Comparison of four Sensitivity Analysis Methods of HBV Conceptual Model Parameters in Karkheh Basin and its Sub-basins

The HBV (Hydrologiska Byråns Vattenbalansavedlning) is a conceptual model widely used for hydrological forecasting and water resource studies. In this study, sensitivity analysis of parameters of the HBV model is investigated for Karkhe basin and its sub-basins for four different periods 1, 5, 10 and 25 years with four methods including FAST (Fourier Amplitude Sensitivity Test), RSA (Regional Sensitivity Analysis), Sobol and regression. After determining the most sensitive parameters, the model is calibrated using Nondominated Sorting Genetic Algorithm (NSGA) method. In all statistical periods, one year has been used for warm-up to eliminate the effects of initial conditions. In this study, the MOUSE Toolbox is used to analyze the sensitivity of the HBV model parameters. This software is based on Java programming language. To analyze the sensitivity of the HBV model parameters based on the Monte Carlo sampling method and the Halton sequence method for each of the samples (time periods) in each sub-basin separately, 1000 samples are taken for the set of input parameters with a specified range for each parameter taken. Objective functions for evaluating performance of model are NSE, RMSE, RSR and BIAS. The results of sensitivity analysis of the parameters show that Sobol and RSA are more reliable methods because of variability in time intervals and different sub-basins. Fast and regression methods in the Karkheh basin and its sub-basins for different time periods show similar results that considering the change in hydroclimate conditions in this basin, isn't practical and the results of these methods can not be used for investigating sensitivity of parameters and their identification in the studied basin. The most sensitive parameters of HBV model for Karkheh basin and its sub-basins in soil routine is maximum soil moisture content (Fcap) and in the response routine is the storage of soil surface moisture content (hl1). These parameters have shown the most sensitive factor in minimum fluxes. The snow routine parameters, especially the threshold temperature for ice freezing (ttlim), are sensitive in the sub-basins of Ghare Sou and Kashkan in short periods (1 and 5 years). For a specific sub-basin, the sensitivity of the parameters in different time periods is not completely stable and a little variability has been observed in different periods. But the most sensitive parameters (hl1 and fcap) have maintained their sustainability almost in all periods. Parameters of response and soil routines are more sensitive to the parameters of snow and routing routines. The results of the interaction between the parameters using the Sobol method in different sub-basins indicate that the strongest interactions are between the soil routine parameters, especially Fcap, with the response routine parameters and also the response routine parameters with each other. The time variability of parameters indicates that the soil routine and response parameters in the minimum discharge show the most sensitivity. Other parameters are more sensitive in the dry season of the basin (summer and autumn). The HBV model has the ability to simulate runoff in the Karkhe basin and its sub-basins with high precision. This study shows that selection of shorter period of calibration gives better simulation results. For one year's period the best NSE, RSR and RMSE are in Gamasyab sub-basin respectively 0.95, 0.21 and 1.4 and the best BIAS is in Kashkan sub-basin and Karkhe basin with 0.13.