1. U.S. Forest Carbon Budget
Richard A. Birdsey
USDA Forest Service
Northeastern Research Station
Newtown Square, PA
Linda S. Heath
Durham, NH
Ken Skog
Forest Products Laboratory
Madison, WI

2. Outline of talk
Inventory data (Rich)
Forest carbon model (Linda)
Carbon in wood products (Ken)

3. This map of forest type groups of the United States is based on satellite data, classified according to the definitions used by the national forest inventory. Forests of the United States include great diversity of productivity, use, and disturbance. The Southeast includes the most intensively managed forests, the highest percentage of softwood forest plantations, and very high productivity. Other areas of the East contain maturing hardwoods and hardwood/softwood mixed forests. The Interior West and Southwest includes dry forests currently affected by high occurrence of wildfire, and large forest areas interspersed with rangeland. The Pacific Northwest includes most of the remaining old-growth forests, and areas of intensive management. Interior Alaska includes large areas dominated by natural processes. Hawaii and Puerto Rico (not shown) represent tropical and subtropical island vegetation.

4. Multi-tier Monitoring Design
Tier One – Remote Sensing and Mapping
Land cover and change; biomass density
Wall-to-wall coverage; stratification
Tier Two – Extensive Inventories and Surveys
Management, productivity, disturbance, ownership, land use
Representative statistical sample
Tier Three – Medium-intensity Sample (new)
Ecosystem C pools and changes, soil CO2 flux (proposed)
Representative sample of condition classes
Tier Four –Intensive Areas
Soil-plant-atmosphere processes; complete C budgets
Relatively small number of specific sites
Multi-tier (multiphase) monitoring designs are efficient ways to collect data and estimate variables at different scales. These designs have been used for decades in land inventories.
FIA citation…

5. Tier One – Remote Sensing
Source: NAPP 1:40,000 color infrared photography
Sample Points: 16 photo points located systematically
over the “effective area” of each photo.
Measurement: land cover
Note: in process of shifting to
satellite data
The national forest inventory is in the process of converting from aerial photography to satellite remote sensing for the first sampling phase.

6. Tier Two – Field Sampling
Year One
Year Two
Year Three
Year Four
Year Five
Five-Year Panel 1 2 3 4 5
Historically, field sample plots have been remeasured on a cycle of about 10 years. The current goal is to have a 5-year cycle for all forest lands, although some forest areas that change slowly or are not strongly affected by human disturbance will have a longer remeasurement cycle.
Sample Intensity = 1 sample location
per 6,000 acres of land
Inventory Cycle Length = Five years or
20 percent of the sample locations
each year

7. Tier Two Sample Location Design
Data collected at field sample locations forms the measurement base for monitoring changes in carbon stocks and attributing those changes to various factors such as harvesting or stand age.
Plot measurements: age, disturbance, owner, physiography, etc.
Tree measurements: species, dimensions, damage, etc.

8. Tier Three – Forest Health Monitoring
Data collected at forest health monitoring plots will form the basis for monitoring forest ecosystem carbon stocks by expanding the information collected at phase two sample plots.
FHM and FIA sample locations are co-located
Additional data: crown condition, soils, understory,
coarse woody debris, etc.

9. Tier Four – Intensive Sites
Detailed observations at small number of sites
(LTER, AmeriFlux)
Complete ecosystem C stocks and fluxes
Used to develop models to aid in large-scale estimation

10. Some Inventory Considerations
Tier 1 – New remote sensing technology being applied
Tier 2 - Forest remeasurement is sparse in some areas (e.g. Alaska)
Tier 2 - Not completely consistent with NRI (gaps, overlaps, independent sampling frames)
Rangeland/forestland interface
Developed lands (urban/suburban)
Tier 2 - Public rangelands?
Tier 3 - FHM only partially implemented
Tier 4 - linkage is being pilot tested
All tiers - measurements not optimized for carbon

11. Area by land class (reconciled with NRI)
Forest Inventory Estimates as a Basis for Carbon Analysis (Trends by State and Region)
Area by land class (reconciled with NRI)
Area by forest type, owner, age class
Tree volume by species and size class
Tree biomass by species and size class
Data from the national forest inventory is made available in public data bases and can be aggregated to become a basis for estimating changes in carbon stocks for large areas.

12. Gross Growth per Acre of Timberland, U.S. by Region, 1952-1997

13. Basic estimation of carbon stocks and stock changes
Carbon stock = CARBON/AREA times AREA
Carbon stock change =
C stock at time 2 minus C stock at time 1
Divide by length of period = carbon/year
 Estimated values can be obtained from measured data or from using models
One of the ways to estimate carbon inventory, also called carbon stock, is by multiplying the area of forest by the amount of carbon per area. Carbon per area may differ substantially in two different forest stands. If the area of the two stands is known, and an average carbon per area is know for each area, a carbon stock could be estimated for each stand and then summed. We are interested most in carbon stock change, which is also called flux. We can estimate a carbon stock change by subtracting the carbon stock at time 1 from the carbon stock at time 2. Carbon stock changes are easy to use when they are displayed on a per year basis. This means dividing the carbon stock estimate by the length of the period between inventories. For example, if the carbon stock at time 2 is measured 10 years after the first measurement, the carbon stock change would be divided by 10 to produce a carbon change per area per year estimate. A final basic fact is that carbon stock could be measured directly, or may be estimated by using models. Soil carbon in a sample is measured directly, and these values are expanded for the appropriate area. Tree carbon may be based on the measurement of a tree diameter, which is used in a biomass equation to predict biomass. The biomass estimate is then converted to carbon. Another common variable used to estimate forest carbon is tree volume.

14. Forest sector carbon pools and flows
ATMOSPHERE
decay
Growth
decay
HARVESTED
CARBON
Removals
BIOMASS
Above and Below
STANDING
DEAD
Mortality
Recycling
processing
Litterfall,
Mortality
Harvest
residue
Treefall
decay
burning
DOWN
DEAD
WOOD
FOREST
FLOOR
PRODUCTS
disposal
burning
Humification
burning
ENERGY
Decomposition
LANDFILLS
SOIL
Imports/
Exports

15. US managed forest C pools, avg. annual 2008
Positive value indicates sink

16. Individual pools of forest C, per area
Live & standing dead tree biomass – FIA measurements and equations
Down dead wood – some data, model
Leaf litter -- model
Soil carbon –data from STATSGO
Harvested carbon – modeled, Ken Skog, FPL

17. Area by land class (reconciled with NRI)
Forest Inventory Estimates as a Basis for Carbon Analysis (Trends by State and Region)
Area by land class (reconciled with NRI)
Area by forest type, owner, age class
Tree volume by species and size class
Tree biomass by species and size class
Data from the national forest inventory is made available in public data bases and can be aggregated to become a basis for estimating changes in carbon stocks for large areas.

18. NOTE: Energy and emissions are releases of C to the atmosphere
Example: Two rotations of pine on a high site in SE Forest C and disposition of C in harvested wood (1995)
Carbon (T/ha)
Forests are unique in that carbon in forest ecosystems is not the only part of the story. The forests of the US are quite productive, and much forest is managed for harvest. Carbon in harvested wood continues to be stored as wood in products in use and discarded wood products in landfills. Some carbon is emitted after harvest, either as logging residue or as mill residue, discarded at the time of processing. Some wood is burned for energy, as a substitute for fossil fuel; other wood may be burned and carbon emitted without capturing the energy. This graph shows the same regional average planted pine stand in the Southeastern US, but the forest is cut and harvested at age 40, and the forest is replanted at that time and allowed to grow for another 40 years. Thus, the carbon stock estimates for the first 40 years are exactly the same as those in slide 14. The fate of the harvested wood is shown for illustration. Carbon in harvested wood removed from the site is delineated into one of four categories: carbon in products in use such as lumber in housing, paper, wood in furniture; carbon in wood in products discarded into landfills or waste wood; carbon from wood burned for energy, and carbon emitted from harvested wood. Wood in logging residue in the forest is a separate category tallied with forest ecosystem carbon. Note that wood burned for energy and emissions from wood are releases of C to the atmosphere, but they are shown here in a way to illustrate the size of that component. After 40 years we estimate that a substantial amount of harvested carbon continues to remain sequestered. 20 40 60 80
Age
NOTE: Energy and emissions are releases of C to the atmosphere

19. Forest sector system of models and data for C estimates and projections of managed U.S. forests-- Timber Assessment
The previous slide showed the general system we are trying to model. This slide shows the models and the data sources we use to produce carbon estimates. The red items illustrate data. FIA stands for Forest Inventory and Analysis, a group within the USDA Forest Service Research & Development branch which collects data in US forests. NRCS is an agency within the US Department of Agriculture, the Natural Resource Conservation Service. NRCS handles the soils data which was discussed in a previous slide. NRI is the National Resource Inventory, also conducted by the NRCS. This are the official land use estimates of the United States. Sources of economic data include the USDA Forest Service and Department of Commerce. These models of the forest sectors are used for projections because much forest in the US is managed. Harvest is the major disturbance, and economics plays a key role in the harvest level. Area change is also affected by economics. The models also provide a consistent framework for handling current estimates of carbon. The Timber Assessment Market Model basically models sawtimber demand and supply, the North American Pulp and Paper model captures the pulp and paper dynamics. WOODCARB converts the products into carbon. The AREA model provides area change projections, including changes between forest type and management intensity, the Aggregated TimberLand Assessment System keeps track of the forest inventory, growing forests while TAMM requests wood supply, asking ATLAS for harvested wood. The FORCARB model models all forest carbon pools, and converts the forest inventory information on trees to carbon. The following references can be viewed as a starting point for these models:
AREA:  Alig, R.J Modeling area changes in forest ownerships and cover types in the Southeast. Res. Pap. RM Ft. Collins, CO: U.S. Department of Agriculture, Forest Service, Rocky Mountain Forest and Range Experiment Station. 18 p.
ATLAS: Mills, J. R; Kincaid J. C The aggregate timberland assessment system‑‑ATLAS: A comprehensive timber projection model, Gen. Tech. Rep. PNW‑281. Portland, OR: U. S. Department of Agriculture, Forest Service, Pacific Northwest Research Station p.
FORCARB: Birdsey, Richard A., Heath, Linda S Carbon changes in U.S. forests. In: Productivity of America's Forests and Climate Change, Ed. by Linda A. Joyce. General Technical Report RM-271. U.S. Department of Agriculture, Forest Service, Ft. Collins, CO ;
Heath, L.S., and R.A. Birdsey Carbon trends of productive temperate forests of the coterminous United States. Water, Air, and Soil Pollution 70: 279‑293.
NAPAP: Ince, P. J Recycling and long‑range timber outlook. Gen. Tech. Rep. RM‑242. Fort Collins, CO: U. S. Department of Agriculture, Forest Service, Rocky Mountain Forest and Range Experiment Station. 66 p.
TAMM: Adams, D. M.; Haynes, R.W The Timber Assessment Market Model: Structure, projections, and policy simulations. Forest Science Monog. No Bethesda, MD. Society of American Foresters. 64 p.
WOODCARB: Skog, Kenneth E., and Geraldine A. Nicholson Carbon cycling through wood products: the role of wood and paper products in carbon sequestration. Forest Products Journal 48: 75-83
An overall reference for the forest sector models is Haynes, RW; Adams, DM; Mills, J The 1993 RPA timber assessment update. Gen. Tech. Rpt. RM-259. Ft. Collins, CO: US Department of Agriculture, Forest Service, Rocky Mountain Forest and Range Experiment Station. 66 p.

20. ATLAS model yield example SE, Planted pine, High site, MI 5
Linking to the ATLAS model means having forest type, ownership, region, age, or volume available for use as predictor variables for carbon

21. Fitted equation and data points for live trees Maple-Beech-Birch, NE region
400
300
Biomass (Tons/ha dry wt.)
200
100
100
200
300
400
Growing stock volume (m /ha) 3

22. Litter layer carbon accumulation, decay, and total--Southern Pines30 20
Litter layer C (Tons/ha) 10
Mixed or
unknown
age 25 50 75
Years
Source: Smith and Heath (in review)

23. Down dead wood by type and basal area
We are analyzing data from down dead wood sampled in conjunction with FIA surveys of ME, NH, and VT.

24. Projected inventory of privately owned managed forests of the United States, 2000
This is an example of the kind of output we produce using out models. Note that this is carbon stock on privately owned managed forests of the US. With this information we can easily determine the bounds of whatever confidence level we choose: 95% and 80% are noted here. Also note that including this information in this form is not easily usable. We suggest dividing the difference between the bounds and the measure of central tendency, assuming symmetry, and expressing the values in terms of percent. An example is 22,400 plus/minus 4% million metric tons. Using percentages however does have its disadvantage as illustrated in the next slide.
Smith and Heath. In press. Considerations for interpreting probabilistic estimates of uncertainty of forest carbon. In: Joyce LA and R. Birdsey, eds. The Impact of Climate Change on America's Forests, (ed.) USDA Forest Service, General Technical Report RMRS-GTR-59.
Source: (Smith and Heath, 2001)

25. Tree carbon per hectare by U.S. county, 1997

26. Managed forest lands, US, 2008-2012 Avg. annual C stock change
C taken up by trees in managed forests
381.9
C released by harvesting trees
-276.0
Net C taken up in Soil
52.4
Net C taken up in Floor
12.8
Net C taken up in Understory
0.7
Net C accrued in live biomass & soil
171.8
C increase in logging residue
26.1
C in products in use
39.1
C in products in landfills
51.3
C stored in products & landfills
90.4
Net C removals related to managed forests
288 +/- 15% MMT/yr
NOTE: the model is currently being updated, and we expect these estimates to change. However, all indications are that the trends and relative size of the estimate will be similar. The update is expected to be completed by spring of (Most of the work is writing manuscripts about the model for peer-review.) This table shows the carbon flux estimate by component for the first commitment period. Positive values indicate carbon is being removed from the atmosphere and stored in the forest or transferred into another forest-related pool; a negative value indicates a release of carbon to the atmosphere or a transfer to another forest-related pool. The modeling system produces carbon stock projections for the years 2005, 2010, 2015, etc. Carbon stock change is calculated as the difference between inventories. Estimates for the period are weighted averages of the periods (1 January) (end of December), and The amount of net carbon accrued in live biomass & soil is the subtotal of the first five rows. Of the C released by harvesting trees (-276 Million Metric Tons/yr), 26.1 MMT/yr is transferred to logging residue, and 90 MMT/yr is transferred into products in use or landfills. The sum of the subtotals indicates 288 MMT per year is sequestered in the forest sector during The uncertainty is estimated at plus/minus 15% at the 80% confidence interval.

27. Trend of carbon sequestration on managed forests, U.S.
Data
Projections
NOTE: the model is currently being updated, and we expect these estimates to change. See slide 39. This slide shows the trend of carbon stock change over time. Note the periods are of different lengths. The first year refers to 1 January of that year, the second year refers to 31 December. Note the trend over time is downward. Part of the downward trend is due to starting to account for carbon in harvested wood in Doing this ignores the decay of carbon from previously harvested wood. Another reason for the downward trend is that growth is relatively flat to slowly increasing, while harvests increase at a higher rate over time.

28. Modeling current and projected carbon storage in wood and paper
Ken Skog
USDA Forest Products Lab
Madison, WI

29. Modeling current and projected carbon storage in wood and paper
Model Framework
Accounting for Imports and Exports
Data needs/ uncertainty
Results

30. Cycling of Carbon Through Wood and Paper Products
Atmosphere
Burned
Decayed
Forests
Harvest
Products
Recycling
Wood-in-
use
Methane and CO2
Landfills

31. The System Framework – Harvested wood carbon Stock change and Atmospheric Flow approaches
Emissions
Imports
Products in use
Products in landfills
Harvest
Points
If you focus on measuring flows and emissions you can still get estimates of change in stocks.
If you focus on measuring change in Stocks you can still get estimates of emissions.
Darkar report suggested incorrectly that you could not get an estimate of stock change if you only have emissions data.
Exports
ΔProduct stocks = Harvest + Imports – Exports – Emissions
Net Flow to system = Harvest - Emissions
Atmospheric Flow = - ΔProduct stocks - Net Imports

32. ΔProduct stocks = Harvest – Emissions
The System Framework – Harvest wood carbon stocks and flows for Production approach
Emissions
Products in use
Products in landfills
Harvest
Exports
Points
The production approach focuses on estimating change in domestic product stocks and export product stocks
The production approach has a “production emissions” counterpart which would focus on emissions of a countries wood harvest.
Production approach would require estimates of how a countries exported products decay in countries where they are shipped.
ΔProduct stocks = Harvest – Emissions
Atmospheric Flow = Harvest - Emissions

33. Data needs Low uncertainty (±10%) Wood and paper product production
Product exports
Product imports
Product use by end use (e.g. construction)
Medium / High uncertainty (±20%+)
Waste when using products and disposition of waste
Years products are in use
Disposition of products after use – burn, decay, landfill
Rate of decay in landfills
Maximum decay in landfills (% ever emitted)
Data on wood and paper products production, and trade and end uses is fairly accurate for the U.S. and is available for all countries from FAO. FAO fuelwood data is less accurate than indurtrial product data
Data is less accurate on lifetimes of wood in use, disposition after use, landfill decay rates and limits

34. Decay of wood and paper in landfills
Material type in landfill
Maximum released
Sold wood (lignin attached) 3%
Newsprint
18%
Coated paper
20%
Boxboard
35%
Office paper
43%
One key factor that limits U.S. decay in our calculations is the estimate that decay of wood and paper in landfills if limited.
While oxygen is available white-rot fungus can decay lignin.
After Oxygen is depleted, lignin is not decayed by anaerobic bacteria and any cellulose incased in lignin is not decayed. Exposed cellulose will be decayed by anaerobic bacteria.

35. Disposition of 1990 harvested wood used for paper through 2050 (Tg)
1990 harvested wood that goes into paper and paperboard (or imports) goes through it’s life cycle shifting from product in use to emissions and to landfills.
Paper goes out of use quickly but is retained in landfills – after 50 year 24% is still retained in landfills.
24% in stocks in 2050
In use
In landfills

36. 54% in stocks in 2050 In landfills In use
Disposition of 1990 harvested wood used for fuel and solid wood products in the U.S. through 2050 (Tg)
54% in stocks in 2050
After 50 years 54% of harvest used for fuel and solid wood products is estimateed to still be in use or landfills due to long used life and limited decay in landfills.
In landfills
In use

37. 41% in stocks in 2050 In landfills In use
Disposition of 1990 wood harvest used for fuel and all products in the U.S. through 2050 (Tg)
41% in stocks in 2050
For 1990 U.S. harvest used in U.S. products 41% is estimated to still be in stocks in 60 years later.
In landfills
In use

38. Sensitivity to use life and decay rates -- Year 2000 Harvested wood carbon in 2050
58% 76% Emitted
The persistence of carbon is stocks is particularly sensitive to estimates of use life and limits of decay in landfills.
But even if we cut use life estimates in half, assume all paper and half of wood in landfills decays 20% of harvested wood carbon is still in stocks after 60 years.

39. U.S. Net Annual Changes to Stocks, and Emissions from Harvested Wood*
Emitted
Energy
We have estimated flows of harvested carbon since 1910 with projections to Landfill accumulation is estimated to have increased in recent years. Since the 1980s wood and paper goes to landfills rather than dumps where much was burned and decay was greater.
Landfill
Products
*Includes net imports. SOURCE: Skog and Nicholson, 2000

40. Managed forest lands, US, 2008-2012 Avg. annual C stock change
C taken up by trees in managed forests
381.9
C released by harvesting trees
-276.0
Net C taken up in Soil
52.4
Net C taken up in Floor
12.8
Net C taken up in Understory
0.7
Net C accrued in live biomass & soil
171.8
C increase in logging residue
26.1
C in products in use
39.1
C in products in landfills
51.3
C stored in products & landfills
90.4
Net C removals related to managed forests
288 +/- 15% MMT/yr
NOTE: the model is currently being updated, and we expect these estimates to change. However, all indications are that the trends and relative size of the estimate will be similar. The update is expected to be completed by spring of (Most of the work is writing manuscripts about the model for peer-review.) This table shows the carbon flux estimate by component for the first commitment period. Positive values indicate carbon is being removed from the atmosphere and stored in the forest or transferred into another forest-related pool; a negative value indicates a release of carbon to the atmosphere or a transfer to another forest-related pool. The modeling system produces carbon stock projections for the years 2005, 2010, 2015, etc. Carbon stock change is calculated as the difference between inventories. Estimates for the period are weighted averages of the periods (1 January) (end of December), and The amount of net carbon accrued in live biomass & soil is the subtotal of the first five rows. Of the C released by harvesting trees (-276 Million Metric Tons/yr), 26.1 MMT/yr is transferred to logging residue, and 90 MMT/yr is transferred into products in use or landfills. The sum of the subtotals indicates 288 MMT per year is sequestered in the forest sector during The uncertainty is estimated at plus/minus 15% at the 80% confidence interval.

41. Modeling current and projected carbon storage in wood and paper
Current studies
Develop generic national method using FAO product statistics
Assess effects of uncertainty on U.S. estimates
Project carbon storage associated with 2000 RPA Timber Assessment projections