Coupled atmosphere-ocean-land-sea ice climate models (AOGCMs) are often tuned using physical variables like temperature and precipitation with the goal of minimizing properties such as the root-mean-square error. As the community moves towards modeling the earth system, it is important to note that not all biases have equivalent impacts on biology. Bioclimatic classification systems provide means of filtering model errors so as to bring out those impacts that may be particularly important for the terrestrial biosphere. We examine one such diagnostic, the classic system of K¿ppen, and show that it can provide an "early warning" of which model biases are likely to produce serious biases in the land biosphere. Moreover, it provides a rough evaluation criterion for the performance of dynamic vegetation models. State-of-the art AOGCMs fail to capture the correct K¿ppen zone in about 20--30% of the land area excluding Antarctica, and misassign a similar fraction to the wrong subzone.