PARAMETER ESTIMATION AND VALIDATION OF THE PROPOSED SWAT BASED RAINFALL-RUNOFF MODEL – METHODS AND Outcomes
N. Milivojević, Z. Simić, A. Orlić, V. Milivojević, B. Stojanović
Abstract
Calibration of model parameters is a process used to set the parameters values with which the
model produces the lowest deviation relative to relevant observed values. Calibration procedure
includes identification of acceptable parameter ranges, sensitivity analysis related to parameter
change (i.e. identification of the method and intensity of a parameter change impact on certain
model results) and, finally, a simultaneous variation of parameters incorporating their mutual
impact. The final step of the calibration procedure is performed automatically by the means of a
computer. Calibration quality is defined by virtue of objective functions used to make the
calibration process converge to the optimum solution, i.e. to the minimum deviation from the
observed hydrograph. Graphical comparison of computed and real elements of the hydrological
system is also useful during the calibration process. In this paper the sensitivity analysis and
reliability analysis of the SWAT model input data have been used to select the groups of the
most important parameters being calibrated, as well as parameters within the groups with their
possible ranges of values. The procedure and algorithm of model parameters calibration by the
parallel genetic algorithm (PGA) have been also described. Calibration of the SWAT model
parameters was performed for many sub-catchments of the River Drina catchment. The results
of the calibration of SWAT model parameters have been presented for the relevant subcatchment.