Using artifical intelligence methods to determinewater budget compenentes

Knowing the quantity and changes in water resources in the context of sustainable integrated watershed management, is fundamental. When creating a water budget by evaluating surface and groundwater together, there are measurement or calculation-based uncertainties in determining budget elements (Pecipitation, Evaporation, Flow, Infiltration). In recent years, artificial intelligence techniques have been effectively and successfully applied to model daily and monthly evaporation rates. Due to their ease of application, simple architecture, and surprisingly positive results, artificial intelligence techniques have become a promising research method in hydrology and water resources engineering. Currently, pre-coded and ready-to-use Artificial Intelligence models such as Linear Regression, Extremely Randomized Trees, Random Forest, Extreme Gradient Boosting, HGBoost, Decision Tree Regression and Deep Learning are used to calculate daily pan evaporations. Data was sourced from institutions like the State Hydraulic Works (DSİ) and the General Directorate of Meteorology, selecting data from the period 1990-2010. Some of the data was used for training, another for validation, and a part for prediction. By using climatic data such as monthly average temperature (°C), humidity (%), wind speed (m/s), pressure (hPa), solar radiation (cal/cm²), and sunshine duration (hours), the obtained results were compared. The results indicated that all the methods used were acceptably successful in predicting evaporation, but the Extremely Randomized Trees method yielded the best results with the available data (R² = 0.76). Optimization did not increase the prediction accuracy. The Decision Tree Regressor method, which was tried later, gave better results (R2= 0.86). Key Words: Water Budget, Evaporation, Artificial Intelligence Methods, Decision Tree Regression