儘管台灣降雨豐沛，但由於地形崎嶇、逕流快速以及降雨分布不均，大部分雨水未經有效利用就流入海中，導致經常面臨缺水問題。此外，面積狹小也限制了水庫儲水的效果。因此，台灣在擁有豐沛降水量的前提下，仍非常容易遇到缺水問題。在減少旱災風險的過程中，彰顯了有效水資源管理的重要性。近年來，地下水已成為台灣重要的水源，使用量已超過水庫，占總用水量的百分之三十以上。由於水庫在台灣有限的空間內無法發揮最佳效果，因此了解地下水位的變化非常重要。
然而，影響地下水位的因子多且複雜，傳統模型的建立通常需要大量知識和人力。本研究將利用機器學習方法，通過純資料驅動的特性，使模型自動發現輸入資料之間的關係，嘗試預測未來的地下水位。研究區域為台灣中部濁水溪沖積扇平原一帶，我們將收集多個地下水位站的日資料作為目標因子，同時收集其他資料如降雨量、溫度、測站經緯度及地下水井高程等作為輸入因子。
經過數據整理後，首先將多個地下水位站的目標因子輸入K-MEANS模型進行集群分析，並以該分群結果對資料進行標準化處理。然後，將標準化後的數據輸入長短期記憶模型（Long Short-Term Memory，LSTM），以預測未來數天的地下水位，並通過均方根誤差（root-mean-square error，RMSE）和納許效率係數（Nash-Sutcliffe efficiency coefficient，NSE）等評估指標進行模型驗證。

Although Taiwan receives ample rainfall, the country continues to face persistent water scarcity due to its challenging topography, rapid runoff, and uneven distribution of precipitation. This issue becomes even more critical in the context of disaster risk management, highlighting the urgent need for effective water resource strategies to mitigate the effects of natural hazards. In recent years, groundwater has become an essential water source, surpassing reservoir usage and now accounting for over thirty percent of the nation's total water consumption, as sedimentation has reduced reservoir capacity. Understanding the variations in groundwater levels is therefore crucial. This study focuses on Taiwan's central region, particularly the Zhuoshui River basin. We gathered data from multiple monitoring stations, capturing groundwater levels along with meteorological factors such as temperature and rainfall. After refining and validating the data, we applied the Long Short-Term Memory (LSTM) deep learning model to predict groundwater level dynamics. To address the significant spatial variability within the Zhuoshui River basin, we employed a zonal model. This approach involves dividing the study area into multiple zones using the K-means clustering, which allows the model to better account for regional differences in groundwater levels. By processing data separately for each zone before feeding it into the LSTM model, we achieved improved prediction accuracy. The zonal model, combined with data integration techniques, proved particularly effective, with the model’s performance metrics such as Nash-Sutcliffe Efficiency (NSE) showing substantial improvement. This enhanced modeling approach not only allows for more precise groundwater level predictions but also provides insights into regional groundwater behaviors, which are crucial for sustainable water resource management.

Goap, A., Sharma, D., Shukla, A. K., & Rama Krishna, C. (2018). An IoT based smart irrigation management system using Machine learning and open source technologies. Computers and Electronics in Agriculture, 155, 41–49. https://doi.org/10.1016/j.compag.2018.09.040
Greenspan, H., Van Ginneken, B., & Summers, R. M. (2016). Guest Editorial Deep Learning in Medical Imaging: Overview and Future Promise of an Exciting New Technique. In IEEE Transactions on Medical Imaging (Vol. 35, Issue 5, pp. 1153–1159). Institute of Electrical and Electronics Engineers Inc. https://doi.org/10.1109/TMI.2016.2553401
Hochreiter, S., & Schmidhuber, J. (1997). Long Short-Term Memory. Neural Computation, 9(8), 1735–1780. https://doi.org/10.1162/neco.1997.9.8.1735
Khosravi, K., Pham, B. T., Chapi, K., Shirzadi, A., Shahabi, H., Revhaug, I., Prakash, I., & Tien Bui, D. (2018). A comparative assessment of decision trees algorithms for flash flood susceptibility modeling at Haraz watershed, northern Iran. Science of the Total Environment, 627, 744–755. https://doi.org/10.1016/j.scitotenv.2018.01.266
Le, X.-H., Ho, H. V., Lee, G., & Jung, S. (2019). Application of Long Short-Term Memory (LSTM) neural network for flood forecasting. Water (Switzerland), 11(7). https://doi.org/10.3390/w11071387
Lu, H., & Ma, X. (2020). Hybrid decision tree-based machine learning models for short-term water quality prediction. Chemosphere, 249. https://doi.org/10.1016/j.chemosphere.2020.126169
Najah Ahmed, A., Binti Othman, F., Abdulmohsin Afan, H., Khaleel Ibrahim, R., Ming Fai, C., Shabbir Hossain, M., Ehteram, M., & Elshafie, A. (2019). Machine learning methods for better water quality prediction. Journal of Hydrology, 578. https://doi.org/10.1016/j.jhydrol.2019.124084
Wang, Z., Lai, C., Chen, X., Yang, B., Zhao, S., & Bai, X. (2015). Flood hazard risk assessment model based on random forest. Journal of Hydrology, 527, 1130–1141. https://doi.org/10.1016/j.jhydrol.2015.06.008
Xiang, Z., Yan, J., & Demir, I. (2020). A Rainfall-Runoff Model With LSTM-Based Sequence-to-Sequence Learning. Water Resources Research, 56(1). https://doi.org/10.1029/2019WR025326
Yaseen, Z. M., Sulaiman, S. O., Deo, R. C., & Chau, K. W. (2019a). An enhanced extreme learning machine model for river flow forecasting: State-of-the-art, practical applications in water resource engineering area and future research direction. In Journal of Hydrology (Vol. 569, pp. 387–408). Elsevier B.V. https://doi.org/10.1016/j.jhydrol.2018.11.069
Yaseen, Z. M., Sulaiman, S. O., Deo, R. C., & Chau, K.-W. (2019b). An enhanced extreme learning machine model for river flow forecasting: State-of-the-art, practical applications in water resource engineering area and future research direction. Journal of Hydrology, 569, 387–408. https://doi.org/10.1016/j.jhydrol.2018.11.069
Zhang, J., Zhu, Y., Zhang, X., Ye, M., & Yang, J. (2018). Developing a Long Short-Term Memory (LSTM) based model for predicting water table depth in agricultural areas. Journal of Hydrology, 561, 918–929. https://doi.org/10.1016/j.jhydrol.2018.04.065
中興工程顧問股份有限公司，濁水溪沖積扇地面地下水聯合運用管理模式建立與機制評估總報告，2007
方熙廷，類神經網路於地下水位預測之時空分析，工程環境會刊，第三十七期，112-128，2017。
王士榮 濁水溪流域地下水資源開發與管理，臺灣水利，第 67 卷，第 1 期，43-67 2019。
巨廷工程顧問股份有限公司，109年度《集集攔河堰監測及安全檢查-環境生態監測》，2021
李心惟，結合HEC-RAS與MODFLOW模式評估濁水溪沖積扇地下水位及地層下陷，臺灣水利，第 67 卷，第 1 期，91-102，2019。
周嫦娥，臺灣水資源利用效率初探，土木水利，第四十八卷，第四期，23-32，2021。
葉明生，地下水環境限制水位與可用水量分析-以濁水溪沖積扇為例，農業工程學報，第66卷 第2期，43-51，2020。
廖玲琬，以頻譜分析法評量地下水位時空變動，工程環境會刊，第二十八期，1-14，2012。
鮑康元，地下水補注聯合營運之研究-以高屏大湖人工湖補注效益及土庫農場地下水可取水量分析，國立臺灣大學生物資源暨農學院生物環境系統工程學研究所碩士論文，台北市，2011。