A coupled machine-learning and sensitivity analysis framework to link dust activity in the Tigris-Euphrates basin to climatic and human-induced drivers.

This study analyzes the impact of climate-related stressors and water resources development in the Tigris-Euphrates basin on regional dust storm intensity and frequency. To this end, we first utilize remote sensing data on Aerosol Optical Depth (AOD), wind speed, surface temperature, soil moisture, precipitation intensity and inter-arrival time, and vegetation cover to explore long-term trends. To identify dust hotspots, we introduce a normalized intensity-frequency index, which combines the daily average values of AOD intensity and frequency exceeding a predefined threshold. Finally, we use a coupled machine learning and sensitivity analysis framework to quantify the relative contribution of anthropogenic and climate-driven stressors of dust activity. To improve the accuracy of the sensitivity analysis, we employ a Gaussian copula to generate correlated samples that capture the natural dependencies between input variables. The results indicate that the AOD intensity-frequency index in northern Iraq and eastern Syria (NI-ES) showed a surge in dust activity from 2007 to 2012, doubling compared to other periods. Additionally, over 50 % of western Iran (WI) recorded AOD greater than 0.2 from 2007 to 2012, which aligned with peak dust emissions in NI-ES during the same period. Meanwhile, internal dust hotspots in WI constitute less than 0.4 % of its total area, which have an insignificant impact on dust emissions compared to external sources. Finally, the sensitivity analysis results identify soil surface temperature, desertification in NI-ES, and water development projects in southeastern Turkey (SET) as the primary contributors to dust events in WI, with total effect of 53 %, 36 %, and 34 %, respectively. In contrast, precipitation in NI-ES has the lowest total effect on dust events, contributing only 9 %.