“…For instance, the model output or response surface is the product of the interplay between model structure, parameters, objective function, data information content, and modeler's decisions. Objective functions characterize the model performance as an aggregated measure of the matching between modeled and observed; either as metrics of model residuals (Bennett et al, 2013;Davtalab et al, 2017;Fowler et al, 2018;Murphy, 1988) or as signatures of similarity (Addor et al, 2018;Fowler et al, 2016;Gupta et al, 2008;Kelleher et al, 2017;Pfannerstill et al, 2014;Sawicz et al, 2014;Schaefli, 2016;Yilmaz et al, 2008); whether a scalar metric/variable (single criterion) or a vector of metrics/variables (i.e., multiple criteria/multivariable; Efstratiadis & Koutsoyiannis, 2010;Gupta et al, 1998;Stisen et al, 2018); and whether aggregated or distributed (Koch et al, 2016;Koch et al, 2017). Performance metrics reduce the complex behavior of a systemoften the integrated response of the catchment system, that is, discharge-from a higher dimension (e.g., a time series) to a single, or a few, point values; thus information loss is inevitable (Gong et al, 2013;Gupta & Nearing, 2014;Nearing & Gupta, 2015).…”