Pixel Level Aerosol Optical Depth Bias Estimates for Suomi NPP and NOAA-20 Visible Infrared Imaging Radiometer Suite

Abstract:

Aerosol optical depth (AOD) is an important atmospheric parameter that is used for monitoring environmental conditions like smoke, dust storms, and urban/industrial pollution. Numerical models that predict weather and air quality assimilate satellite observations of AODs to initialize the models. The accuracy of satellite retrieved AODs is very critical so model performance is not degraded by data artifacts. In order to help the Numerical Weather Prediction (NWP) centers with their aerosol data assimilation, we explored a way to provide observational errors at every pixel where there is an AOD retrieval. The first phase of the work involved analyzing the data for summer 2019 over the continental United States (CONUS). The SNPP and NOAA-20 VIIRS AOD matches with Aerosol Robotic Network (AERONET) were found for 76 stations and an exponential fit was applied to the binned biases between AERONET and VIIRS AOD data; the reference AERONET data was taken to be ground-truth. To evaluate the efficacy for the estimate of the bias, a fit for the prediction interval (PI) was found and applied to the fit of the mean bias. The mean of future binned AOD retrievals is expected to fall, with a given probability, within the range defined by the PI centered on the fit line of bias. The fits for bias and PI were then applied to test days when weather was clear, cloudy, and contained dust or smoke. Fits in the bias for NOAA-20 and S-NPP were found to have root mean squared errors of 0.013, and 0.013, respectively. The mean bias with the 95% PI contained all the binned AOD points and their standard deviations except the last bin for both NOAA-20 and S-NPP. The pixel-level bias estimates for summer 2019 AOD data will be provided to users to test their utility in their assimilation studies. Future work includes extending this work by using a global data-set.