宜蘭河洪水流量持續性分析

Yilan River Flood Flow Persistent Analysis

石廣英(Guang-Ying Shih)

指導教授 ： 鄭克聲

國立臺灣大學/生物資源暨農學院/生物環境系統工程學研究所/碩士(2017年)

https://doi.org/10.6342/NTU201703280

摘要

在水文研究中，洪水流量預報是個常探討的議題。在過去文獻中，洪水流量的預測模式建立與評估此模式之下，其評估結果大多呈現良好。影響評估結果主要因素為河川流量的持續性。水文過程為時間序列資料，在本研究中，藉由偏自我相關係數（partial autocorrelation coefficient function, PACF）做模式鑑定並建立自回歸模式（Autoregressive processes, AR）來當預測模式，可得量化持續性的指標CIR（Cumulative impulse response）與評估模式指標效率係數（Coefficient of efficiency, CE）、持續係數（Coefficient of persistence, CP）和流量峰值誤差百分比（Error in peak flow in percentages, ）。然河川流量的持續性也受集流時間影響，根據流量與雨量逐時資料之相關性可估算集流時間。
本研究區域為宜蘭河流域，其流量站為中山橋、員山大橋和新城橋。各測站之觀測資料做模式建定並建立ㄧAR模式。在每測站中有一AR模式，並套配於其本身測站之各水文事件中，得出這些水文事件之CIR值、CE值、CP值和 值。並由結果顯示各測站之預測模式良好。在各測站的CIR指標皆高於7，表示流量持續性高。在集流時間上新城橋比員山大橋快，故員山大橋的CIR指標比新城橋的CIR指標大。

關鍵字

洪水預報 ； 流量持續性 ； 時間序列 ； 模式評估 ； 宜蘭河流域

並列摘要

Flood forecasting is an essential issue in hydrological studies. In the literature, many flood forecasting models were shown to perform well. However, it has also been recognized that, due to flood flow persistence, even simple models could also achieve good performance. In this study, two model performance criteria, namely the coefficient of efficiency (CE) and coefficient of persistence (CP) were used to evaluate performance of flood forecasting models. Flood flow data at three stations in the Yilan River Basin were represented by autoregressive (AR) models. An asymptotic theoretical relationship between CE and CP, which is dependent on the lag-k autocorrelation coefficient, was derived and used to demonstrate why the simple naïve forecasting model could achieve good performance, in certain cases, even outperform more complicated models.

並列關鍵字

Flood forecasting ； Flow persistence ； Time series ； Model performance evaluation ； Yilan River Basin

參考文獻

2. 連以婷(2010)，「水文模式之參數不確定性分析」，國立台灣大學生物環境系統工程研究所研究論文。

4. Andrews, D.W.K., Chen, H.Y. (1994) Approximately median-unbiased estimation of autoregressive models. J Bus Econ Stat, Vol. 12, No. 2, pp. 187–204.

7. Chen, P.A., Chang, L.C., Chang, F.J. (2013) Reinforced recurrent neural networks for multi-step-ahead flood forecasts, Journal of Hydrology, Vol. 497, pp. 71–79

8. Cheng, K.S., Lien, Y.T., Wu, Y.C., Su, Y.F. (2016). On the criteria of model performance evaluation for real-time flood forecasting. Stochastic Environmental Research and Risk Assessment, doi: 10.1007/s00477-016-1322-7

9. Chiang, Y.M., Hsu, K.L., Chang, F.J., Hong, Y., Sorooshian, S. (2007) Merging multiple precipitation sources for flash flood forecasting. Journal of Hydrology, Vol. 340, pp. 183–196.