HOLISTIC ASSESSMENT OF TEMPERATURE IMPACTS ON WATER AVAILABILITY: INTEGRATING SENTINEL-2 IMAGERY, GROUND SENSORS, AND ADVANCED PREDICTIVE MODELING

Abstract

Water level forecasting is important for water resource management, flood control, and the sustainability of aquatic systems. This study aims to provide a comprehensive assessment of the impact of temperature changes on water availability using Sentinel-2 imagery and ground sensors. The technique applied in the study involves a multistage process of data collection on temperature, rainfall, forest cover, and water resources. NDVI and LCM are applied to identify the changes in the forest cover. In this study, regression analysis is applied to establish the correlation between forest cover and water deficit as depicted by the trend analysis using GIS tools and time series analysis. Exploratory Factor Analysis (EFA) is applied for variable reduction and determining the number of factors. Random Forest (RF) and Gradient Boosting Regression (GBR) are applied for predictive analytics. Parrot Optimization (PO) is applied for tuning the hyperparameter of the prediction model to improve the water level prediction system. The rainfall prediction is performed using Long Short-Term Memory (LSTM) networks. Temperature scenarios are ranked using TOPSIS, Compound Factor (CF), and VIKOR methods to identify the impacts on water resources and decision-making. The optimization techniques enhance the quality of the model and find the solution for water resource management. The performances of the developed model are evaluated with the existing techniques in the context of MSE, MAE, and RMSE. The proposed model achieves the lowest MAE of 0.26, MSE of 0.27, and RMSE of 0.23 in the 80/20 split.