Recent patterns and mechanisms of carbon exchange by terrestrial ecosystems

David Schimel¹, Jo House¹, Kathy Hibbard², Phillipe Bousquet³, Philippe Ciais³, Phillipe Peylin³, Mike Apps⁴, David Baker⁵, Alberte Bondeau⁶, Rob Brasswell¹, Josep Canadell⁷, Galina Churkina¹, Wolfgang Cramer⁶, Scott Denning⁸, Chris Field⁹, Pierre Friedlingstein³, Christine Goodale¹⁰, Martin Heimann¹, R.A. Houghton¹⁰, Jerry Melillo¹⁰, Berrien Moore III¹¹, Daniel Murdiyarso¹², Ian Noble¹³, Steve Pacala¹⁴, Colin Prentice¹, Mike Raupach¹⁵, Peter Rayner¹⁶, Bob Scholes¹⁷, Will Steffen¹⁸, Christian Wirth¹

¹Max Planck Institute für Biogeochemie, Jena, Germany; ²IGBP/GAIM, University of New Hampshire, Morse Hall, Durham, NH 03824, USA; ³LSCE Unité mixte CEA-CNRS; Bat 709, CE L'Orme des Merisiers, 91191, Gif sur Yvette France; ⁴Natural Resources Canada, Canadian Forest Service Northern Forestry Center, 5320 122 Street, Edmonton, Alberta, Canada; ⁵NCAR, 1850 Table Mesa Drive, Boulder 80303, USA; ⁶Potsdam Institute for Climate Impact Research, Telegrafenberg C4, 14473 Potsdam; ⁷GCTE International Project Office, CSIRO Sustainable Ecosystems, PO Box 284, Canberra, ACT, 2601, Australia; ⁸Department of Atmospheric Science, Colorado State University, Fort Collins, CO 80523-1371, USA; ⁹Carnegie Institute of Washington, Department of Plant Biology, 260 Panama Street, Stanford, CA 94305, USA; ¹⁰Woods Hole Research Center, P.O. Box 296, Woods Hole, MA 02543, USA; ¹¹University of New Hampshire, Center for the observations of Earth, Oceans and Space, Morse Hall, Durham, NH 03824, USA; ¹²GCTE Impacts Center for Southeast Asia, Jalan Raya Tajur Km 6, POB 116, Bogor, Indonesia; ¹³Ecosystem Dynamics, RSBS, Australian National University, POB4 Acton, Canberra, ACT 0200 Australia; ¹⁴Department of Ecology and Evolutionary Biology, Princeton University, Princeton NH 08544-1003, USA; ¹⁵CSIRO Division of Land and Water, GPO Box 1666, Canberra ACT 2601 Australia; ¹⁶CSIRO-DAR, PMB #1, 3195 Aspendale, Australia; ¹⁷Environmentek, CSIR, PO Box 395, Pretoria 0001, South Africa; ¹⁸IGBP Secretariat, Box 50005, S104-05, Stockholm, Sweden.

We synthesize recent information about continental and global patterns of terrestrial ecosystem carbon exchange, drawing on atmospheric inverse models, ground-based inventories, process studies and remote sensing. Atmospheric oxygen data confirm that the land was a net sink in the 1990s with components of the uptake in North America, Eurasia and the Tropics, despite continuing emissions due to deforestation. New inverse calculations suggest the Northern Hemisphere sink is distributed between the two continents and are compared to satellite and modeled indices of ecosystem activity. Process study and inventory data suggest that the relative importance of land management and changing forest growth rates differ between North America and Eurasia.

The global fluxes of carbon dioxide are dramatically affected by human activity. New analyses of atmospheric measurements, ecosystem inventories, satellite observations and ecosystem models provide a clearer view of both regional and continental fluxes, as well as of underlying mechanisms.

The global carbon budget

High-precision atmospheric observations of CO₂ and O₂:N₂ ratios make it possible to partition carbon fluxes between the land and ocean with increased confidence. Global carbon budgets have been updated in the most recent IPCC assessment¹ using this approach (Table 1). The net terrestrial flux between the land and atmosphere was, on balance, a sink of about -0.2 Gt C per year in the 1980s and -1.4 in the 1990s (the negative sign denotes a flux from the atmosphere). The net terrestrial flux is the difference between terrestrial uptake (sinks) and sources (mainly deforestation). Estimates of land use change suggest emissions in the range of +0.6 to +2.5 Gt C per year for the 1980s, largely from deforestation in the tropics¹. If land use change emissions were of a similar size in the 1990s, this would imply a residual terrestrial sink of between -2.0 and -3.9 Gt C per year.

Spatial patterns of carbon uptake

Since the late 1980s, researchers have identified a significant extratropical Northern Hemisphere contribution to the global sink^2-4. Evidence has accumulated since then supporting the existence of a significant Northern land sink from continued analysis of land inventory data^5-7, atmospheric inverse model analyses^8-11, studies of land use change¹² and ecosystem process models^13,14. Figure 1 shows zonal results from an ensemble of global inverse calculations, resulting in a range of estimates of the Northern net land sink from -0.6 to -2.3 Gt C per year in the 1980s¹¹.

Longitudinal regional partitioning of fluxes remains less certain than broad latitudinal partitioning, however, recent inverse modelling studies show the Northern sink is distributed between North America and Eurasia^9,10,15-17, in contrast to earlier reports of a dominant sink in North America¹⁸. Given currently available observations, the results are sensitive to the atmospheric model and spatial and temporal resolution, thus the estimates of the North American sink presented in Table 2 vary from 0 to 88% of the total Northern sink depending on the approach. On average, Eurasia is estimated as a sink about twice the size of North America.

On per unit land basis, the mean estimate implies similar uptake rates, of –30 to -35 gCm^-2y^-1 on both continents in the early 1990s. This rises to approximately -45 gCm^-2y^-1 when evaluated on a vegetated area basis, as both continents include sparsely vegetated desert and barren regions. This uptake, Net Biome Productivity (NBP)¹⁹, incorporates all carbon fluxes between land ecosystems and the atmosphere. The NBP values are from the period 1990-1994, when global uptake was high relative to the rest of the decade (Figure 2). Using estimated NBP and modeled Net Primary Productivity (NPP) suggests that 10 to 20% of Northern Hemisphere annual plant growth is transformed into NBP (Table 2). This can be compared to inventory-based estimates over large regions: for example, Nilsson et al²⁰ estimated this ratio for Russian ecosystems as 5% in forests, 10% in wetlands and 16% in grass/shrublands.

Carbon fluxes are sensitive to growing season length. Weighting area by growing season length (Table 2) provides an integrative index of ecosystem activity. North America has 60% of the GSW-area of Eurasia, compared to the ratio of mean NBPs (47%). North America’s total area and vegetated area are 60% and 55% of Eurasia’s. Rates of NBP per growing season day are 0.32 (range of 0.0 to 0.8) and 0.40 (0.1 to 0.6) gCm^-2 in North America and Eurasia. If the NBPs per m² growing-season-day differ dramatically, this would suggest an asymmetry in mechanism. Some differences are known to exist: nitrogen deposition is 2-4 times higher in Europe compared to the US²¹ and land management practices and history differ. The present results suggest slightly higher activity per unit ecosystem area and activity in Eurasia compared to North America. Given the extreme uncertainty in partitioning of fluxes (Table 2), this cannot yet be regarded as significant. Growing season fluxes derived from the global eddy covariance flux network range from 0.6 to 5 g m2 per day²². The eddy fluxes are higher than the NBP fluxes derived from the global atmospheric measurement network which incorporate losses due to winter respiration and, ultimately, disturbance

In the tropics, atmospheric inversions do not detect a large carbon dioxide source, as would be expected from deforestation, implying a sink that at least partly balances the deforestation source^10,11. Inverse model results (Figure 1) show variable results clustering around zero, suggesting a balance between uptake and deforestation in the tropics. If the deforestation source averages +1.6 Gt C y ^{–1 23}, then a net sink of –0.4 Gt C y ^–1 (Table 2) implies -2 Gt C y ^–1 uptake by tropical ecosystems (probably an upper bound) or -28 gCm^-2y^-1 Net Ecosystem Production. Local studies show carbon uptake in a range of mature tropical forest types, but it is not possible to extrapolate these to the entire tropical region^24,25. Because of sparse atmospheric and ecological sampling, and complex meteorology, estimates of tropical fluxes have high uncertainty.

Temporal variability

The year-to-year variability of the average annual growth in atmospheric CO₂ concentrations is high²⁶. Both top-down atmospheric calculations^8,27,28 as well as eddy covariance flux observations, long-term NPP studies and modelling^14,29-35 highlight year-to-year variability in terrestrial metabolism (Figure 2). Long term processes such as rising CO₂ and land-use history are major drivers of the mean fluxes, but probably have a small impact on year-to-year variations³⁶. These variations are likely caused by the effect of climate on carbon pools with short lifetimes (foliage, plant litter, soil microbes) through variations in photosynthesis, respiration, nutrient cycling and, possibly, fire.

The net terrestrial sink appears to have increased from the 1980s to the 1990s (Table 1). Links between climate system variability and carbon cycling are becoming clearer. Statistical and modeling links between Northern Hemisphere temperature variability, the El Nino-Southern Oscillation cycle and Monsoon rainfall/cloudiness and atmospheric CO₂ have been established ^27,28,37. Globally, there appears to be a net release of carbon to the atmosphere during warm and dry years, and net uptake during cooler years^28,38.

Controls over terrestrial carbon exchange

Recent studies make it clear that changing land management is a major contributor to Northern Hemisphere carbon uptake^12,14,30,39. While Table 2 suggests roughly similar rates of carbon uptake per unit ecosystem activity on the two continents, the relative importance of processes varies. In the United States, studies indicate that much (possibly most) of the sink is due to changing land management^13,14,40. European studies show large effects of both land use change^7,41,42 and increased tree growth possibly due to CO₂ fertilization and N deposition. The size and causes of tropical sinks remain basically unknown.

The importance of the land-use and management-related sinks in the Northern Hemisphere implies that sinks of today’s magnitude cannot be counted on to operate steadily into the future as forests reach maturity and opportunities for further sinks diminish. The current net terrestrial sink may disappear altogether in the future if the various uptake mechanisms decline and/or negative climate impacts and land use changes increase^1,43,44.

Rapid environmental changes may already be having an impact on terrestrial ecosystem processes in some regions⁴⁵. Much of Siberia has been warming at ~0.5^oC/decade 1960-2000 and substantial water stress increases have been documented in Alaska^46,47. Warming and associated recent increases in wildfire and insects appear to have caused a weakening sink or even a small source in Siberian and North American Boreal forests^20,47,48.

Key issues for future research

Ecosystem responses to local forces, such as land use, nitrogen deposition rate, or rainfall and temperature anomalies (which affect areas of <10⁶ km) control carbon fluxes and are not predictable from global averages, thus a regional observing system is needed. The present results suggest potential differences in ecosystem activity between regions and do not resolve the question of underlying mechanisms for ecosystem carbon storage.

While local flux measurement networks (e.g., eddy covariance) are valuable, the carbon storage consequences of patterns of land use and disturbance cannot be integrated from flux data alone. The key quantity, NBP, can only be estimated combining local fluxes with inventories of episodic and disturbance fluxes Improved atmospheric CO₂ measurements can also help reduce uncertainty: the sparse atmospheric measurement network over the continents and tropics can be substantially improved using CO₂ concentrations measured from aircraft or on tall towers.

Estimates of the residual terrestrial sink come from an overall balance (Table 1). Land use change emissions account for a large (approximately 20%) but uncertain fraction of anthropogenic CO₂ emissions¹. The lack of a clear atmospheric signal of tropical deforestation, and the implied large tropical sink, causes a major uncertainty in our ability to balance the terrestrial carbon cycle. Also, atmospheric inverse calculations simultaneously derive ocean and land fluxes and so uncertainty in terrestrial and marine fluxes are not independent. Understanding and stabilization of the Earth System will require these uncertainties to be addressed by efforts linking terrestrial ecology with the other Earth Sciences.

Acknowledgements

This paper arose from a workshop entitled "International Workshop on Integrated Carbon Research and Observations" held 16-20 October, 2000, in Durham, New Hampshire, USA. This workshop was co-sponsored

by the National Oceanographic and Atmospheric Administration, the National Science Foundation, the National Aeronautic and Science Administration, the Department of Energy, the Department of Interior/United States Geological Survey and the United States Department of Agriculture and the International Geosphere Biosphere Programme. The senior author acknowledges additional support from the National Aeronautics and Space Administration and the Bundesministerium für Bildung und Forschung (Germany).

Correspondence and requests for materials should be addressed to D.S. Schimel (email: dschimel@bgc-jena.mpg.de)

Table 1 Contemporary carbon budgets for the decades of 1980 and 1990 GtC

	1980s ^a	1990s ^a
Emissions (fossil, cement)	5.4 ± 0.3	6.4 ± 0.6
Atmospheric increase	3.3 ± 0.1	3.2 ± 0.2
Ocean-atmosphere flux	-1.9 ± 0.5	-1.7 ± 0.5
Land-atmosphere flux	-0.2 ± 0.7	-1.4 ±0.7
Land use change emission	1.7 (0.6 to 2.5)^b	Assume 1.6 ± 0.8^c
Residual terrestrial sink	-1.9 (-3.8 to 0.3)	-3.0 (highly uncertain)

a) From IPCC Third Assessment Report 2001¹

b) The range of estimates available to IPCC 2001¹

c) Based on the early 1990s only and not the full decade²³

negative values denote flux from the atmosphere, i.e. ocean or land uptake

Table 2. Estimated distribution of net atmosphere-land carbon fluxes (NBP) between North America and Eurasia from an ensemble of inverse experiments¹⁰.

	NBP (Gt C/y)	% of Northern uptake	Total land area^a (10^9m^2)	NBP per unit area (gCm^-2y^-1)	Vegetated area^b (10^9m^2)	NBP per unit vegetated area (gCm^-2y^-1)	Growing-season-weighted area (10^15m^2d^-1)	Modeled NPP (Gt C/y)	NBP per unit NPP
North America	-0.8 (-2.1 to +0.1)	0 to 88%	25	-32	20	-40	2.5	-5 to -9	9 to 16%
Eurasia	-1.7 (-2.5 to -0.2)	13 to 100%	42	-39	36	-46	4.2	-8 to -15	11 to 21%
Northern extratropical total	-2.4 (-4.3 to -1.5)	NA	67	-36	56	-44	6.7	-13 to -23	11 to 18%
Tropical and Southern Temperate	-0.4 (-1.2 to +0.8)	NA	70	-5^c	47	-8^c	12.5	-28 to -45	~1%^c
Global total	-2.8 (-4.3 to -1.7)	NA	137	-20