Document Type : Research Paper
Authors
Department of Water Engineering, Faculty of Agriculture, Razi University of Kermanshah, Iran
Abstract
Keywords
Main Subjects
Runoff simulation with HEC-HMS model and sensitivity analysis of flood hydrograph trending parameters using differential evolution algorithm (case study: Merck River catchment)
EXTRACTED ABSTRACT
When the flow rate exceeds the throughput capacity of the river, due to the overflow on the left and right banks of the river, we see a flood event. This increase in flow rate can be due to heavy and short-term rains or long and heavy rains. Simulation of runoff and hydrograph production is widely used in the analysis of basin behavior against precipitation, calculation of flood volume and its peak, amount of casualties and the possibility of designing the dimensions of structures. The HEC-HMS model is very capable of simulating runoff and hydrograph production.But hydrograph production requires calibration and sensitivity analysis of parameters that dependent on the model.
In this study, to simulate and calibrate the runoff precipitation of the Merck. HEC-HMS model and HEC-GeoHMS extension have been used in Arc GIS. To prepare CN from land use and soil hydrological map, And for rainfall data from 6-hour rainfall at Kermanshah station, and daily rainfall of Mahidasht station, and for flow data from Khersabad, hydrometric station was used. to simulate runoff from three flood events used from February 25 to March 1, 27-31 March 2016 and 8-15 November 2015. Then, to calculate rainfall losses and convert the surplus rainfall to flow, the American soil protection method (SCS) was used, Monthly fixed method was used for base flow, For the flow process of the method Muskingum was used. After the physical model was built in the HEC-GHMS tool, In the HEC-HMS4.11 simulator, the recall and data of three extreme flood events and the corresponding precipitations of meteorological stations were entered into the model as a time series and after execution, the calibration operation was performed. In this research, for calibration, K was first calculated from Clark's relationship and X coefficient, which remained unchanged for the defined model and the calibration was done by trial and error by other methods. In the trial and error method, by the constant change of CN and the links related to that limited time, concentration time and precipitation losses, the model was placed in the appropriate range. The initial value for the curve at the basin level was assumed to be 70, which reached between 62 and 70 from the optimal CN calibration at the basin level. After the optimization is done by trial and error method and the optimal value of CN, lag time and precipitation losses are obtained, To Optimizing of Muskingum's K and X coefficients, Differential Evolutuion was used.
In the calibration by trial and error method for the November 2015 flood event (NSE=0.665, PBias=26.62, RMSE=0.6, NRMSE=3.59) in March 2016 (NSE=0.430, PBias=29.78, RMSE=0.8, NRMSE=1.62) and in February 2020 (NSE=0.235, PBias=7.57, RMSE=0.9, NRMSE=2.13) were obtained. In optimizing the muskingum coefficients by differential evolution algorithm and a 50 replication stop, for the November 2015 flood event (NSE=0.871, PBias=25.52, RMSE=0.4, NRMSE=2.63) in the March 2016 event (NSE=0.731, PBias=28.82, RMSE=0.5, NRMSE=1.01) and in February 2020 (NSE=0.834, PBias=7.96, RMSE=0.4, NRMSE=0.95) were calculated. In calibration by the algorithm of differential evolution of the range of changes for parameter K, the positive and negative coefficients were 30% to 50% of the initial value and the range of changes of the X parameter was between 0.1 and 0.35. If the upper and lower limit is not defined for the K and X values, the algorithm will be able to adapt the most computational and observations, So that in the event of February 2020, statistical results (NSE=0.977, PBIAS=-4.33, RMSE=0.2, NRMSE=0.47) will be obtained.
In this study, precipitation simulation was performed by the HEC-HMS model and the use of the HEC GEOHMS plugin, by a different method, first calibration by trial and error, and then by differential evolution algorithm, the sensitivity analysis of the parameters involved in the simulation process was performed. By analyzing the sensitivity of flow trending parameters and obtaining optimal values of K and X, Muskingum showed the ability of differential evolution (DE) algorithm in calibration.
)