Document Type : Research Paper
Senior Technical Expert, Tehran Regional Water Distribution Company
Assistant Professor, Shahrud University of Technology
Flood frequency analysis method is employed to predict the design floods needed for the implementation of hydraulic structures along a river. In this method, the magnitudes of floods are made related to their frequencies using statistical distribution functions. The accuracy of flood estimation depends on some factors including fitting the best probability distribution function along with an estimation of its parameters, period of available data, commonplace errors in measurement and an existence of outlier data. Calibrated hydrological models can also be applied to simulate large flood events. Uncertainties arising from the hydrological modeling, e.g. the modeling assumptions and the uncertainties in inputs and in parameters which are used in the models, are the factors which affect the simulated peak discharges. In the course of this paper, flood frequency analysis was performed using a hydrological modeling approach with the results compared with traditional flood frequency analysis as based on the observed data. The method was applied to Tangrah Watershed. The average rainfall values were initially calculated for different durations and for different return periods. For each rainfall value, a Monte Carlo scheme was employed to consider the uncertainties in rainfall spatial distribution and Antecedent soil Moisture Condition (AMC). Using different combinations of rainfall depths along with their spatial distributions over watershed and also AMC conditions as the inputs of hydrologic model (HEC-HMS), peak discharges of different return periods were found out. Besides the modeling approach, the best probability distribution was fitted as based on "Anderson–Darling" goodness-of-fit test with flood peak discharges being calculated while using the observed data. It was deduced that the outlier rainfall exerts a significant effect on rainfall – runoff simulation results and this effect increases with increase in return period. For instance, the simulated peak discharge of 50% probability reduces from 1297 to 338 cm2 following an omission of the outlier for the case of 200 year return period of 6hr duration rainfalls. The spatial rainfall distribution and antecedent soil moisture conditions largely influence the peak discharge of different return periods. Also, an observed peak discharge is seen as greater than the results obtained from rainfall simulation. As an example, for a return period of 200 years, hydrologic modeling results are found out as lower than the frequency analysis results by about 30.4%. An incorrect selection of statistical distribution plays a major role in increasing the difference between the results obtained from these two methods.