Supporting Information

Files in this Data Supplement:

SI Text

The geographic scope of this study was the managed forest of Canada (240 million ha). For the purposes of United Nations Framework Convention on Climate Change (UNFCCC) reporting, Canada has defined the managed forest by using an area-based approach (1), where forests that are either managed for timber or nontimber forest values or are under intensive protection against natural disturbances (e.g., fire suppression) are considered to be subject to forest management. Further description of the managed forest area is provided in Canada's 2007 national greenhouse gas inventory submission to the UNFCCC, available online (http://unfccc.int/national_reports/annex_i_ghg_inventories/national_inventories_submissions/items/3929.php). Each year, inventory data are reevaluated and improvements are made, so the estimate of the total area of managed forest is refined over time.

The majority of forest lands in Canada fall under provincial and territorial jurisdiction. Provincial and territorial forest resource management agencies provided forest inventory and growth and yield information for this study either directly, or by the Canadian Forest Inventory (CanFI 2001), a national compilation of inventory data for Canada (2).

Disturbance monitoring information used in this study included statistics for the period 2000-2005 on area annually burned, area infested by insects (mountain pine beetle data were available for 2000-2006), and area and wood volumes harvested. We used national datasets for fire (3) and harvest (4). Provincial and territorial forest management agencies provided insect monitoring data, additional data on fire and harvesting, and information used to formulate simulation rules.

Other disturbances, such as windthrow, mortality after drought, and insect outbreaks that are considered to be less important in the national context [e.g., spruce beetle (Dendroctonus rufipennis Kirby) in the Yukon Territory and western spruce budworm (Choristoneura occidentalis Freeman) in northern British Columbia] were not included in this analysis.

We used a Monte Carlo risk assessment approach to project future greenhouse gas (GHG) emissions and removals from Canada's managed forest. This enabled us to evaluate both the range of potential future outcomes and their probabilities. We developed separate regional probability density functions (PDFs) for each of six disturbance agents (fire, spruce budworm, mountain pine beetle, aspen defoliators, jack pine budworm, and hemlock looper) based on 1959-2000 natural disturbance statistics. These PDFs were used to generate time series of future area disturbed for each of 24 modeling regions (SI Fig. 3). We used projections of future harvest levels from provincial and territorial timber supply planning processes to estimate future harvest rates. We then estimated the managed forest GHG balance distribution by using a Monte Carlo approach to generate 100 model outputs with use of data for the period 2000-2005 or 2006, harvest projections to 2022, and 100 disturbance time series for fire and insects as input to CBM-CFS3. Last, we resampled combinations of the regional results to generate 500 national estimates per time step for annual distributions (Fig. 1) and 5,000 national estimates for the 2008-2012 average annual distribution (Fig. 2).

Area burned PDFs were calibrated for modeling regions (SI Fig. 3) with data from the 1959-1999 period (5). We fit the PDFs to the data such that the probability of burning an area greater than the worst fire year on record would be no more than 2% and no less than 1%. We further constrained future fires to be no larger than twice the maximum observed area burned in the 1959-1999 record. We assumed no spatial autocorrelation between regions and no temporal correlation between years because only weak correlations could be demonstrated, even in adjacent strata.

The future area infested by a given insect disturbance agent in any year was strongly correlated with the area infested by that insect in previous years. We constructed future insect outbreaks by using a number of parameters, each described by using a PDF parameterized based on available information on past outbreak dynamics (6,7), provincial data, and expert judgment. Future insect outbreaks were constructed by describing for each insect and region: (i) the duration of the interval between outbreaks, (ii) the total area infested during outbreak, (iii) the duration of outbreak in years, and (iv) the rate of expansion and contraction of area infested (outbreak shape). This approach was used to project the future dynamics of ongoing outbreaks and of outbreaks that are projected to start in the future.

Mountain pine beetle outbreak dynamics were projected based on the characteristics of the remaining host (i.e., pine trees of suitable age) and on the expert judgment of regional entomologists. The carbon impacts of the beetle outbreak include both direct impacts (mortality of infested trees) and subsequent indirect impacts (decay of killed biomass, removal of carbon by increased salvage logging activity, and associated impacts on forest succession and age-class distributions). We simulated the impacts by using a series of impact severity classes that target host forest areas according to the projected outbreak areas for each modeling region. Our projections were found to be consistent with volume loss projections produced by the British Columbia Ministry of Forests using detailed beetle population and host dynamics modeling (8).

Spruce budworm outbreak dynamics were defined by using the Spruce Budworm Decision Support System (SBWDSS) (9), which provided detailed projections of future impacts of defoliation, including growth loss and mortality of defoliated host trees. Impacts projected by the SBWDSS were calibrated based on projected host characteristics and the observed impacts of the previous outbreak, which started in the early 1970s (10). We assumed (i) that the impacts of the next outbreak will be of similar magnitude to those of the previous outbreak, and (ii) that spruce budworm outbreak dynamics will continue to follow the ~35-year outbreak cycle that has been observed in the past (11), which means that the next outbreak can be anticipated to begin very soon. We assumed that the outbreak will begin no later than 2011, with defoliation peaking several years into the outbreak. Impacts of spruce budworm outbreak include reduced growth in defoliated stands and partial mortality of host species in more severely defoliated stands.

The Carbon Budget Model of the Canadian Forest Sector (CBM-CFS3) is an empirically driven stand- and landscape-level forest carbon simulation model. Earlier versions of the model were documented by Kurz et al. (12) and by Kurz and Apps (13). The current version of the model builds on the functionality of CBM-CFS2, extending the model's capability to simulate partial mortality and improving the simulation of dead organic matter, soil organic carbon dynamics, and fire impacts. Dead organic matter and soil carbon dynamics have been reparameterized by using data from the Canadian Inter-site Decomposition Experiment (14) and the Canadian Forest Ecosystem Carbon Database (15); fire impacts have been reparameterized by using information from BORFIRE (16). CBM-CFS3 simulates biomass carbon dynamics by using standard merchantable volume yield curves. Thousands of such curves have been developed for Canadian forests by government resource management agencies and industry for the purpose of timber supply analysis. Growth curves are derived from millions of tree measurements taken at permanent and temporary sample plots located throughout the managed forest. In CBM-CFS3, merchantable volume yields are converted to biomass carbon yields by using a national system of stand-level above-ground biomass estimation models (17), below-ground biomass estimation models (18), and a biomass carbon content of 0.5 t C t^-1 biomass (1). Future growth and decomposition rates were projected by using empirical data and were assumed to be unresponsive to changes in climate during the simulation period.

1. Penman J, Gytarsky M, Hiraishi T, Krug T, Kruger D, Pipatti R, Buendia L, Miwa K, Ngara T, Tanabe K (2003) Good Practice Guidelines for Land Use, Land-use Change and Forestry (IPCC National Greenhouse Gas Inventories Programme, Hayama, Japan).

2. Power K, Gillis M (2006) Canada's Forest Inventory 2001 (Natural Resources Canada, Canadian Forest Service, Victoria), BC-X-408.

3. Lee J, Alexander ME, Hawkes BC, Lynham TJ, Stocks BJ, Englefield P (2002) Comput Electron Agric 37:185-198.

4. Canadian Council of Forest Ministers (2005) Compendium of Canadian Forestry Statistics Available at: http://nfdp.ccfm.org).

5. Stocks BJ, Mason JA, Todd JB, Bosch EM, Wotton BM, Amiro BD, Flannigan MD, Hirsch KG, Logan KA, Martel DL, et al. (2002) J Geophys Res Atmos 108:FFR5.1-FFR5.12.

6. Simpson R, Coy D (1999) in An Ecological Atlas of Forest Insect Defoliation in Canada 1980-1996 (Natural Resources Canada, Canadian Forest Service, Fredericton), Information Report M-X-206E.

7. GeoConnections, Insect monitoring datasets for 1980-2000. Available at: http://geodiscover.cgdi.ca.

8. Walton A, Hughes J, Eng M, Fall A, Shore T, Riel B, Hall P (2007) Provincial-Level Projection of the Current Mountain Pine Beetle Outbreak (British Columbia Ministry of Forests and Range, British Columbia).

9. MacLean DA, Erdle TA, MacKinnon WE, Porter KB, Beaton KP, Cormier G, Morehouse S, Budd M (2001) Can J For Res 31:1742-1757.

10. Blais JR (1983) Can J For Res 13:539-547.

11. Royama T (1984) Ecol Monogr 54:429-462.

12. Kurz WA, Apps MJ, Webb TM, McNamee PJ (1992) Carbon Budget of the Canadian Forest Sector Phase I (Natural Resources Canada, Canadian Forest Service, Northwest Region, Northern Forestry Centre, Edmonton), Information Report NOR-X-326.

13. Kurz WA, Apps MJ (1999) Ecol Appl 9:526-547.

14. Trofymow JA, CIDET Working Group (1998) The Canadian Intersite Decomposition Experiment (CIDET): Project and Site Establishment Report (Natural Resources Canada, Canadian Forest Service, Victoria), Information Report BC-X-378.

15. Shaw CH, Bhatti JS, Sabourin KJ (2005) An Ecosystem Carbon Database for Canadian Forests (Natural Resources Canada, Canadian Forest Service, Northern Forestry Centre, Edmonton), Information Report NOR-X-403.

16. de Groot WJ, Landry R, Kurz WA, Anderson KR, Englefield P, Fraser RH, Hall RJ, Banfield E, Raymond DA, Decker V, et al. (2007) Int J Wildland Fire 16:593-606.

17. Boudewyn P, Song A, Magnussen S, Gillis MD (2007) Model Based Volume-to-Biomass Conversion for Frested and Vegetated Land in Canada (Natural Resources Canada, Canadian Forest Service, Victoria), Information Report BC-X-411.

18. Li Z, Apps MJ, Kurz WA, Banfield E (2003) Can J For Res 33:2340-2351.

Fig. 3. Fire and insect modeling regions. Canada's managed forest land area was stratified into regions (colored areas) by using the boundaries of the Terrestrial Ecozones of Canada and the provincial and territorial borders. Provincial and territorial borders were used because each province and territory was run separately through the modeling analyses. Ecozone boundaries were used to delineate regions with common natural disturbance regimes - similar fire return intervals and insect outbreak histories - within provinces and territories. Northern limits of the managed forest were also used to define modeling region boundaries. The areas shown in gray in Canada's north were not included in the analyses because these areas have no managed forests. The areas shown in gray in southern Ontario and southern Quebec where included in the analyses, but no future natural disturbances were projected due to the high density of human settlement.