Ecological Applications, 17(7), 2007, pp.

1982–1988
Ó 2007 by the Ecological Society of America

PREDICTING LEAF PHYSIOLOGY FROM SIMPLE PLANT AND CLIMATE

ATTRIBUTES: A GLOBAL GLOPNET ANALYSIS
PETER B. REICH,1,3 IAN J. WRIGHT,2 AND CHRISTOPHER H. LUSK2
1
Department of Forest Resources, University of Minnesota, 1530 Cleveland Avenue North, St. Paul, Minnesota 55108 USA
2
Department of Biological Sciences, Macquarie University, Sydney, New South Wales 2109 Australia

Abstract. Knowledge of leaf chemistry, physiology, and life span is essential for global
vegetation modeling, but such data are scarce or lacking for some regions, especially in
developing countries. Here we use data from 2021 species at 175 sites around the world from
the GLOPNET compilation to show that key physiological traits that are difficult to measure
(such as photosynthetic capacity) can be predicted from simple qualitative plant
characteristics, climate information, easily measured (‘‘soft’’) leaf traits, or all of these in
combination. The qualitative plant functional type (PFT) attributes examined are phylogeny
(angiosperm or gymnosperm), growth form (grass, herb, shrub, or tree), and leaf phenology
(deciduous vs. evergreen). These three PFT attributes explain between one-third and two-
thirds of the variation in each of five quantitative leaf ecophysiological traits: specific leaf area
(SLA), leaf life span, mass-based net photosynthetic capacity (Amass), nitrogen content (Nmass),
and phosphorus content (Pmass). Alternatively, the combination of four simple, widely
available climate metrics (mean annual temperature, mean annual precipitation, mean vapor
pressure deficit, and solar irradiance) explain only 5–20% of the variation in those same five
leaf traits. Adding the climate metrics to the qualitative PFTs as independent factors in the
model increases explanatory power by 3–11% for the five traits. If a single easily measured leaf
trait (SLA) is also included in the model along with qualitative plant traits and climate metrics,
an additional 5–25% of the variation in the other four other leaf traits is explained, with the
models accounting for 62%, 65%, 66%, and 73% of global variation in Nmass, Pmass, Amass, and
leaf life span, respectively. Given the wide availability of the summary climate data and
qualitative PFT data used in these analyses, they could be used to explain roughly half of
global variation in the less accessible leaf traits (Amass, leaf life span, Nmass, Pmass); this can be
augmented to two-thirds of all variation if climatic and PFT data are used in combination with
the readily measured trait SLA. This shows encouraging possibilities of progress in developing
general predictive equations for macro-ecology, global scaling, and global modeling.
Key words: leaf life span; nitrogen; phosphorus; photosynthesis; plant functional type; specific leaf area.

INTRODUCTION Many ecosystem process models simplify real vegeta-

tion by dividing species into categories called ‘‘plant
Foliage attributes such as leaf structure, nutrient
functional types’’ (PFTs). Leading models include the
content, and net photosynthetic capacity are key
Sheffield, LPJ, and NCAR dynamic global vegetation
determinants of carbon dioxide and water vapor fluxes
models (Woodward et al. 1995, Bonan et al. 2003, Sitch
between vegetation and the atmosphere at every
et al. 2003, Woodward and Lomas 2004) and biogeo-
temporal and spatial scale and of biogeochemical cycles
graphic and biogeochemical models such as BIOME4
that link soil, climate, and atmosphere at the same (Kaplan et al. 2003) and BIOME-BGC (White et al.
scales. Thus, the ability to characterize key leaf 2000). In these models, each PFT has a particular set of
functional traits such as photosynthetic capacity for traits and makes up a particular proportion of the
species and communities at regional, continental, and vegetation at a site. But recent progress in understanding
global scales is important for a variety of scientific ecological strategy variation across plant species (Reich
disciplines, including global biogeography and macro- et al. 1997, 2003, Wright et al. 2004) suggests
ecology (Diaz et al. 2004), as well as for vegetation, possibilities for building new vegetation schemes that
carbon balance, and land surface models (e.g., Haxeltine are conceptually cleaner, computationally easier, and
and Prentice 1996, Bonan et al. 2003, Sitch et al. 2003) underpinned by richer data and that express trait
such as those used to predict responses to changes in variation more satisfactorily.
land use, atmospheric chemistry, and climate. Variation across species in most ecologically impor-
tant traits is naturally continuous rather than divided
Manuscript received 30 October 2006; revised 9 March 2007;
into classes. Further, although traits such as leaf life span
accepted 12 April 2007. Corresponding Editor: H. P. Schmid. or mass-based leaf nitrogen content differ on average
3 E-mail: preich@umn.edu between herbs, grasses, and woody plants, there is wide
1982
October 2007 PREDICTING LEAF FUNCTIONAL TRAITS 1983

spread within categories and broad overlap between were published in Wright et al. (2004). Only site-based
them (Reich et al. 1997, Wright et al. 2004). Similarly, data sets were used, i.e., those to which we could
comparisons between habitats often show overlapping reasonably attach climate data. The total database
ranges of trait values, despite different averages. Thus, consists of 2548 species–site combinations from 175
an alternative to using PFTs in vegetation models would sites: 2021 different species in total, with 342 species
be to describe trait variation among sets of coexisting occurring at more than one site (data sources and the
species with a mean and an index of spread for each trait. data set itself are available in appendices associated with
Although a wealth of gas exchange data has been Wright et al. [2004]). Site mean annual temperature
published, there is an inevitable bias towards econom- (MAT) ranged from 168C to 27.58C, and mean annual
ically important species in developed regions of the rainfall ranged from 133 to 5300 mm/yr. This covers
world, with few data available from some less developed most of the range of MAT/rainfall space in which higher
regions (Wright et al. 2004, but see Han et al. 2005, He plants are found. We focus on the following leaf traits,
et al. 2006). This imbalance is a potential limitation to all defined as in Wright et al. (2004): photosynthetic
the generality of global models of land surface processes. capacity per unit leaf mass (Amass); leaf nitrogen and
Our objective in this paper is to advance our capacity to phosphorus concentration per unit mass (Nmass and
predict patterns of variation in leaf traits, with a view to Pmass); leaf life span; and SLA, defined here as the one-
improving coverage for regions where ecophysiological sided projected area of foliage per unit dry mass
data are scarce. We ask whether well-known qualitative (Cornelissen et al. 2003).
characteristics of species, which are often used to define Species were grouped in PFTs by the simplest possible
plant functional groups, provide a useful foundation for groupings: phylogeny (contrasting gymnosperms and
making such predictions, particularly when combined angiosperms), growth form (grasses, forbs, shrubs, and
with other widely available data, such as climate data. trees), and leaf habit (deciduous vs. evergreen). The
Second, we ask whether these qualitative plant charac- original GLOPNET data set included data for vines and
teristics are more poorly, similarly, or better related to a number of other types. There was insufficient
plant ecophysiological traits than are climate variables, replication of these, so only the four main types were
which are also generally more widely available than included herein.
ecophysiological data. Third, we ask whether the Site climate was considered in terms of temperature,
addition of information about an easily obtained leaf rainfall, vapor pressure deficit (VPD), and solar
trait, specific leaf area (SLA), to the PFT and climate radiation annual, summed or averaged over annual
data allows substantial improvements in ability to periods as well as for the growth season period only.
predict the less accessible traits such as gas fluxes, Details on sources and calculations of climate data were
chemistry, and leaf life span. If so, this would suggest provided in Wright et al. (2004). Results using yearly
that coordinated programs to measure SLA could be and growth season climate indices or indices of
used to improve global data bases about leaf physiology seasonality were similar; hence for brevity we only
in general. We focus on SLA because this trait, besides report results relating to yearly climate averages.
being easy to measure, is a strong (positive) correlate of Climate variables were cross-correlated to an extent.
the photosynthetic capacity and potential relative Across the 175 sites, VPD and solar irradiance were
growth rate of plants and inversely related to the degree more closely associated with MAT than with mean
of physical defense of a leaf (Reich et al. 1991, 1997, annual rainfall although, clearly, both MAT and rainfall
Wright and Westoby 2002, Cornelissen et al. 2003). affect a property such as VPD.
We used the Global Plant Trait Network (GLOP- Where traits were reported separately for sun leaves
NET) database (Wright et al. 2004, 2005) to assess and shade leaves in the source publications, only the
several alternative approaches to estimating photosyn- former were used. If data were presented separately for
thetic capacity, leaf life span, leaf nitrogen content, and recently matured and old leaves, recently matured leaves
leaf phosphorus content from more easily obtained and were used. That is, where there was a choice, we used
more widely available data. This database covers 2222 data from leaves closer to their ‘‘peak’’ physiological
species from 175 sites on six continents. The present stage, prior to significant age- or light-related decline in
paper builds on previous work that has used these data nutrient contents and photosynthetic capacity (Reich et
to assess the generality of scaling relationships among al. 1991).
quantitative leaf ecophysiological traits that define
trade-off surfaces (e.g., Wright et al. 2004) or the Data analysis
relationship of such surfaces to large-scale climate All leaf traits were approximately log-normally
variability (Wright et al. 2005). distributed across the data set, as were site rainfall and
METHODS VPD. Accordingly, these variables were log10-trans-
formed prior to analyses. Mean annual temperature and
Leaf and climate data solar radiation were left untransformed since their
Data were compiled from both published and distribution was approximately normal. Simple correla-
unpublished sources, although all quantitative data tion and multiple regression analyses were used for
 Ecological Applications
1984 PETER B. REICH ET AL.
Vol. 17, No. 7

TABLE 1. Multiple regression analyses of five ecophysiological traits (mass-based net photosynthetic capacity [Amass], phosphorus
content [Pmass], nitrogen content [Nmass], leaf life span, and specific leaf area [SLA]) in relation to three plant functional types
(PFT) attributes.

Phylogeny Growth Leaf

Trait Mean 6 SE R2 n F form F habit F
Amass 1.973 6 0.019 0.51 747 94.9*** 32.5*** 199.5***
Pmass 1.102 6 0.046 0.37 736 4.3* 29.5*** 153.7***
Nmass 0.217 6 0.010 0.33 1931 28.0*** 76.6*** 290.2***
Leaf life span 0.965 6 0.021 0.67 723 77.9*** 48.9*** 637.9***
SLA 1.991 6 0.012 0.40 2164 52.3*** 92.9*** 511.0***
Notes: We report the mean of the logarithm of each parameter and the standard error (SE) of the predicted values; n is sample size.
The F values and their significance are shown for each of the dependent variables. All whole models were significant at P , 0.001.
* P , 0.05; *** P , 0.001.

quantifying relationships between single leaf traits and contrast, individual species can be classified according
PFT attributes, climate variables, SLA, and their to combinations of major groupings (such as herb vs.
combination. Inclusion of interaction terms did little trees and deciduous vs. evergreen) that differ in
to improve the variance explained (typically by 1–4%) predictable ways in their average leaf traits. Therefore,
compared to models without interaction terms (those despite appreciable variation in individual leaf traits
shown in Tables 1–4). Moreover, models with interac- within any individual PFT grouping, these differ
tions (including those with all interaction terms or the sufficiently on average among PFTs (Fig. 1) that their
best models following backwards stepwise regression) (three-way) combination explains one-third to two-
had Akaike’s Information Criterion values that were thirds of global variation among species in the five leaf
similar to or usually greater than the simpler models (no traits.
interactions) with fewer terms, and thus the latter All three PFT groupings (angiosperm/gymnosperm,
models were considered the best. Thus, interactions functional type, or leaf habit) were significant predictors
were uniformly omitted from all presented models of all five ecophysiological traits in the ‘‘PFT alone’’
(Tables 1–4, Appendices A–C). All statistical procedures models (Table 1, Appendix A). As generally observed
were carried out with JMP Statistical Software 5.0.1.a previously, species that are gymnosperms, evergreen, or
(SAS Institute, Cary, North Carolina, USA). woody on average occupy positions closer to the ‘‘slow
metabolism’’ end of the leaf trait gradient than species
RESULTS AND DISCUSSION that are angiosperms, deciduous, or herbaceous (Fig. 1).
Easily available qualitative PFT information ex- The ‘‘slow metabolism’’ end of the leaf economics
plained a substantial portion of the total variation in spectrum is associated with low Amass, Nmass, Pmass,
all five leaf functional traits (Table 1, Appendix A). The and SLA and persistent leaves (Reich et al. 1997, 1999,
PFT data by itself explained 33%, 37%, 40%, 51%, and Wright et al. 2004). In the ‘‘climate-alone’’ models, from
67% of the variation in Nmass, Pmass, SLA, Amass, and one to four of the climate metrics were significant
leaf life span, respectively (these represent whole-model predictors of each of the five ecophysiological traits, and
r2 values with phylogeny, growth form, and leaf habit each climate metric was significant in models for 2, 3, or
included). In contrast, climate data (MAT, annual 4 of the ecophysiological traits (Table 2).
precipitation, mean VPD, and solar irradiance) collec- Although most evergreen species have longer-lived
tively explained between 5% and 20% of variation in the leaves than most deciduous species, there is a class of
same five leaf traits (Table 2, Appendix B). Thus, species that are both evergreen and characterized by
variation among species within sites is sufficiently large short-lived foliage (Reich et al. 1997, 1999, Wright et al.
that climate alone predicts only a small fraction of leaf 2004). This group of species is numerically small globally
functional trait variation (Wright et al. 2005). In (e.g., ;4% of the woody plants for which we know

TABLE 2. Summary of multiple regression analyses of five ecophysiological traits (as in Table 1) in relation to climate metrics:
mean annual temperature (MAT), annual precipitation (PPT), mean annual vapor pressure deficit (VPD), and yearly mean
daily-summed solar radiation (RAD).

Trait R2 n MAT F PPT F VPD F RAD F

Amass 0.05 764 0.07 6.8*** 1.8 2.8
Pmass 0.19 737 19.0*** 0.4 8.5*** 83.6***
Nmass 0.12 2026 5.6* 5.1* 0.07 91.9***
Leaf life span 0.10 744 0.08 0.29 0.57 14.3***
SLA 0.20 2331 19.3*** 61.3* 4.9* 99.4***
Note: All whole models were significant at P , 0.001.
* P , 0.05; *** P , 0.001.
October 2007 PREDICTING LEAF FUNCTIONAL TRAITS 1985

TABLE 3. Whole-model R2 values for multiple regression analyses of five ecophysiological traits (as in Table 1) in a series of models
with increasing numbers of independent variables.

Leaf
Model Amass Pmass Nmass life span SLA
VEGETATION 0.51 0.37 0.33 0.67 0.40
CLIMATE 0.05 0.19 0.12 0.10 0.20
VEG þ CLIMATE 0.54 0.43 0.37 0.68 0.51
VEG þ CLIMATE þ SLA 0.66 0.65 0.62 0.73
VEG þ CLIMATE þ SLA þ N 0.73 0.76 0.75
VEG þ CLIMATE þ SLA þ N þ leaf life span 0.80 0.76
Notes: The models labeled ‘‘VEGETATION’’ included phylogeny, growth form, and leaf habit; ‘‘CLIMATE’’ included mean
annual temperature, rainfall, mean vapor pressure deficit, and mean solar irradiance; ‘‘VEG þ CLIMATE’’ included the first two
sets combined; ‘‘VEG þ CLIMATE þ SLA’’ included the prior set plus SLA; ‘‘VEG þ CLIMATE þ SLA þ N’’ included the prior
set plus percentage of nitrogen. See Tables 1, 2, and 4 for more details regarding the models in the first, second, and fourth rows,
respectively.

evergreen/deciduous status in the GLOPNET survey), atory power by 3–11% for the five quantitative leaf traits
but includes a species type, woody evergreen pioneers, (Table 3), thus accounting for between 37% (Nmass) and
that is important in tropical forests and that has traits 68% (leaf life span) of global variation in these five traits.
similar to woody deciduous pioneers, including short If the most easily measured of the quantitative traits
leaf life span, high nutrient concentrations, and high (SLA) is also included in the model, an additional 5–
metabolic rates (Reich et al. 1991, 1997, 1999). In the 25% of the variation is explained for each of the four
current analyses, our simple division of all taxa into other leaf traits, with the result that between 62% and
deciduous vs. evergreen classes thus includes species 73% of global variation in Nmass, Pmass, leaf life span,
with leaves with deciduous-like characteristics in the and Amass can be explained (Tables 3 and 4, Appendix
evergreen class. Does this weaken our models? We C). The PFT and climate metrics generally remain
assessed this in two ways. First we created a classifica- significant in these models: for instance, Amass was
tion that lumped species into one of two groups: (1) positively related to MAT and solar radiation and was
evergreen species with leaf life span 8 months and (2) negatively related to rainfall and VPD (Table 4,
deciduous species plus evergreen species with leaf life Appendix C).
span ,8 months (similar to dividing all species into Adding Nmass to predictive models may also be a
those with leaf life span 8 months vs. ,8 months). viable option. Although not as easy to determine as
Secondly, we simply divided species into those consid- SLA, the total cost of analyzing, say, 1000 samples for N
ered ‘‘pioneer’’ species and those that are not. The first content would still be 10-fold lower than the cost of
classification improved most model fits by 2–5% purchasing an infrared gas analysis system and associ-
(compared to Table 1) but requires data quite difficult ated chambers for making gas exchange measurements.
to obtain. The second classification improved model fits Measuring canopy Nmass from remote sensing may also
marginally, if at all. Thus, the existence of species with become operationally feasible in the future (Smith et al.
the relatively unusual trait combination of evergreen but 2003). When Nmass was added to regression models
short-lived leaves only modestly lessens the predictive already containing VEG, CLIMATE, and SLA, an
power of the simple models based on simple and widely additional 7% and a total of 73% of total variation in
available classifications. Amass was explained (Table 3), or 75–76% of total
Can we increase explanatory power by combining variation in leaf life span and Pmass.
qualitative plant attributes and climatic data? Adding To test whether these multiple regression relationships
the climate metrics to the PFT data increased explan- based on the entire data set would yield reliable

TABLE 4. Multiple regression analyses of four ecophysiological traits (mass-based net photosynthetic capacity [Amass], phosphorus
content [Pmass], nitrogen content [Nmass], and leaf life span) in relation to plant functional type (PFT) information, climate
metrics (mean annual temperature [MAT], annual precipitation [PPT], mean annual vapor pressure deficit [VPD], yearly mean
daily-summed solar radiation [RAD]), and specific leaf area (SLA).

Phylogeny Growth Leaf

Trait R2 n F form F habit F MAT F PPT F VPD F RAD F SLA F
Amass 0.66 741 35.4*** 22.8*** 48.4*** 8.8** 39.6*** 5.9* 8.8** 274.4***
Pmass 0.65 724 24.58*** 12.7*** 7.6** 0.6 31.2*** 16.1*** 23.8*** 423.9***
Nmass 0.62 1852 0.03 26.9*** 4.5* 0.2 49.9*** 4.4* 7.5** 1151.2***
Leaf life span 0.73 671 49.4*** 27.4*** 259.2*** 0.5 0.6 13.2*** 9.9** 114.3***
Notes: The F values and their significance are shown for each of the dependent variables. This provides details of the fourth
model in Table 3. All whole models were significant at P , 0.001.
* P , 0.05; ** P , 0.01; *** P , 0.001.
 Ecological Applications
1986 PETER B. REICH ET AL.
Vol. 17, No. 7

predictions, we derived the same relationships as shown

for the entire data set (Tables 1–4) using one-half of the
data (randomly chosen) and then predicted the quanti-
tative leaf traits for the other half of the data set. The fits
(Fig. 2) were generally very close to those generated for
the entire data set.
There is close coordination, physiologically and
evolutionarily, between the five quantitative leaf traits
measured in this study (Reich et al. 1997, 1999, Wright
et al. 2004), and thus a sizeable fraction of total
variation in one trait can be explained by other traits.
However, all of these traits require time, effort, and
funds to obtain, and all but SLA also require substantial
equipment, analytical, or time costs. Thus, other than
SLA the other traits do not offer an easy, simple, cheap
surrogate index for functional leaf traits, and obtaining
SLA data itself is not without time, effort, and cost
(Cornelissen et al. 2003). In contrast, the combination of
PFT information and simple climate metrics, both
generally and freely available, explain a similar fraction,
or roughly half (mean of 51% for the five leaf traits;
Table 3), of all variation in the selected leaf traits, as do
individual leaf traits in relation to one another (mean of
48% for the 10 bivariate relations of the five quantitative
leaf traits presented in this paper; Wright et al. 2004).
Given the much greater availability of simple PFT
information and databases on vegetation distribution
than of physiological data, these offer promise for
incorporation into predictive models as well as models
driven by remotely sensed information. Moreover, the
combination of PFT, climate, and SLA explain approx-
imately two-thirds of global variation in the other leaf
traits. Given that we do not have enough data to fully
explore interactions between PFTs and climate variables
and that we use only linear effects, there is ample
opportunity for further refinements of such analyses and
for development of predictive models that are statisti-
cally more sophisticated than used here. We view this as
a promising beginning, given that a greater quantity of
and better metrics for each of these variables can be
obtained, and as well, other metrics may become
available (e.g., soils data) that additionally can be used
with these in future models.
Improved ability to predict leaf attributes at a species
level will be particularly advantageous in multispecies
canopies. Given the heterogeneity among species within
sites (Reich et al. 1999, Wright et al. 2004), the
importance of species and functional-group heterogene-
ity and diversity to ecosystem processes, including
responses to global environmental change (e.g., Reich
et al. 2004), and the likelihood that the physiological
FIG. 1. Box plots of five leaf traits: (a) specific leaf area
(SLA), (b) leaf life span, (c) mass-based N concentration
(Nmass), (d) mass-based P concentration (Pmass), and (e) mass- distribution of points for each variable and group. The ends of
based net photosynthetic capacity (Amass), by qualitative plant the box are the 25th and 75th percentiles or quartiles. The gray-
functional type (PFT) classifications that separate species by shaded area between the quartiles is the interquartile range. The
phylogeny (angiosperm vs. gymnosperm), leaf habit (Ever, line across the middle of the box is the median value. The lines
evergreen; Dec, deciduous), and growth form (G, grass; H, extending from the ends of the box denote the maximum and
herb; S, shrub; T, tree). The box plots summarize the minimum values.
October 2007 PREDICTING LEAF FUNCTIONAL TRAITS 1987

FIG. 2. Relationships between observed and predicted (a) mass-based net photosynthetic capacity (Amass), (b) leaf life span, (c)
mass-based N concentration (Nmass), and (d) mass-based P concentration (Pmass). Predictive equations were generated from a
randomly selected half of the data set, using plant functional type (PFT) information, climate data, and specific leaf area, and were
applied to the remaining data. Fits (R2) of the relationships were 0.61, 0.71, 0.59, and 0.66, respectively, for panels (a)–(d) (all P ,
0.0001), and slopes were near 1:1.

response of a community canopy based on mean values the U.S. National Science Foundation (0322057, 0080382, and
is not necessarily the same as the responses of a real 0128958). The authors are aware of no conflicts of interest
canopy made up of species differing widely in intrinsic relative to this work.
leaf traits, it may prove necessary at some point to LITERATURE CITED
consider this functional diversity in models (Garnier et Bonan, G. B., S. Levis, S. Sitch, M. Vertenstein, and K. W.
al. 2004). In many, but certainly not all, areas of the Oleson. 2003. A dynamic global vegetation model for use
globe there is some information about the compositional with climate models: concepts and description of simulated
make-up of vegetation communities, even when there is vegetation dynamics. Global Change Biology 9:1543–1566.
little or no information about ecophysiology. In such Cornelissen, J. H. C., et al. 2003. A handbook of protocols for
cases, predicting the traits of the species in those standardised and easy measurement of plant functional traits
worldwide. Australian Journal of Botany 51:335–380.
communities and then aggregating those traits to the Dı́az, S., et al. 2004. The plant traits that drive ecosystems:
community as a whole weighted by their relative evidence from three continents. Journal of Vegetation
abundances will provide not only a mean but a notion Science 15:295–304.
of the functional diversity of the community. In Garnier, E., J. Cortez, G. Billès, M.-L. Navas, C. Roumet, M.
contrast, predictions based solely on climate or a Debussche, G. Laurent, A. Blanchard, D. Aubry, A.
dominant functional type would only provide a mean Bellmann, C. Neill, and J.-P. Toussaint. 2004. Plant
functional markers capture ecosystem properties during
value. As models improve, one would expect that land secondary succession. Ecology 85:2630–2637.
surface and vegetation dynamics models would incor- Han, W., J. Fang, and D. Guo. 2005. Leaf N and P
porate heterogeneity of vegetative properties as well as stoichiometry across 753 terrestrial plant species in China.
the average of those properties. New Phytology 168:377–385.
Haxeltine, A., and I. C. Prentice. 1996. BIOME3: an
ACKNOWLEDGMENTS equilibrium terrestrial biosphere model based on ecophysio-
We thank the many researchers whose work contributed to logical constraints, resource availability and competition
the GLOPNET data compilation. This study was partially among plant functional types. Global Biogeochemical Cycles
supported by grants from the Australian Research Council and 10:693–709.
 Ecological Applications
1988 PETER B. REICH ET AL.
Vol. 17, No. 7

He, J.-S., Z. Wang, X. Wang, B. Schmid, W. Zuo, M. Zhou, C. Thonicke, and S. Venevsky. 2003. Evaluation of ecosystem
Zheng, M. Wang, and J. Fang. 2006. A test of the generality dynamics, plant geography and terrestrial carbon cycling in
of leaf trait relationships on the Tibetan Plateau. New the LPJ dynamic global vegetation model. Global Change
Phytology 170:835–848. Biology 9:161–185.
Kaplan, J. O., et al. 2003. Climate change and arctic ecosystems Smith, M. L., M. E. Martin, L. Plourde, and S. V. Ollinger.
II: modeling, paleodata-model comparisons, and future 2003. Analysis of hyperspectral data for estimation of
projections. Journal of Geophysical Research–Atmospheres temperate forest canopy nitrogen concentration: comparison
108(D19):8171.
between an airborne (AVIRIS) and a spaceborne (Hyperion)
Reich, P. B., D. S. Ellsworth, M. B. Walters, J. Vose, C.
sensor. IEEE Transactions on Geoscience and Remote
Gresham, J. Volin, and W. Bowman. 1999. Generality of leaf
traits relationships: a test across six biomes. Ecology 80: Sensing 41:1332–1337.
1955–1969. White, M. A., P. E. Thornton, S. W. Running, and R. R.
Reich, P. B., D. Tilman, S. Naeem, D. Ellsworth, J. Knops, J. Nemani. 2000. Parameterization and sensitivity analysis of
Craine, D. Wedin, and J. Trost. 2004. Species and functional the BIOME–BGC terrestrial ecosystem model: net primary
diversity independently influence biomass accumulation and production controls. Earth Interactions 4:1–85.
its response to CO2 and N. Proceedings of the National Woodward, F. I., and M. R. Lomas. 2004. Vegetation
Academy of Sciences (USA) 101:10101–10106. dynamics—simulating responses to climatic change. Biolog-
Reich, P. B., C. Uhl, M. B. Walters, and D. S. Ellsworth. 1991. ical Reviews 79:643–670.
Leaf lifespan as a determinant of leaf structure and function Woodward, F. I., T. M. Smith, and W. R. Emanuel. 1995. A
among 23 amazonian tree species. Oecologia 86:16–24. global land primary productivity and phytogeography model.
Reich, P. B., M. B. Walters, and D. S. Ellsworth. 1997. From Global Biogeochemical Cycles 9:471–490.
tropics to tundra: global convergence in plant functioning. Wright, I., P. B. Reich, and M. Westoby, and GLOPNET
Proceedings of the National Academy of Sciences (USA) 94:
researchers. 2004. The worldwide leaf economics spectrum.
13730–13734.
Reich, P. B., I. Wright, J. Cavender-Bares, J. Craine, J. Nature 428:821–827.
Oleksyn, M. Westoby, and M. B. Walters. 2003. The Wright, I. J., and M. Westoby. 2002. Leaves at low versus high
evolution of plant functional variation: traits, spectra, and rainfall: coordination of structure, lifespan and physiology.
strategies. International Journal of Plant Sciences 164:s143– New Phytologist 155:403–416.
s164. Wright, I. J., et al. 2005. Modulation of leaf economic traits and
Sitch, S., B. Smith, I. C. Prentice, A. Arneth, A. Bondeau, W. trait relationships by climate. Global Ecology and Biogeog-
Cramer, J. O. Kaplan, S. Levis, W. Lucht, M. T. Sykes, K. raphy 14:411–421.

APPENDIX A
Model parameters for the models shown in Table 1 (Ecological Archives A017-078-A1).

APPENDIX B
Model parameters for the models shown in Table 2 (Ecological Archives A017-078-A2).

APPENDIX C
Model parameters for the models shown in Table 4 (Ecological Archives A017-078-A3).