Development of probabilistic warning water level estimation model–A case study of Keelung River watershed
|關鍵字:||警戒水位;SOBEK;多變量蒙地卡羅模擬法;不確定性與風險分析;預警可靠度;Warning level;SOBEK;Multivariate Monte Carlo Simulation;Uncertainty and risk analysis;Early warning reliability|
In Taiwan, flood early warning can be issued by Water Resources Agency (WRA) in accordance with the observed water stage exceeding the thresholds (named warning water level) specific river-stage gauges. In general, warning water-levels can be determined by using the hydrological and hydraulic models with historical hydrologic data based on the control elevation of dikes or hydraulic structures. However, climate change and extreme rainstorm events with high occurrence frequency significantly have led to variations in the hydrological and physiographic data in the watersheds which should impact the reliability of hydrological and hydrological models. There exists, accordingly, the uncertainty in resulting thresholds of water levels for flooding and inundation. Therefore, this study aims to develop a probabilistic model for the estimation of warning water levels in the fluvial system by taking into account uncertainties in hydrological and physiographic factors,of which uncertainty factors of interest, hydrological factors include the rainfall characteristics (i.e. rainfall duration, depth and storm pattern) and tide depth and the roughness in the river bed is regarded as the physiographic factor in terms of Manning’s n coefficient. Since the variation of hydrological data should influence the reliability of calibrated parameters of rainfall-runoff models, the parameters of rainfall-runoff models are also treated as the uncertainty factor. In detail, the proposed probabilistic model for the warning water levels is developed by using uncertainty and risk analysis in conjunction with the rainfall-runoff model and river routing model. Noted the in this study, the SOBEK model and SAC-SMA (Sacramento soil moisture accounting) are used in the estimation of runoff and water levels respectively. Eventually, the advanced first-order and second moment (AFOSM) approach is applied in quantifying the reliability of warning water levels attributed to uncertainty factors of interest. The Keelung river watershed is selected as the study area and the hydrological data (i.e. rainfall, discharge and tide depth during typhoon events),which were recorded in associated rainfall and water-stage gauges are adopted in the model development and application. The results from model application indicate that the proposed probabilistic model can not only quantify the reliability of warning water levels at stage stations in the case of combination of variations in the uncertainty factors, but also evaluate the effect of uncertainty factors considered on the warning water levels. In addition, through the proposed probabilistic model, it reveals that issued water levels by WRA at stage gauges along Keelung River are in association with high reliability in early warning and they can enhance the performance of disaster prevention operation.