A combination of MODIS measurements that are atmospherically corrected with simultaneous AERONET sky radiometer estimates of aerosols from AERONET-based Surface Reflectance Validation Network (ASRVN) shows that the current empirical relations between the VIS/SWIR ratios and the Modified Vegetation Index (MVI) being used in MODIS C005 for different urban/suburban areas are qualitatively and quantitatively in error. In addition, we explore the relationships for multiple urban sites and illustrate a general correspondence between the surface reflection ratios and biases in AOD retrieval. To explain this qualitative behavior, we use simple mixing models suitable for urban areas and their nearby surroundings. These mixing models are based on direct measurements from the ASTER spectral library which has a large database of measured spectra covering from 400nm to 2500nm. Finally, we show that taking the modified surface albedo into account results in more realistic and spatial distributions of the pollutant field in comparison to ground based optical sensors as well as TEOM PM2.5 monitors.