Assessment of severe aerosol events from NASA MODIS and VIIRS aerosol products for data assimilation and climate continuity

Abstract

While the use and data assimilation (DA) of operational Moderate Resolution Imaging Spectroradiometer (MODIS) aerosol data is commonplace, MODIS is scheduled to sunset in the next year. For data continuity, focus has turned to the development of next-generation aerosol products and sensors such as those associated with the Visible Infrared Imaging Radiometer Suite (VIIRS) on Suomi NPOESS Preparation Project (S-NPP) and NOAA-20. Like MODIS algorithms, products from these sensors require their own set of extensive error characterization and correction exercises. This is particularly true in the context of monitoring significant aerosol events that tax an algorithm's ability to separate cloud from aerosol and account for multiple scattering related errors exacerbated by uncertainties in aerosol optical properties. To investigate the performance of polar-orbiting satellite algorithms to monitor and characterize significant events, a level 3 (L3) product has been developed using a consistent aggregation methodology for 4 years of observations (2016–2019) that is referred to as the SSEC/NRL L3 product. Included in this product are the AErosol RObotic NETwork (AERONET), MODIS Dark Target, Deep Blue, and Multi-Angle Implementation of Atmospheric Correction (MAIAC) algorithms. These MODIS “baseline algorithms” are compared to NASA's recently released NASA Deep Blue algorithm for use with VIIRS. Using this new dataset, the relative performance of the algorithms for both land and ocean were investigated with a focus on the relative skill of detecting severe events and accuracy of the retrievals using AERONET. Maps of higher-percentile aerosol optical depth (AOD) regions of the world by product identified those with the highest measured AODs and determined what is high by local standards. While patterns in AOD match across products and median to moderate AOD values match well, there are regionally correlated biases between products based on sampling, algorithm differences, and AOD range – in particular for higher AOD events. Most notable are differences in boreal biomass burning and Saharan dust. Significant percentile biases must be accounted for when data are used in trend studies, data assimilation, or inverse modeling. These biases vary by aerosol regime and are likely due to retrieval assumptions in lower boundary conditions and aerosol optical models.