Diagnosing Anthropogenic and Climatic Drivers of Lake Urmia Desiccation Using a Probabilistic Neural ODE Digital Twin

The environmental collapse of terminal lake basins necessitates a rigorous distinction between climatic variability and anthropogenic pressures to inform restoration policies. This study aims to diagnose the primary drivers of desiccation in the Lake Urmia Basin through a continuous-time modeling approach. A Probabilistic Neural Ordinary Differential Equation (Neural ODE) framework was developed as a "Digital Twin" of the basin, calibrated on the natural period (۱۹۸۱–۱۹۹۶) using multi-source satellite imagery and ERA۵-Land reanalysis data. By projecting these learned natural dynamics through ۲۰۲۵, counterfactual simulations were employed to quantify the relative contributions of climate warming and water withdrawal. Results indicate that while climate warming (representing temperature and evaporation shifts) accounts for approximately ۳۷% of the surface area decline, direct anthropogenic intervention is responsible for ۶۳% (±۵%) of the total deficit. Furthermore, seasonal decomposition reveals a critical water deficit during winter months, consistent with the interception of runoff by upstream infrastructure. Identifying this seasonal timing of water scarcity is highly critical for drought risk management and restoration policy, as it distinguishes baseline evaporative losses from direct infrastructure-driven runoff interception, thereby guiding targeted reservoir release strategies. These findings demonstrate that the basin retains significant hydro-climatic potential for recovery. The study concludes that restoration efforts should prioritize the reform of winter reservoir release policies and the enforcement of agricultural consumption caps rather than focusing solely on climate adaptation.

کلیدواژه ها:

Drought ، Endorheic systems ، Hydrological modeling ، Water resource management ، Runoff interception