石門水庫入流量推估不確定性分析

Abstract

台灣由於地形及氣候條件的限制，造成降雨時空間分布不均。在台灣，只有20%左右的水可以被人民使用，因此水資源的分配已成為政府的一個重要議題。由於水資源的短缺，水庫是台灣最重要的水利設施，如果可以提供未來水庫入流量的模擬值，那麼對於水資源的調控將有會很大的幫助。因此有必要建立一個基於降雨逕流模型的水庫入流量推估模型。由於水文模擬的過程充滿各種不確定性，所以勢必以序率 (stochastic) 的架構來討論水庫入流量。 本文的研究區域為石門水庫，為台灣北部重要的水庫。主要提供農業用水、公共用水、發電和防洪等，是一個多功能的水庫。每年的11月至隔年5月為台灣枯水期，而一期稻作的耕種期間則為每年的1月至7月，由於兩者時間重疊，常常造成上半年水情吃緊，所以本研究將針對枯水期的水庫入流量進行探討。 本研究以SWAT模式 (Soil and Water Assessment Tool) 進行石門水庫入流量的推估；使用氣象繁衍模式生成單點的降雨量，並利用Copula函數建構的石門水庫上游雨量站的多變量分布，模擬上游集水區的降雨分佈。兩者搭配使用，可以提供水庫入流的不確性區間。在此不確定性的架構下，水庫入流量可被表示為一個機率密度函數。該入流量的機率密度函數，可以提供政府機關制定決策時的依據，對於未來可能發生的洪旱災提早做準備。期望本研究對於台灣未來的水資源調控有所幫助。

Due to the limitation of topography and climatic conditions, Taiwan has uneven spatial distribution during rainfall. In Taiwan, only about 20% of water can be used by people, so the distribution of water resources has become an important issue for the government. Due to the shortage of water resources, the reservoir is the most important water conservancy facility in Taiwan. If it can provide the simulated value of the future reservoir inflow, then the regulation of water resources will be of great help. Therefore, it is necessary to establish a reservoir inflow estimation model based on rainfall runoff model. Since the process of hydrological simulation is full of uncertainties, it is imperative to discuss reservoir inflows in a stochastic architecture. The research area of this paper is Shimen Reservoir, which is an important reservoir in northern Taiwan. It mainly provides agricultural water, public water, power generation and flood control, and is a multifunctional reservoir. From November of each year to May of the next year, it is the dry season of Taiwan, while the cultivation period of the first phase of rice cultivation is from January to July of each year. Due to the overlapping of the two times, the water situation in the first half of the year is often tight, so this study will target the dry season. The reservoir inflow is explored. In this study, the SWAT model (Soil and Water Assessment Tool) was used to estimate the inflow of Shimen Reservoir; the meteorological reproduction model was used to generate a single point of rainfall, and the Copula function was used to construct the multivariate distribution of the upstream rainfall station of Shimen Reservoir to simulate the upstream Rainfall distribution in the catchment area. The combination of the two can provide an uncertainty interval for reservoir inflow. Under this uncertainty structure, the reservoir inflow can be expressed as a probability density function. The probability density function of the inflow can provide a basis for government agencies to make decisions and prepare for the possible floods and droughts in the future. It is expected that this study will help Taiwan's future water resources regulation.