Predictive groundwater levels modelling by Inclusive Multiple Modelling (IMM) at multiple levels

Abstract

An investigation is presented in this paper to predict Groundwater Level (GWL) by Inclusive Multiple Modelling (IMM) practices, introduced recently by the authors, as a strategy by organising multiple models at three hierarchical levels: Artificial Neural Networks (ANNs), Sugeno Fuzzy Logic (SFL), Neuro-Fuzzy (NF), Support Vector Machine (SVM) and Gene Expression Programming (GEP). The IMM novelty of the study is to investigate a modelling strategy at three hierarchical levels, such that any base models at Level 1 are not reused as the combiner model at Levels 2 and Level 3 and this leads to a number of strategies. The results provide some evidence that (i) combining base models at Levels 2 and 3 enhance the performances compared with those of individual base models at Level 1; and (ii) the results at Level 3 become defensible and thereby suitable for the development of management scenarios. The decline in GWLs is investigated through management scenarios, which show that water use has higher impacts on groundwater level variations in the study area than those by climatic variabilities and this underpins the evidence for the necessity of management plans and strategies for the study area.