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.
