Calibration et analyse de sensibilité du modèle de simulation continue de l'écoulement STORM

Calibration and sensitivity analysis of the STORM continuous flow simulation model

¹ Ingénieur du Génie rural, 1er assistant, École Polytechnique Fédérale de Lausanne, Suisse
² Professeur de Génie hydraulique Purdue University, West Lafayette, Indiana, U.S.A.

Abstract

This study relates to the hourly continuous runoff simulation model STORM. This model provides a substantial improvernent over the elementary rainfall-runoff transformations, without being so sophisticated as to become cumbersome, in the practice by consulting engineer offices of small size. Using the composite runoff coefficient method in an urban watershed and in addition, the Soil Conservation Services method for the pervious areas in a semi-urban watershed, a methodology is developed to calibrate the parameters characterizing the equations relating to each of these two methods. A sensitivity analysis of the model is done around the calibration values of the parameters. In an urban area, the runoff coefficient for impervious surfaces is a decisive parameter. Usually the relative error in the runoff is of the same magnitude as the relative error in this parameter. The other parameters require less care in their determination. In semi-urban areas the decisive parameter is the maximum soil moisture retention capacity. Its determination requires a laboratory analysis. Some empirical relations and a soil map provide good estimates for the other parameters. The two methods of analysis show that the yearly total amount of runoff, the total amounts of runoff by storm event and the peaks by event are all good and stable estimates. The choice of the method depends on the percent of imperviousness. Above 30 % imperviousness the first method is recommended and around 30 % of imperviousness, the use of both methods is suggested. A goodness-of-fit and a confidence interval calculation show that the amounts of runoff by event and the runoff peaks can be fitted by a Gamma distribution. In addition the model STORM calculates hourly hydrograph ordinates. However, for the watershed under study it was not possible to obtain reliable results before the peak time, because of the short time of concentration which was approximately 30 minutes. This study should provide the user with additional assistance and confidence in choosing the input parameters which require some degree of engineering judgment.