Influência da seleção das funções objetivo na calibração do modelo hidrológico distribuído MGB-IPH

Abstract

In the automatic multiobjective calibration of hydrological models, the parameters of the model are adjusted within the search space, according to two or more objectives, and the choice of functions is an important step. This type of calibration usually leads to better model performance, but it is still not well understood what criteria should be used to select objective functions (OF). In this sense, the main objective of this study was to evaluate the influence of different OFs on the automatic multiobjective calibration of the distributed hydrological model MGB-IPH and the approach of alternative methods of calibration applied in the semiarid region.The application of the model covers part of the hydrographic basin of the Piancó river with an area of 4603.39 km². The MGB-IPH model was run with step time daily, with calibration in the period from Jan/01/1970 to Dec/31/1990 and validation Jan/01/1991 to Dec/31/2011. First, a sensitivity analysis of 36 OF was performed in ten tests using series of synthetic flows. Subsequently, 111 calibrations were performed with different OF combinations. Were also applied two alternative calibration methods: the inlet hydrograph fraction calibration method and by period of year calibration method (wet and dry). The results show that it was possible to identify in which fractions of the hydrograph the OF are more sensitive through the sensitivity tests. With the analysis of multiobjective calibrations with different OF, it was observed that the MGB-IPH model satisfactorily simulated the flow regime observed in most calibrations, with a relatively good adjustment of the hydrographs for the twelve best NS calibrations between 0.70 to 0.75, KGE between 0.76 and 0.84, VE between 0.39 and 0.44, among other measures of performance. The alternative automatic calibration methods demonstrated relatively good performance when compared to the traditional calibration method with satisfactory values of the performance measures used. It was concluded that the automatic multiobjective calibration demonstrated sensitivity regarding the choice of OF in the calibration results. The best sets of OF were represented by the calibrations CALIB 92 (EGK - NSDlog - VE) and CALIB 111 (NSD - NSlog - FCPerm).