“…However, the success of the second approach depends critically on the choice of the model comparison metric. Often the comparison metric is purely a quantitative measure of the goodness of fit, such as the coefficient of determination, R 2 , calculated through linear regression between model output and observed data [e.g., Katul et al, 2000;van Wijk et al, 2000;Moriana et al, 2002;Misson et al, 2004]. However, like the transpiration models used here, semiempirical models are often nonlinear, and all of the models included in a comparison may not use the same predictor variables, therefore, the most appropriate way to calculate R 2 uniformly for all the models is not always clear [Healy, 1984;Kvålseth, 1985;Anderson-Sprecher, 1994;Mitchell, 1997].…”