Evaluation of Techniques for the Estimation of Pixel-Level AOD Uncertainties in Satellite Aerosol Remote Sensing

Getting the most use out of remotely-sensed data such as aerosol optical depth (AOD) requires a robust characterization and understanding of retrieval errors on a pixel-level scale. This is particularly important for applications like data assimilation. Various techniques are being used for pixel-level uncertainty estimation in current standard satellite AOD products. This study evaluates several of them, through comparisons of predicted uncertainties vs. actual retrieval errors at selected AERONET validation sites. The techniques evaluated are (1) global diagnostic envelopes (by MODIS Dark Target), (2) empirical prognostic expressions (by MODIS Deep Blue), (3) weighted dispersion of possible solutions (by MISR), and (4) Optimal Estimation (by AATSR ORAC, SEVIRI CISAR, and MODIS BAR). Understanding the strengths and limitations of these techniques is important to develop plans for algorithms and validation of data from forthcoming missions, such as NASA's Plankton, Aerosol, Cloud, ocean Ecosystem (PACE) mission (https://pace.gsfc.nasa.gov/).