Towards robust methods to couple lumped rainfall-runoff models and hydraulic models: A sensitivity analysis on the Illinois River

JOURNAL OF HYDROLOGY

卷:418

Article

Lerat, Julien^1,3  Perrin, Charles¹  Andreassian, Vazken¹  Loumagne, Cecile¹  Ribstein, Pierre² 

[1] Irstea, Hydrosyst Res Unit, F-92163 Antony, France

[2] Univ Paris 06, UPMC, UMR 7619 Sisyphe, F-75252 Paris 05, France

[3] CSIRO, Acton, ACT 2601, Australia

关键词: Hydrological model;    Hydraulic model;    Model coupling;    Lateral inflows;    Semi-distribution;   

DOI  :  10.1016/j.jhydrol.2009.09.019

来源: Elsevier

【 摘 要 】

Accurate prediction of a flood inundation area constitutes an essential part of a flood forecasting system. When a river reach receives significant lateral inflows, flood inundation modelling requires the joint application of a hydrological model to calculate lateral inflows and a hydraulic model to calculate water levels along the river reach. In this study, we compared different strategies to couple the GR4J lumped rainfall-runoff model and the linearised diffusive wave propagation model. These strategies introduced variations in the nature of the connections between the two models using combinations of point and uniformly distributed lateral inflows. The coupled model was then applied to the Illinois River case study for which 10 years of hourly data were made available within the Distributed Model Intercomparison Project, now in phase 2. The simulations were assessed at the downstream end of the reach and at two interior points considered to be ungauged during the calibration process. The results first show that including uniformly distributed inflows made the coupled model more robust and stable compared to only using point flow input. A similar level of performance was reached with models using point inflows only, but at the cost of more uncertain parameters and less stable model performance when changing test periods. Second, identifying the optimal number of tributaries to be modelled individually by the hydrological model was easier when a combination of uniformly distributed and point inflows was used. In this case, model performance was less sensitive to the number of tributaries used and the inclusion of two or three tributaries appeared sufficient to obtain satisfactory performance for the simulations on the main channel. Last, the results on the main channel differed from those obtained on tributaries: overall performance was better on the main channel and required a lower degree of lateral inflow resolution, which suggests that upstream flow measurements provide valuable information for flow simulation on ungauged points. (C) 2009 Elsevier B.V. All rights reserved.

【 授权许可】

Free