Managing Invisible Water: Artificial Intelligence-Based Technologies for Capacity Building of Groundwater Governance in Taiwan

From the world to Taiwan, groundwater is an important water resource that supports human development. However, the overdraft of groundwater due to a lack of appropriate management has led to environmental disasters, such as land subsidence, seawater invasion and depletion of rivers linked to the aquifer. This study synthesized the studies relevant to artificial intelligence (AI) technologies on groundwater management in Taiwan through a content analysis. We discuss the potential of deep learning to improve the capacity building of groundwater governance under the third wave of AI. We demonstrated that the deep learning-based computer vision technology could investigate and manage enormous unregulated private pumping wells. The case study revealed the evolution of AI applications in groundwater management, transitioning from machine learning to deep learning. While machine learning methods have shown its ability in predicting complex and nonlinear groundwater level changes, but they may also face the problems such as overtraining and low generalizability. Given the fact that traditional machine learning applications are limited to numerical data, deep learning-based approaches can capture important features from big data variety by incorporating images, videos, texts, etc. As such, deep learning technologies can facilitate regulatory and fiscal functions of groundwater administration via the improvement of information management functions. As AI continues to evolve, the multi-modal machine learning, which incorporates data diversity, can further improve groundwater governance.