MAPSS and BIOME3 produce similar vegetation maps under current climate, but there are some differences. Some of the discrepancies between MAPSS and BIOME3 under current climate (Figure C-1) are due to questions of classification, especially in the drier types. For example, the Sahel region in Africa is labeled as 'shrub-steppe' in the original BIOME3 classification, but as various grassland types in MAPSS. The MAPSS grassland types do allow some shrubs, but the shrub density is usually reduced by the fire model, which assumes that there has been no reduction in fuel due to grazing. Were such land-use constraints included, the two models would be in better agreement on the classification. This classification difference between the models occurs over many of the drier parts of the world.
The models each appear to be better calibrated to their 'home' continents than either is to other continents. For example, MAPSS over-estimates the distribution of Temperate Evergreen Forests (conifers) in western Europe; while, BIOME3 overestimates the distribution of Temperate Mixed Forests (broadleaf) in western North America (Table C-1); yet, the two models are generally in agreement on the amount and location of temperate forests.
One area of significant departure from observed vegetation is the Pampas of southern South America. Both models simulate forests where grasslands are generally predominant. Various hypotheses for this discrepancy include unique soils, fire disturbance, rainfall seasonality and interannual variability of rainfall (VEMAP Members, 1995; Neilson, 1995; Neilson and Marks, 1994) and represent a focus for future research. Other local to regional errors in the MAPSS and BIOME3 classifications will be apparent to the knowledgeable reader. Reasons for these errors are many, but include 1) possible errors in the interpolated climate, 2) grazing, harvest, fire and other disturbances, and 3) missing or weak representation of some processes in the models. Globally, both models are reasonably accurate and are generally considered to be more accurate under altered climates than previous, empirical approaches (VEMAP Members, 1995). Empirical approaches cannot simultaneously simulate changes in vegetation distribution and changes in vegetation density and hydrology. Nor can they examine the sensitivity of the system to altered CO2 concentrations. However, as the focus shifts to ever smaller regions or locales, the model uncertainty and the likelihood of error increases.
Both MAPSS and BIOME3 produce large shifts of cold-limited vegetation boundaries into higher latitudes and elevations. However, the water-controlled boundaries may exhibit any direction of change, depending on the interaction of several variables including among others, the relative changes in temperature and precipitation, and whether or not the direct effects of CO2 have been incorporated.
The older GCM scenarios tend to be hotter than the newer ones and produce a more dramatic change in vegetation distribution. The MAPSS results under GFDL-R30 serve to illustrate one of the older simulations. MAPSS and BIOME2 (Haxeltine et al., 1996) were similar over the U.S. under this scenario (VEMAP Members, 1995), if both incorporated the direct CO2 effects. MAPSS was far more xeric in response than BIOME2 without the direct CO2 effects (ibid).
The Tundra decreases by as much as 1/3 to 2/3 of its present size, as does the Taiga/Tundra, under all scenarios and with both ecological models (Table C-1, Figures C-2, C-3, C-4 and C-5). The boreal forest expands in size under all scenarios ranging from 108% to 133% of its present size (MAPSS only). Since BIOME3 includes the Taiga/Tundra, which contracts under all warming scenarios (MAPSS simulations), with the Boreal Conifer Forest, which expands under all scenarios (MAPSS simulations), the net change simulated by BIOME3 usually indicates a loss of Boreal Forest. However, the aggregation of the two types in BIOME3 hides the observation that the two vegetation types (as defined above) tend to change in opposite sign with respect to area, i.e., Taiga/Tundra decreasing, Boreal Conifer increasing. The two models are quite consistent in the simulated response of the combined biomes (Table C-1). Temperate forests (inclusive of both types) increase in area (107% to 158%). Tropical forests could either expand or contract, largely dependent on the inclusion of the direct CO2 effect, but also dependent on the severity of the scenario. Savanna/woodlands expand or contract, depending on whether or not they are encroached upon by neighboring forests or semi-arid lands, again reflecting whether or not direct CO2 effects are considered and on the scenario. BIOME3 shows a competitive displacement of tropical savannas by neighboring forests, due to the superior competitive ability in the model of the C3 trees over the C4 grasses under elevated CO2. The total area of grasslands and shrublands in these simulations remains largely unchanged or expands by as much as 27%, depending on the CO2 effect and the scenario. If the direct effects of CO2 are included, arid lands tend to contract in all scenarios, shifting to less arid types (Table C-1, Figures C-2, C-3, C-4 and C-5). Without the direct CO2, arid lands either expand or contract in area, depending on the climate scenario and the ecological model.
Other reports in this collection