Fine-root production dominates response of a

deciduous forest to atmospheric CO2 enrichment

Richard J. Norby*†, Joanne Ledford*, Carolyn D. Reilly*, Nicole E. Miller*‡, and Elizabeth G. O’Neill*
*Environmental Sciences Division, Oak Ridge National Laboratory, Oak Ridge, TN 37831-6422; and ‡Department of Biology, Loyola University,
Chicago, IL 60626

Communicated by William H. Schlesinger, Duke University, Durham, NC, May 17, 2004 (received for review April 1, 2004)

Fine-root production and turnover are important regulators of the replication), spatial and temporal variation are not confounded.
biogeochemical cycles of ecosystems and key components of their We assumed that the root system in this closed-canopy forest
response to global change. We present a nearly continuous 6-year stand had fully occupied the soil volume and was no longer
record of fine-root production and mortality from minirhizotron expanding, and an initial hypothesis for the FACE experiment
analysis of a closed-canopy, deciduous sweetgum forest in a was that full soil occupancy would constrain the N cycle (5). Here
free-air CO2 enrichment experiment. Annual production of fine we present fine-root dynamics for six growing seasons and
roots was more than doubled in plots with 550 ppm CO2 compared evaluate their role in the integrated response of the forest to CO2
with plots in ambient air. This response was the primary compo- enrichment of the atmosphere.
nent of the sustained 22% increase in net primary productivity.
Annual fine-root mortality matched annual production, and the Methods
mean residence time of roots was not altered by elevated CO2, but Experimental Site. The experimental site is a planted sweetgum
peak fine-root standing crop in midsummer was significantly (Liquidambar styraciflua L.) forest stand, which was established
higher in CO2-enriched plots, especially deeper in the soil profile. in 1988 at the Oak Ridge National Environmental Research Park
The preferential allocation of additional carbon to fine roots, which in Roane County, TN (35°54⬘N, 84°20⬘W). One-year-old, bare-
have a fast turnover rate in this species, rather than to stemwood rooted sweetgum seedlings were planted at a spacing of 2.3 ⫻
reduces the possibility of long-term enhancement by elevated CO2 1.2 m. Herbicide was used in 1989 and 1990 to control compe-
of carbon sequestration in biomass. However, sequestration of tition from weeds; no fertilizer has been added. The soil at the
some of the fine-root carbon in soil pools is not precluded, and site, which is classified as an Aquic Hapludult, developed in
there may be other benefits to the tree from a seasonally larger and alluvium washed from upland soils derived from a variety of
deeper fine-root system. Root-system dynamics can explain differ- rocks including dolomite, sandstone, and shale, has a silty clay
ences among ecosystems in their response to elevated atmospheric loam texture, and is moderately well drained. The soil is slightly
CO2; hence, accurate assessments of carbon flux and storage in acidic (water pH, ⬇5.5–6.0) with high base saturation largely
forests in a globally changing atmosphere must account for this dominated by exchangeable Ca. The mean annual temperature
unseen and difficult-to-measure component. (1962–1993) is 13.9°C, and the mean annual precipitation is 1,371
mm, with a generally even distribution of precipitation through-
out the year (11).
R esearchers who attempt to quantify C pools and fluxes in
terrestrial ecosystems often find it expedient to measure only
the aboveground components despite the wide recognition that
Five 25-m-diameter plots were laid out in 1996, and FACE
apparatus (8) was assembled in four of them; site disturbance was
minimized during construction. When pretreatment measure-
a substantial fraction of net primary production occurs below-
ments were made in 1997, the trees (⬇90 per plot) were 12 m tall
ground (1). Fine roots (ephemeral, nonwoody roots generally of
with an average diameter of 11 cm and a stand basal area of 28
diameters ⬍1 mm) are important regulators of biogeochemical
cm2䡠m⫺2 (11). The trees were in a linear growth phase, leaf area
cycling (2), and their turnover is a key component of C seques-
index was 5.5 m2䡠m⫺2, and the canopy was no longer expanding
tration in soil (3). Root production often is enhanced when the
(12). Exposure to elevated CO2 commenced in two plots in April
plants are grown in a CO2-enriched atmosphere (4, 5). However, 1998 and has continued during the growing season (April to
the difficulty in measuring these roots, which cannot generally be November) since then. The average daytime CO2 concentrations
seen, are not easily sampled, and can grow and die rapidly and during the 1998–2003 growing seasons were 528, 538, 546, 548,

ECOLOGY
unpredictably, means that they are left out of both retrospective 552, and 549 ppm in the two CO2-enriched plots, including
(6) and prospective (7) studies of the potential for terrestrial periods when the exposure system was not functioning, and
ecosystems to sequester additional C in response to a changing 384–398 ppm in ambient plots. The standard deviation of 1-min
atmosphere and climate. Fine-root production and distribution averages in the CO2-enriched plots was 60 ppm. The site
in forests exhibit high spatial and temporal variation, making and experimental design has been fully described in ref. 11,
their characterization especially difficult, and most of the ex- and operating and meteorological conditions are documented
perimental evidence on tree root responses to elevated CO2 at http:兾兾face.ornl.gov and http:兾兾cdiac.ornl.gov兾programs兾
comes from static or short-term observations that are strongly FACE兾ornldata兾ornldata.html.
confounded with plant developmental trajectories (5). Free-air
CO2 enrichment (FACE) experiments (8) provide an opportu- Minirhizotron Analysis. Five minirhizotron tubes were installed in
nity for longer-term observations under realistic conditions in each FACE plot in July 1997. Cellulose acetate butyrate mi-
fully developed forests stands. nirhizotron tubes (Bartz Technology, Santa Barbara, CA) were
We used minirhizotron observation tubes and video imaging installed in each plot at a 60° angle from vertical and to a depth
(9, 10) to quantify fine-root production and mortality in a FACE of 60 cm by using a hand auger (10). The tubes were stabilized
experiment in an established deciduous forest. Minirhizotrons
permit observations of the production and disappearance of
individual fine roots over frequent sampling intervals without Abbreviations: FACE, free-air CO2 enrichment; MRT, mean residence time; NPP, net primary
the repeated disturbance effects associated with coring tech- productivity; SRL, specific root length.
niques. Although the problem of high spatial variability remains †To whom correspondence should be addressed. E-mail: [email protected].
(and the labor-intensive nature of the analysis precludes optimal © 2004 by The National Academy of Sciences of the USA

www.pnas.org兾cgi兾doi兾10.1073兾pnas.0403491101 PNAS 兩 June 29, 2004 兩 vol. 101 兩 no. 26 兩 9689 –9693
with rebar and wrapped above the soil surface with black foam peaks occurring throughout the year (Fig. 1B), but generally
insulation, the upper ends sealed with a rubber stopper. Video mortality was highest in late summer and fall. Annually, mor-
images of the root systems were recorded beginning in February tality (Fig. 2B) matched production in both ambient and ele-
1998 and continued biweekly during the growing seasons vated CO2.
through November 2003. Video images were collected with a The net result of seasonal patterns of productivity and mor-
BTC-2 minirhizotron camera with a Smucker handle (Bartz tality is the fine-root standing crop (Fig. 1C). Peak standing crop
Technology). Images were not collected during the winter in ambient CO2 varied from June to early September in different
because shrinkage of the tubes in cold soil precluded movement years, but the peak in elevated CO2 always occurred later by
of the camera in the tubes. Each tube had 90 viewing windows 15–42 days. The effect of CO2 on peak standing-root mass was
(12.4 ⫻ 18.0 mm) for a total viewing area of 0.0201 m2. From highly significant (Fig. 2C). Root turnover, the fraction of the
1998 to 2002, individual frames on the videotape were captured population that is replaced during a year, was calculated as
by using a Targa⫹ video board (True Vision, Indianapolis) and annual production divided by the maximum standing crop (14).
digitized by using ROOTS software (Michigan State University, Averaged over the 6 years of observation, root turnover was 1.7
Lansing). In 2003, digital images were captured in the field by years⫺1, corresponding to mean residence time (MRT) of 0.62
using the I-CAP system (Bartz Technology) and analyzed with years, and this was unaffected by CO2 treatment. An indepen-
ROOTRACKER software (Duke University, Durham, NC). The dent estimate of MRT from cohort analysis of roots ⬍1.0 mm in
length and width of each root segment were measured, and the diameter ranged from 0.81–1.4 years and did not vary with CO2
incremental growth, death, or disappearance was recorded. treatment. In a third approach, MRT of C in sweetgum roots in
Root-length production, mortality, and standing crop were elevated CO2 plots was calculated to be 1.25 years based on the
calculated as described in ref. 9. Roots were not included in replacement of pretreatment roots with new roots with the
mortality until they disappeared; although a few roots that were distinct 13C signature of the added CO2 (3). The absence of an
most likely dead were observed in the minirhizotron windows, effect of CO2 on turnover rate from the first two methods
they always disappeared quickly, and we equate ‘‘disappearance’’ indicates that the increase in root mortality and concomitant
with ‘‘mortality.’’ input of root C into the soil was a direct result of increased root
Cohort analysis was used to estimate fine-root longevity. We production and not an alteration of root physiology.
measured survivorship on an annual basis for all roots produced Elevated CO2 can increase standing root crop by accelerating
between the first filming of the year to June 1 of that year. Roots whole-plant development, as in a young, developing tree stand
present at the first filming were excluded, as were all roots ⬎1.0 (15), through a specific change in plant allocation or indirectly
mm in diameter. Survival data were analyzed (13) and tested for through its effects on transpiration and soil moisture (16).
significance by using Peto and Peto’s log-rank test (with 1° of Separation of these often confounded mechanisms can be dif-
freedom and ␣ ⫽ 0.05) for censored data, where not all ficult, but it is important for prediction of long-term responses
individuals followed begin the treatment at the same time. in fully developed stands (5). Our initial assumption was that the
Half-lives for the root cohorts were calculated from regressions sweetgum trees had fully occupied the site. This assumption was
of log-transformed data by year and CO2 treatment. verified above ground in that peak leaf area index did not
increase from 1998 to 2003 (12), but the root system may have
Soil Cores. Five soil cores were extracted from each plot in July still been expanding during this period. Peak root length (Fig.
2002 by using a 10-cm-diameter hole saw. The cores were 15 cm 1C) and mass (Fig. 2C) in ambient CO2 were 150% greater in
deep and collected in 5-cm increments. Roots were picked from 2003 than in 1998, although the trend through time was not
the cores first by hand, and then the soil was washed through a statistically significant (R2 ⫽ 0.19, P ⫽ 0.07). However, there was
590-␮m sieve. Roots were sorted into diameter classes (⬍0.5 little change in peak length during the last 3 years in either
mm, 0.5–1.0 mm, 1.0–2.0 mm, and ⬎2.0 mm). A subsample of ambient or elevated CO2, so the effect of CO2 cannot be ascribed
different diameter classes from each plot and depth was removed simply to an acceleration of ontogeny. Indirect effects of CO2
and kept cold and moist; the remaining roots were oven-dried through soil moisture can be discounted. Although stomatal
and weighed. The subsampled roots were laid out individually on conductance of upper canopy leaves was usually lower in ele-
a minirhizotrom tube in the laboratory and filmed and digitized vated CO2, canopy conductance and stand-level evapotranspi-
to determine length by using the same techniques as used in the ration were little affected (17). Hence, soil moisture was not
field. These roots were then dried and weighed, and specific root significantly different between plots. Our results suggest that
lengths (SRLs) (length兾mass) were determined. Additional soil CO2 enrichment significantly increased peak-standing root crop
cores (two per plot) were collected in October 2003 to compare by altering allocation such that the potential for root occupancy
a direct measurement of fine-root mass with the estimate from of the soil volume was increased. This response was manifested
minirhizotron analysis. especially in the deeper distribution of roots in the soil profile
(Fig. 3). In 1998 most of the root length was in the upper soil
Results and Discussion (40% in the top 15 cm and 79% in the top 30 cm). Five years later
The seasonal pattern of fine-root production varied from year to 63% of the root length was still at 0–30 cm depth in ambient CO2,
year, and peaks of activity occurred throughout the growing but the root distribution was distinctly different in elevated CO2
season (Fig. 1A). Generally, root productivity was higher in plots. CO2 enrichment increased root length in the upper profile,
March to June than in July to October. The large majority of root but the largest increases occurred in deeper soil: 3-fold more
length (⬇80%) observed in the minirhizotrons was in roots ⬍0.5 length at 30–45 cm and 4-fold more at 45–60 cm. This response,
mm in diameter, and ⬍5% was in roots ⬎1 mm in diameter. This although statistically significant, was highly variable: 3 of the 10
pattern did not change with time or CO2 treatment. CO2 minirhizotron tubes in elevated CO2 accounted for 90% of the
enrichment began to have an effect on root productivity during root length deeper than 45 cm. Soil cores collected in October
the third year, and the response was strongest in midsummer 2003 confirmed the presence of fine roots at 60 cm, but not
(June to August). The CO2 effect on annual production (Fig. 2A) enough cores were collected to verify the effect of CO2 on depth
was highly significant, with production 2.2-fold higher in CO2- distribution given the high spatial variability. High variability in
enriched plots from 2000 to 2003. This large and significant root distribution is not surprising and cannot be dismissed as a
effect of CO2 on fine-root productivity would not have been seen sampling artifact: morphological plasticity of root systems, in-
if the measurements had been maintained for only the first 2 cluding spatially explicit root proliferation, is considered a major
years of the experiment. Mortality was highly variable, with mechanism whereby plants cope with the heterogeneity of soil

9690 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0403491101 Norby et al.

Fig. 1. Seasonal dynamics of fine roots were determined by analysis of minirhizotron images collected biweekly during the growing seasons from February
1998 to November 2003. (A) Production during an observation interval was calculated as the length of new roots plus increases in length of existing roots. (B)
Mortality is the disappearance of roots during an interval, with corrections made for roots that subsequently reappear. (C) Standing crop is the total length of
root visible; changes in standing crop generally reflect production minus mortality. Data are expressed as the length of root per square meter of observation
window. Pooled SEs for each year are shown.

resources (2, 18). Other reports of effects of CO2 on root corresponded closely with that measured directly in soil cores
distribution include an upward shift in a grassland ecosystem collected at the same time, lending confidence to our scaling
(16), a downward shift in a developing Populus stand (15), assumptions. The mass of roots in soil cores (0–15 cm in depth)
and stimulation of both surface and deep roots in a scrub oak collected in July 2002 and averaged across all plots was 103
system (19). g䡠m⫺2, compared with 95 g䡠m⫺2 calculated from minirhizotron

ECOLOGY
The mass of fine roots produced in a year was a significant data for the same date and depth. Fine-root mass in cores
fraction (11–34%) of net primary productivity (NPP) (Fig. 2D). collected in October 2003 was 228 g䡠m⫺2 (0–30 cm in depth),
Converting minirhizotron data to units comparable with compared with 239 g䡠m⫺2 from minirhizotron data.
aboveground production data is problematic, requiring assump- The additional C taken up and converted to organic matter by
tions about the volume of soil observed (10) and SRL of roots trees in CO2-enriched plots (average 22% increase in NPP) was
that vary over a wide range. Root length (meters per square allocated primarily to fine-root production (Fig. 2D). After the
meter viewing area) was converted to length per unit land area first year of treatment, there was no significant increase in
by dividing by the 0.002-m effective depth of field (10) and aboveground woody dry matter production (stem growth). Be-
multiplying by 0.6 m3 of soil per m2 of land area to account for cause fine-root turnover is much faster than turnover of woody
the soil volume observed to a depth of 60 cm. This value was then stems, the preferential allocation to fine roots should signifi-
expressed as grams of root per square meter by dividing by the cantly reduce the potential for additional C sequestration in trees
SRLs specific to each plot and year. SRL of roots extracted from in elevated CO2 (20). However, sequestration of some of that C
soil cores did not vary significantly between FACE plots, but it in the forest remains a possibility. As fine roots die, their C enters
was very different for the different diameter classes: 15.5 m䡠g⫺1 the soil system where there is the potential for movement into
for roots ⬍0.5 mm in diameter, 4.4 m䡠g⫺1 for roots 0.5–1.0 mm long-lived organic matter pools. Although the efflux of CO2
in diameter, and 1.5 m䡠g⫺1 for roots 1.0–2.0 mm in diameter. A from the soil surface has been higher in the CO2-enriched plots
composite SRL was calculated for each plot and year based on (21), the response is inconsistent and smaller than the increase
the diameter distribution when the standing crop was maximum. in fine-root production and mortality. Soil analysis indicates that
The average standing crop calculated from minirhizotron data there is increased accumulation of new C in CO2-enriched plots

Norby et al. PNAS 兩 June 29, 2004 兩 vol. 101 兩 no. 26 兩 9691
Fig. 2. Fine roots comprised an increasing fraction of NPP in CO2-enriched trees. Annual budgets of production (A), mortality (B), and maximum standing crop
(C) were expressed in grams of dry matter per square meter of ground. Increases in root mass between the last measurement in the fall and the first measurement
in the spring of the next year were attributed to production in the spring, whereas decreases were attributed to mortality in the fall. Data are the means ⫾ SE
of three ambient or two elevated CO2 plots. Analysis of variance indicated the main effect of CO2 on production, mortality, and peak standing crop to be
significant at P ⬍ 0.013, P ⬍ 0.068, and P ⬍ 0.010, respectively. Probability levels for the effect of year were P ⬍ 0.157 (production), P ⬍ 0.057 (mortality), and
P ⬍ 0.028 (standing crop); CO2–years interaction was never significant (P ⬎ 0.20). (D) The contribution of fine-root production to total annual NPP (grams of C
per square meter of ground area). NPP was calculated through allometric relationships, leaf litter trap collections, and the data in A, following methods described
in ref. 20. Conversion of dry matter units to C units was based on measured C concentration values of 47.1% in wood (open), 46.3% in leaf litter (vertical hatching),
and 39.6% in fine roots (diagonal hatching). The percentage increase in NPP is shown above the bars. The effects of CO2 on leaf litter C mass and NPP are
statistically significant (P ⬍ 0.001 and P ⬍ 0.002, respectively), but, excluding 1998, there is no statistically significant difference in wood C mass production.

(E. Hanlon and W. M. Post, personal communication), partic- important that the greatest increases in root production in
ularly in microaggregate fractions (at 0–5 cm in depth) that elevated CO2 occur in deeper soil, where sequestration into
facilitate movement of C into pools with long residence times (J. longer-lived pools may be more likely (22). Construction of a
Jastrow, personal communication). It may become especially complete C budget for the stand is a clear need and a formidable
challenge.
These results contrast with those from the similar FACE
experiment in a loblolly pine stand (23). There, the enhancement
in NPP was allocated primarily to aboveground stem increment.
The C in pine fine roots has a 3-fold-higher MRT than in
sweetgum roots (3) (similar to the difference in MRT of the
evergreen vs. deciduous canopy); therefore, annual fine-root
production and mortality rates are much lower than in our
sweetgum stand, and the pine root system is less responsive to
elevated CO2 (24). There has been no evidence for increases in
soil C content in the pine stand (25). Hence, the inherent
difference in fine-root dynamics between these two contrasting
tree species is a key ecosystem attribute controlling how C flux
and storage respond to rising CO2.
The substantial response of fine-root productivity to CO2
enrichment in the sweetgum stand may be a mechanism for
increasing N uptake to meet the increased demand associated
Fig. 3. Distribution of fine-root length by depth in soil at the time of peak with additional C uptake in elevated CO2. Long-term effects of
standing-root length in 1998 and 2003. Data are the means ⫾ SE of three
elevated CO2 on tree growth and C sequestration in forests are
ambient and two elevated plots, based on analyses of five minirhizotron tubes
per plot and expressed as length per square meter of observation window. CO2 highly dependent on the availability and cycling of N (7, 26), and
had a significant effect on root length at 0 –15 cm (P ⬍ 0.035), 30 – 45 cm (P ⬍ N availability is considered a major factor regulating fine-root
0.001), and 45– 60 cm (P ⬍ 0.013) but did not at 15–30 cm. The patterns in production in elevated CO2 (27). In grasslands, microbial N
intervening years were intermediate to those shown here. processes that control the supply of N to plants have been

9692 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0403491101 Norby et al.

 increased fine-root production may represent a futile mechanism
for improving N nutrition in this forest stand. The increased C
supply in CO2-enriched plots was used to support more root
production, but the resulting increase in N uptake has so far
supported only increased root production and not whole-tree N
nutrition or aboveground growth (31).
The CO2-induced increase in fine-root standing crop in summer
could also be an important mechanism for conferring increased
resistance to late-season droughts. Because leaf area index of this
stand did not respond to CO2 enrichment (12), there was a large
increase in the ratio of fine-root length to leaf area. This result
implies a reduction in hydraulic resistance at the rhizosphere,
making more soil water available (32). The stimulation of root
growth in deeper soil could be particularly important in buffering
trees against seasonal droughts. However, we have not observed any
interactions between CO2 and drought in this stand.
The potential for CO2 enhancement of root productivity
mandates that analyses of ecosystem responses to atmospheric
Fig. 4. Relationship between N uptake and root-length duration. N uptake
change take a whole-ecosystem approach. Analyses based solely
per plot for each year (1998 –2002) was calculated as the N content (dry matter
production times N concentration) of woody increment, leaf litter, and annual
on aboveground production will possibly miss a significant
fine-root production plus net foliar leaching (31). Root-length duration is the fraction of C and underestimate the potential of the ecosystem
area under the plots of fine-root standing crop (Fig. 1C). Regression line: N for additional C storage. Deciduous forest ecosystems with
uptake ⫽ 5.74 ⫻ root-length duration ⫺ 36.3; R2 ⫽ 0.80. highly dynamic root systems will exhibit different relationships
between C uptake, allocation, storage, and cycling than forests
with less dynamic root systems, and generalizations about the
responsive to elevated CO2, with N availability increasing be- potential for forests to sequester additional C and provide a
cause of an indirect effect of CO2 on soil moisture (28) or negative feedback to the rising atmospheric concentration must
declining because of reduced quality of litter inputs (29). Mi- account for these differences. Although current methods for
crobial N cycling and N availability have not responded to quantifying root dynamics are cumbersome and not appropriate
elevated CO2 in this or other forest stands (30); therefore, the N for documenting compliance to C sequestration management
needed to meet the requirement for increased NPP must be met (33), methodological limitations should not be an excuse for
through increased access to soil N. To assess this possibility, we ignoring this critical ecosystem response to global change.
calculated fine-root-length duration as an integrative measure of
the capacity for N uptake by the root system during the growing We thank Stan Wullschleger, Roser Matamala, Julie Jastrow, and four
season. N uptake has been linearly associated with fine-root- anonymous reviewers for helpful comments on the manuscript. This
length duration (Fig. 4) in support of the premise that the work was supported by the Oak Ridge Institute for Science Education
(C.D.R.), Summer Undergraduate Research Experience of the Depart-
fine-root response is important in the interaction between C and
ment of Energy Global Change Education Program (N.E.M.), and
N cycles. Our hypothesis that full soil occupancy in this forest the Department of Energy’s Office of Science, Biological, and Environ-
stand would constrain the N cycle is not supported. Elevated mental Research. Oak Ridge National Laboratory is managed by
CO2 increased the capacity of roots to occupy the soil volume, UT-Battelle for the Department of Energy (Contract DE-AC05-
and there was a concomitant increase in N uptake. However, 00OR22725).

1. Jackson, R. B., Mooney, H. A. & Schulze, E.-D. (1997) Proc. Natl. Acad. Sci. 18. Hodge, A. (2004) New Phytol. 162, 9–24.
USA 94, 7362–7366. 19. Day, F. P., Weber, E. P., Hinkle, C. R. & Drake, B. G. (1996) Glob. Change
2. Pregitzer, K. S., DeForest, J. L., Burton, A. J., Allen, M. F., Ruess, R. W. & Biol. 2, 143–148.
Hendrick, R. L. (2002) Ecol. Monogr. 72, 293–309. 20. Norby, R. J., Hanson, P. J., O’Neill, E. G., Tschaplinski, T. J., Weltzin, J. F.,
3. Matamala, R., Gonzàlez-Meler, M. A., Jastrow, J. D., Norby, R. J. & Hansen, R. T., Cheng, W., Wullschleger, S. D., Gunderson, C. A., Edwards,
Schlesinger, W. H. (2003) Science 302, 1385–1387. N. T. & Johnson, D. W. (2002) Ecol. Appl. 12, 1261–1266.

ECOLOGY
4. Rogers, H. H., Runion, G. B. & Krupa, S. V. (1994) Environ. Pollut. 83, 155–189. 21. King, J. S., Hanson, P. J., Bernhardt, E., DeAngelis, P., Norby, R. J. &
5. Norby, R. J., Wullschleger, S. D., Gunderson, C. A., Johnson, D. W. & Pregitzer, K. S. (2004) Glob. Change Biol. 10, 1027–1042.
Ceulemans, R. (1999) Plant Cell Environ. 22, 683–714. 22. Gill, R. A. & Burke, I. C. (2002) Plant Soil 241, 233–242.
6. Caspersen, J. P., Pacala, S. W., Jenkins, J. C., Hurtt, G. C., Moorcroft, P. R. 23. Hamilton, J. G., DeLucia, E. H., George, K., Naidu, S. L., Finzi, A. C. &
& Birdsey, R. A. (2000) Science 290, 1148–1151. Schlesinger, W. H. (2002) Oecologia 131, 250–260.
7. Oren, R., Ellsworth, D. S., Johnsen, K. H., Phillips, N., Ewers, B. E., Maier, C., 24. Matamala, R. & Schlesinger, W. H. (2000) Glob. Change Biol. 6, 967–979.
Schäfer, K. V. R., McCarthy, H., Hendrey, G., McNulty, S. G. & Katul, G. G.
25. Schlesinger, W. H. & Lichter, J. (2001) Nature 411, 466–469.
(2001) Nature 411, 469–472.
26. Hungate, B. A., Dukes, J. S., Shaw, M. R., Luo, Y. & Field, C. B. (2003) Science
8. Hendrey, G. R., Ellsworth, D. S., Lewin, K. F. & Nagy, J. (1999) Glob. Change
302, 1512–1513.
Biol. 5, 293–309.
27. Pregitzer K. S., Zak, D. R., Curtis, P. S., Kubiske, M. E., Teeri, J. A. & Vogel,
9. Hendrick, R. L. & Pregitzer, K. S. (1993) Can. J. For. Res. 23, 2507–2520.
C. S. (1995) New Phytol. 129, 579–585.
10. Johnson, M. G., Tingey, D. T., Phillips, D. L. & Storm, M. J. (2001) Environ.
Exp. Bot. 45, 263–289. 28. Hungate, B. A., Chapin, F. S., Zhong, H., Holland, E. A. & Field, C. B. (1997)
11. Norby, R. J., Todd, D. E., Fults, J. & Johnson, D. W. (2001) New Phytol. 150, Oecologia 109, 149–153.
477–487. 29. Gill, R. A., Polley, H. W., Johnson, H. B., Anderson, L. J., Maherali, H. &
12. Norby, R. J., Sholtis, J. D., Gunderson, C. A. & Jawdy, S. S. (2003) Oecologia Jackson, R. B. (2002) Nature 417, 279–282.
136, 574–584. 30. Zak, D. R., Holmes, W. E., Finzi, A. C., Norby, R. J. & Schlesinger, W. H.
13. Pyke, D. A. & Thompson, J. N. (1986) Ecology 67, 240–245. (2003) Ecol. Appl. 13, 1508–1514.
14. Dahlman, R. C. & Kucera, C. L. (1965) Ecology 46, 84–89. 31. Johnson, D. W., Cheng, W., Joslin, J. D., Norby, R. J., Edwards, N. T. & Todd,
15. Lukac, M., Calfapietra, C. & Godbold, D. L. (2003) Glob. Change Biol. 9, 838–848. D. E., Jr. (2004) Biogeochemistry, in press.
16. Arnone, J. A., Zaller, J. G., Spehn, E. M., Niklaus, P. A., Wells, C. E. & Körner, 32. Sperry, J. S., Hacke, U. G., Oren, R. & Comstock, J. P. (2002) Plant Cell
C. (2000) New Phytol. 147, 73–86. Environ. 25, 251–263.
17. Wullschleger, S. D. & Norby, R. J. (2001) New Phytol. 150, 489–498. 33. Brown, S. (2002) Environ. Pollut. 116, 363–372.

Norby et al. PNAS 兩 June 29, 2004 兩 vol. 101 兩 no. 26 兩 9693