Assessment of Potential Artificial Recharge Area Using 3D Geological Model Based on Multi-Geoelectrical Data and Machine Learning Approach

The present study involved the coalescing of multi-geoelectrical data such as Transient Electromagnetic (TEM), Electrical Resistivity Imaging (ERI), Vertical Electrical Sounding (VES), and Normal Borehole Resistivity (NBR) through a machine learning approach to evaluate the potential groundwater recharge area in the proximal fan of the Choushui River Alluvial Fan. 77 TEM sites and 33 ERI survey lines were collected in the field, while 13 VES data and 15 NBR were obtained from the Central Geological Survey of Taiwan database. All the geoelectrical data were inverted independently, then assimilated the data to cope with the scale and resolution problem before 3D modeling. Furthermore, we applied Hierarchical Agglomerative Clustering (HAC) in machine learning to interpret the resistivity model into the geological model and evaluated it with the Silhouette Index (SI). Thus, we were able to transform the 3D resistivity model into the 3D geological model. Finally, we determined the potential recharge area in reference to the accumulated gravel and clay thickness distribution.