水資源是人類不可或缺的資源，而乾旱事件的發生不只會在農業上產生損失，工商產業上也會因生產停頓及減產造成經濟上的損失。該如何管理及分配水資源，儼然成為現今一個重要的議題。氣候預報資訊可以預先掌握未來降雨趨勢預多寡，能在乾旱發生之前就先掌握未來降雨多寡落點，提早擬定決策以降低乾旱所造成的衝擊。然而預報存在風險，故本研究希望能量化預報的經濟價值及風險。 因此本研究集合季尺度氣象預報資訊與SIA-CWBI系統，模擬繁衍出符合預報的未來石門區農業供水量，搭配農業計畫需水量將會推估農業缺水量。換算成農業缺水下的政府所支付的補助金和休耕補助金比較，加入決策樹來探討氣候預報資訊對於休耕決策的經濟價值跟風險。 另一方面，本研究將農業灌溉管理模式(CROPWAT)模擬出種植稻米的需水量搭配週預報推估的降雨求得種植稻米的缺水量，計算農民負擔因缺水造成的可能損失。搭配農民用水決策，可推估週預報於農業用水的經濟價值及風險。 將季預報加入決策樹得到的研究結果發現，庫容在上限到嚴重下限之間會產生EVPI（完全資訊價值）及EVII(不完全資訊價值)。因為預報會比無預報更接近未來會發生降雨的趨勢，故會產生預報的經濟價值。此種庫容下加入預報所做的決策在每種預報組合不一定皆會與無預報相同，補助期望值不與無預報相同，所以會產生具有風險預報的經濟價值。 本研究建構了一個適逢乾旱時週尺度的對外引水與否決策機制。爾後欲對外購買別種水來補充農業缺水量，可依照本研究架構做決策。然而，由於結合週尺度及季尺度預報應用於農業用水決策還在測試階段，敏感係數會隨不同作物、不同地區而有所不一樣；引水實際購買成本也會因地而有所不同。待未來可以做出更貼近實際狀況的模擬與推估，並且得到更準確的預報經濟價值。 ;Water resource is necessary for human. The drought causes economic losses not only in agriculture but also in industry and commerce. Climate forecasts help us to predict rainfalls before droughts happen. Policymakers can make decisions in advance to reduce the drought-induced impact on agriculture, industry or commerce. However, there are uncertainties in climate forecasts. Quantification of the risk of decisions that are made based on climate forecasts and the corresponding information value of climate forecasts becomes an important issue of water resource management. In this study, I used the seasonal climate forecasts and the SIA-CWBI system to simulate the agricultural water supplies in Shimen, estimated the amounts of the agricultural water shortages which were calculated from the differences between the agricultural water demands and supplies, found out the compensation costs and the water tariffs, and got the information value of the climate forecasts with the decision trees. In addition, I used the CROPWAT system to simulate the water demands for rice, estimated the amounts of the water shortages for rice which equal to the amounts of the water demands for rice minus the water supplies for rice, converted the amounts of the water shortages into the economic losses, and got the information value of the climate forecasts with the decision trees. Finally, I evaluated the expect value of perfect information (EVPI) and the expect value of imperfect information (EVII) of the short-term climate forecast on the fallow decision. The results show that EVII increased as the degree of the accuracy of the forecasts increased. In addition, the maximum EVPI/EVII occurred when the fallow decision based on the initial water storage was difficult to make. I also constructed a framework of agricultural water usage with the help of monthly climate forecasts. By following the framework, decision-makers can reduce the damage caused by droughts and quantify the information value of the monthly climate forecasts.
