Development of optimized model based on evidential belief function for groundwater mapping

Groundwater is one of the most important natural resources in any nation serving as a major source of water to communities, industries and agricultural purposes. In recent years, groundwater resources in Malaysia due to high demands of local water system and especially prolonged drought period has become a serious issue in the Klang Valley, Malaysia. Spatial data integration and analysis for prediction of groundwater potential were conducted on the available datasets of Langat basin, Malaysia. In the search of groundwater potential areas, borehole data are essential as an indicator for directing exploration activities. Geographic information system (GIS) is a rapid, useful and low cost technique tool for implementing of groundwater mapping. The main objective of this study is to identify an optimized model for groundwater potential mapping. For that reason, some statistical methods including both bivariate and multivariate models,such as frequency ratio (FR), logistic regression (LR) and evidential belief function (EBF), were applied and tested. Evidential belief function model has not been applied in groundwater mapping. This contribution is novelty of this study. Then the developed model was compared and validated with well-known techniques such as FR and LR models. The processes of the method application include (i) identification of groundwater conditioning factors using data which obtained from available maps, remotely sensed imagery and related databases. The conditioning parameters are, elevation, slope,curvature, topographic wetness index, stream power index, river density, lineament density, lithology, land use, normalized difference vegetation index, soil and rainfall.(ii) The probabilistic of each conditioning factor was estimated using statistical weighting methods and a thematic map was produced for each conditioning factor. The optimized groundwater conditioning factors were then integrated to produce groundwater potential map. Then, the most indicative groundwater potential map was validated using the groundwater occurrence locations that were not used for generating the map. The resultant maps derived from integration of each method separately, were verified by the groundwater well locations for the study area. The AUC for the prediction curve of the groundwater potential map through three type of modelling was at 0.720, 0.720, and 0.779 of prediction accuracy for, FR, LR and EBF methods,respectively. The validation results demonstrate that integration of all evidential maps give satisfactory result for groundwater potential mapping. Both advantages and drawbacks of implementation for the proposed prescriptive approach are illustrated in the thesis. Recommendations for the study area are indicated within the perspective of the existing water supply systems. In summary, the results of this study suggests a comprehensive evaluation of groundwater exploration development and environmental management for future planning by related agencies in Malaysia which provided an effective method and reduce cost as well as less time consuming