Integrating Machine Learning and Ground-Based Data to Reconstruct ERA۵ Precipitation Maps in the Urmia Lake Basin, Iran

Accurate precipitation data is essential for hydrological modeling and water resources management, particularly in environmentally sensitive regions like Iran's Urmia Lake Basin. This study improves the accuracy of ERA۵ precipitation estimates by integrating ground-based station data with advanced machine learning (ML) techniques. Five ML models (XGBRegressor, RandomForestRegressor, Support VectorRegression, KNeighbors Regressor, and GradientBoostingRegressor) were assessed using daily precipitation data from ۲۴ stations (March ۲۰۱۹-March ۲۰۲۲) and corresponding ERA۵ data from Google Earth Engine. After preprocessing, RF and GBR outperformed others, reducing the Root Mean Square Error (RMSE) from ۲.۹ mm to ۲.۷ mm on average, with some stations achieving RMSEs as low as ۲ mm. These models also significantly improved the detection of heavy precipitation events, which is critical for managing extreme weather impacts. The enhanced precipitation accuracy can improve hydrological modeling and inform more effective water management strategies in the Urmia Lake Basin. This research highlights the benefits of combining satellite and ground-based observations with ML techniques. Future studies could further refine these models by incorporating additional spatial features, such as elevation and land cover data, to capture regional topography. Additionally, integrating temporal dynamics and multi-source satellite data may yield even more precise precipitation estimates.

کلیدواژه ها:

Precipitation ، ERA۵ ، Urmia Lake Basin ، Ground-Based Data ، Machine Learning