Prediction of Ground Water Level using SVM-WOA Approach: A Case Study

Abstract

Reliable and accurate estimation of Groundwater Level (GWL) fluctuations is essential and vital for sustainable water resources management. Due to uncertainties and interdependencies in hydro-geological processes, GWL prediction is complex by the fact that fluctuation of GWL is extremely nonlinear and non-stationary. Utilising novel methods for accurately predicting GWL is of vital significance in arid regions. In present work, Support Vector Machine (SVM), in combination with Whale Optimisation Algorithm (SVM-WOA), is applied to forecast GWL in Bhubaneswar region (Odisha University of Agricultural Technology). Three quantitative statistical performance assessment indices, coefficient of determination (R2), Mean Squared Error (MSE), and Wilmott Index (WI), is used to assess model performances. Based on the assessment with conventional SVM and RBFN models, the performance of hybrid SVM-WOA model is preeminent. SVM-WOA is capable of predicting nonlinear behavior of GWLs. Proposed modelling technique can be applied in different regions for proper management of groundwater resources and provides significant information, at a short time scale, to estimate variability in groundwater at local level.