In the companion to the present paper, the soil model associated with CLASS (Canadian Land Surface Scheme) was outlined. In this paper, the accompanying vegetation model is described. This model includes physically based treatment of energy and moisture fluxes from the canopy as well as radiation and precipitation cascades through it, and incorporates explicit thermal separation of the vegetation from the underlying ground. Seasonal variations of canopy parameters are accounted for. The morphological characteristics of the ‘composite canopy’ associated with each grid square are calculated as weighted averages over the vegetation types present. Each grid square is divided into a maximum of four separate subareas: bare soil, snow-covered, vegetation-covered, and snow-and-vegetation covered.
Test runs were done in coupled mode with the Canadian Climate Centre GCM, to evaluate the performance of CLASS compared with that of the simpler land surface scheme previously used. Two versions of CLASS were run: one with ponded surface water saved between time steps, and one with it discarded. For the seasons of June—July—August and December—January—February, diagnostic calculations showed that the old scheme underestimated the globally averaged land surface screen temperature by as much as 3.0°C, and overestimated the globally averaged precipitation rate over land by up to 1.0 mm day−1. CLASS, on the other hand, produced screen temperature anomalies, varying in sign, of 0.2–0.3°C, and positive precipitation anomalies of 0.6–0.7 mm day−1. The relatively poor performance of the old model was attributed to its neglect of vegetation stomatal resistance, its assumption that the contents of the soil moisture ‘bucket’ had to be completely frozen before the surface temperature could fall below 0°C, and its use of the force-restore method for soil temperatures, which systematically neglects long-term thermal forcing from the soil substrate. The assumption made in most GCMs that excess surface water immediately becomes overland runoff is shown to result in substantial overestimates of surface screen temperatures in continental interiors.