Assessment and Prediction of Groundwater using Geospatial and ANN modeling

Abstract

Abstract In semi-arid regions the deterioration in groundwater quality and drop in water level upshots the importance of spatio-temporal mapping with geospatial and advanced modeling techniques. In present study, changes in water level, water quality trend patterns and future scenarios of groundwater in 171 villages of Phagi tehsil, Jaipur district was assessed using eight years (2012-2019) groundwater data. Spatial interpolation maps were drawn using kriging method for pre-monsoon season and integrated with three different time series forecasting models (Simple Exponential Smoothing, Holt's Trend Method, ARIMA) and Artificial Neural Network models to ascertain the optimal prediction for groundwater level and quality parameters. Results reveal that the use of ANN model can describe the behavior of groundwater level and quality parameters more accurately than time series forecasting models. In addition, different ANN algorithms were tested to select the best-performing algorithm and ANN15 is found the most accurate one in simulating the magnitude and patterns of pre-monsoon water level data for year 2019 with R2 = 0.98, and NSE = 0.81. The change in groundwater table was observed with more than 4.0m rise in 81 villages during 2012-2013 whereas ANNpredicted results of 2023-2024 infer no rise in water table (>4.0m). Water level drop of more than 6.0m was observed in 16 villages of Phagi tehsil based on predicted results of 2024. Assessment of groundwater quality parameters like Total dissolved solids, chloride, fluoride and nitrate indicate chemically unsuitable groundwater for drinking purpose in most part of the Phagi. ANNpredictions point out excess nitrate content in 58% villages however, Water quality Index reveals unfit groundwater in 74% villages for human consumption in 2024. This time series and projected outcome of groundwater at village level can assist the planners and decision-makers for proper management of groundwater risk areas.