[A-HW17] Near Surface Investigation and Modeling for Groundwater Resources Assessment and Conservation

[AHW17-03] Applying Deep Learning for Unconsolidated Sediment Classification: An Exploratory Study

*Chun-Wei Huang¹、Si Ying - Yau²、Sheng-Sheng Huang³ (1.Ming Chi University of Technology, General Education Center、2.National Taiwan University, Department of Geography、3.Ming Chi University of Technology, Department of Electronic Engineering)

キーワード：Hydrogeology, Groundwater, Artificial Intelligence

Hydrogeological exploration forms the basis for comprehending groundwater distribution. However, manually identifying hydrogeological conditions is both time-consuming and subjective. This study delves into the potential of 1-D Convolutional Neural Networks (1-D CNNs) to capture the geophysical features of unconsolidated sediments. We utilized geo-big data, including Electrical Resistivity (ER), Self-Potential (SP), and Gamma Ray (GR) data in Taiwan. The results demonstrated that our CNN-based model achieved an overall accuracy of 83.14% in identifying gravel, sand, and clay. Nevertheless, the precision and recall in identifying sand need improvement due to limited support in sample size.