Chapter 7 - Spatial Prediction of Groundwater Level Using Models Based on Fuzzy Logic and Geostatistical Methods

Abstract

Spatial analysis of groundwater levels is presented using four techniques: two artificial intelligence methods—Sugeno fuzzy logic (SFL) and Mamdani fuzzy logic (MFL), and two geostatistical methods—ordinary kriging (OK) and co-kriging (CK). The performances of all four techniques were then compared. Both SFL and MFL are bottom-up and data-driven techniques, but OK and CK use techniques with restrictive assumptions. The data requirements for these techniques comprise coordinate specifications and ground elevations from the observation stations, and groundwater levels from 18 of the observation wells; their outputs comprises the groundwater levels for all of the observation wells. The performance metrics of the four models showed that SFL performs better than the three remaining models, although the performance metrics of MFL were close to those of SFL. Based on the overall modeling results, the performances were ranked as SFL, MFL, CK, and OK.