SPATIO-TEMPORAL GROUNDWATER DROUGHT ASSESSMENT BASED ON ANN MODEL AND GIS FOR A SUB-BASIN OF BHARATHAPUZHA

Abstract

Groundwater is one of the most precious and significant sources of water in the world. Understanding the effects of both natural and human-made factors on groundwater reserves and exploitation is crucial for developing appropriate management strategies to deal with unsustainable use. The understanding of groundwater level variability and trend is crucial for water resource planning in a region. The groundwater level fluctuation is a non-linear phenomenon. Hence, Artificial Neural Networks (ANN) proves to be one of the best tools for modelling non-linear relationship between input and output datasets. Once the groundwater level is modelled, it is easy to assess the groundwater drought conditions of a region using Standardized Groundwater level Index (SGI). Therefore, systematic information about likely occurrence and distribution of drought may assist in preparedness and mitigation of drought disasters. The present study was conducted in Kalpathypuzha sub-basin of Bharathapuzha to analyze the variability and trend of groundwater level, to develop ANN model for groundwater level prediction and to assess the groundwater drought using Standardised Groundwater level Index (SGI). Twelve observation wells evenly distributed in the blocks of Kuzhalmannam, Palakkad, Malampuzha and Chittur were selected. The groundwater level variability was analyzed by various descriptive statistics such as mean, standard deviation, coefficient of variation, skewness and kurtosis. The groundwater level trend was estimated using Mann- Kendall test and Sens slope estimator. ANN models were developed separately for each well to predict the groundwater level using MATLAB R2016a software. The input parameters used were precipitation, maximum and minimum temperature and output data used was groundwater level collected for a period of 15 years from 2007 to 2021. SGI values were estimated for both observed and predicted groundwater level data to assess the groundwater drought scenario of the study area and to develop spatio-temporal groundwater drought map. Monthly groundwater level and drought conditions were predicted for the year 2023 using the developed ANN model.

Results of trend analysis showed a decreasing pre-monsoon groundwater level trend in three wells, well 129 of Palakkad block and wells 133 and 142 of Malampuzha block while decreasing post-monsoon groundwater level trend in well 139 of Chittur block. But there was no trend in all other wells for both pre-monsoon and post-monsoon. Feed forward ANN models were developed for all the twelve wells in the study area and the performance indicators correlation coefficient r (0.93 to 0.74), Root Mean Square Error RMSE (0.11 to 0.45 m), and coefficient of determination R2 (0.87 to 0.69) were found in the acceptable range. The best model performance for training was for the well PKD S-4 with model configuration 3-10- 1 and r = 0.92 whereas, during testing it was found for the well 129 with model configuration 3-14-1 and r = 0.93. ANN predicted groundwater level was found in close agreement with that of the observed groundwater level in this study. Hence the model developed could be safely and effectively applied in the study area. The SGI was estimated for pre-monsoon months Jan, Feb, Mar, Apr and May of the study period from 2007 to 2021 for all the twelve wells as drought was more severe during these months. SGI values ranged from -3.7 to 1.1 indicated exceptional to no drought condition in the study area. The computed SGI values indicated that the years 2013, 2016, 2017 were the severe drought years of the study area. According to Spatial distribution of SGI values for the years 2013, 2016 and 2017 Chittur and Malampuzha block were the most drought affected areas followed by Kuzhalmannam and Palakkad block. Hence the study revealed that the majority of Kapathypuzha sub-basin is drought prone and immediate measures are to be adopted to prevent the extend of severity in the area.