台灣地區豐枯水季差異明顯，在枯水季期間若遭逢蓄水量及入流量不足，不只會對經濟上的造成損失，亦會使民生用水造成衝擊，若能有效應用氣候預報及早擬定因應對策，則能適度降低乾旱所造成的影響。停灌休耕是乾旱時期水資源調配的重要策略之一，然而決定停灌時間點的拿捏往往考驗決策者的能力。為了量化短期氣候預報對於不同的決策時間點的影響，本研究將氣象局季短期氣候預報，帶入氣象合成模式與水文模式模擬石門水庫集水區之入流量，搭配水庫歷史超越機率入流量與系統動力模式模擬水庫供需水量以及各時期水庫水位，決策者可依據預報的水位決定停灌與否。將氣象預報準確率使用貝式定理反推預報不確定性，進一步評估在相同情境下，不同時期預報可能的風險。本研究比較無預報、完全準確預報以及非完全準確預報對於決策的影響，其中非完全準確預報是以目前氣象局提供的預報資料為主。結果顯示，預報確實能有效降低停灌休耕決策之風險。完全準確的預報決策失敗的風險會隨決策時間越晚有下降趨勢。但是受限於預報準確度，非完全準確預報失敗風險並沒有隨著決策時間往後而下降。;During drought periods, the shortage of water resource causes economic losses as well as the domestic water useage. The impact of droughts can be reduced by applying short-term climate forecasts. The spring land fallow is an important strategy to reduce the impacts of water shortage on both industrial and domestic water usages. Therefore, the timing of the fallow decision becomes an important issue of the water management during droughts. In this study, the weather data is derived from meteorological synthesis mode and the data of the seasonal climate forecast form Central Weather Bureau. Then, input the weather data into the hydrological models to simulate the inflow of the Shihmen Reservoir watershed. Use system dynamics simulation model to simulate the water of supply and demand and the water level of reservoir. Combine the level of reservoir with historical exceedance probability inflow to calculate the shortage rate of probability agricultural water. The forecast uncertainty is estimated by using Bayesian theorem with accuracy of weather forecast. We investiage the effect of seasonal forcast data on the fallow decision based on three scenarios: no forecast, 100% accurate forecast and forecast with uncertainty. The result displays that the risk of decision failure in three scenarios tends to decrease with the delay of decision. The benefit of accurate forecast on the decision of fallow is significant. It can reduce the risk of the decision failure. However, the decrease of the risk of decision failure under forecast with uncertainty is insignificant because of the high uncertainty of itself.
