The Canadian Land Surface Scheme (CLASS) was tested by comparing its simulated soil watercontents with those measured from two potato sites in the Quebec City region for the growing seasonof 1993. Being a physically based model, most of CLASS input parameters can be either measured orobtained from the literature. However in the simulation of soil water contents, CLASS was found tobe sensitive to one particular soil texture empirical parameter b, which is used to define theunsaturated soil hydraulic conductivity and soil water retention functions. Therefore, someadjustment of this parameter was done. To improve the performance of CLASS for the entire season,a new evaporation model over bare soil was introduced and its performance was clearly superior tothat of the standard Philip's model. With the adjusted b values and the new evaporation model, thevalues of mean bias error (MBE) of soil water content (averaged for the two sites) were -0.09 byvolume for the depth of 0 to 0.45 m and -0.02 by volume for the depth of 0 to 0.90 m, respectively.The corresponding values of mean absolute error (MAE) were 1.39 and 0.96 by volume. Theseerrors were considerably less than those obtained with two other models (SWACROP andSUBSTOR)using the same data set. These models had MAE values of 2.4 and 2.8 by volume, respectively, forthe depth of 0 to 0.45 m. Considering the fact that the adjusted b values lie within the expected rangefor the soil types at the sites, we believe that physically based models such as CLASS have promisingpotential for agricultural and meteorological applications.
展开▼
