Incorporating Likelihood information into Multi-objective Calibration of Conceptual Rainfall-Runoff Models

In this paper, the incorporation of parametric likelihood information into NSGAII algorithm has been attempted in order to preserve solutions with more overall likelihood. The crowded comparison operator in NSGA-II, which is used to select the potential solutions for the next generation, is substituted by likelihood comparison operator which includes the consideration of the likelihood information about the potential solutions rather than their distance from each other. As a result the potential solution with higher overall likelihood measure has more chance to be selected in the next generation. Three different scenarios for the estimation of overall likelihood measure are presented. The modified algorithm is used for calibration of two different conceptual rainfall-runoff models in 24 USA MOPEX catchments. The results show that the new modification results to different searching process which can be compared with NSGA-II from different perspectives.