Dynamic Global Vegetation Models

It has long been recognized that climate exerts a general control on the vegetation zones of the Earth (Woodward and Williams, 1987). An early approach to modeling this effect was analogous to bioclimatic modeling of individual species. Models that correlated vegetation types with certain climatic conditions were shown to reproduce global vegetation zonation with reasonable accuracy. Such correlative models could also be used to project future vegetation types using projections of future climates from GCMs.

More recently, correlative models have been largely replaced by mechanistic models using plant physiology to simulate vegetation patterns. These models are known as "dynamic global vegetation models'' (DGVMs). These models use equations describing basic plant physiological processes—such as photosynthesis and respiration—to determine net amounts of carbon available for plant growth and the allocation of that carbon (Prentice et al., 1992; Woodward and Beerling, 1997). The results of these calculations are expressed as "plant-functional types''—for example, "evergreen needleleaf forest'', "grass savanna'', or "deciduous broadleaf forest''. Plant-functional types in these applications are direct analogs to Box's (1981) "life forms''.

Numerous authors have contributed to the development of over half a dozen DGVMs that are being actively tested and refined. SDGVM (Woodward and Lomas, 2004), TRIFFID (Cox 2001), IBIS (Foley et al., 1996), LPJ (Sitch et al., 2003), VECODE (Brovkin et al., 1997), MC1 (Bachelet et al., 2001), and HYBRID (Friend et al., 1997) are examples. Widespread increases or decreases in forest cover projected in response to CO2 rise and climate change may therefore indirectly contribute to additional contributions to the regional and global climate changes through alterations to land surface properties. Furthermore, the net changes in terrestrial carbon stocks in the tropics and elsewhere may influence the rise in CO2 itself. Ecosystems may therefore have different impacts on climate change, both at the regional and global scale.

Cramer et al. (2001) present the results of a model intercomparison effort conducted with several of these major DGVMs. DGVMs have great utility in assessing the impacts of climate change, fire, and other environmental drivers on gross vegetation structure and physiognomy, especially at large scales. They are less useful for studies on individual species or biodiversity assessments at small scale. DGVMs may also be integrated into climate models to assess influences of vegetation on carbon cycles and global climatic change.

Responses to climate change and elevated CO2 modeled by DGVMs show broad similarities but also substantial inter-model variation. Cramer et al. (2001) used an ensemble of six DGVMs to make projections of global vegetation responses to transient climate change simulated with the HadCM2 GCM under the IS92a greenhouse gas and sulphate aerosol concentration scenario. The DGVMs generally overestimated the amount of tropical forest for current climates, especially in Africa but also in South America and Southeast Asia (Cramer et al., 2001). Tropical dry forest tends to be over-predicted in Southeast Asia and under-predicted in Africa. TRIFFID and HYBRID particularly over-predict tropical wet forest, while results from VECODE and SDGVM more closely approximate tropical moist and dry deciduous vegetation classifications derived from satellite images.

A particular feature of this study was projection for a major reduction in forest cover in the eastern half of Amazonia, due to significantly reduced precipitation and increased temperature. A drying of the Amazonian climate emerges in a number of GCMs, generally associated with an El Nino like pattern of global climate warming, but it is important to note that not all GCMs show this response. All six DGVMs showed a tendency toward reduction in forest cover due to drier conditions, and a drop in Amazonian biomass by 2100 (Figure 14.1; Cox et al., 2004). Variability in outcomes is influenced by assumptions about photosynthetic response and more efficient water use by the vegatation due to increased CO2 concentrations. While all models project a reduction in forest cover in northeastern Amazonia, considerable variability is evident in the models for other Amazonian outcomes.

DGVMs are rather difficult to test against independent data, in no small part due to their scales in time and space. The Foley et al. (1996) IBIS model has been tested globally against the flows of the major rivers of the world and regionally for the flows of the Amazon and its tributaries. The rationale is that the models compute evapotranspiration as one of its dynamic variables and water flow of a basin can be taken as the difference between rainfall and evapotranspiration when corrected for soil and ground-water storage. This regional- and global-scale DGVM testing is a healthy development particularly in that it uses data that are independent of model development and parameterization. Most DVGM model testing has been for consistency with overall patterns in the parameterization data or in the form of model comparisons (as opposed to independent data comparisons).

0 0

Post a comment