MODELLING GEOTHERMAL RESOURCE ASSESSMENT FROM GEOPHYSICAL AND REMOTE SENSING DATA USING GA-ANFIS-ALGORITHM IN PARTS OF SOUTHWESTERN NIGERIA

Abstract

The application of machine learning algorithms to geophysical and remote sensing parameters in establishing geothermal resource management was investigated. This study explored the efficacy of a GIS-based machine learning algorithm to analyze remote sensing and geophysical datasets to ascertain this task. The acquired remote sensing dataset was processed to derive surface-induced Geothermal Potential Conditioning Factors (GPCFs) viz: Lineament Density, Land Use Land Cover (LULC), Land Surface Temperature, Slope percent, and Normalized Difference Vegetation Index (NDVI). Spectral analysis was carried out on aeromagnetic data to derive sub- surface induced GPCFs, including Curie point depth (CDP), Heat Flow (HF), and Geothermal Gradient (GG). Thematic maps were produced in a GIS environment from the robust geospatial analysis of the aforementioned GPCFs. The Adaptive Neuro- fuzzy Inference System (ANFIS) models involving metaheuristic optimization algorithms, namely: Genetic Algorithm (GA), Invasive Weed Optimization (IWO), and Particle Swarm Optimization (PSO), were implemented on 2500 fishnet (observation) points spread across the produced GPCFs' thematic maps in the MATLAB program environment from which the optimized weights were processed. The results of these algorithms' weightage evaluation were processed on the GIS platform to produce geothermal potential predictive index (GPPI) maps for the study area. Furthermore, the performance evaluations of the developed machine learning predictive models' output were compared with the applied results of the Fuzzy Analytical Hierarchy Process (FAHP), and Analytical Hierarchy Process (AHP) knowledge-based multi-criteria techniques via validation approaches of Receiver Operating Characteristics (ROC) technique, the Root mean square error (RMSE) technique and multi-faceted geologywithin the study. The ROC-AUC results showed 73.5%, 77.2%, 81.1%, 77.5%, 52.6%, iv52.6%, and 50% for ANFIS, IWO-ANFIS, GA-ANFIS, PSO-ANFIS, FAHP-Extent analysis method, FAHP-Geometric method, and AHP, respectively. The overall result showed that the GA-ANFIS based model is a good alternative for the prediction of geothermal potential zones. Thus, the produced map could form part of the decision- making mechanisms for geothermal exploitation and management.