Aishwarya Raman, Avelino F. Arellano
Department of Hydrology and Atmoshperic Sciences, The University of Arizona, Tucson, Arizona
Particulate matter (PM) concentrations are one of the fundamental indicators of air quality. Earth-orbiting satellite platforms acquire column aerosol abundance that can in turn provide information about the PM concentrations. One of the serious limitations of column aerosol retrievals from low earth-orbiting satellites is that these algorithms are based on clear sky assumptions.
In this study, we demonstrate a method to fill in gaps in Moderate Imaging Resolution Spectroradiometer (MODIS) aerosol optical depth (AOD) retrievals based on ensembles generated using an analog-based kalman filter approach (KFAN). It provides a probabilistic distribution of AOD using historical records of model simulations of meteorological and chemical predictors such as AOD, relative humidity, mass concentrations of chemical species, and past observational records of MODIS AOD at a given target site. We use simulations from two models: 1) a coupled community regional weather forecasting model with chemistry (WRF-Chem) run at 36km, and 2) a global community atmosphere model with chemistry (CAM-Chem) run at a coarser resolution. Analogs selected from the model simulations and corresponding observations are used as a training dataset. Then, missing AOD retrievals in MODIS pixels in the last two weeks of the selected period are estimated. We use MODIS retrievals that were not used for optimization and an independent set of AOD retrievals from AERONET stations for evaluating analog estimates. KFAN is an efficient approach to generate an ensemble as it involves only one model run and provides an estimate of uncertainty that complies with the physical and chemical state of the atmosphere.