臺灣淹水災害對水稻生育期災損評估

Abstract

近幾年台灣強降雨事件造成了許多生命財產的損失，在極端降雨日趨嚴重下，對我國的農業也帶來了不小衝擊。其中一期水稻的成熟期及二期水稻的營養生長期與台灣主要的降雨季節重疊，為了分散風險，政府辦理了天然災害救助與農作物保險等措施，但卻面臨缺乏可靠的災損數據，導致了災損認定上的爭議與農作物保險設計的困難。本研究旨在建立災損函數模型，以彰化縣2017年6月1日豪雨事件為例，並探討模式估計金額與實際損失金額間的關係。本研究將彰化縣劃分為20×20公尺網格，依據每個網格內的情況判斷損失情形，並使用淹水潛勢圖作為預測豪雨發生時淹水深度之依據，經計算網格內損失後，以鄉鎮市為單位加總範圍內的損失金額得出估計損失金額。而後利用Pearson相關係數與差異比方式找出可能的極端值，接著以最小平方法將估計損失金額與實際損失金額進行線性迴歸，對模式整體解釋力作分析；本研究亦將針對無法有效估計的地區分析其潛在原因，並且透過空間上的關係找出可能影響的因素。研究結果顯示，模式可以估計彰化縣七成的鄉鎮市損失金額與救助金額；然而，模式本身仍會受到其他條件的限制，如品種差異、產地價格及違反淹水潛勢圖假設等因素，造成在部分鄉鎮市存有高估與低估的問題。儘管如此，在將極端值刪除並使用估計值與實際值進行迴歸分析後，其修正自由度的迴歸判定係數最高達到0.926，顯示模式仍然具有良好的解釋力，可提供災損估計時另一參考方向。In recent years, intense rainfall in Taiwan have caused serious damages to our properties. With the increasingly serious extreme rainfall, it has also brought a lot of impact on Taiwan’s agriculture. The maturity period of the first-stage rice and the vegetative growth period of the second-stage rice overlap with the main rainfall seasons in Taiwan. To spread the risk, the government has implemented the natural disaster relief and the agricultural insurance but faced a lack of reliable disaster loss data. This solves the controversy on the identification of disaster damage and the difficulty of agricultural insurance design. This study aims to establish a disaster loss function model, taking the intense rainfall event in Changhua County on June 1, 2017 as an example, and discussing the relationship between the estimated amount of the model and the actual amount of loss.In this study, Changhua County is divided into several 20×20-meter grids, the loss is judged according to the situation in each grid, and the Inundation Potential Map is used as the basis for predicting the flooding depth when intense rainfall occurs. After calculating the grid After the internal loss, take the town as the unit to add up the loss amount within the scope to get the estimated loss amount. Then use the Pearson correlation coefficient and the difference ratio method to find the possible extreme values, and then use the least square method to linearly regression. This research will also target areas that cannot be effectively estimated analyze its potential causes and find out possible influence factors through spatial relationships.The result of this research shows that the model can estimate the amount of loss and the amount of assistance in 70% of the town in Changhua County; However, the model itself is still limited by other conditions, such as rice variety difference, price of producing area and violation of the hypothesis of Inundation Potential Map, which leads to overestimation and underestimation in some towns. Nevertheless, after regression analysis of the estimated value and the actual value, the adjusted R-square is up to 0.926. The display model still has a good explanatory power and can provide another reference direction for damage estimation.