Assessment of three dimensionless measures of model performance. (November 2015)

• Record Type: Journal Article
• Authors: Willmott, Cort J.; Robeson, Scott M.; Matsuura, Kenji; Ficklin, Darren L.
• Is Part Of: Environmental modelling & software. Volume 73(2015:Nov.)
• Page Start: 167
• Page End: 174
• Publication Date: 2015-11
• Subjects: Dimensionless accuracy indices -- Model-performance statistics -- Streamflow model evaluation; Environmental monitoring -- Computer programs -- Periodicals; Ecology -- Computer simulation -- Periodicals; Digital computer simulation -- Periodicals; Computer software -- Periodicals; Environmental Monitoring -- Periodicals; Computer Simulation -- Periodicals; Environnement -- Surveillance -- Logiciels -- Périodiques; Écologie -- Simulation, Méthodes de -- Périodiques; Simulation par ordinateur -- Périodiques; Logiciels -- Périodiques; Computer software; Digital computer simulation; Ecology -- Computer simulation; Environmental monitoring -- Computer programs; Periodicals; Electronic journals; 363.70015118
• DOI: 10.1016/j.envsoft.2015.08.012
• Languages: English
• ISSNs: 1364-8152

Abstract

Abstract: Pertinent characteristics of three dimensionless and comparable model-performance or model-efficiency measures are examined and compared. Our model-assessment recommendations apply to many types of environmental models. Representing measures based on sums-of-squared errors or "quadratic measures" (by far, the most widely used class) is Nash and Sutcliffe's (1970) well-known efficiency measure ( E ). Our assessments of E, by and large, also apply to similar-in-form benchmark measures (Seibert, 2001). Legates and McCabe's (1999) version of E ( E 1 ) and Willmott et al.'s (2012) refined index of agreement ( d r ) represent the less-often-employed but more interpretable class of measures based on sums of error magnitudes. Conceptual and algebraic arguments are used in conjunction with assessments of many parameter sets of the Soil and Water Assessment Tool (SWAT) hydrologic model, over the period 1950–2005. Our findings suggest that, of the three measures, d r has the broadest utility, followed in order by E 1 and E . Highlights: Properties of a refined index of agreement ( d r ) are characterized and explained. Two other widely used indices ( E 1 and E ) are compared with the refined index. Our findings indicate that, of the three indices, d r has the broadest utility. Parameterizations of a streamflow model are used for evaluation and comparison.