Simulation of Water Table Depth Using Hybrid CANFIS Model: A Case Study

Abstract

Long-term water table depth (WTD) prediction in agricultural areas proves to be a challenging task. These regions have heterogeneous and complex hydrogeological characteristics, human activities, and boundary conditions; also, there are nonlinear interactions between these aspects. Machine learning (ML) approaches have been broadly implemented for WTD forecasting because of their capability of modelling nonlinearities amongst GWL and its conditional factors. A new ML model was developed known as the co-active neuro-fuzzy inference system combined with the firefly algorithm (CANFIS-FA) for estimating monthly WTD of Nuapada watershed located in Odisha state, India. Prediction results of CANFIS-FA model presented good performance with mean squared error of 1.084–3.709; the correlation coefficient is >0.98, demonstrating that the hybrid model is appropriate to assess multifaceted groundwater systems. Therefore, it is evident that proposed model can assist as an alternate method in WTD prediction, particularly in regions where hydrogeological data are challenging to acquire.