Section 126.96.36.199 raises the question of whether organisms will need to migrate under climate change or whether many species will be able to survive in situ under new climatic conditions (Woodward and Beerling, 1997). Some species do occupy sites that are on the limits of their physiological tolerance, and if climate change takes local climate beyond that threshold, clearly they will not be able to persist at that site (see Section 5.6 for examples). However, there is mounting paleoecological evidence of vegetation types persisting through significant climate changes.
Lavoie and Payette (1996) conclude that the stability of the black spruce (Picea mariana [Mill.] BSP.) forest boundary during warm and cold periods of the late Holocene [warm approximately 2,000 years before present (BP), medieval times, and this century; cold approximately 3,000 years BP, 1,300 years BP, and the Little Ice Age] demonstrates that mechanisms that allow forest boundaries to advance or retreat are not easily triggered by climatic change. The black spruce old growth forest persisted for more than 1,500 years through many variations in climate probably because of the buffering effect of the trees on the local environment (Arseneault and Payette, 1997). Only a fire in 1568 AD broke this forest influence on microclimate and local growth conditions and caused the forest vegetation to shift to krummholz. Such studies emphasize the importance of effectively incorporating sufficient biophysical detail to capture climate-ameliorating effects and realistic disturbance regimes.
No projection of the future state of the earth's ecosystems can be made
without taking into account past, present, and future human land-use patterns.
Human land-use will endanger some ecosystems, enhance the survival of others,
and greatly affect the ability of organisms to adapt to climate change via migration.
Leemans (1999) used IMAGE2 to forecast possible global shifts in vegetation types and land-use change. Using a mid-level scenario of human responses to global change, he forecasts that nondomesticated land (his proxy for biodiversity) will decrease from 71 to 62% of the land area between 2000 and 2025, then remain approximately stable until the end of the 21st century. Losses in Africa and parts of Asia from 2000 to 2025 may be as much as 20-30% of remaining nondomesticated land.
It is clear that ecosystem change models still contain assumptions that are
not fully tested, and most models inevitably work better for the geographical
regions and time periods for which they were constructed (Hurtt et al., 1998).
This derives from a mixture of fitting parameters to available data and deliberate
and subconscious bias in selecting processes to include in a model. However,
testing of ecosystem response models is gradually improving. Bugmann and Solomon
(1995) tested the behavior of a model developed for European ecosystems by running
it for a comparable North American site. The results were broadly in agreement,
with useful indicators about where the model could be improved.
Another test is to compare different approaches to modeling for a particular
purpose, as Yates et al. (2000) did for correlative models such as the
Holdridge climatic correlation model and more mechanistic models of vegetation
distribution. The strengths, weaknesses, and appropriate areas of application
can be determined. Another approach is to conduct sensitivity studies in which
the sensitivity of model outputs to changes in the input data and assumptions
is evaluated. Hallgren and Pitman (2000) have carried out a sensitivity analysis
of BIOME3 (Haxeltine and Prentice, 1996); they conclude that parameters that
affect photosynthesis, water use, and NPP change the competitive interactions
between specific plant groups (e.g., C3 versus C4
plants). Numerous studies have found that the outputs of ecosystem response
models are sensitive to the precise treatment of water availability and water
use by vegetation (Gao and Yu, 1998; Churkina et al., 1999; Hallgren
and Pitman, 2000; Lassiter et al., 2000). This points to areas in which
further development is needed.
Beerling et al. (1997) used a climate change experiment (CLIMEX) that
exposed an entire catchment of boreal vegetation to elevated CO2
and temperature for 3 years to test their Sheffield dynamic global vegetation
model. There generally was a good match between observations and predictions,
but longer runs of such experiments are needed to test such models thoroughly.
As paleoecological data sets improve, there is increasing opportunity to test
models against these data. However, this is a multiple test because the validity
of paleoclimate models themselves also is under test. For example, Kohfeld and
Harrison (2000) tested the ability of GCMs to describe changes in data collected
for the environments of the last glacial maximum (21,000 years BP) and/or mid-Holocene
(6,000 years BP). They conclude that better land-surface (including vegetation)
response models are needed to capture the detail observed in the paleoecological
In summary, recent studies show that the potential for significant disruption of ecosystems under climate change remains. Further development of simple correlative models that were available at the time of the SAR points to areas where ecosystem disruption and the potential for ecosystem migration are high. Observational data and newer dynamic vegetation models linked to transient climate models are refining the projections. However, the precise outcomes depend on processes that are too subtle to be fully captured by current models.
Other reports in this collection