Evaluation of Uncertainty in Satellite Monthly Aerosol Products Associated with Spatial Sampling

Large amounts of data are frequently missing from global satellite Level 3 daily aerosol products, due to limitations of both the satellite sensor and data retrieval algorithms. Missing data may be caused by orbital gaps, or the impact of cloud, and snow/ice, etc. For example, the spatial completeness (i.e. valid data point ratio) of global 1ox1o resolution daily aerosol optical thickness (AOT) from MODIS (Dark Target algorithm) is about 25-40%, while SeaWiFS (Deep Blue algorithm) is typically 15-23%, and MISR is only ~ 5%. Spatial completeness of AOT varies geographically and seasonally. Wave-like artifacts were observed in MISR monthly products due to large data gaps associated with its narrow orbit. To evaluate uncertainty in AOT caused by spatial incompleteness, using data from NASA Goddard Chemistry Aerosol Radiation and Transport (GOCART) model, data gaps in satellite level 3 daily aerosol products from SeaWiFS (Deep Blue), MODIS (Dark Target), and MISR are simulated. Errors and uncertainty due to spatial sampling are analyzed by comparing the original and filtered model data. The uncertainty of global daily mean AOT has a negative relationship to the spatial completeness, i.e., uncertainty increases as the spatial completeness of the data set decreases. Significant negative bias is found in the global mean of filtered AOT from MODIS, in particular March-September, but not in SeaWiFS and MISR. This is largely due to the lack of coverage over deserts in the MODIS Dark Target data set. Monthly AOT are generally composed from daily data - we found that the differences between monthly data composed from original and filtered model daily data are significant in some regions. This presentation will show a statistical analysis of uncertainty in monthly AOT data due to spatial sampling.