Groundwater Modelling in Nashik District using Artificial Neural Network (ANN) and Support Vector Machine (SVM)
Loading...
Date
2021-09
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
G.B. Pant University of Agriculture and Technology, Pantnagar - 263145 (Uttarakhand)
Abstract
The present study was undertaken in Nashik District of Maharashtra to investigate the groundwater behavior and to assess the groundwater utilization development stage; and to develop the groundwater models using Artificial Neural Network (ANN) and Support Vector Machine (SVM) to predict the seasonal depth to water table. The input parameter used were net recharge, net discharge, recharge due to rainfall, recharge due to return flow of irrigation, recharge due to seepage from canals, discharge due to draft for irrigation use, industrial use, domestic use and livestock use and previous year water table depth for the period w.e.f. 1998 to 2018. Four groundwater models were developed in which model 1 and model 2 were developed using annual data whereas model 3 and model 4 were developed using seasonal data. MATLAB R2019a was used to develop the Artificial Neural Network (ANN) based models and e1071 package in R-4.0.1 and R Studio – 1.3.1093 was used to develop the Support Vector Machine (SVM) based models. During the study period of 21 years, out of 181 hydrograph stations, the water table trend at 56 hydrograph stations was found to be as rising, whereas neither rising nor falling water table trend was observed at 78 hydrograph stations and 47 hydrograph stations were found under falling water table trend during pre-monsoon season. In the course of post-monsoon season, 41 hydrograph station were found to be on rising water table trend, 54 hydrograph stations were on falling water table trend whereas about 84 hydrograph stations fell under neither rising nor falling water table trend. There was significant increase in the number of open wells, slight decrease in the number of pump set on bore wells. The area under maize crop, spices, fruit crops and vegetable crop was found to be continuously increasing during the study period. The groundwater balance studies indicated that during the study period, out of 15 talukas, one taluka namely Dindori transformed from lowest category (Safe) to higher category (Semi-critical category); one taluka namely Yeola transformed from higher category (Over-exploited) to lowest category (Semi-critical). However remaining thirteen talukas remained in the same category of groundwater utilization development stage. The value of performance indicators such as r, R2, NSE, MAE, RMSE, MAPE, RMSPE, RRSE and RAE were calculated to evaluate the performance of ANN and SVM based models. Based on the global raking obtained from the values of performance indicators, out of four ANN based models ANN- Model 1 and ANNModel 2 were selected for the prediction of the depth to water table during pre-monsoon and post-monsoon
seasons, respectively whereas out of four SVM based models, SVM-Model 1 and SVM-Model 2 were selected for the prediction of the depth to water table during pre-monsoon and post-monsoon seasons, respectively. On comparing ANN and SVM based models, it was found that SVM based model was better than ANN model to predict pre-monsoon depth to water table while ANN based model was found to better than SVM to predict post-monsoon depth to water table. It was concluded that, both ANN and SVM based models were not able to predict the depth to water table precisely however depth to water table predicted by ANN and SVM models followed the trend of observed water table precisely and accurately. Recharging structures such as percolation tanks, cement plugs/bunds, contour trenches, gabion structure, nala bunds, village ponds, underground bandharas or sub surface dykes and KT Weirs were also suggested in problematic regions of the study area.