Assessing Machine Learning Models for Precipitation Prediction in the Upper Indus Basin: A Comparative Analysis

Precipitation plays a critical role in the effective management of water resources and the maintenance of reservoir water levels. However, climate change has significantly altered precipitation patterns, leading to extreme hydrological events such as droughts and floods, which have profound socioeconomic and environmental impacts. This study focuses on predicting precipitation events in the Upper Indus Basin (UIB) using machine learning models. In this study three widely used machine learning algorithms Support Vector Machine (SVM), K-Nearest Neighbor (KNN), and Random Forest (RF) were employed to forecast precipitation events in the UIB. The dataset was divided into training (80%) and testing (20%) subsets for model evaluation. Among the algorithms tested, KNN demonstrated the best predictive performance, yielding a mean absolute error (MAE) of 2.662, a root means square error (RMSE) of 16.3, and an R² score of 0.879, with an overall accuracy of 83.16%. The results indicate that the KNN algorithm is the most effective machine-learning model for precipitation prediction in the UIB. The findings of this study contribute to improving early warning systems and facilitating efficient water resource management in the face of climate variability and extreme weather events.