daily Mean Surface Downward SW Flux (W m−2)
Error of Daily Surface Flux (W m−2) (Retrieval–BSRN observation)
Original model is referenced version 1; improved model is referenced version 2. Table 5 was computed for the days in year 2001 if daily cloud fraction is less than 20% (not for dry season only).
62.28 ± 28.74
58.24 ± 23.37
Alice Springs, Australia
−20.50 ± 19.40
4.81 ± 13.76
Solar Village, Saudi Arabia
−2.67 ± 20.86
14.41 ± 21.12
Desert Rock, USA
−38.14 ± 28.61
−4.21 ± 10.34
−19.56 ± 26.46
−3.71 ± 15.18
 Great improvement in the estimation of global irradiance for the relatively clear episodes at Alice Springs (Australia) were achieved by using the version 2 of the aerosol treatment; at this site, low aerosol loading (τ0.55 μm is about 0.1) leads to small diffuse fluxes (about 20 W m−2). At Solar Village, another clear-sky-dominated station, improvements were obvious in the first 3 months; however, overestimation of total fluxes is persistent between April and October. Generally, large discrepancies in the clear-sky cases can be found in the diffuse flux, which can be explained in a similar way as the Ilorin case, namely, either aerosols are not absorbing enough (assumed ω0 0.55μm is about 0.95) or asymmetry parameter is underestimated (assumed g0.55 μm is about 0.70).
 For Ilorin, scatterplots of surface irradiance against measurements are presented in Figure 12 with different colors associated with three time periods (January to February, March to October, and November to December). The positive bias for the global irradiance, which is above possible sampling and measurement uncertainties, can be attributed to the large overestimations during dry season. Lower direct and higher diffuse fluxes are evident here for the cloud-dominated cases (March to October). For the dry season, overestimation of direct irradiance is obvious with the most serious cases within the November to December period, which might be explained by the underestimation of monthly mean τ0.55 μm (0.17 for November and 0.47 for December compared with 0.80 for January and 1.04 for February). The overestimation of both direct and diffuse fluxes raises the suspicion that assumed aerosol type may not be sufficiently absorbing (estimated ω0 0.55μm is about 0.92) or/and less forward scattering (estimated g0.55μm is about 0.69). Another explanation could be cloud contamination, which results from mistakenly identifying enhanced aerosol episodes as cloudy scenarios. Such misclassification could also miss certain parts of the atmospheric absorption and generate overestimation of surface irradiance. The lesser improvement in version 2 at the Ilorin site as compared to the other sites leads us to believe that the problem might be related to mixture of aerosols from different sources (biomass and dust) as well as their vertical distribution (not the case at the other locations). For estimating the level of possible errors due to such factors, additional simulations have been performed using the unique information collected during the DABEX flights. Simulation of TOA SW upward and surface downward flux for the case of DABEX Flight B162 is presented in (Figure 13, red dot). Adopting the measured aerosol vertical profile (Figure 9) and total column aerosol optical depth (0.82) from collocated AERONET measurement at Ilorin, smoke and dust aerosol optical depth at 0.55 μm is estimated as 0.33 and 0.49, respectively. The solid line represents the change of flux at the surface and TOA with aerosol optical depth for dust aerosols only. For the case when the DABEX measurement is taken (the red dot in Figure 13) retrieval of the surface downward SW flux from the satellite observation under the assumption of dust only would result in smaller aerosol optical depth (∼0.4) and, subsequently, an overestimation of surface flux by ∼81 W m−2. Figure 13 shows that adding the smoke layer on top of dust would reduce the TOA SW upward flux by ∼5 W m−2. If only one layer of dust aerosol is assumed, surface flux would be underestimated by ∼81 W m−2 from the retrieval using TOA satellite measurement.
 It has been demonstrated that radiative fluxes (surface and TOA) can be derived from satellite observations at satisfactory accuracies for addressing many climate issues. The satellite-based fluxes have been evaluated against numerous ground observations of high quality [Pinker et al., 2003; Gupta et al., 2004]. The data have been found useful in climate research such as evaluation of numerical and weather prediction models [Rodriguez-Puebla et al., 2008], in Land Data Assimilation work [Mitchell et al., 2004], oceanic applications [Grodsky et al., 2009], hydrological modeling [Lohmann et al., 2004], and investigations of long-term trends in SW radiation [Pinker et al., 2005; Hinkelman et al., 2009]. Incorporation of realistic information on the effect of aerosols into the inference schemes is still a challenge. Recent satellite systems such as AVHRR [Geogdzhayev et al., 2002; Mishchenko et al., 2007], MODIS [Kaufman et al., 1997], and MISR [Kahn et al., 2005] as well as Chemical Transport Models such as GOCART [Ginoux et al., 2001; Chin et al., 2002] provide part of the needed information, yet the various aerosol products differ from each other. They account only for the total columnar aerosol loading, while no information is available on their vertical distribution and only limited data are available on aerosol optical properties (e.g., absorption). Another complication arises due to mixture of different aerosol types and their vertical distribution. This latter issue is addressed in this paper. As documented by Haywood et al.  from observations of vertical profiles of aerosol scattering and extinction from aircraft [Johnson et al. 2008b] and lidars [Heese and Wiegner, 2008; Derimian et al., 2008] and from the analysis of wind fields from the Met Office Mesoscale model, the mineral dust is transported into the Sahel from the Sahara desert by east to north easterly flow at low altitude. This cool stable air mixes with and undercuts the warmer smoke laden air to the south of the region, forcing the aerosols from the biomass to rise above the dusty air. The biomass burning laden air is advected to the north and is detectable several hundreds of kilometers north of the areas of active fires as well as to the south of the fires over the Gulf of Guinea. It is anticipated that the most recent satellite observing systems like CALIPSO and CALIOP [Anderson et al., 2005; Winker et al., 2007] and CloudSat [Stephens et al., 2002] will provide the necessary information to study this issue in more detail and determine its overall impact on estimates of global scale trends in surface radiative fluxes.
 The support of NASA under grants NNG05GB35G and NNG04GD65G to the University of Maryland is greatly appreciated. The observations at the Ilorin site were made under support from NASA EOS Validation grant NAG56464 to the University of Maryland. Thanks are due to the University of Ilorin Administration for support of the observational activity at Ilorin and the ground support team for overseeing the ground observations. This work would not have been possible without the vision of the late Tony Slingo, who made RADAGAST happen and extended the DABEX flights over the Ilorin site. We are indebted to him and the teams that participated in RADAGAST and DABEX and kindly made their data available to us.