A novel approach for predicting spring locations using machine learning algorithms in Indian Himalayan Region

Due to the concerning effects of climate change, groundwater will be one of the significant sources of water for both primary and secondary use in the future. Therefore, identifying the spatial patterns of groundwater distribution might help implement practical water resources management projects. Springs are a potential source of groundwater in the Indian Himalayan Region. The main objective of the current study is to explore a novel methodological approach that utilizes the Variance Inflation factor (VIF) to perform a feature selection procedure and most used machine learning (ML) algorithms, including Random Forest (RF), Gradient Boosting Machine (GBM), and Neural Network (NN) for generating a groundwater spring potential map of the Ravi Basin in Himachal Pradesh, India. Used, 1834 spring and non-spring locations were selected from the field and split into two groups. Of 1834 samples, 70% (1283) were used for model training, and 30% (551) were used for model validation. The model’s overall accuracy of 0.89, 0.87, and 0.88 for RF, GBM, and NN, respectively, around 10% area, has a very high potential for spring occurrence. The novel methodology can be employed to find the initial information for GW exploitation for inaccessible areas and the lack of data sources in this area.