由於雲林地區地層下陷情形嚴重,除了對當地自然環境造成影響,如海水入侵陸地及土壤鹽化,也影響社會經濟活動,例如農作物產量減少,與高鐵高架軌道下陷風險。本研究探討雲林地區受到地層下陷所造成之影響,以農業活動之金額損失與臨近重大基礎建設之潛在損失來評量脆弱程度。考量年平均地下水位差,水井數量,月平均雨量,月平均溫度,經濟所得,稻作產量,醫療據點數量,與不透水層面積等八項影響因子,為造成雲林脆弱度的潛在影響因子,透過隨機森林模式的特徵篩選方法-Boruta,篩選最具影響力的四個因子:稻作產量,水井數量,年平均地下水位差,以及月平均雨量,同時計算四大因子的權重,建立脆弱度指標推估模式。以此模式繪製研究地區各村里的脆弱度指數分布圖,並利用地理加權迴歸模式,透過空間統計分析,檢視不同因子如何影響對各村里的脆弱度指數。研究成果顯示在脆弱度指數與四大因子具有顯著關聯性,而地層下陷也與各因子具有空間關聯性。為降低雲林地區因地層下陷造成的脆弱度影響,除了透過降雨量增加,減少地下用水使用之外,應考慮減少用水量極高的稻作產量,減少水井開鑿,並持續監控年平均地下水位差,以期維護雲林地區的永續發展。
Land subsidence has been a severe environmental issue in Yunlin, Taiwan. From the physical perspective, it induces to seawater intrusion and causes soil salinization at coastal areas. From the social perspective, land subsidence makes farmland unable to cultivate plants and reduce the crop yield. It also causes safety issues of Taiwan High Speed Rail. To examine how land subsidence affects the assessment of vulnerability in Yunlin County, this study first determines economic loss from agricultural products and Taiwan High Speed Rail as the measurement of vulnerability. Eight candidate factors are chosen due to the relevance of land subsidence, and they are: 1. annual difference of water table, 2. the number of wells, 3. average monthly precipitation, 4. average monthly temperature, 5. household income, 6. rice yield, 7. the number of hospital, and 8. aquiclude. Boruta, a feature selection method from Random Forest modeling, is introduced to select the most significant factors of vulnerability. The four selected factors: 1. annual difference of water table, 2. the number of wells, 3. average monthly precipitation, and 4. rice yield, are weighted as the key variables to calculate the vulnerability index of Yunlin County. In addition, geographically weighted regression is adopted to observe the significance of spatial correlation between each variable and vulnerability. The result shows if residents in Yunlin County reduce the cultivation of rice, and use fewer wells to access ground water, the process of land subsidence tend be slower, and the vulnerability will be eased.