Validation and empirical correction of MODIS AOTand AE over ocean

Abstract. We present a validation study of Coll. 5 MODIS level 2 Aqua and Terra AOT and AE over ocean by comparison to coastal and island AERONET sites for the years 2003–2009. We show that MODIS AOT exhibits significant biases due to windspeed and cloudiness of the observed scene, while MODIS AE although overall unbiased, exhibits less spatial contrast on global scales than the AERONET observations. The same behaviour can be seen when MODIS AOT is compared against marine AERONET data, suggesting that the spatial coverage of our datasets does not preclude global conclusions. Thus, we develop empirical correction formulae for MODIS AOT and AE that signifcantly improve agreement of MODIS and AERONET observations. We show these correction formulae to be robust. Finally, we study random errors in the corrected MODIS AOT and AE and show that they mainly depend on AOT itself, although small contributions are present due to windspeed and cloud-fraction in AOT random errors and due to Ångström exponent and cloud fraction in AE random errors. Our analysis yields significantly higher random AOT errors than the official MODIS error estimate (0.03 + 0.05 τ), while random AE errors are smaller than might be expected. We interpret these findings in reference to the MODIS retrieval algorithm. This new dataset of bias-corrected MODIS AOT and AE over ocean is intended for aerosol model validation and assimilation studies, but also has consequences as a stand-alone observational product. For instance, the corrected dataset suggests that much less fine mode aerosol is transported across the Pacific and Atlantic oceans.