Atmospheric Aerosol Prediction over Egypt with LSTM-RNN using NASA’s MERRA-2

Abstract

Aerosol optical depth (AOD) is one of the most critical indicators for air quality. Estimation of accurate AOD needs to include both dust and chemical reactions in the calculations which are expensive from a computational point of view. In this work, we present a novel and simple model to estimate and predict the temporal trend of AOD based on the well-known algorithm long-short term memory (LSTM). Five domains are the core of this study, Four popular cities Cairo, Alexandria, Aswan, and Hurghada are selected. In addition to one sub-domain which includes one of the most important and internal dust sources for Egypt, Qattara depression. We applied the LSTM algorithm to NASA’s MERRA-2 monthly AOD datasets as training and validation data-set. The algorithms showed a lower root mean square error. The trained models after validation are used to predict the temporal trend of AOD for the period 2020-2022 over the five selected domains.