Evaluation of Dynamic Aerosol Model Selection over Land in the VIIRS AOD Algorithm

The VIIRS Aerosol Optical Depth (AOD) algorithm performs a simultaneous retrieval of surface reflectance and AOD using pre-computed Look-Up Tables (LUTs) for five different aerosol models with known optical properties and empirically derived spectral surface reflectance ratios. The algorithm computes AOD using all five LUTs dynamically but reports the AOD for a model that provides minimum residual between computed and observed reflectance. The models include: Oceanic, Dust, Generic, Urban, and smoke. In order to correctly measure optical depth, it could be argued that selecting the model that most closely represents reality is necessary for accurate results. To test this, the algorithm's outputs are analyzed and compared with other measurements in a variety of cases throughout the globe. In particular, dust and smoke events are compared with other ground and satellite measurements. Additionally, different types of surfaces and viewing geometries are used to test the algorithm's choice of aerosol model. Conclusions from this study can help identify areas for improvement for the algorithm such as forcing aerosol model selection in areas of known aerosol types. Results show that it is not necessary for the aerosol model to match the expected aerosol type to retrieve optical depth that is consistent with other observations.