經驗動態建模於季長期天氣展望與乾旱預警系統之應用-以濁水溪流域為例

Abstract

台灣降雨時空分配不均，豐枯水期降雨量差距大，當枯水期降雨不足，往往會導致水資源缺乏而無法滿足供水量。近年來，受到氣候變遷影響下，旱澇事件更加頻繁地交替發生，豐枯水期差距更加懸殊，又因為工業園區的增設，導致需水量上升而更容易產生供水缺口，使得政府不得不採取限水、停灌等政策，造成龐大的經濟損失以及人民生活上的不便。為了填補氣候變遷下缺水風險的增加，勢必要訂定對應的調適策略，然而興建新興水源既花錢又曠日費時，欲解決乾旱問題所投資的成本過高。若能及早掌握未來缺水情勢，從水資源調配及管理著手，以乾旱預警系統作為調適手段，將能有效強化水資源系統之調適能力並減少乾旱風險。 本研究應用中央氣象局季長期天氣展望資料，並同時以經驗動態建模(Empirical Dynamic Modeling, EDM)─Simplex projection & S-map預測方法，結合標準化降雨指標(SPI)，發展EDM季長期天氣預報方法，並以濁水溪流域為例，透過氣象合成模式(WGEN)、水文模式(GWLF)、水資源系統動力模式評估季長期的供水缺口，並建立乾旱預警系統以預測未來水資源之供需情勢，作為科技部氣候變遷調適科技整合研究計畫（TaiCCAT）氣候調適六步驟中第四步驟「界定與評估調適選項」之調適選項。研究結果顯示，以2012年至2015年間發生的乾旱事件為例，季長期天氣展望雖能合理預測降雨變化趨勢，在預測乾旱有時候會有些許高估情形產生，導致無法準確預測缺水事件發生；Simplex projection & S-map雨量預報以月尺度SPI-1及旬尺度SPI-1預報最能有效預測缺水事件，適合作為乾旱預警系統的預報方法，以提供決策者及早進行水資源的調配。

The temporal and spatial distribution of rainfall is extremely uneven in Taiwan. There is a significant difference between cumulative rainfall in dry seasons and in wet seasons. If the rainfall was insufficient in dry seasons, the water demand couldn’t be satisfied due to the lack of water resourses. Because of the influence of climate change, the frequency of drought or flood events increases gradually. The rainfall difference between wet and dry seasons will be more drastic in comparison to past years. In addition, the development of industrial areas led to increasing water demand and deteriorate water shortage problem. It will force the government to adopt water ratinoning or fallow policy, which would cause a significant economic loss and inconvenience to people. To respond the increasing risk of drought under climate change, corresponding adaptation strategies must be estabalished. However, building new reservoirs is espensive and time-consuming. If the water shortage in the future can be predicted, water resources management and allocation are feasible to prevent the water shortage. The prediction can be also used to develop a drought early warning system as an adaptation measure, which will strengthen the adaptive capacity of the water supply system and reduce the impacts of drought events. The case study in this research is based on the water supply system of the Chuoshui River basin in Taiwan. This study applies the seasonal weather outlook released from Taiwan Central Weather Bureau (CWB) to establish a drought warning system, which is used to estimate the water shortage in the future. In the meantime, a EDM seasonal weather forecast has been developed in this study by applying empirical dynamic modeling predicting method─Simplex projection & S-map combining with Standard Precipitation Index (SPI). The CWB seasonal weather outlook and the EDM seasonal weather forecast are conducted as an adaptation measure of the fourth step of the six-step decision support tool developed by the Taiwan integrated Research Program on Climate Change Adaptation Technology (TaiCCAT). Also, the weather generator (WGEN), Generalized Watershed Loading Functions (GWLF), and the water supply system dynamic model were integrated in this predicting process. The results show that the CWB seasonal weather outlook can rationally predict the trend of rainfall, but overestimate streamflow value during drought event sometimes. The EDM seasonal weather forecast shows that the predictions by monthly SPI-1 and ten-days SPI-1 can effectively meet the real situation of water shortage. These predictions can be regarded as appropriate methods to inform the drought early warning system, allowing the decition-maker to allocate water resources at an early date.