Integrating Seasonal Climate Forecast to Develope Novel Reservoir Rule Curves – A Case Study of the Shihmen Reservoir

由於台灣降雨分配不均、乾濕季分明，為求能蓄洪濟枯，使乾季也有足夠的水資源使用，因此水庫之蓄水便為不可或缺的水資源來源，而為求水庫之水資源能有效調配且利用，建立一合適之水庫運用規線就成為一個重要的課題。隨著全球氣候變遷，極端降雨事件的發生日趨頻繁，降雨分配不均的情況也日益嚴重，導致台灣近幾年發生枯旱情形的頻率也隨之增加，而乾旱是一個緩慢且不易察覺的現象，因現行之水庫運用規線只依據當月水位進行水資源調配之判別，所以往往使水資源管理者錯失最佳調配水資源的時機，為提早因應未來的枯旱潛勢並掌握供水策略調整的先機，本研究將中央氣象局季長期氣候預報融入於水庫運用規線的操作中，以降低旱災對社會、經濟的負面衝擊。 本研究以石門水庫為例，建立同時考慮水庫有效蓄水量和未來兩個月之水庫入流量之新型水庫運用規線，並定義水庫有效蓄水量和未來兩個月水庫入流量的線性疊加函數為資訊流指標，以取代傳統運用規線依水位決定水庫供水量的供水規則並與石門水庫運用規線M5做比較。研究結果顯示在完美流量預報下，新型運用規線可比M5規線提前1至3個月分散未來缺水事件之缺水量，進而降低極端缺水的嚴重性，達到水庫水資源調配之目標，然而，若使用季長期氣候預報繁衍之流量預報操作水庫供水量時，季長期氣候預報的不確定性和不準確性會對未來水文情勢造成誤判，可能會導致新型運用規線喪失其優勢，影響水庫供水之穩健性。Due to significant seasonal rainfall variations, reservoirs and their flexible operational rules are indispensable to Taiwan. Furthermore, with the intensifying impacts of climate change on extreme climate, the frequency of droughts in Taiwan has been increasing in recent years. Drought is a creeping phenomenon, the slow onset character of drought makes it difficult to detect at an early stage, and causes delays on making the best decision of allocating water. For these reasons, novel reservoir rule curves using projected seasonal climate are proposed in this study, which can potentially reduce the adverse effects of drought. This study dedicated establishing new rule curves which consider both current available reservoir storage and anticipated monthly inflows with leading time of two months to reduce the risk of water shortage. The monthly inflows are projected based on the seasonal climate forecasts from Central Weather Bureau (CWB), which two disaggregation models, weather generator(WGEN) and semiparametric weather generator (SWG), are used to produce daily weather data for the hydrological component of the GWLF. To incorporate future monthly inflow projections into rule curves, this study designs an information flow index which is a linear combination of current available storage and inflow projections with leading time of 2 months. By optimizing linear relationship coefficients of decision flow index, the shape of rule curves and the percent of water supply in each zone, the best rule curves to decrease water shortage risk and impacts can be developed. The Shimen Reservoir in the northern Taiwan is used as a case study to demonstrate the proposed method. Existing rule curves of Shimen Reservoir (M5 curves) are compared with two cases of new rule curves, including hindcast simulations and historic seasonal forecasts. The results show new rule curves can allocate shortage amount to preceding 1 to 3 months to avoid extreme shortage events under perfect forecasts. However, the accuracy and uncertainties in historic forecasts would result unnecessary discounts of water supply, it may cause worse performs than M5 curves during droughts