rj
\(\overline{f(r)}\) is mean of all observed values.
Different interpretations are drawn from the above formulated
performance measures. Coefficient of determination is adopted very
commonly as a performance measuring criterion. Li and Heap (2008) has
discussed that R2 solely cannot represent the
performance of any model; hence other measures are also used to compare
the performances of models used in this study. As per Hallak and Pereira
(2011), MBE does not provide a clear picture about the individual
errors; this is because the positive and negative error gets cancelled
out. To overcome this MAE is calculated, Fox (1981) introduced MSE is
another performance measure which is used to check the accuracy of
models to predict, but according to Ponce-Hernandez (2006) MAE shows the
error on large scale because the errors are squared and errors get
amplified. RMSE is another parameter which is commonly used for
measuring the performance of models. Willmot (1982) found that RMSE is
one of the best tools to measure the error of a model as it gives us a
summary about the average difference of values between observed ones and
predicted ones. Model efficiency (ME) is also taken into consideration
for evaluation of the models. According to Nash and Sutcliffe (1970) ME
can take any value in the range of (-∞, 1] and best model is the one
which is closer to unity. For the model having ME as zero, it can be
inferred that the predicted values are just the mean of the observed
values in search radius. For ME less than zero it should be inferred
that mean of observed values are better estimates than the predicted one
(Krause et al., 2005).