Climatic Factors in Forecasting Water Level in Angat Reservoir through Artificial Neural Network

This study explores the relationship between climatic variables and water level fluctuations in the Angat Water Reservoir, a key source of potable water, irrigation, and hydroelectric power in the Philippines.Using historical data from 2012 to 2023, the research investigates how average temperature, rainfall, and relative humidity influence reservoir levels and evaluates the predictive capacity of an Artificial Neural Network (ANN) model.Climatic data were gathered from PAGASA synoptic stations in Clark, CLSU, and Science Garden, while water level records were obtained from the National Water Resources Board (NWRB).The ANN model, developed with three hidden layers and optimized through grid search and 5-fold cross-validation, demonstrated strong forecasting performance with an average RMSE of approximately 55 meters.Rainfall emerged as the most significant climatic factor, while temperature and relative humidity showed lesser or even negative contributions to prediction accuracy.Historical analysis also revealed a warming trend and high variability in rainfall patterns, further reinforcing the importance of robust forecasting tools.The model's predictions closely mirrored actual water levels, validating ANN's utility in hydrological forecasting.The results offer actionable insights for water resource management, enabling better planning for water allocation, flood control, and drought preparedness.Integration of ANN models into early warning systems and operational planning by agencies like NWRB and MWSS is recommended.Overall, this study supports the growing role of machine learning in sustainable and climate-resilient water resource management.