Comparison of surface water balance performance using two reservoir models in the Karun -4 watershed

Water balance looks like a simple function, but its estimation at the watershed level is a complex process. For this purpose, water balance models could be a useful tool for simulating water balance components. The aim of this study was to simulate the surface water balance of the Karun 4 watershed using a model with minimum input data, because one of the problems in mountainous areas is lack of data. Many water balance models have been used in Karun watershed. Therefore, the Exp-Hydro model, which has not been used in Iran so far and has a simple structure and minimum input data, was selected. Exp-Hydro is one of the watershed-scale daily hydrological models that make use of daily data including precipitation, air temperature and potential evapotranspiration as input and simulates the daily streamflow at the watershed outlet. Due to its simple structure in this study, the open-source model, written in Python was used to simulate the daily runoff of the Karun 4 watershed.

In this study, the spatially lumped version of the Exp-Hydro model was used to simulate daily runoff. The script code of the model was written in Python. Python software, NumPy, SciPy, and Matplotlib library functions were installed to run this model. The input data of the model were prepared in separate files in text format. Then the results of Exp-Hydro model were compared with a daily water balance model that was developed in this study. This model was prepared in Excel software. The statistical period of this study was from 2000 to 2020, where two-thirds of the data were used for the calibration period and one-third for the validation period. The WHAT software and recursive digital filter method were used to separate the baseflow from the observed daily flow data. The Exp-Hydro model was calibrated automatically using PSO optimization algorithm and the daily water balance model that developed in this study was calibrated using Goal Seek method. The efficiency of the models were evaluated in this study by means of Kling Gupta criteria (KGE), Nash-Sutcliffe efficiency coefficient (NSE), Correlation coefficient (R), Root Mean Square Error (RMSE) and Mean Absolute Error (MAE).

In the separation process of the base flow values from the total flow of Armand hydrometric station, was done in such a way that in the minimum flows, the separated flow rate corresponds to the minimum values, which in this case, the values of the two constant parameters of the filter parameter and base flow index (ratio of baseflow to the total flow) was chosen as 0.9 and 0.36, respectively. Evaluation of the developed daily water balance model's performance using the KGE, NSE, R, RMSE and MAE showed that these objective functions were 0.76, 0.69, 0.83, 0.25, and 0.11 respectively during the calibration and 0.76, 0.62, 0.85, 0.19, and 0.09 respectively during the validation period. This means the developed daily water balance model in this study has produced good and satisfactory outputs. Also, the evaluation of Exp-Hydro model's performance using the KGE, NSE, R, RMSE and MAE showed that these objective functions were 0.69, 0.37, 0.69, 0.35, and 0.14 respectively during the calibration and 0.6, 0.19, 0.71, 0.28, and 0.11 respectively during the validation period. This means the developed daily water balance model in this study has produced good and satisfactory outputs. This means the Exp-Hydro model has an intermediate performance according to the KGE coefficient in both calibration and validation periods. In addition, the developed daily water balance model in this study estimated the low flows with higher accuracy and both models had underestimated regarding flood flows. Sensitivity analysis of Exp-Hydro model was done manually by changing each parameter of the model and comparing the changes between observation and simulation runoff hydrographs, and the value of efficiency coefficients was also controlled. In this analysis, it was found that since f is the parameter that controls the reduction of runoff depending on the reserve, this parameter has an inverse relationship with simulation runoff value; So that by decreasing the f value, the simulation runoff values increase and become more than the observation runoff values. Also, this parameter was the most sensitive parameter in the Exp-Hydro model. In the developed daily water balance model in this study, m in calculating maximum runoff generated in the catchment bucket using the Reservoir performance exponential function and e in calculating snow melt were the most sensitive parameters. These parameters were the most sensitive parameter in this model; so that by reducing their values, the simulated runoff values of the model decrease and become lower than the observed runoff values. In general, the performance of the developed daily water balance model in this study was better than Exp-Hydro model.