Assessment of Land Subsidence Hazards in the Choshui Delta Under Climate Change Scenarios Using Artificial Intelligence

Abstract

In recent years, global climate change has significantly impacted fluctuations in groundwater levels, with Taiwan experiencing a severe water shortage crisis during the 2021 century drought. The intensive depletion of groundwater from aquifer systems has led to a significant increase in land subsidence in the Choshui delta of central Taiwan. In this study, we propose the assessment of land subsidence hazards in the Choshui delta under climate change scenarios using artificial intelligence. This study first compiles basic data such as groundwater levels, pumping, precipitation, drilling data, land-use patterns, and other historical monitoring data from the past 20 years. The geographic information systems are then utilized to establish spatial layers of geographical information factors. Subsequently, principal component analysis is employed to reduce the dimensionality of the above factors, analyzing the coupled relationships and correlations between land subsidence and relevant factors. This process aims to identify factors highly correlated with land subsidence, serving as input data for the proposed neural network model. The proposed model is then used to predict land subsidence and validated against historical land subsidence events. This facilitates the subsequent development of a neural network-based analysis model for land subsidence hazards. Results indicate a high correlation between groundwater levels and extraction quantities with the impact on land subsidence. Furthermore, the neural network model developed in this study demonstrates high precision during both training and prediction stages, with root mean square error values of 10-3 and 10-1, respectively. The future application of the developed neural network model enables continuous prediction of land subsidence hazard potential under various climate change scenarios in the Choshui delta.