[A-HW22] River Channel Morphology, Water Resource Management, and Advanced Techniques

[AHW22-01] Integrating Machine Learning for Advanced Groundwater Level Estimation and Drought Vulnerability Assessment in a Changing Climate

*Wei Sun¹, Hang-Yeh Lin¹, Fi-John Chang¹ (1.National Taiwan University)

Keywords:Convolution Neural Network-Back Propagation (CNN-BP), Water resource management, Regional groundwater level estimations, Drought prediction, Climate change scenarios

Climate change has amplified the challenges posed by extreme weather events, with increasing drought frequency becoming a critical concern for global water resource management. During droughts, reduced rainfall and surface water availability make groundwater an increasingly critical source of water supply. This study introduces an innovative approach to integrate regional groundwater level estimation from a Convolutional Neural Network-Back Propagation (CNN-BP) model with geographic and socio-economic parameters for providing a comprehensive assessment of drought vulnerability. The CNN-BP model is trained on approximately 20 years of meteorological and groundwater level data, incorporating climate change scenarios to enhance the accuracy of regional groundwater level estimations. In addition to groundwater level estimations, geographic factors such as elevation, slope, and land use, along with socio-economic variables including population projections, are utilized to evaluate drought vulnerability using a Comprehensive Drought Vulnerability Indicator (CDVI).
Climate change scenarios, ranging from optimistic (SSP1-2.6) to pessimistic (SSP5-8.5), are applied to analyze groundwater variation trends for drought vulnerability assessment. The results reveal that drought vulnerability is expected to rise across all scenarios, with groundwater levels projected to decline, further exacerbated by the increasing frequency of drought events. These findings highlight the urgent need for adaptive water resource management strategies to address the impacts of climate change effectively. By combining deep learning techniques with multi-dimensional data, this study provides a robust framework for predicting groundwater dynamics and assessing drought vulnerability. It offers valuable insights to support sustainable groundwater management and policy development in the face of escalating climate change challenges.