Climate envelope models have been used since at least the 1980s (e.g., Box,
1981; Busby, 1988). Usually they are based on describing the climate or environment
encompassing the current distribution of a species or ecosystem (the environmental
envelope or climatic envelope), then mapping the location of this same envelope
under a climate change scenario. Sometimes, other forms of bioclimatic correlations
or categorization systems are used, but they are based on the same assumption
of a close correlation between climate and ecosystems associated with it. These
models are useful as a first screening device to point to potential significant
changes in ecosystem composition resulting from climate change (e.g., Huntley,
1995, and Huntley et al., 1995 for European plants and birds; Somaratne and
Dhanapala, 1996, for Sri Lankan plants; Brereton et al., 1995, for Australia
mammals; Gignac et al., 1998, for Canadian peatlands; Eeley et al., 1999, for
South African woody plants; Iverson and Prasad, 1998, Iverson et al., 1999,
for eastern U.S. trees).
A basic problem with these models is that every species has a "potential niche"that is, a location in which it could survive and reproduce under the climatic conditions (Kirschbaum et al., 1996). What is observed and used as the basis for the model, however, is the "realized niche," which is a more limited area in which the species is found given the effects of competitors, predators, and diseases. If climate change results in species remaining within their potential niche and the competitors and predators change, the species may be able to survive in situ. In some models, some locations are predicted to have climates that are not encompassed by any of the climate envelopes of any of the vegetation types (Lenihan and Neilson, 1995). Used cautiously, however, the approach can point to priority areas for further study. For example, Lassiter et al. (2000) analyzed data for 200 woody plant species in the eastern United States and found the usual result: a northward shift in predicted range. Excluding water limitations (i.e., allowing increased WUE under elevated CO2 to compensate for a drier climate), however, most species survive in their current locations. This implies the need to better understand water-use and water-availability relationships under climate change. If species are able to persist, there may be little spread into new regions and thus little vegetation change for many decades.
In a study of 80 species of trees in the United States, using 100,000 plots
to describe current distribution and importance, nearly half of the species
assessed showed the potential for ecological optima to shift at least 100 km
to the north (Iverson et al., 1999). Whether these species will be able to achieve
these potential distributions will depend on their migration rates through fragmented
landscapes. When Iverson et al. (1999) incorporated a migration model, they
found severely limited migration in regions of high forest fragmentation, particularly
when the species is low in abundance near the range boundary.
Kirilenko and Solomon (1998) also have developed a bioclimate correlative model
that incorporates a migration component. They demonstrate the importance of
incorporating migration by showing that in simulations in which tree migration
is delayed, the estimated global terrestrial carbon stock decreases by 7-34
Gt C, in contrast to an increase in carbon stock projected under nonlimiting
migration (Solomon and Kirilenko, 1997).
Climate envelopes continue to be a useful tool in identifying potential changes in species distributions but are limited by the foregoing problems. emphasis is shifting to more mechanistically based dynamic models of vegetation change that are linked directly to transient outputs from GCMs.
Since the SAR, there have been significant changes in vegetation models and
GCMs. A major modeling comparison projectVegetation/Ecosystem Modeling
and Analysis Project (VEMAP) compared six equilibrium models of vegetation
distribution and biogeochemistry (VEMAP Members, 1995). In the IPCC Special
Report on the Regional Impacts of Climate Change, Neilson et al. (1998) compared
two leading vegetation models (MAPSS and BIOME3) run against transient outputs
from recent GCMs.
In MAPPS and BIOME3, potential vegetation is simulated by first calculating
carbon flux (plant growth) on the basis of climate and hydrology information
derived from the GCM. The models calculate a leaf area index that represents
the capacity of the site to support plant canopy. Physiologically based rules
are then used to classify the site into a vegetation type (e.g., forest, shrub,
or grassland; evergreen or deciduous), leading to allocation to one of 45 vegetation
types in MAPSS or 18 in BIOME3. These vegetation models are equilibrium models
in that they calculate which vegetation type might be most suited to the climate
and do not consider how the existing vegetation type might change to that new
type. Details of the model outputs and maps are presented in IPCC (1998) and
Neilson et al. (1998).
Neilson et al. (1998) compared the more recent combination of vegetation and
climate models with those being used in the SAR. They found that the newer,
transient GCMs produce a cooler climate than earlier versions, mostly because
lag effects are simulated and the temperature does not increase to its equilibrium
value within the model runs. This means that changes in predicted vegetation
tend to be less than those reported in the SAR. Nevertheless, there are significant
poleward shifts in cold-limited vegetation types. Vegetation types that are
limited by water availability showed more complex changes, depending on the
balance between precipitation change, hydrological balance, and physiological
adjustment under higher CO2.
The newer modeling combinations continue to predict significant potential changes
in the distribution of most ecosystem types and increases in the area of tropical
and temperate forests, but it cannot be determined whether this potential will
be observed until transient (or dynamic) global vegetation models are developed.
These models simulate the change in abundance of important species or "functional
groups" of species on a year-by-year (or similar) basis in response to
the output of the GCM (Cramer et al., 1999).
Such models are being developed and used for assessments of overall carbon
storage potential of the land biosphere (Cramer et al., 2001), but at this stage
it is too early to place much reliance on the outputs for specific biomes or
ecosystems. The results show the sensitivity of ecosystems to the treatment
of water use and especially the balance between changes in water availability
resulting from climate change (often decreased availability in a warmer climate)
and response to higher CO2 concentrations in the atmosphere (often
increased water-use efficiency). This means that model output can vary significantly,
depending on the GCM used, because these models have tended to produce different
interannual variability in precipitation and thus water availability. Other
challenges are to simulate loss of vegetation from disturbances such as fire,
blowdown, or pest attacks and migration of species or groups of species to new
Van Minnen et al. (2000) have shown that by modifying the IMAGE2 model to include unlimited migration, limited migration and no migration result in significantly different patterns of vegetation change, especially in high-latitude regions.
Other reports in this collection