Abstract
Once abundant and readily available, groundwater (GW) is now dwindling at an alarming rate. This vital resource is under growing pressure from both natural and human-induced factors. Groundwater Level (GWL) is closely related to Groundwater Storage (GWS) thus the decline in GWL creates a shortage in GWS. This research developed a robust predictive model for GWS in Rajshahi district, Bangladesh, for the period 2001–2022 using six climatic variables, namely, Mean Temperature, Cloud Coverage, Humidity (percent), Solar Radiation, Sunshine, and Wind Speed. Three Machine Learning (ML)-based regression models- Random Forest (RF), Support Vector Machine (SVM), and Gradient Boosting Machine (GBM) were applied for this purpose. Results showed that the accuracy level was quite high while RF regression was plugged into the observed dataset (R2 = 0.80). Moreover, among the six climatic variables, cloud coverage, humidity, and wind speed contributed 87.4% altogether to predict the GWS. These findings offer valuable insights not only for understanding the GWS dynamics in Rajshahi district but also for informing sustainable management strategies. By providing decision-makers with a clear understanding of the key climatic drivers and their impact, this research empowers them to implement effective interventions and conservation measures to ensure the long-term availability of this critical resource.