The surface reflectance ratio between the visible (VIS) and shortwave infrared (SWIR) radiation is an important quantity for the retrieval of the aerosol optical depth (τ_a) from the MODIS sensor data. Based on empirically determined VIS/SWIR ratios, MODIS τ_a retrieval uses the surface reflectance in the SWIR band (2.1 µm), where the interaction between solar radiation and the aerosol layer is small, to predict the visible reflectances in the blue (0.47 µm) and red (0.66 µm) bands. Therefore, accurate knowledge of the VIS/SWIR ratio is essential for achieving accurate retrieval of aerosol optical depth from MODIS. We analyzed the surface reflectance over some distinct surface covers in and around the Mexico City metropolitan area (MCMA) using MODIS radiances at 0.66 µm and 2.1 µm. The analysis was performed at 1.5 km×1.5 km spatial resolution. Also, ground-based AERONET sun-photometer data acquired in Mexico City from 2002 to 2005 were analyzed for aerosol depth and other aerosol optical properties. In addition, a network of hand-held sun-photometers deployed in Mexico City, as part of the MCMA-2006 Study during the MILAGRO Campaign, provided an unprecedented measurement of τ_a in 5 different sites well distributed in the city. We found that the average RED/SWIR ratio representative of the urbanized sites analyzed is 0.73±0.06 for scattering angles <140° and goes up to 0.77±0.06 for higher ones. The average ratio for non-urban sites was significantly lower (approximately 0.55). In fact, this ratio strongly depends on differences in urbanization levels (i.e. relative urban to vegetation proportions and types of surface materials). The aerosol optical depth retrieved from MODIS radiances at a spatial resolution of 1.5 km×1.5 km and averaged within 10×10 km boxes were compared with collocated 1-h τ_a averaged from sun-photometer measurements. The use of the new RED/SWIR ratio of 0.73 in the MODIS retrieval over Mexico City led to a significant improvement in the agreement between the MODIS and sun-photometer AOD results; with the slope, offset, and the correlation coefficient of the linear regression changing from (τ_aMODIS=0.91τ_{a sun-photometer}+0.33, R²=0.66) to (τ_aMODIS=0.96 τ_{a sun-photometer}−0.006, R²=0.87). Indeed, an underestimation of this ratio in urban areas lead to a significant overestimation of the AOD retrieved from satellite. Therefore, we strongly encourage similar analyses in other urban areas to enhance the development of a parameterization of the surface ratios accounting for urban heterogeneities.