1

1

Direct estimates of forest C uptake--which reflect the interacting effects of rising CO₂, N fertilization, climatic changes, as well as reforestation and regrowth--indicate that forests in 28 eastern U.S. states during the late 1980s to early 1990s had an estimated net C uptake of only 0.17 Pg year¹ aboveground (2). This estimate includes the southeastern forests, which are among the most productive forest in the United States. European and Russian forest inventories (3) suggest comparable C uptake, which contradicts the North American distribution of the sink proposed in the report (1).

Anthropogenic N deposition stimulates C uptake in North America by 0.29 to 0.35 Pg C year¹ with significantly more uptake in Eurasia (0.67 to 0.86 Pg C year¹) according to model calculations (4). Moreover, recent field experiments by Nadelhoffer et al. (5) suggest that even this estimate of fertilized CO₂ uptake may be too high. Modeled CO₂ and climate effects generate a Northern Hemisphere sink with no significant east-west bias and a magnitude of ~0.58 Pg C year¹ (6).

A robust understanding of the global carbon budget requires reconciliation of ecological mechanisms with inverse estimates of the spatial distribution of the so-called missing sink. Fan et al. concede that their analysis approaches the limits of uncertainty, but the lack of corroboration from independent observations suggest that they may have overextended these limits.

Elisabeth A. Holland
National Center for Atmospheric Research,
Atmospheric Chemistry Division,
1850 Table Mesa Drive,
Boulder, CO, 80307-3000, USA, and
Max Planck Institut für Biogeochemie,
Tatzendpromenade 1a,
07745 Jena, Germany
E-mail: eholland{at}ucar.edu
Sandra Brown
Winrock International,
1611 North Kent Street, Suite 600,
Arlington, VA 22209, USA
E-mail: sbrown{at}winrock.org

REFERENCES AND NOTES

1. S. Fan, et al., Science 282, 442 (1998) [Abstract/Free Full Text] .
2. The estimate is based on U.S. Department of Agriculture inventory data and accounts for mortality and harvesting as well as changes in pools of live and dead mass and wood products. S. L. Brown and P. E. Schroeder, Ecolog. Applic., in press; spatial patterns of aboveground production and mortality of woody biomass based on inventory data for eastern U.S. forests.
3. O. N. Krankina and R. K. Dixon, World Resources Rev. 6, 88 (1994) ; P. E. Kauppi, K. Mielikäinen, K. Kuuslea, Science 256, 70 (1992) [Abstract/Free Full Text] .
4. E. A. Holland et al., J. Geophys. Res. 102, 15,849 (1997).
5. K. Nadelhoffer, et al., Ecolog. Applic. 9, 72 (1999) .
6. M. Cao and F. I. Woodward, Nature 393, 249 (1998) [CrossRef] .

The report by S. Fan et al. (1) has generated an important debate about their inference of an annual C sink as large as 1.7 ± 0.5 Pg in terrestrial ecosystems of North America. A terrestrial sink flux of this magnitude could completely offset a continental emission source from fossil fuel of 1.6 Pg C year¹. Fan et al. applied atmospheric constraints in their inverse model approach and used monthly C fluxes from the "equilibrium" version (without year-to-year variability) of the Carnegie-Ames-Stanford Approach (CASA) biosphere model mainly to estimate the magnitude of the seasonal "rectifier effect" in the atmosphere. We took a fundamentally different approach to study these questions (2). We used newly derived terrestrial C fluxes, predicted directly from forward-modeling CASA simulations in a nonequilibrium mode (using observed interannual variability for surface climate and satellite imagery). Our results (2) imply a much different conclusion from that of the report (1) with respect to the North America land sink during the late 1980s.

In accordance with the findings by Fan et al. at the global scale, we calculate that the worldwide C sink from net ecosystem production (NEP) can vary between 0.4 and 2.6 Pg C year¹ in the terrestrial biosphere. However, by making NEP estimates at 1° resolution directly from the CASA biosphere model, we found that C sink fluxes in terrestrial ecosystems of the United States and Canada totaled only 0.12 and 0.10 Pg C in 1987, respectively, and 0.05 and 0.17 Pg C in 1988. The land area covered by states of the former Soviet Union shows a larger NEP sink of 0.4 to 0.6 Pg C year^-1 during 1987 to 1988. These years may in fact represent two of most favorable of the 1980s for North American sink fluxes, owing to temperature warming trends.

We also included in the CASA model routines to determine aboveground biomass with the use of typical residence times in pools of wood and leaves to estimate changes in forest regrowth and deforestation C fluxes. This analysis, which relies on country-by-country changes in global forest cover for the years 1990 to 1995 as compiled by the United Nations (3), suggests that net forest regrowth can add about 0.09 Pg C annually to the North America C sink. Thus, it appears from forward CASA ecosystem modeling that total annual accumulation of atmospheric C in terrestrial ecosystems of North America could offset only about 20% of continental C source from fossil fuel burning in the late 1980s.

While analysis of CASA's satellite-driven net primary production (NPP) and soil heterotrophic CO₂ fluxes suggests that regional warming in Northern Hemisphere forests can enhance ecosystem production notably, our prediction of the regional distribution for this NEP sink over North America and Eurasia is not in agreement with the geographic patterns reported by Fan et al. for the late 1980s. It seems unlikely that the intercontinental balance of a large terrestrial C sink could shift so rapidly over just a few years, as suggested by results in the report. In any case, the potential for high interannual variability implies merely a transient sink pattern in North America. We infer that (in relation to results from the CASA model) climate variability, recent forest regrowth, and increased greenness as recorded in the satellite data capture the major processes of a terrestrial ecosystem sink.

Christopher S. Potter
Ecosystem Science and Technology Branch,
National Aeronautics and
Space Administration,
NASA-Ames Research Center,
Mail Stop 242-4,
Moffett Field, CA 94035, USA
E-mail: cpotter{at}mail.arc.nasa.gov
Steven A. Klooster
Earth System Science and Policy,
California State University at Monterey Bay,
Seaside, CA 93955, USA
E-mail: sklooster{at}gaia.arc.nasa.gov

REFERENCES AND NOTES

1. S. Fan, et al., Science 282, 442 (1998) .
2. C. S. Potter and S. A. Klooster, Climat. Change, in press.
3. Food and Agriculture Organization of the United Nations (FAO), State of the World's Forests 1997 (FAO, Rome, 1997).

Response: Our report (1) confirmed previous estimates of the size of a terrestrial C sink in the mid-latitude Northern Hemisphere and suggested that most of the sink occurred in North America from 1988 to 1992. Partitioning of the terrestrial sink among the continents is controversial because the CO₂ observations are sparse and the models are imperfect. In our report we mentioned possible causes of the sink, but we did not identify it with any specific mechanism because our method does not allow us to do so. Potter et al. and Holland and Brown [in these comments and elsewhere (2) with colleagues] estimated terrestrial C uptake by invoking specific causes. The difference between these estimates and ours is the subject of intense research and may be resolved by a growing network of atmospheric and ecological observations.

Current research on the sources and sinks of atmospheric CO₂ consists of four major activities: (i) long-term continuous measurement of C fluxes between the atmosphere and the biosphere by eddy correlation, (ii) repeated measurements over time of C inventories in terrestrial ecosystems and in the ocean, (iii) long-term monitoring of atmospheric CO₂ in a global air sampling network, and (iv) development of mechanistical ecosystem models that predict past, present, and future C cycles on land and in the sea. These activities have been conducted separately in different laboratories and have produced independent estimates of the global carbon budget.

The estimates of oceanic uptake of CO₂ by different approaches appear to converge. For instance, a global uptake of about 2 Pg C year¹ is estimated for the 1980s separately by ocean circulation and biogeochemistry models (from changes in dissolved inorganic C over time) and by measurements of the air-sea difference of CO₂ partial pressure. The estimated air-sea exchange fluxes show similar spatial distributions. More recent measurements of O₂/N₂ and ¹³CO₂ in the atmosphere are consistent with the ocean sink estimates.

However, estimates of terrestrial C sequestration by different methods are disparate in both their magnitudes and spatial distributions. Terrestrial ecosystem models (such as the CASA model discussed by Potter) relate the NEP to environmental parameters (light; air; soil temperature, moisture, and N; and the ambient CO₂ mixing ratio) and to ecological properties (land use history, stand age, species of vegetation, and leaf area index), and predict C fluxes resulting from succession after disturbance, interannual changes in climate pattern, increasing CO₂ in the atmosphere, and input of N from atmospheric deposition. Some of these models predict large global NEP or C uptake rates as a result of CO₂ fertilization. Most of this uptake is predicted to occur in the tropical forests, with a small contribution from the mid-latitude Northern Hemisphere (for example, 3, 4). But the response of plant photosynthesis to ambient CO₂ concentration is uncertain in the natural environment. The treatment of land-use changes and natural disturbances is difficult in a time-dependent terrestrial ecosystem model because of a lack of historical data in most of the world. As a result, we cannot assess the reliability of NEP estimates from any of the global ecosystem models including that of Potter and Klooster (4).

A fertilization of the ecosystems by atmospheric N deposition may also cause a terrestrial C sink. Nitrogen deposition is heaviest in Europe and eastern North America. Earlier estimates of this sink are as large as 1.5 Pg C year¹ on a global scale, if one assumes a 100% utilization of the N for plant CO₂ assimilation (5). However, recent studies indicate that only one-third of the deposited N may be used by plants, while the rest may be tied up in soil organic matter (6). Plant tissues have a much higher ratio of C:N than does soil organic matter.

Measurements of forest C inventory can provide a direct estimate of forest C uptake. The estimate of Brown and Schroeder (as cited in the comment by Holland and Brown) is similar to previous analyses of forest inventory data: the forest ecosystems in North America sequester carbon at a rate < 0.2 Pg C year¹. The global C sequestration is estimated to be 0.8 Pg C year¹ by forest ecosystems, more than half of the estimated sink is located in Europe and Russia.

In contrast, atmospheric observations of CO₂ and the ratios of O₂/N₂ and ¹³CO₂/¹²CO₂, combined with numerical models of atmospheric transport, have consistently implied the presence of a terrestrial C sink in the mid-latitude Northern Hemisphere of 1 to 3 Pg C year¹ (Table 1). Our report supports the many previous estimates of this sink (1.4 ± 0.2 Pg C year¹ by SKYHI and 2.2 ± 0.2 Pg C year¹ by GCTM). Table 1 compares various estimates of the terrestrial C uptake in the mid-latitude Northern Hemisphere. The uptake rate implied by atmospheric and oceanic CO₂ data and models tends to be much larger than estimates based on forest inventory data and mechanistic terrestrial ecosystem models.

Table 1. Estimated terrestrial C uptake (Pg C year1) in the mid-latitude Northern Hemisphere, based on various data and models. Period Uptake Constraints Reference (A) Estimates based on atmospheric and oceanic data and models 1981-1987 2 to 3 CO2 data Tans et al. (1990), (9) 1992-1993 2.5 to 3.5 CO2 and 13CO2 data Ciais et al. (1995), (10) 1991-1994 1.9 ± 0.9 CO2 and O2/N2 data Keeling et al. (1996), (11) 1988-1992 1.4 ± 0.2 CO2 data and models Fan et al. (1998), (1)* 1988-1992 2.2 ± 0.2 CO2 data and models Fan et al. (1998), (1)* (B) Estimates based on forest inventory data and mechanistic terrestrial ecosystem modes  1980s 0.6 to 1.0 Forest and land use data Dixon et al. (1994), (12)  1980s 1.0 to 1.2 N deposition model Holand et al. (1997), (13) ~1990  ~0.6 Climate and ecosystem model Cao and Woodward (1998), (3) * Two atmospheric models were used by Fan et al. (1): SKYHI and GCTM, respectively. Enting et al. (14) also estimated a large terrestrial uptake in the mid-latitude Northern Hemisphere for 1989-1990, and a smaller uptake for 1986-1987. Keeling et al. (15) estimated a Northern Hemispheric land sink of 0.6 Pg C year1; however, they imposed a large pre-industrial North Atlantic uptake (~1 Pg C year1), which reduces terrestrial uptake accordingly and was laterestimated to be much smaller (~0.3 Pg C year1, Keeling and Peng, 16).

We stated in our report that the spatial distribution of the mid-latitude terrestrial C sink is not as well constrained by the atmospheric CO₂ data. As shown in figure 2 in our report, we cannot reject the presence of a substantial Eurasian sink. One-half of the estimated Northern Hemisphere terrestrial sink is attributable to Eurasia at the limit of 1 standard deviation (SD), and two-thirds at the limit of 2 SD. However, we estimate a very low probability for the North American sink to be as small as 0.2 Pg C year¹ for 1988 to 1992, as suggested by Potter and Klooster (4) and by Brown and Schroeder (2).

The intercomparison between our inverse estimates (based on atmospheric and oceanic data and models) and estimates based on mechanistic models and ground data must take into consideration the large temporal variability of the terrestrial C cycling (7). The growth of atmospheric CO₂ decreased from 4.9 to 1.4 Pg C year¹ from 1988 to 1992, while the global fossil CO₂ emission remained nearly constant (5.9 Pg C year¹ in 1988 to 6.1 Pg C year¹ in 1992). This implies an increase of the global ocean and terrestrial C sink from 1.0 Pg C year¹ in 1988 to 4.7 Pg C year¹ in 1992. According to model calculations, the oceanic uptake could not have changed by more than 1 Pg C year¹. The large interannual CO₂ variability must have been caused mainly by terrestrial ecosystem productivity and respiration.

The estimate of Potter and Klooster covers a different period (1987 and 1988). Given the large interannual variability of terrestrial NEP, their estimate should not agree with ours even if both are correct. Also, it is possible that our estimate, if correct, might apply to the restricted years studied and might disagree with correct long-term estimates from inventory data.

Last, but not least, direct CO₂ flux measurements by eddy correlation have shown evidence of significant C uptake in a deciduous forest in New England that had its last major disturbance by a hurricane some 60 years ago (8). A network of the tower-based measurements has been implemented in North America (AmeriFlux) and Europe (EuroFlux), and is being extended to South America (and remains to be extended to Africa and Aia). Long-term C flux and ecological data collected from these tower sites will provide critical information on the size and causes of the terrestrial C sink from landscape to global scales. This is an exciting development in C cycle research that we hope will soon resolve the discrepancies between the various estimates of terrestrial C sequestration.

S. Fan
Carbon Modeling Consortium,
c/o Atmospheric and Oceanic
Sciences Program,
Princeton University,
Princeton, NJ 08544, USA
E-mail: cmc{at}princeton.edu
M. Gloor
Department of Ecology and
Evolutionary Biology,
Princeton University
J. Mahlman
Geophysical Fluid Dynamics Laboratory,
National Oceanic and Atmospheric Administration,
Princeton University,
Post Office Box 308,
Princeton, NJ 08542, USA
S. Pacala
Department of Ecology and
Evolutionary Biology,
Princeton University
J. Sarmiento
Atmospheric and Oceanic
Sciences Program,
Princeton University
T. Takahashi
Lamont-Doherty Earth Observatory,
Columbia University,
Palisades, NY 10964, USA
P. Tans
Climate Modeling and
Diagnostics Laboratory,
National Oceanic and
Atmospheric Administration,
Boulder, CO 80303, USA

REFERENCES

1. S. Fan, et al., Science 282, 442 (1998) .
2. S. L. Brown and P. E. Schroeder, Ecolog. Applic., in press.
3. M. Cao and F. I. Woodward, Nature 393, 249 (1998) .
4. C. S. Potter and S. A. Klooster, Climate Change, in press.
5. E. A. Holland et al., J. Geophys. Res. 102, 15,849 (1997).
6. K. Nadelhoffer, et al., Ecolog. Applic. 9, 72 (1999) .
7. T. J. Conway et al., J. Geophys. Res. 99, 22,831 (1994); R. J. Francey, et al., Nature 373, 326 (1995) ; C. D. Keeling, T. P. Whorf, M. Wahlen, J. van der Plicht, ibid. 375, 666 (1995) [CrossRef].
8. S. C. Wofsy, et al., Science 260, 1314 (1993) [Abstract/Free Full Text] ; M. L. Goulden, J. W. Munger, S.-M. Fan, B. C. Daube, S. C. Wofsy, ibid. 271, 1576 (1996) [Abstract].
9. P. P. Tans, I. Y. Fung, T. Takahashi, ibid. 247, 1431 (1990).
10. P. Ciais, P. P. Tans, M. Trolier, J. W. C. White, R. J. Francey, ibid. 269, 1098 (1995).
11. R. F. Keeling, S. C. Piper, M. Heimann, Nature 381, 218 (1996) .
12. R. K. Dixon, et al., Science 263, 185 (1994) [Abstract/Free Full Text] .
13. E. A. Holland et al., J. Geophys. Res. 102, 15,849 (1997).
14. I. G. Enting, C. M. Trudinger, R. J. Francey, Tellus 47B, 35 (1995) [CrossRef].
15. C. D. Keeling, S. C. Piper, M. Heimann, in Aspects of Climate Variability in the Pacific and Western Americas, AGU Monograph 55, D. H. Peterson, Ed. (American Geophysical Union, Washington, DC, 1989), pp. 305-363.
16. R. F. Keeling and T. H. Peng, Philos. Trans. R. Soc. London Ser. B 348, 133 (1995) [CrossRef] .