Innovative AI Strategies for Groundwater and Subsidence Management in Taiwan

The Choshui Delta in Taiwan is experiencing accelerated land subsidence due to excessive groundwater extraction and the impacts of climate change. Addressing this challenge requires advanced predictive tools to monitor and forecast subsidence over time. This study proposes a novel artificial intelligence (AI) framework combining Deep Neural Networks (DNNs) with Principal Component Analysis (PCA) for time-series land subsidence prediction. PCA is utilized to analyze eight critical factors influencing subsidence, reducing their complexity by extracting principal components. These components are then used as input features for the DNN model, enabling it to effectively capture the intricate, multi-factorial dynamics of subsidence. Validation of the model was conducted by comparing reconstructed groundwater level data with historical measurements, demonstrating high reliability and accuracy. The integration of DNN and PCA delivers precise predictions of subsidence patterns, offering a robust and scalable solution for managing subsidence risks in rapidly sinking regions like the Choshui Delta. This AI approach provides valuable insights for sustainable groundwater management and infrastructure protection in vulnerable areas.