Comparisons
of model performances was done by comparing values of root mean square errors
(RMSE) from cross validation for the two numerical models and regression analysis
for the linear regression model. It follows the idea that error occur where
prediction deviates from the observation for each of the 42 points. The value
of RMSE provides a reliable and quantifiable way of comparison between models.
The smaller the value, the better it is the model predictive power is. However,
it has to be assumed that predictive power of each model is not limited to the
data points available in the dataset but to the entire surface (i.e. those grid
cells between the data points). If this assumption does not hold true, the comparison
of RMSE values will not mean much.