台灣的地形及氣候容易誘發崩塌，其中深層崩塌與地下水位有直接關係，本研究以類神經網路進行崩塌地地下水預測，首先收集梨山地滑地之歷史降雨與地下水位資料，再以Hong(2017)研發之類神經網路模式為基礎，選取一場暴雨，進行模式之參數校準及驗證，並應用於之後發生之另一場暴雨。分析結果發現，可以精準預測一小時、二小時後之地下水位，作為坡地崩塌預警系統建置之參考依據。 Taiwan's topography and climate are prone to induce landslide. Deep-seated landslide is directly related to groundwater level. In this study, the neural network was used to predict groundwater in deep-seated landslide areas. First, the historical rainfall and groundwater level data of the Lishan Landslide were collected. Based on neural network models developed by Hong (2017), a heavy rainfall was selected to perform parameter calibration and verification of the model, which was applied to predict the groundwater level that occurred later. Analysis found that the groundwater level can be accurately predicted one hour and two hours later, and used as a reference for the establishment of a deep-seated landslide warning system.
