Predicting of dust storm source by combining remote sensing, statistic-based predictive models and game theory in the Sistan watershed, southwestern Asia

Dust storms in arid and desert areas affect radiation budget, air quality, visibility, enzymatic activities, agricultural products human health. Due to increased drought land use changes recent years, the frequency of dust occurrence Iran has been increased. This study aims identify source Sistan watershed (Iran-Afghanistan borders)—an important regional for southwestern Asia, using remote sensing (RS) bivariate statistical models. Furthermore, this determines relative importance factors controlling emissions ratio (FR) weights evidence (WOE) models interpretability predictive game theory. For purpose, we identified 211 sources area generated a distribution map—inventory map—by potential index based on RS data. In addition, spatial maps topographic affecting including soil, lithology, slope, Normalized difference vegetation (NDVI), geomorphology were prepared. The performance two (WOE FR) was evaluated under curve (AUC) receiver operating characteristic curve. results showed that slope exhibited greatest influence areas. 55.3% (according 62.6% WOE) total classified as high very sources, while both displayed acceptable accuracy with subsurface levels 0.704 FR 0.751 WOE, although they predict different fractions classes. Based Shapley additive explanations (SHAP), three factors, i.e., NDVI have highest impact model’s output. Overall, combination statistic-based (or data mining models), theory techniques can provide accurate semi-arid regions, which be helpful mitigation negative effects storms.