水庫供水量可靠度推估模式之建立-以石門水庫為例

Abstract

本研究考量各類水文因子(降雨特性、基流量、蒸發損失、水庫及水庫起始水位)、水庫操作規線因子及降雨-逕流模式參數之不確定性而發展一供水可靠度模式用探討水庫供水量之可靠度，也就是reliability assessment model for water supply from the reservoir (RA_WS_Res)。RA_WS_Res模式係以Sacramento Soil Moisture Accounting (SAC-SMA)模式為入庫流量推估模式，配合蒙地卡羅模擬法(Multivariate Monte Carlo Simulation Method, MMCS)，衍生多組不確定性因子模擬值，採用水資源調配模式Ribasim (River Basin Simulation Model)進行水庫水資源預測及調配模擬，應用多變數迴歸分析，推估供水量與其他不確定性因子之相關式，並藉由敏感度分析量化影響供水量之敏感因子，再利用改良一階二矩法(Advanced First- Order Second-Moment method, AFOSM )及羅吉斯迴歸分析，建立各分析節點供水量超越機率計算方程式。綜合上述，本研究所發展的水庫供水量可靠度推估模式主要包括5個部份：不確定因子之衍生、入庫流量之推估，供水量之模擬、供水量風險之量化及供水量超越機率計算方式之建立。因此，期使可透過模式進行不確定性因子(水文、水庫因子及降雨逕流模式參數)之變異程度對於供水量可靠度之影響分析。 本研究以石門水庫供水系統為研究區域，並挑選出6個供水節點，採用1987年至2014年之時降雨序列及相關水文資料，應用上述模式發展分析架構，建立石門水庫供水量可靠度推估模式。由模式應用結果可發現，不確定性因子中的水庫初始水位、當旬之平均降雨強度、基流量及水庫操作規線間距(Range_T-F)為影響供水量可靠度之重要不確定性因子。其中，水庫初始水位平均值變化率之影響依旬遞減；平均降雨強度與基流量皆呈正相關，顯示水庫供水量除來自降雨外，基流量亦為供水量重要來源之一。此外，水庫操作規線間距(Range_T-F)在各旬對供水量可靠度有程度不一之影響。由上述結果可知，本研究建立之供水量可靠度推估模式可合理反應各不確定性因子對供水量超越機率之變化趨勢，因此未來可透過集水區降雨特性分析，由所建置之關係式推得石門水庫下游節點之供水量超越機率，亦可應用於因極端氣候所造成之水文量變異性，評估水庫下游供水量之可靠度，作為水資源規劃與分析之參考。

This study aims to develop a reliability assessment model for water supply from the reservoir (RA_WS_Res) in order to quantify the risk of water supply due to the uncertainty factors. The uncertainty factors group into three types: hydrological factors, reservoir operation rules, and parameters of rainfall-runoff model (i.e. Sacramento Soil Moisture Accounting, SAC-SMA). Specifically, the hydrological factors includes the rainfall characteristics, baseflow, evaporation, and initial water level of dam; the reservoir operation rules involve the flood level, target level and firm storage level. In detail, the proposed RA_WS_Res model primarily employs the multivariate Monte Carlo simulation (Wu et al., 2006) to generate the uncertainty factors in order to produce dam inflow by incorporating with the SAC-SMA model. After that, the water supply at the demand nodes of interest can be obtained from the water-resource allocation model (River Basin Simulation model, Ribasim). Using the simulated uncertainty factors and corresponding estimated water supply, the resulting exceedance probability (i.e. insufficient risk) can be calculated by using the uncertainty and risk analysis (i.e. advanced first-order and second moment, AFOSM). Eventually, this study carries out the logistic regression analysis to establish the relationship between the exceedance probability and average rainfall intensity for various 10-day periods at the specific nodes. This relationship is named the exceedance probability calculation equation. In summary, the proposed RA_WS_Res model is composed of five components: simulation of uncertainty factors, estimation of dam inflow, estimation of water supply, quantification of insufficient risk, and establishment of the exceedance calculation equation. It is expected that the proposed RA_WS_Res model can quantify the effect of variation in uncertainty factors, especially due to climate change, on the reliability of water supply from the reservoir. Shihmen Reservoir watershed is selected as the study area and six demand locations within the watershed are selects as the study nodes. In addition, the hourly rainfall data from 1987 to 2014 and associated hydrological data (i.e. evaporation and baseflow) as well as the operation rules are used in the model development and application. The results indicate that, amond these uncertainty factors, the reliability of water supply is more sensitive to uncertainties in the initial water level, average rainfall intensity in each 10-day period, baseflow and the range between the target level and firm storage level. In detail, the effect of variation in the initial water level to water supply gradually reduces with time (ten days) from the 1st ten-day to the 15th ten-day period in Shihmen Reservoir. Moreover, the average rainfall intensity and baseflow are positively related to water supply; this implies that the baseflow plays an important role in the estimation of water supply, besides the average rainfall intensity. In particular, the range between the target and firm storage levels has significantly influence on the reliability of water supply. This reveals that regulating the operation rules for the reservoir not only focuses on the flood mitigation and prevention, but also takes into account the water-supply reliability. Consequently, the proposed RA_WS_Res model can effectively and reasonably quantify the insufficient risk (i.e. exceedance probability) of water supply attributed to variation in uncertainty factors due to climate change; this can be useful for the water-resource allocation and analysis.