Groundwater-level forecasting is an important component of water-resource monitoring because groundwater conditions vary with environmental drivers and local hydrogeological characteristics. This paper presents a groundwater prediction and visualization system that combines historical groundwater observations from the Central Ground Water Board (CGWB), India, with meteorological variables retrieved from NASA’s Prediction of Worldwide Energy Resources (POWER) service. Five consecutive annual observations of groundwater level, rainfall, temperature, and humidity-related data are used to predict the groundwater level for the following year. The system compares Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Hybrid Spatio-Temporal GRU models. In the reported model comparison, GRU obtains an RMSE of 2.1616 m below ground level (mbgl), an MAE of 1.2821 mbgl, and an R² of 0.9242. The hybrid model, which combines a temporal branch with station latitude and longitude, obtains an RMSE of 2.1885 mbgl, an MAE of 1.3178 mbgl, and an R² of 0.9223. An interactive Leaflet-based map links predictions with station metadata. The results indicate that the GRU model provides the strongest aggregate performance among the evaluated deep-learning models under the reported test protocol.

Jahanara Shaik, Srujan Amaragatti, Indrajith A S, Tejjas A R, Sanketh Kulkarni, "Deep Learning-Based Groundwater Level Prediction Using Multivariate Time-Series Analysis and Spatial Visualization" International Journal of Scientific Research and Engineering Development, V9(6): Page(82-89) November - December 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.