Delineation of spring potential zones (SPZs) using multi-criteria decision making (MCDM), frequency ratio (FR), and weights of evidence (WofE) methods

Abstract

The apparent demand for phreatic water necessitates the construction of potential zones as a long-term strategy for groundwater identification, conservation, and management. An approach for delineating potential spring zones utilizing the Multi-Criteria Decision Making (MCDM), Frequency Ratio (FR), and Weights of Evidence (WofE) techniques are demonstrated for the Sikkim region. A spring potential map is created using the 17 influencing layers, namely Elevation, Normalized Difference Vegetation Index, Land-use/ Land Cover, Slope, Average Annual Rainfall, Aspect, Geomorphology, Drainage Density, Soil, Lithology, Stream Power Index, Curvature, Terrain Ruggedness Index, Distance to Road, Plan Curvature, Distance to River, and Profile Curvature. The overlay weighted sum approach merges all thematic feature maps to build a spring potential map. The map is categorized into three sub-divisions, namely (a) Poor, (b) Moderate, and (c) Good. A total of 202 springs are scattered over the study area, with 141 springs (70%) considered for training purposes and 63 springs (30%) considered to validate groundwater spring potential zones. The validation is done for < 5 lpm and > 5 lpm discharge using the FR and WofE methods. The SPI value ranges from 0.13 to 0.27 for MCDM, 3.06 to 49 for FR (< 5 lpm discharge), 24.78 to 96.16 for WofE (< 5 lpm discharge), 3.21 to 47.40 for FR (> 5 lpm discharge) and 15.35 to 105.20 for WofE (> 5 lpm discharge). The accomplishment of MCDM, FR, and WofE was appraised by dealing with the Receiver Operating Characteristic (ROC) curve and the Area Under the Curve (AUC). The MCDM method utilizing the training dataset had an AUC value of 0.776 and a validation dataset of 0.767. The AUC value of the FR (< 5 lpm) method is 0.939, WofE (< 5 lpm) is 0.968, FR (> 5 lpm) is 0.895, and WofE (> 5 lpm) is 0.879. Sensitivity analyses have been carried out for MCDM, FR, and WofE, which signifies the importance and contribution of each layer in spring potential maps. The adopted methods can be implemented in different locations with or without modification. This work also shows that the “Poor” zone is visible in the northern portion, while the “Good” zone surrounds the southern portion and Weights of evidence provides the best result among all methods used to identify SPZs.