optimization of the hydrological tank model by downhill simplex method

The hydrological tank model was used in this study to simulate the rainfall-runoff relationship in the flint river basin near carsonville, georgia, usa. the model’s structure consisted of two tanks (upper and lower), and it depended primarily on a set of eight parameters to simulate daily stream flow components (surface, subsurface, and base flows) from daily precipitation while taking into account the effects of daily evapotranspiration, the water storage height, and the infiltration rate. the parameter set was calibrated, optimized, and validated with data spanning a period of 10 years (oct. 1st, 2007–sep. 30th, 2017). model calibration and validation were performed by using the first and second 5-year periods of water flow data, respectively. the model performance was verified using the coefficient of determination (r2) and the nash–sutcliffe efficiency (nse) coefficient. the results of one- and two-parameter sensitivity analyses showed that the model performance was highly dependent on the parameters ko and k1 (the coefficients of the surface and subsurface flows, respectively). the downhill simplex method (dsm) was used to find the optimum parameter set via the maximization of r2 and nse. the optimum parameter set produced by maximizing the nse values was adopted in this study because it produced a very good correlation between the simulated and observed stream flows for both the calibration and validation time periods. using the adopted optimum parameter set, the maximum values of r2 and nse were 0.818 and 0.812 in the calibration time period, respectively, and 0.756 and 0.750 in the validation time period, respectively. moreover, the dsm has proven to be very reliable in locating global optimums and requires reasonable computational efforts in the calibration of a hydrological model.