Surface Radiative Fluxes from GOES-E over the Amazon Basin: Model Comparison

In this study reported are results from an algorithm intercomparison initiative aimed at the development of improved estimates of surface radiative fluxes from satellite observations over the Amazon Basin. Three algorithms are used: (UMD-SRB, University of Maryland; GL1.2, INPE, Brazil; and Brasil-SR, INPE and University of Santa Catarina, Brazil). The algorithms are physically based, yet differ in their implementation and the way they address issues specific to this region, such as aerosols from biomass burning. Two fifteen day periods in 2005 were selected representing the rainy and dry seasons. The same satellite observations from GOES E were used by all the models. Ground truth from existing stations in the Amazon as well as from a new solar monitoring network of high quality have been used in evaluation. Using daily mean values for the March rainy season, it was found that: 1) the Brasil-SR and UMD-SRB estimates bear a close resemblance; 2) higher irradiances for Petrolina (semi-arid region in Northeast Brazil) are best described by the UMD-SRB and Brasil-SR, probably due to better assessment of water vapor column and absorption parameterization; 3) the GL1.2 results shows a systematic deviation, underestimating daily mean by about 20 Wm-2, but have lower dispersion than UMD-SRB or Brasil-SR; 4) irradiance interval 180 < E < 250 Wm-2 seems better described by GL1.2. This last behavior may be related to better assessment of cloudiness under partial coverage situations. September is characterized by intensive biomass burning in several Brazilian regions, particularly in the Amazon. The Northeast region is not affected by aerosols and estimates from all three models are in close agreement and have similar characteristics to those of March. For the Amazon sites: 1) lower irradiances (for overcast days) are correctly assessed; 2) UMD-SRB and Brasil-SR overestimate solar radiation, especially for higher irradiances (lower cloudiness); 3) GL1.2 model does not include aerosol effects and also overestimates the surface fluxes, but has better agreement with observations for lower irradiances. These preliminary results on surface radiative fluxes from all three models show similar accuracies when evaluated against ground observations. The advantages of each model will be synthesized to produce a version that produces best results in the Amazon Basin.