A simple automatic calibration routine for the HBV model

Abstract

A simple, but efficient, method for automatic calibration of the conceptual HBV rainfall-runoff model was developed. A new criterion, which combines the commonly used efficiency criterion R2 and the relative volume error was introduced. Optimising this combined criterion resulted in R2 values nearly as high as those for optimising only R2, but with much smaller volume errors. An earlier automatic calibration method for the HBV model relied on the use of different criteria for different parameters. With the simplification to one single criterion, the optimum search method could be made more efficient. The optimisation is made for one parameter at a time, while the others are kept constant. This one-dimensional optimisation is repeated in a loop for all parameters. A new loop is performed as long as there is a sufficiently large improvement since the last one. After each loop a search is made in the direction which is defined by the differences in parameter values between the two latest loops. The calibration routine was developed for, and tested with, the HBV model, but it should be general enough to be applicable to other models as well.