Canadian Forest Service Publications

Large scale modelling of Canada’s forest ecosystem responses to climate change. Final report to Government of Canada: Climate Change Impacts and Adaption Program, Project A636, May 2006. 2006. Price, D.T.; Scott, D. CCIAP - Climate Change Impacts and Adaptation Program, Environment Canada, Ottawa, Ontario. 53 p.

Year: 2006

Issued by: Northern Forestry Centre

Catalog ID: 35660

Language: English

CFS Availability: PDF (download)

Mark record


Simulating the effects of a changing climate on North American forests.

The VINCERA ("Vulnerability and Impacts of North American Forests to Climate Change: Ecosystem Responses and Adaptation") project involved vegetationmodelling groups from Canada, the UK and the USA, each using a different Dynamic Global Vegetation Model (DGVM). The three DGVMs and their respective team leads were: IBIS - David Price of CFS, Northern Forestry Centre in Edmonton; SDGVM - Ian Woodward at University of Sheffield; and MC1 - Ron Neilson at Oregon State University. Each team ran their respective DGVM with a consistent data set for historical climate and six climate change scenarios that were developed by CFS (Price and McKenney). The suite of six climate change scenarios used three different Global Climate Models (CGCM2, HadCM3, CSIRO Mk2) and two Inter-governmental Panel on Climate Change (IPCC) emission scenarios (SRES A2 and B2) in order to reflect uncertainty in future climate conditions. The combinations of DGVMs and climate change scenarios were used to investigate the sensitivity of North American forest ecosystems to projected changes in climate, through key ecosystem parameters that included: Net Primary Productivity (NPP), Net Biome Productivity (NBP), total vegetation biomass and soil carbon, dominant vegetation type, and area burned.

In attempts to include IBIS, Price's group encountered several problems which prevented them from obtaining acceptable results for present-day conditions. It was therefore unreasonable to use IBIS to project future responses. We continue to work to resolve these problems and we hope to have a complete set of IBIS results for North America available in the near future. Recent development work on IBIS is documented in Appendix I. In the mean time this report focuses on results obtained from the SDGVM and MC1 models, though some earlier results obtained from IBIS for southern Canada are presented for comparison.

Major findings

Applied to North America, the two dynamic vegetation models, MC1 and SDGVM, were remarkably similar in their simulation of year-to-year variations in 20th century productivity including net primary production (NPP) and NBP (= NPP minus losses due to disturbances and decomposition of dead material). Closer inspection showed that although these variations were strongly correlated, the NBP (i.e., net ecosystem carbon exchange) simulated by SDGVM was significantly higher (and generally positive) with most of this additional uptake accumulating in the soil C pools. On the other hand, MC1 maintained average NBP much closer to zero, with more carbon accumulating in vegetation biomass (such that simulated total ecosystem carbon densities were rather similar between the two models).

For the 21st century, the two models diverged greatly in their responses to the suite of forcing climate scenarios, with SDGVM simulating a strong positive effect of increasing CO2 concentration on NPP, such that NBP remained positive and total biomass and soil carbon continue to increase. Hence, according to SDGVM, North America's forests would remain as a net carbon sink throughout the 21st century. Conversely, MC1 generally shows an initial increase in NBP (albeit somewhat slower than SDGVM, because of lower sensitivity of NPP to increasing CO2 concentration), followed by a serious decline from about 2030 onwards-due to increasing occurrence of droughts and, to a limited extent, increased areas lost to wildfire. By 2100 these projected differences in NBP lead, on the one hand, to significant gains in biomass carbon and total ecosystem carbon when simulated by SDGVM, but to significant losses when simulated by MC1 (with the CGCM2-A2 climate scenario causing the biggest declines). Neilson has described these as the "green-up" and "green-up followed by brown-down" projections, respectively.

The divergence between the two DGVMs far outweighed any differences in the responses of each model to the range of climate forcings produced by each of three GCMs and the two IPCC SRES emissions scenarios used to drive them. At the least, this finding indicates that ecosystem responses to climatic change are much less certain than the uncertainty implicit in the GCM climate scenarios. More broadly, this result demonstrates that the DGVMs are not yet reliable and require more development and testing. The differences between the two DGVMs (i.e., in their respective responses to the entire set of climate scenarios) appear to be caused primarily by the way in which they respond to the interacting effects of increased drought and increasing atmospheric CO2 concentration.

The MC1 projections show the greatest carbon losses occurring in three major forested regions of North America, broadly grouped by Neilson et al. (2006), as western USA, eastern USA and the boreal (covering Alaska and much of western and central Canada). It is worth noting that a significant decline in forested cover and total carbon was also projected in earlier IBIS simulations for the western Canadian boreal region forced by the CGCM2 IS92A scenario (which closely resembles the CGCM2 SRES A2 used in VINCERA). More detailed discussion of regional changes in the ecosystem indicators identified above, as modeled by the different DGVM and climate change scenario combinations, is presented in Section 4 of the report. Numerous maps are presented, comparing results for each model in the year 2000 with those for each of the six climate scenarios in 2100.

At this stage it is difficult to determine which of the two sets of model projections is more believable. Recent review of CO2 responses of stands at FACE ("Free-Air CO2 Enrichment") sites by Norby et al. (2005), suggests there is a generally positive response of tree-level NPP to increasing CO2 concentration, but with the caveat that FACE rings are generally located in immature stands. On the other hand, Körner et al. (2005) provide strong evidence that mature stands (i.e., at or close to site carrying capacity) do not exhibit a sustained positive response to elevated CO2. Interestingly, SDGVM shows a positive NPP response very consistent with the Norby et al. observations, due to its use of the well-accepted Farquhar leaf photosynthesis algorithm. Conversely, MC1 uses a much simpler empirical formulation for photosynthesis, and exhibits a much weaker NPP response. A second important difference is that MC1 has a relatively sophisticated fire model which responds to interannual climate variations and to spatial differences in fuel densities, whereas SDGVM has a very simplistic fire model that removes biomass with very little dependence on fuel densities or sensitivity to climate. However, SDGVM burns on average about three times the area annually compared to MC1. Moreover, although SDGVM productivity is generally higher, biomass densities are maintained lower. This suggests that high fire in SDGVM may keep biomass density below observed carrying capacities and hence causes consistently positive productivity responses. A third possibly important factor relates to how simulated vegetation responds to drought years: water use efficiency (WUE) is known to increase with increasing CO2 concentration, but it differs greatly between the two models, evidently causing vegetation to die back much more in MC1 than in SDGVM. This mortality then feeds into litter production and fuel build-up, and may help to trigger more simulated fires in MC1.

The collaboration between the three DGVM teams and other partners initiated by the VINCERA project will continue as the teams work to better understand the different results produced by the DGVMs. As refined projections become available they will provide indictors of the "relative health" of Canada's forests over periods of a few years to several decades. Interpreting model results in this way will provide a fundamentally new assessment of the impacts of climate change for forestry, protected areas and other forest users (i.e., to inform decision makers where and when significant changes in Canada's forests are most likely, and how rapidly they may occur). This continued work will culminate in forthcoming publications that will be forwarded to the Climate Change Impacts and Adaptation Program as they become available.

Future work - resolving the disagreement

More work is clearly required to determine the true nature of the response of NPP to rising CO2 concentration for a range of forested ecosystems. I.e., is the positive response presently seen at FACE experiments likely to be maintained as stands reach maturity? Other questions requiring attention include:

  1. What are the critical relationships among climate, vegetation productivity and area burned in temperate and boreal forests of North America, and can this be adequately captured in dynamic vegetation models? What are the likely impacts of soil water deficits, as a function of soil hydrology, and CO2 concentration effects on WUE, on forest vegetation under plausible scenarios of future climate?