獨立成分與小波多分辦分析識別地下水抽水成分之研究－以濁水溪流域為例

本研究應用獨立成分分析（Independent Component Analysis, ICA）與小波多分辨分析（Wavelet Multiresolution Analysis, WMRA）識別地下水抽水成分。應用ICA於地下水觀測系統可分離降雨及河川刺激所產生之地下水位成分，對於抽水所引起之水位反應，因其除有高頻成分（1 cycle/day）外，亦含低頻（1 cycle/year）載波特性，故本研究以WMRA分解抽水獨立成分，進一步解釋各頻率特徵。本研究以濁水溪沖積扇淺層觀測井作為分析標的，並以兩個案例討論ICA結合WMRA於抽水獨立成分之應用：一、局部案例：以虎尾鎮87口農業抽水井於2016年之直接抽水量觀測資料，驗證此抽水特徵之地下水位，並解釋此時間序列之低頻成分與農業灌溉時程、降雨特性等因素有關，高頻成分則與人為作息密不可分。二、區域案例：以濁水溪沖積扇58口觀測井之抽水獨立成分解釋WMRA的分析過程，所得抽水特徵可作為後續區域水資源管理之參考。

關鍵字

獨立成份分析；小波多分辨分析；傅立葉轉換；地下水

並列摘要

This study employed independent component analysis (ICA) and Wavelet Multiresolution Analysis (WMRA) to identify the pumping features of the study area. ICA can successfully separate the features of head variations caused by rainfall and river stage; however, the explanations of the pumping features still need to be further improved since the pumping features consist of high and low -frequency signals. Accordingly, this study employed WMRA to analyze the pumping components from ICA to extract and explain the low and high -frequency signals. The proposed method was employed to the shallow aquifer of Zhuoshui River alluvial fan, and two cases were used for examining the capability of the proposed method: (1) Local case study: The proposed method was applied to the seven groundwater observation wells located in the Huwei Town, and the derived pumping features were compared with the pumping records of 87 pumping wells in this area in 2016. The results show that the low -frequency pumping features highly correlates with rainfall and agriculture pumping schedule, while the high-frequency pumping feature highly correlates to human’s lifestyle. (2) Regional case study: The head observations of the 58 groundwater monitoring wells were analyzed by the proposed method, and the progress of analyzing the pumping components using WMRA was explained here. The derived pumping features based on the proposed method are valuable references for the regional groundwater resources management.