Improvement of MODIS retrieval of aerosols over urban areas using a regionally tuned ground albedo model

The collection 5 MODIS aerosol algorithm was designed to be more robust in handling surface reflection by differentiating different surfaces. When compared to the old collection 4 algorithm, we found that although there was some improvement, the MODIS retrievals were still biased high. This result was traced to the fact that the correlation coefficient in collect 5 was still too low compared to results from Hyperion. To address this issue, a regional surface albedo model is obtained based on matchups of CIMEL sunphotometer and MODIS retrievals. Given the optical depth measurments and utilizing the urban aerosol phase function (tuned by AOD) , we obtained a set of reflectance measurements which were than regressed against the scattering angle. Using this surface model, we recalculated the AOD using MODIS over a different set of measurments and saw a significant reduction in bias.