The long awaited development of the ABI sensor on the GOES-R satellite system encompasses the benefits of high temporal resolution (~ 5 minutes) together with the multispectral inversion retrieval capacity of present day MODIS. Current Algorithms being developed for GOES-R ABI for AOD retrieval over land surfaces rely heavily on the MODIS heritage which makes heavy use of the 2.1 micron channel together with quasi-empirical models for the surface reflection for the Blue / Red channels to isolate the atmosphere signal from the ground contamination. One issue with these algorithms is the strong over bias often seen in the MODIS AOD retrievals due to an underestimation of the surface reflection. These overbiases as well as other systemic biases can clearly affect the use of AOD measurements for surface PM2.5 estimates. On the other hand, the current GOES Aerosol and Smoke Product (GASP) attempts to retrieve the aerosols using assumptions for both the aerosol type (i.e phase function) as well as the assumed aerosol loading that represents the “clearest” sky pixel. In this talk we first make detailed correlative studies between the PM2.5 loading for the New York Region and the AOD measurements from MODIS C005 , GASP and AERONET. Most important is the seasonal performance of the correlations which significantly improve for summer months achieving values ~ 0.8 during the summer. To understand the reason for this improvement, correlative studies between the AOD and AERONET instruments shows similar seasonal improvements although the AERONET AOD / PM2.5 correlations are significantly higher than the satellite AOD / PM2.5 cases. These observations illustrate first that in general, summer based observations will always be more accurate due to the more homogeneous and larger mixtures of aerosols within the PBL. However, it is also clear that improving the accuracy of current AOD retrievals by removing some of the inherent biases from satellites should further improve PM2.5 observations. To illustrate these ideas, we first explore the biases of MODIS surface reflection on PM2.5 retrieval by reprocessing the AOD retrievals for summer using a more realistic surface model based on local tuning of the surface. In particular, the use of this regional refinement of the surface removes the AOD retrieval biases and improves the PM2.5 retrievals. On the other hand, we also explore possible improvements in the GASP product. This includes a seasonal correction for the minimum AOD ascribed to the clear sky as well as regionally defined aerosol models. Finally, we explore the potential of using MODIS derived aerosol models directly into the GASP data's processing stream. This approach can be thought of as a first approximation to the ultimate GOES-R retrieval which extract the underlying aerosol mixture during the retrieval process