E cology Letters, (2 0 1 1 )
doi: 10.1111/j.l461-0248.2011.01658.x
REVI EW AND
SYNTHESI S
Relationships am ong n e t primary productivity, nutrients and
clim ate in tropical rain forest: a pan-tropical analysis
A b strac t
Cory C. C leveland,^* Alan R.
T ow nsend,^ Philip Taylor,^ Silvia
Alvarez-Clare,'^ M ercedes M. C.
B ustam ante,^ G eorge Chuyong,®
Solom on Z. Dobrowski,^ Pauline
Grierson,® Kyle E. Harms,® Benjamin
Z. Moulton,^® A lison Marklein,^®
William P a rto n ," S tep h en
Porder,^^ Sasha C. Reed,^® Carlos A.
Sierra,^'^ W hend ee L. Silver,^®
Edmund V. J. Tanner^® and William
R. Wieder®
Tropical rain forests play a dom inant role in global biosphere-atm osphere C O 2 exchange. A lthough climate and
nutrient availability regulate net primary production (NPP) and decom position in aU terrestrial ecosystems, the
nature and extent o f such controls in tropical forests rem ain poorly resolved. We conducted a meta-analysis o f
carbon-nutrient-chm ate relationships in 113 sites across the tropical forest biome. O ur analyses showed that
m ean annual tem perature was the strongest predictor o f aboveground N P P (ANPP) across aU tropical forests,
bu t this relationship was driven by distinct tem perature differences betw een upland and lowland forests. W ithin
lowland forests (< 1 0 0 0 m), a regression tree analysis revealed that foliar and sod-based measurem ents o f
phosphorus (P) were the only variables that explained a significant p roportion o f the variation in A N PP,
although the relationships were weak. H owever, foliar P, foliar nitrogen (N), litter decom position rate (k), sod
N and sod respiration were ad directly related w ith total surface (0-10 cm) soil P concentrations. O ur analysis
provides som e evidence that P availabdity regulates N P P and other ecosystem processes in lowland tropical
forests, but m ore im portantly, underscores the need for a series o f large-scale nutrient manipulations —
especially in lowland forests - to elucidate the m ost im portant nutrient interactions and controls.
Keyw ords
C arbon cycle, cdmate, decom position, meta-analysis, nitrogen, nutrient dmitation, phosphorus, primary
production, tropical forest.
Ecology letters (2011)
INT R O D U C T IO N
M ore than a quarter century ago, V itousek (1984) w rote that,
‘[inferential] analyses suggest that phosphorus (P) b u t n o t nitrogen (N)
availabdity limits dtterfad in a substantial subset o f intact tropical
forests.’ Since then, several global syntheses o f b o th fodar and dtterfad
nutrient data have indicated that P appears to be less abundant than N
(relative to plant demand) in many tropical forest sites (M cGroddy
et al. 2004a; Reich & Oleksyn 2004; T ow nsend et al. 2007). W hether
these patterns indicate P dm itation o f net primary production (NPP)
or not, however, remains an open question, as tropical forests often
have sufficient labde P in surface soils to satisfy the grow th
requirem ents o f N P P (Johnson et al 2003). This incom plete under­
standing has global implications: tropical forests contain c. 25% o f the
w orld’s terrestrial biom ass and sod C (Jobbagy & Jackson 2000),
exchange m ore w ater and carbon dioxide (CO 2) w ith the atm osphere
than any other biom e (Foley et al 2005), and account for at least onethird o f global annual terrestrial N P P (Grace et al 1995; Phddps et al
1998). Thus, identifying the relationships betw een nutrient availabdity
and C cycdng is critical for accurately predicting how these forests may
^ D epartm ent o f Ecosystem a n d C onservation Sciences, U niversity o f M ontana,
^ °D epartm ent o f Land, A ir a n d W ater Resources, U niversity o f California,
M issoula, M T 59812, USA
Davis, CA 95616, USA
^ D epartm en t o f E cology a n d E volutionary B iology, E nviron m ental Stu dies
Program a n d INSTAAR, U niversity o f C olorado, Boulder, CO 80309, USA
N atural R esource E cology L aboratory, C olorado S ta te University, Fort Collins,
CO 80523-1499, USA
^ D epartm en t o f E cology a n d E volutionary B io lo g y a n d INSTAAR, U niversity o f
^^D epartm ent o f E cology a n d E volutionary B iology, B row n U niversity,
C olorado, B oulder, CO 80309, USA
Providence, Rl 02912, USA
^ D epartm en t o f B io lo g y & School o f N atural Resources & E nvironm ent,
^^U.S. G eological Survey, S o u th w e s t B iological Science C enter, 229 0 S.W.
U niversity o f Florida, Gainesville, FL 32611, USA
R esource Blvd., M oab, U T 84532, USA
^ D epartam en to d e Ecologia, U niversidade d e Brasilia, Brasilia-DF, Brazil
^^D epartm ent o f B iological Processes, M ax-Planck-lnstitute fo r B io g eo ­
^ D epartm en t o f Plant a n d A n im a l Sciences, U niversity o f Buea, POB 63,
chem istry, 07745 Jena, G erm an y
Buea, C am eroon
^ D epartm en t o f F orest M a n a g em en t, U niversity o f M ontana, M issoula,
D e p a rtm e n t o f E nviron m ental Science, Policy, a n d M a n a g em en t, U niversity o f
California, B erkeley, CA 94720, USA
M T 59812, USA
^^ D epartm en t o f Plant Sciences, U niversity o f C am bridge, D ow n in g Street,
^School o f Plant B io lo g y M 090, The U niversity o f W estern A ustralia,
C am bridge, CB2 3EA, UK
3 5 Stirling H ighw ay, Crawley, WA 6016, Australia
* C orrespondence: E-mail: cory.clevelan d@ u m on tan a.edu
^ D epartm en t o f B iological Sciences, Louisiana S ta te University, B aton R ouge,
LA 70803, USA; a n d Sm ithsonian Tropical Research Institute, R epu blic o f Panama
© 2011 Blackwell Publishing L td /C N R S
2 C. C. Cleveland et al.
respond to (and control) many aspects o f natural and hum an-caused
environm ental change (Hungate et al 2003; T h o rn to n et al 2007;
B onan & Levis 2010; Z hao & Running 2010; T ow nsend et al. 2011).
In tem perate and high latitude ecosystems, a wealth o f direct
evidence (i.e. data from nutrient m anipulation experiments) suggests
w idespread N or N /P co-limitations to N P P (Aber et al 1991; G ough
et al 2000; Elser et al 2007), and som e data also suggest that N
lim itation may be com m on in m ontane tropical forests o n minimally
w eathered soils (Tanner et a l 1998). D irect evidence o f nutrient
lim itation in lowland tropical forests is rare, b u t indirect proxies
indicate an abundance o f N relative to P. These include: high foliar
(M cGroddy et al 2004a; Reich & Oleksyn 2004; Tow nsend et al 2007)
and litter (Vitousek 1984) N : P ratios; potentially high N fixation
(Cleveland et al. 1999; H o ulton et al 2008) and nitrification rates
(R obertson 1984); large N trace gas emissions (R obertson & Tiedje
1988; M atson & V itousek 1990) and N 2 fluxes (H oulton et al 2006);
elevated soil and foliar
values (MartineUi et a l 1999; N ardoto
et al. 2008); rapid internal N cycling (Silver et al. 2005; T em pler et al.
2008); and hydrological export o f biologically available N (H edin et al
2009). These empirical data have been used in conjunction w ith the
theoretical fram ew ork provided by W alker & Syers (1976) — w hich
predicts P poverty on highly w eathered soils such as the Oxisols and
Ultisols that dom inate the lowland tropics —to argue that P is likely to
Hmit N P P in lowland tropical forests.
Nevertheless, there is some reason to question the prevailing m odel
o f ‘N -rich’ and ‘P -poor’ lowland tropical systems. F or example, the
W alker & Syers (1976) m odel may n o t be appropriate for the tropics as
a whole, w hich vary widely in soil age (Porder & Hflley 2010), and
virtually every other factor th at could affect nutrient availability
(Townsend et al 2008). Walker & Syers (1976) focused exclusively on
soil age and the idea o f a steadily declining source o f rock-derived
nutrients through time, b u t m ore recent w ork has show n th at this
construct is too simplistic: w et sites can experience nutrient depletion in
a relatively short tim e (i.e. 10"*^—10^ years; Chadwick et al 1999),
different substrates may lose P at very different rates (Porder et al
2007), and thus tim e should be considered in concert w ith other state
factors to explain nutrient depletion in tropical soils (Porder & Hflley
2010). In addition, erosion can rem ove highly w eathered soils and
expose fresh parent material to weathering, thereby maintaining a
source o f new rock-derived elements to the ecosystem (Bern et al 2005;
Porder et al 2007). This phenom enon may help explain m ore
pronounced responses to N than to P additions in tropical m ontane
forests, w here relatively rapid rates o f erosion may replenish P, b u t n o t
N (e.g. T anner et al 1998), although erosion rates can be rapid in
lowland settings as well. Likewise, the deposition o f putatively rockderived nutrients via b o th marine aerosols and dust may provide large
o r even dom inant inputs o f P to som e systems (Chadwick et a l 1999;
O kin et al 2004; Pett-Ridge et al 2009). T hese data support the
argum ent that the tropics exhibit trem endous biogeochemical heter­
ogeneity (Townsend et al 2008), and suggest that a blanket m odel o f P
poverty in the lowland tropics may be an oversimplification.
Ultimately, the best tests o f potential nutrient limitation emerge
from direct manipulations o f nutrient supply, and analysing results
from multiple experiments can lend greater insight into large-scale
patterns. F or example, a recent global meta-analysis o f fertilisation
experiments concluded that limitation by b o th N and P is m ore
com m on across terrestrial ecosystems than is limitation by either
elem ent alone (Elser et al 2007). However, the Elser et al (2007)
meta-analysis included results from only one fertilisation experim ent
© 2011 Blackwell PubHshlng L td /C N R S
Review a n d Synthesis
conducted in a primary forest in the lowland tropics in w hich N and P
were added in a fuU-factorial design, and w here aboveground N PP
(ANPP) (i.e. fine HtterfaU a n d /o r biomass grow th increment) was
included as a response variable (M irmanto et al 1999). T here are a few
m ore studies from upland and secondary lowland forests (Tanner et al
1998; D avidson et al 2007), b u t the paucity o f data from fertilisation
experiments conducted in tropical forests precludes a meta-analysis o f
how N P P responds to direct m anipulations o f potentially limiting
elements.
G iven the lack o f data from fertilisation experiments in tropical
forests, here we used meta-analyses to ask. How does N P P vary with
climate, N and P availability across the tropical forest biome? In addition,
given the fact th at P is widely hypothesised to Hmit N P P in many
lowland tropical forests (Vitousek 1984; Lambers et al 2008; H edin
et a l 2009), we were particularly interested in exploring possible
relationships betw een soil P avaflabflity and multiple C cycHng metrics
am ong lowland forest sites. Like other researchers w ho have explored
this question over the past three decades, we were forced to explore
potential nutrient constraints using indirect metrics. Correlational
analyses are n o t direct tests o f nutrient Hmitation, and thus
conclusions draw n using such approaches have inherent weaknesses.
However, indirect analyses can provide im portant insight and help
identify w here our understanding is poor, and w here direct tests are
m ost needed (e.g. V itousek 1984; V itousek & Sanford 1986).
In addition, we note that elements other than N o r P may also be
im portant in regulating ecosystem function in tropical forests (Cuevas
& M edina 1988; B arron et a l 2008; K aspari et al 2008), but we
focused o n relationships betw een N and P and the m ajor features o f
the tropical C cycle for three reasons. First, recent global syntheses
underscore a long-held beflef that N , P o r bo th o f these elements Hmit
multiple processes in many terrestrial ecosystems w orldwide (Elser
et a l 2007; LeBauer & T reseder 2008). Second, given the dom inant
role o f the tropics in the global C cycle, multiple studies have
higlflighted the need to understand how N and P regulate tropical C
storage and exchange (Bonan & Levis 2010; T ow nsend et al 2011).
Third, data describing N and P concentrations in plants and soils are
relatively com m on, whereas those describing other nutrients are m uch
m ore Hmited.
M ETH O D S
W e assembled data from the literature and accessible databases and
conducted a meta-analysis o f relationships betw een cHmate, nutrient
status and multiple metrics o f b o th above- and below -ground C
cycHng in upland and lowland m oist and w et tropical forests. AU data
were obtained either from the literature o r from accessible electronic
databases (e.g. N ational Center for Ecological Analysis and Synthesis
[NCEAS], O ak Ridge N ational L aboratory D istributed Active Archive
Center [O RNL DAAC]). T o locate published data sources, we
conducted searches using onHne databases (e.g. the ISI W eb o f
Science) or search engines (e.g. G oogle Scholar) using the search
terms: tropical forest, nutrient, nutrient Hmitation, rfltrogen, ph o s­
phorus, primary productivity, decom position and soil respiration.
O nce relevant papers were located, references therein were checked in
an attem pt to locate others that may n o t have been accessible using
the ISI database. W e also included som e data published in previous
syntheses (Elser et al 2007; Tow nsend et al 2007; LeBauer & Treseder
2008; Q uesada et a l 2009), and examined relevant references cited
therein for additional data.
Nutrients, climate a n d tropical NPP 3
Review a nd Synthesis
In some cases, values for a given characteristic (e.g. foliar P) for
each site represent means o f aU available data for th at characteristic.
F or example, w ithin a site, w hen foliar nutrient concentrations were
assessed for multiple species, we used those data to generate a simple
(site) m ean value for th at characteristic. W hen multiple data
were available for a single species characteristic (e.g. foliar nutrients),
those data were averaged to generate a species mean, and species
m eans were used to calculate the site m ean for th at characteristic.
W e did n o t weight the foliar nutrient data by the relative abundance o f
species. D ata collected multiple times (e.g. seasonally) were averaged
to generate a single mean. Finally, data from multiple sites w ithin a
single geographical region (e.g. spanning topographical, Uthologic or
pedogenic gradients) were treated as separate sites in the database.
In sum, the final database included data from 113 tropical forest
sites, although n o t all data were available from all sites. O u r analysis
included data from b o th upland in — 32) and lowland forests in — 81)
located w ithin tropical latitudes (+ 23.5°N and S), and all sites in the
database were either m oist or w et tropical forests receiving a
m inim um o f 1300 m m m ean annual precipitation (MAP) (See
Table 51). T he majority o f data were from sites in the A m azon Basin
— the largest tract o f rem aining tropical forest in the w orld — with
fewer sites in N o rth and Central America and Asia, and only one site
in Africa and Australia respectively (Fig. 1). D ata were entered on
separate online accessible electronic data entry forms available
through N CEA S and im ported into a single database prior to analysis.
W e selected data that were obtained using consistent sampling (e.g.
soil depth) and m easurem ent techniques, and all data were converted
to com m on units prior to analysis. D ata used in this analysis are
available at the public repository at N CEA S (h ttp ://k n b .ec o in fo rmatics .o rg /k n b / style / skins / nceas /).
T o assess relationships betw een nutrient availability and N P P , we
assembled data on the following C cycle com ponents: (1) total fine
UtterfaU (sum o f leaves, twigs < 2 m m diam eter and reproductive
parts, (2 ) biomass grow th increm ent (based o n measured increases in
tree stem diameters), (3) A N PP, calculated as the sum o f observed
UtterfaU and biomass grow th increm ent, (4) Utter decom position rate
{k, the decom position rate constant, a value that describes the rate o f
decom position o f Utter on the forest floor) and (5) soU respiration (soU
C O 2 flux). F or sites in w hich data describing only one com ponent o f
A N P P were avaUable, A N P P was modeUed using the observed
relationships betw een UtterfaU o r biomass increm ent and A N P P and
equations generated using data from sites w here they were avaUable
(Figure SI).
od*
o
Figure 1 M ap illustrating the global distribution o f th e tropical fo rest sites used in
the meta-analysis. O p e n circles d en o te low land forest sites (< 1000 m) and triangles
d enote upland forest sites (> 1000 m).
N ext, we explored relationships betw een A N P P (the response
variable) and a num ber o f predictors, including cUmate (mean annual
tem perature [MAT| and MAP), soU type, nutrient cycUng variables
(foUar N and P, surface soU [0—10 cm] inorganic N (nitrate [N O 3 ]
and am m onium [NHy^]) content), decom position [k) and total surface
soil N and P content. Estim ates o f ‘avaUable soU P ’ may be m ore
appropriate than total soU P, b u t the diverse indices o f P avaUabUity
m eant th at com parable data were too sparse to include in the m eta­
analysis. M oreover, SUver (1994) found th at total P and indices o f
available P were simUarly related to UtterfaU nutrient use efficiency in a
wide range o f tropical forests, probably due to the exchange betw een
labUe and refractory soU P pools (Richter et al. 2006; Syers et al 2008).
FinaUy, we excluded data from two sites in the HawaUan Islands
(Thurston [300 years] and Laupahoehoe [20 kyr]), due to a com bina­
tion o f very young soU ages, substantial elem ent inputs from rapidly
w eatherable volcanic ash and low overaU tree diversity (Vitousek
2004), and because these tw o sites are in m any ways significantly
dissimUar from the majority o f tropical forest sites used in the analysis.
D ata from m ore successionaUy advanced sites w ith m ore w eathered
substrates in the HawaUan Archipelago (e.g. Maui and K aua’i) were
included in the analysis.
Statistical analysis
T he relationship betw een the response variable (ANPP) and each o f
the predictor variables was first analysed using a classification and
regression tree (CART) analysis (McCune & G race 2002). We
perform ed the regression tree analysis in tw o steps. First, w e related
A N P P to aU predictor variables including a num ber o f broadly defined
variables (e.g. soil type). N ext, we conducted a second CART analysis
in w hich we analysed upland and lowland forests separately (see
below). CART was im plem ented using the rpart Ubrary in R 2.11.1
(R developm ent Core team). Regression trees were iiUtiaUy over-fit,
and optim al tree sizes were determ ined using cost-complexity pruiUng
based o n 10-fold cross-vaUdation (Venables & Ripley 2002).
W e chose to use CART for several reasons. First, CART is ideaUy
suited for ascertaiiUng the relative im portance o f predictors in
explaiiUng variation in the response variable. CART can handle
unbalanced data w ith missing values —a characteristic o f this dataset —
and one th at can preclude analysis using other regression o r structural
equation approaches (D e’ath & Fabricius 2000). Regression trees are
also invariant to m onotonic transform ations o f the predictor variables.
Finally, CART effectively handles data th at contain higher order
interactions am ong variables (Uke the nutrient interactions that have
been hypothesised in tropical forests; T ow nsend et al. 2011).
H owever, CART also has som e Umitations. CART produces a
hierarchical tree-Uke classification o f a dichotom ous form to classify
observations into defined groups, and thus it can be difficult to assess
the functional form (e.g. Unear, curvUinear, etc.) o f the relationship
betw een the response and the predictors at each spUt. In addition,
CART can have difficulty modeUing sm ooth functions, and is prone
to over-fitting such that small changes in traiiUng data can result in a
different series o f spUts (EUth et al. 2008).
G iven this, following the CART analysis, we used regression and
correlation analysis (Pearson-product m om ent) to m ore thoroughly
investigate the relationships th at emerged in the CART analysis o f the
lowland forest site data. W e focused o n lowland forests for several
reasons. First, they play a dom inant role in the global C cycle, yet the
nature and extent o f nutrient Hmitation o f N P P in lowland forests is
>2011 Blackwell Pubkstiing L td /C N R S
4 C. C. Cleveland et al.
Review an d Synthesis
poorly understood. Second, lowland tropical forests represent a m uch
larger proportion o f the tropical forest biom e area than do m ontane
forests (Bruijnzeel et al 2011). FinaUy, the relatively large num ber o f
lowland sites (i.e. 81) in the database aUowed us to confidently explore
multiple possible relationships betw een C cycUng, cUmate and
nutrients. Correlation analyses were perform ed using SPSS version
19 for W indows (SPSS, Chicago, IL, USA), and in aU cases
significance was determ ined at P < 0.05.
RESULTS
T he first regression tree (including cUmate and soil variables from aU
sites in the database) explained 29% o f the variance in tropical A N P P
(Fig. 2A), and M AT was the m ost im portant predictor, explaining the
majority (20%) o f the explained variance in A N P P (Fig. 1). In cool
sites (< 18.1 °C), A N P P averaged 8.92 Mg ha~^ y ear~ \ whereas
A N P P in w arm er sites (> 18.1 °C) was higher ( > 1 4 Mg ha ^ year ^).
In w arm sites, the next spUt, which explained 9% o f the remaining
variance, was based on soU type: one branch included sites with
Ultisols or Oxisols with relatively low A N P P (14.74 Mg ha ^ year ^),
and the other consisted o f forests growing on other soU types (e.g.
Entisols, Inceptisols or Histosols) with relatively high A N P P
(18.76 M g ha“ ' year“ ^) (Fig. 2A).
The emergence o f soU order as a significant predictor o f A N P P
am ong the w arm tropical forest sites in the initial CART analysis
reflects potential influences o f soU fertUity on productivity. In
addition, because o f the strong relationship betw een tem perature
and elevation, and in Ught o f the widely held beUef that P Umits A N PP
in lowland tropical forests, we analysed the data from upland and
lowland forests separately. In classifying ‘lowland’ forests, we chose a
threshold elevation o f 1000 m (sensu Holdridge et al 1971). This
threshold corresponded to an M AT breakpoint o f 20.7 °C (as
opposed to the 18.1 °C suggested by the CART analysis); ‘upland’
forests included aU forests above 1000 m (20.7
M A I). In contrast
to the im portance o f tem perature in explaining variations in A N P P
observed across aU the data (Fig. 1), M AT did n o t emerge as a strong
or significant predictor o f A N P P in either the upland or the lowland
24
I
# y = 2 .4 5 x + 12.65 _
•
r^ = 0.089;
P = 0 .02 5-
22
&
—
I
CO
CL
CL
Z
<
0
1
0.5
1.5
2
2.5
Foliar P (pg g - '')
25
(a)
y = 0.004X + 14.46
MAT f C )
•
d = 0.057;
P = 0.083 -
20 %
I
CO
< 18.1
> i8 .rc
CL
CL
z
<
0
Soil ty p e
200
U ltisols/
O x iso ls
9%
14.74 (70)
o t h e r soil
ty p e s
800
1000
F oliar P
7%
2.5
y = 0 .0 0 1 x + 0.726
• d = 0.296;
P < 0.0001
18.76 (21)
I
D)
D)
I_ _
<0.75
600
T o t a l soil P (g g "
8.92 (22)
(b)
400
>0.75
Q_
••
CO
o
12.82(12)
15.84 (44)
LL
0.5
Figure 2 R egression tree showing: (a) h o w climate and soil type explain variance in
A N P P across th e tropical fo rest biom e; and (b) h o w foliar P explains variance in
A N P P across low land tropical forests. P red icto r variables are depicted at the to p o f
each branch, p red icto r variable thresholds at the side o f each b ran ch and the m ean
0
200
400
600
800
1000
T o t a l soil P ( | j g g~ ^)
A N P P (including the n u m b er o f observations in parentheses) is rep o rted below the
term inal nodes. T h e height o f each bran ch indicates th e relative p ro p o rtio n o f the
Figure 3 R elationships am ong P cycling variables and A N P P in low land tropical
total sum o f squares explained by th a t split.
forests, (a) Foliar P vs. A N P P ; (b) total soil P vs. A N P P ; (c) total soil P vs. foliar P.
© 2011 Blackwell Publishing L td /C N R S
Nutrients, climate a n d tropical NPP 5
Review a nd Synthesis
forest data analysis (Supplementary Figure 52). In the lowlands, the
optim al tree explained c. 7% o f the variance in A N P P and was based
on a single spUt o f foliar P (Fig. 2B). In the low foliar P sites (i.e.
< 0.75 pg P g ^), A N P P averaged 12.82 M g ha ^ year \ whereas
A N P P in the high foliar P sites (> 0.75 pg P g
averaged
15.84 Mg ha~^ year” ^ (Fig. 2B). G iven th at CART allows surrogate
splits to be assessed, we note th at soil P was the next best predictor
for the prim ary spUt o f the lowland forest data and had similar
explanatory pow er to foUar P. This apparent relationship was
confirm ed using regression analysis (Fig. 3), w hich showed a positive
Unear relationship betw een foUar P and A N P P and (P = 0.025) and a
similar (though n o t significant) relationship betw een total soU P and
A N P P (P = 0.083). T here was also a stronger direct relationship
betw een soU P and foUar {P < 0.0001; Fig. 3).
Inform ed by the initial CART analysis (which revealed relationships
betw een P metrics and A N PP), we subsequently conducted a series o f
bivariate com parisons using correlation analyses to explore potential
relationships betw een total soU P and foUar N , total soU N ,
decom position rate (p) and soU respiration in the lowland forest sites
(Table I). T here were significant positive relationships betw een total
soU P and b o th foUar N and total soU N , and across aU lowland forest
sites from w hich data were available, total soU P was strongly related with
b o th decom position rate (k) and soU respiration rate (Fig. 4; Table I).
3.5
2.5
CO
CD
>>
r= 0.70
Corrected P-value = 0.02
0.5
16
14
12
10
8
a.
r = 0 .6 0
Corrected P-value = 0.03
6
DIS CUSSIO N
0
D irect cUmate controls over A N P P are weU estabUshed (e.g. Chapin
et al. 2002; Z hao & R unning 2010), and substantial variations in b o th
tem perature and precipitation across the tropical forest biom e have
been show n to influence multiple aspects o f the tropical C cycle (e.g.
Saleska et al 2003; Schuur 2003; Powers et al 2009). O u r analysis
showed that M A T explains the largest p ro p o rtio n o f the variance in
A N P P across aU tropical forests (Fig. 2a). However, tem perature was
only im portant in broadly distinguishing tropical forest types; the first
SpUt in the regression tree analysis effectively separated upland (i.e.
cool forests w ith low A NPP) from lowland (i.e. warm forests with
higher A NPP) sites (Fig 2A). W hen analysing upland or lowland
forests in isolation, the relationship betw een tem perature and A N PP
was n o t significant (Figure 52). Thus, although cUmate organises plant
productivity globaUy (Chapin et al 2002) and helps explain large-scale
variations in A N P P across tropical forests as a whole, our analysis
shows that w ithin stUl large regions (e.g. upland o r lowland tropical
forest), M A T had Uttle explanatory power. 5imUarly, there was n o t a
significant relationship betw een M AP and A N P P in either the pantropical or the upland/low land tropical forest analyses (Figure 52).
This was som ew hat surprising, given the decUne in N P P in sites with
Table 1 Sum m ary o f Pearso n correlation coefficients b etw een to tal soil P and
covariates, P-values for individual pair-w ise com parisons and adjusted P-values for
m ultiple com parisons. P-values w ere adjusted using H o lm ’s sequential B onferroni
correction (H olm 1979)
T o tal soil P
r
Soil respiration
P-value
C orrected P-value
0.60
0.34
0.03
0.012
0.70
0.44
0.007
0.021
Foliar N
0.004
0.016
Foliar P
0.55
T o tal soil N
k
< 0.0001
0.03
0.024
< 0.00005
200
400
600
Tota l soil P (|jg g
800
1000
)
Figure 4 T o tal soil P vs.: (a) d e com position (M); and (b) soil respiration in low land
tropical forest. 7?-values rep resen t Pearson correlation coefficients.
high M A P previously reported (e.g. 5chuur 2003). However, our
analysis included m ore data from very w et lowland sites (i.e.
> 3500 m m year” ^) — sites th at had higher A N P P than m ost o f the
w et upland forest sites included in 5chuur (2003), perhaps accounting
for the difference (5ee Table 51).
I f M A T and M A P do n o t explain variations in tropical forest
A N PP, w hat does? T he regression tree analysis indicated that none o f
the cUmate or nutrient predictor variables we used explained a
significant prop o rtio n o f the variabUity in upland forest A N PP,
despite direct evidence o f N hm itation in some upland tropical forests
on young substrates (reviewed in T anner et a l 1998). However, it is
im portant to note th at the upland forest database included a relatively
smaU num ber o f sites (i.e. 32), perhaps Umiting our abUity to detect
possible relationships. In addition, other w ork has show n that N
lim itation may n o t exist in high elevation forests th at occupy unusuaUy
stable surface soils (e.g. V itousek & Farrington 1997) or in upland
forests o n older substrates (5Uver et al 2011).
D irect tests o f nutrient limitation in prim ary lowland tropical forests
are rare, b u t in the three experiments that have been done, two
showed th at P Umits fine UtterfaU production (Wright et al 2011). In
this analysis, after MAT, soU type was next in im portance in explaining
variance in A N P P across aU sites (Fig. 2a), w hich is consistent with
data showing th at soU nutrient avaUabUity varies w ith soU develop­
mental stage (Walker & 5yers 1976; Vitousek & Farrington 1997).
T h at said, we note th at the W alker & 5yers (1976) m odel — and
subsequent generaUzations about lowland forest nutrient hm itation —
have focused o n P-constraints w ith soU age. However, total soU P in
: 20f f Blackwell Publishing L td /C N R S
6 C. C. Cleveland et al.
the lowland sites represented in our database varied by m ore than an
order-of-m agnitude, suggesting th at simple assum ptions about nutri­
ent avaUabUity or nutrient lim itation based o n soU developm ental stage
may be overly simplistic (Table Si).
T he only metrics that predicted lowland forest A N P P to any extent
were P cycUng metrics (foUar and soU P), implying that P may play a
role in broadly structuring lowland tropical forest A N P P (Fig. 3). It is
im portant to note, however, th at the relationships were weak. For
example, the best predictor o f A N P P (foUar P) explained < 10% o f
the variation (Fig. 3A) in b o th the CART (Fig. 2) and regression
analyses (Fig. 3). SimUarly, there was an apparent relationship betw een
soU P and A N P P, but that relationship was also weak ( / = 0.057) and
did n o t m eet our threshold for significance (P = 0.083; Fig. 3b).
As w ith past inferential studies (e.g. V itousek 1984; MartineUi et al
1999; T ow nsend et a l 2007), we stress that o ur analyses do n o t
provide evidence — either for or against — nutrient Umitation per se.
T hat said, despite a wealth o f indirect evidence (e.g. V itousek 1984;
M cG roddy et al 2004a; PaoU et a l 2005; PaoU & Curran 2007;
T ow nsend et al 2007; Reich et a l 2009), and som e direct evidence
(M irmanto et al 1999; W right et al 2011) for widespread P Umitation
in the lowland tropics, our analysis revealed only weak relationships
betw een metrics o f P cycUng and A N PP.
Tw o types o f explanations exist for the weak relationships betw een
P avaUabUity and A N P P we observed across the lowland tropics, and
it is Ukely that b o th contribute to som e extent. First, m ethods for
m easuring b o th the predictor (e.g. cUmate o r soU nutrient avaUabUity)
and response variables (i.e. A NPP) have multiple know n Umitations.
F or example, A N P P in tropical forests is notoriously difficult to
measure, m ethods for assessing A N P P vary widely betw een sites, and
A N P P is frequently estim ated from an incom plete set o f measured
A N P P metrics (Clark et al 2001). In addition, measured estimates o f
A N P P frequently reflect fine UtterfaU inputs and biomass increm ent,
and often do n o t acknowledge the potential im portance o f other
aspects o f A N P P (e.g. losses to herbivores, contributions o f coarse
w oody debris), potentiaUy biasing estimates. CompUcating matters, in
m any sites, only one m etric o f A N P P is measured regularly (e.g.
UtterfaU or biomass increm ent), and for m ore than half the sites in our
database, we were forced to estim ate A N P P based o n relatively weak
relationships betw een biomass increm ent o r UtterfaU and A N PP.
M oreover, these methodological concerns are n o t restricted to A N PP;
as has been suggested elsewhere, logistical difficulties in accurately
sampUng and assessing nutrient avaUabUity (Stark & H art 1997) may
com bine to mask stronger associations that actuaUy do exist across
environm ental gradients. F or example, current m ethods for assessing
soU N or P avaUabUity (e.g. episodic m easurem ents using chemical
extractions) may n o t accurately reflect true avaUabUity to either
m icrobes or plants. Such m ethodological concerns can chaUenge
meta-analyses for any biom e, b u t are Ukely to be especiaUy
problem atic in tropical forests, especiaUy for A N P P (see Clark et al
2001 ).
Second, it is possible th at a sizable fraction o f the variance in
tropical A N P P may be explained by factors other than those directly
considered here. F o r example, although few m ulti-nutrient m anipu­
lations exist, in those that do, som e evidence has emerged indicating
control by nutrients other than N or P (Cuevas & M edina 1988;
B arron et al 2008; K aspari et a l 2008; W right et al 2011). Further­
more, the stature and complexity o f tropical forest canopies drives
considerable variation in photosytheticaUy active radiation, and Ught
avaUabUity may directly Umit A N P P across gradients in bo th cUmate
© 2011 Blackwell PubUstiing L td /C N R S
Review a n d Synthesis
and nutrient avaUabUity (W right & van Schaik 1994). Next, although
relatively easy to measure and readUy avaUable in the Uterature, simple
cUmate metrics Uke M A T and M A P may be insufficient to capture the
m ost im portant effects o f cUmate o n A N P P (e.g. the effects o f
seasonal variations in precipitation th at are n o t reflected in MAP).
FinaUy, the high species diversity o f m ost tropical canopies prom otes
local and regional heterogeneity in N and P aUocation and cycUng
(Townsend et al 2008) th at is Unked to a broad suite o f Ufe history and
evolutionary constraints; how those factors, along with variation in
other levels o f ecological organisation (e.g. herbivory and other plantanimal interactions; Fine et a l 2004), com bine to regulate A N P P in the
tropics remains poorly understood. T he existence o f such heteroge­
neity, along w ith multiple possible drivers o f the C cycle at b o th local
and regional scales, impUes th at detecting broad, ecosystem-level
correlates wUl require a thorough systematic sampUng at any given site
(Townsend et a l 2008). Such thorough sampUng is rarely reflected in
available data, including many o f those used in the current database,
bu t our analysis suggests th at such a strategy may be critical.
A lthough P metrics accounted for a smaU p ro portion o f the
observed variance in A N PP, our analysis does provide stronger
evidence o f Unks betw een P avaUabUity and below ground C cycUng.
F or example, we found significant relationships betw een total soU P
and b o th decom position (/fe) and soU respiration rates (Table 1; Fig. 4).
D ata from experiments directly investigating the effects o f nutrients
on decom position in the tropics are nearly as scarce as those on
aboveground responses (but see H obble & V itousek 2000; M cG roddy
et a l 2004b; Cleveland et al 2006; K aspari et a l 2008; Powers et al
2009), b u t previous research at the p lo t scale has show n simUar direct
relationships betw een Utter decom position and P avaUabUity. For
example, W ieder et al. (2009) show ed th at although Utter Ugnin : N
ratios m ost strongly predict decom position rates in tem perate
ecosystems (MeUUo et al 1982), Ugnin: P ratios were the best
predictor o f Utter decom position rates in a P -poor tropical forest in
Costa Rica. In another experiment, K aspari & Yanoviak (2008)
showed that Utter C : P ratios positively correlated with increases in
forest floor orgarUc m atter accumulation, suggesting that decom po­
sition and nutrient turnover slow w ith increasing C : P. SimUarly,
H obble & V itousek (2000) show ed th at P Umits Utter decom position
in forests o n highly w eathered, P -poor sites in HawaU. O ther studies
have observed high P immobiUzation during decom position in P-poor
soUs (M cGroddy et a l 2004b) w ithout simultaneous increases in mass
loss rates, b u t Utter P immobiUzation could translate into m ore rapid
and com plete decom position o f dissolved orgarUc m atter leached
from Utter in P-poor soUs (Cleveland et al. 2002). FinaUy, P avaUabUity
has been show n to Umit soU respiration in low-P tropical soUs
(Cleveland & Tow nsend 2006). A lthough this result could indicate
plant ro o t responses to soU P avaUabUity (e.g. O stertag 2001), these
studies are consistent w ith o ur results showing that decom position
broadly scales w ith P avaUabUity across the lowland tropical forest
sites in the database. G iven the relatively high P dem and and low
stoichiometric plasticity o f decom posers relative to plants (Cleveland
& Liptzin 2007), it is perhaps n o t surprising that P-constraints over
ceUular grow th and metaboUsm may be m ore apparent in metrics o f
decom position than in those used to assess A N PP.
O u r data also provide evidence suggesting th at interactions betw een
the N and P cycles in lowland tropical forests (e.g. T reseder &
V itousek 2001) may also have impUcations for C uptake and loss
(Vitousek et al 2009; Tow nsend et a l 2011). F or example, w e found
correlations betw een total soU P and b o th soU N and foUar N
Nutrients, climate a n d tropical NPP 7
Review a nd Synthesis
(Table 1), and w e suggest these Unks may emerge because o f an
underlying control by P that is expressed in two ways. First, N fixation
rates often increase w ith P availabUity (e.g. Pearson & V itousek 2001;
Reed et al. 2007), and under high P conditions, relatively high N inputs
via N fixation could result in b o th higher soil and foUar N content.
Second, our data suggest that P regulation o f the N cycle goes beyond
its effects on N fixation, and that P exerts a broader level o f control
over soU N turnover. F or example, the observed Unks betw een total
soU P and b o th Utter decom position (/fe) and soU respiration (Fig. 4)
imply that rates o f decom position in sites w here P is abundant may
drive paraUel increases in the rate at w hich N becomes avaUable for
plant uptake, helping to explain the observed relationship betw een soU
P and foUar N (Table 1).
As a whole, our analysis provides som e evidence th at soU P effects on
the tropical C cycle may be manifested in b o th direct and indirect ways,
as has been hypothesised elsewhere (Reich et al. 2008, 2009; V itousek
et al. 2010; Tow nsend et al. 2011). First, the relationship betw een
metrics o f P avaUabUity and A N P P (Fig. 2) are at least consistent with
previous claims o f direct P controls o n A N P P in lowland tropical
forests. In addition, the Unks betw een soU P and b o th decom position
(Fig. 4) and N pools (Fig. 3 and Table 1) imply that P-constraints may
also operate indirectly via controls o n N turnover and N avaUabUity.
G iven the widely recognised im portance o f foUar N content to
photosynthetic rates (Field & M ooney 1986; Reich et al 1997; W right
et al. 2004; W arren et al 2005), the soU P-foUar N correlation suggests
an additional m echanism by w hich P may act as the underlying
constraint on ecosystem C uptake (rcwra V itousek et al 2010). V itousek
et al. (2 0 1 0 ) described differences betw een proxim ate and ultimate
nutrient Umitation, and argued th at in many ecosystems, proxim ate
constraints by N may ‘m ask’ an ultim ate Umitation by P. Based on our
results, low soU P may pose a broad constraint o n Utter decom position
and soU organic m atter turnover in the tropics, w ith the highest rates o f
decom position occurring w here P is relatively abundant. In turn, rapid
decom position o f P-rich organic m atter should lead to simultaneous
increases in rates o f N mineraUsation a n d /o r in relatively high rates o f
N fixation (Reed et al. 2007; H o ulton et al. 2008), prom oting b o th
higher soU N avaUabUity and greater overaU ecosystem N stocks.
SimUarly, elevated N may also enhance P avaUabUity, mainly by meeting
the high N requirem ents o f phosphatase production (Olander &
V itousek 2000; T reseder & V itousek 2001; H o ulton et al. 2008).
However, although our analysis does provide additional evidence
suggesting the im portance o f P in ecosystem processes in lowland
tropical forests, perhaps m ore im portantly, it also highUghts th at our
understanding o f the role o f nutrients (and cUmate) in regulating
ecosystem processes in tropical forests is far from complete. In 1998,
T anner and coUeagues stated: ‘M any m ore experiments, especiaUy in
lowland forests, are needed to test o u r speculation that P usuaUy Umits
productivity in tropical lowland rain forests and that N Umits
productivity in tropical m ontane rain forests.’ M ore than a decade later
and with oiUy a handful o f notable exceptions (e.g. M irm anto et al.
1999; K aspari et al 2008; W right et al. 2011), we stiU lack such
experiments, and w e thus lack the abUity to make broad generaUza­
tions about the nature o f nutrient Umitation in this im portant biome.
Thus, to m ore clearly identify the nature and strength o f nutrient
Umitation in the lowland tropics, and to distinguish the role o f
nutrients from other potential drivers (e.g. cUmate, species traits, soil
physical properties, community-level interactions, etc.), multiple
manipulative experiments th at span im portant environm ental gradi­
ents are urgently needed. These are especiaUy im portant in the
lowlands, given b o th the paucity o f data and the enorm ous
im portance o f lowland forests to the global C cycle and cUmate
(e.g. Z hao & R unning 2010). As a starting point, given the evidence
presented here and elsewhere o f the multiple potential ways in w hich
P may constrain b o th the C and N cycles throughout the lowlands,
we w ould recom m end a series o f fertiUsation experiments that span
the m ore than order-of-m agnitude variation in soil P that exists across
lowland tropical forests (Supplementary Table 51). In addition, a
series o f nutrient manipulations spanning tem perature, precipitation
a n d /o r soil fertUity gradients w ould be particularly valuable in
elucidating how tropical forests may respond to global change. We
also n ote that given the sigrUficant variance in canopy nutrient
contents (FyUas et al. 2009; G leason et al. 2009) and Ufe history
strategies (Denslow 1980) th at occur in m ost tropical forests,
m anipulation experiments th at can capture com plete ecosystem scale
feedbacks and consequences wUl require large plots, thorough
sampUng protocols across species and long-tim e horizons. These
facts pose sigrUficant chaUenges o n b o th financial and logistical fronts,
bu t such efforts are critical to an im proved understanding o f how
tropical forests wUl respond to a suite o f global changes, including
b o th current (e.g. Li et al 2008) and forecasted cUmate and
atm ospheric change (T hornton et al. 2007; Loarie et al. 2009).
ACKNOW LEDGEM ENTS
This paper is a contribution from the Tropical N utrient Lim itation
working group at the N ational Center for Ecological Analysis and
Synthesis (funded by the N ational Science Foundation), the UiUversity
o f CaUfornia and the State o f CaUfornia. We thank the staff o f
NCEA S for logistical and technical support, R. RawUnson for
assistance w ith figure production and the anonym ous many others
th at helped locate data and assemble the database. W e are also grateful
to three anonymous referees and to D . H o o p er for valuable
com m ents o n the subm itted manuscript. C.C. wishes to acknowledge
financial support from the A.W. MeUon foundation through a grant to
investigate nutrient cycUng in tropical forests.
REFERENCES
A ber, J.D ., MeUUo, J.M ., N adelU offer, K.J., Pastor, J. & B oone, R .D . (1991). Factors
controlling nitrogen cycling and nitrogen saturation in n o rth e rn tem perate forest
ecosystem s. EcoL A p p l. 1, 303—315.
B arron, A.R., W urzburger, N ., BeUenger, J.P., W right, S.J., K raepiel, A.M.L. &
Fiedin, L.O . (2008). M olybdenum hm itation o f asym biotic nitrogen fixation in
tropical forest soils. JStat. Geosci., 2, 42—45.
B ern, C.R., T ow nsend, A.R. & Farm er, G.L. (2005). U nexpected dom inance o f
parent-m aterial strontium in a tropical forest o n highly w eathered soils. Ecology.
86, 626-632.
B onan, G .B. & Levis, S. (2010). Q uantifying carbon-nitrogen feedbacks in the
C om m unity L and M odel (CLM4). Geophys. Res. Eett.. 37, L07401.
Bruijnzeel, L.A., Juvik, J., Scatena, F.N ., H am ilton, L.S. & B ubb, P. (Eds) (2011).
Forests in the Mist. Science fo r Conservation and Management of Tropical Montane Cloud
Forests. U niversity o f H awaii Press, H onolulu.
C hadw ick, O ., D erry, L., V itousek, P., H uebert, B. & H edin, L. (1999). C hanging
sources o f nutrients during four miUion years o f ecosystem developm ent. Nature,
397, 491-497.
C hapin, F.S. I ll , M atson, P.A. & M ooney, H .A . (2002). Principles of Terrestrial
Ecoystem Ecology. Springer, N e w York.
Clark, D .A ., B row n, S., K ickhghter, D .W ., C ham bers, J.Q ., T h om h n so n , J.R. &
N i, J. (2001). M easuring n e t prim ary p ro d u c tio n in forests: concepts and field
m ethods. E col A p p l, 11, 356—370.
>2011 Blackwell Pubhshing L td /C N R S
8 C. C. Cleveland et al.
Review a n d Synthesis
C leveland, C.C. & Liptzin, D . (2007). C :N :P stoichiom etry in sod: Is there a
Jobbagy, E. & Jackson, R. (2000). T h e vertical distribution o f soil organic carbon
and its relation to chm ate and vegetation. E col A p p l, 10, 423M-36.
“Redfield ratio” for th e m icrobial biom ass? Biogeochemistty, 85, 235—252.
Cleveland, C.C. & T ow nsend, A.R. (2006). N u trie n t additions to a tropical rain
Jo h n so n , A., Frizano, J. & V ann, D . (2003). B iogeochem ical im phcations o f labile
forest drive substantial soil c arb o n dioxide losses to th e atm osphere. Proc. N a tl
p h o sp h o ru s in forest soils determ ined by the H edley fractionation procedure.
Oecologia, 135, 487—499.
Acad. Sci. USA, 103, 10316-10321.
Cleveland, C.C., T o w n sen d , A.R., Schimel, D .S., Fisher, H ., H ow arth, R.W., H edin,
Kaspari, M. & Y anoviak, S.P. (2008). B iogeography o f h tte r d e p th in tropical
L.O . et a l (1999). G lobal patterns o f terrestrial biological n itro g en (N 2 ) fixation in
forests: evaluating the p h o sp h o ru s grow th rate hypothesis. Funct. Ecol, 22, 919—
natural ecosystem s. Global Biogeochem. Cy., 13, 623—645.
923.
Cleveland, C.C., T o w n sen d , A.R. & Schm idt, S.K. (2002). P h o sp h o ru s hm itation o f
m icrobial processes in m o ist tropical forests: evidence from sh o rt-term labora­
tory incubations and field experim ents. Ecoystems, 5, 680—691.
Cleveland, C.C., Reed, S.C. & T o w n sen d , A.R. (2006). N u trie n t regulation o f o r­
ganic m atter d eco m p o sitio n in a tropical rain forest. Ecology, 87, 492—503.
Cuevas, E. & M edina, E. (1988). N u trie n t dynam ics w ith am azonian forests II. Fine
ro o t grow th, n u trien t availabihty, an d leaf h tter decom position. Oecologia, 76, 222—
(2008). M ultiple nutrients hm it htterfah and decom position in a tropical forest.
E col U tt., 11, 35M 3.
Lam bers, J.H ., R aven, J.A ., Shaver, G.R. & Smith, S.E. (2008). P lant nutrientacquisition strategies change w ith soil age. Trend Ecol E v o l, 23, 95—103.
LeB auer, D .S. & T reseder, K .K . (2008). N itro g e n hm itation o f n e t prim ary p ro ­
ductivity in terrestrial ecosystem s is globahy distributed. Ecology, 89, 371—379.
Li, W .H ., Fu, R., Juarez, R.I.N . & F ernandes, K. (2008). O bserved change o f the
235.
D avidson, E.A ., Reis de Carvalho, C.J., Figuera, A.M ., Ishida, F.Y., O m etto,
J.P.H .B .,
Kaspari, M ., G arcia, M .N ., H arm s, K .E ., Santana, M ., W right, S J. & Y avitt, J.B.
N a rd o to ,
G.B.
et a l
(2007).
R ecuperation
o f n itro g en
cychng
in A m azonian forests foUowing agricultural aban d o n m en t. Nature, 447, 995—
standardized precipitation index, its p otential cause and im phcations to future
chm ate change in the A m azo n region. Philos. Trans. R. Eond. Soc. B, 363, 1767—
1772.
Loarie, S.R., D uffy, P.B., H am ilton, H ., A sner, G .P., Field, C.B. & Ackerly, D .D .
999.
D e ’ath, G . & Fabricius, K .E . (2000). C lassification an d regression trees: a pow erful
yet simple tecnhique fo r ecological data analysis. Ecology, 81, 3178—3192.
(2009). T h e velocity o f chm ate change. Nature, 462, 1052—1055.
M artinehi, L.A., Piccolo, M .C., T ow nsend, A.R., V itousek, P.M ., C uevas, E.,
D enslow , J. (1980). P atterns o f p la n t species diversity d uring succession under
different disturbance regimes. Oecologia, 46, 18—21.
M cD ow eh, W . et al. (1999). N itro g e n stable isotopic c o m position o f leaves and
E hth, J., Leathw ick, J.R. & H astie, T. (2008). A w orking guide to b o o sted regression
M atson, P. & V itousek, P. (1990). E cosystem ap p ro ach to a global nitrous oxide
trees, y.
Ecol, 11, 802—813.
Elser, J., B racken, M .E.S., Cleland, E .E ., G runer, D .S., H arpole, W.S., H ihebrand,
H . et a l (2007). G lobal analysis o f n itro g en and p h o sp h o ru s hm itatio n o f prim ary
producers in freshw ater, m arine and terrestrial ecosystem s. Ecol Eett., 10, 1135—
1142.
Field, C. & M ooney, H . (1986). T h e photo sy n th esis—nitro g en relationship in wild
plants. In: On the Economy of Plant Form and Function (ed. G ivnish, T.J.). Cam bridge
U niversity Press, Cam bridge, pp. 25—55.
Fine, P.V., M esones, A.I. & Coley, P.D . (2004). H erbivores p ro m o te h ab itat speciahzation by trees in A m azonian forests. Science, 305, 663—665.
Foley, J., D eFries, R., A sner, G .P., B arford, C., B onan, G ., C arpenter, S.R. et al
(2005). G lobal consequences o f land use. Science, 309, 570—574.
Fyhas, N .M ., Patino, S., Baker, T.R ., Bielefeld N a rd o to , G ., M artinehi, L.A.,
Q uesada, C.A. et al (2009). B asin-w ide variations in fohar p roperties o f A m a­
zonian forest: phylogeny, soils an d chmate. Biogeosciences, 6, 2677—2708.
G leason, S.M., R ead,}., A res, A. & M etcalfe, D J . (2009). P h o sp h o ru s econom ics o f
tropical rainforest species an d stands across soil contrasts in Q ueensland, A ustraha: und erstanding th e effects o f soil speciahzation and trait plasticity. Funct.
Ecol, 23, 1157-1166.
G ough, L., O senberg, C.W ., G ross, K.L. & C o llins, S.L. (2000). F ertilization effects
soil: tropical versus tem perate forests. Biogeochemistty, 46, 45—65.
budget. Bioscience, 40, 667—672.
M cC une, B. & G race, J.B. (2002). Analysis of Ecological Communities. M jM Softw are
D esign, G leneden Beach.
M cG roddy, M .E ., D aufresne, T. & H edin, L .O . (2004a). Scahng o f C :N :P stoi­
chiom etry in forests w orldw ide: Im phcations o f terrestrial Redfield-type ratios.
Ecolo^, 85, 2390-2401.
M cG roddy, M .E ., Silver, W.L. & C osm e de O hveira, R. (2004b). T h e effect o f
p h o sp h o ru s availabililty o n decom position in a seasonal low land A m azonian
forest. Ecosystems, 1, 172—179.
Melillo, J.M ., A ber, J.D . & M uratore, J.F. (1982). N itro g en and hgnin co n tro l o f
hard w o o d leaf htter decom position dynamics. Ecology, 63, 621—626.
M irm anto, E., P ro cto r,} ., G reen,}., N agy, L. & Suriantata, ?????. (1999). Effects o f
n itrogen and p h o sp h o ru s fertilization in a low land evergreen rain forest. Philos.
Trans. R Eond. Soc. B, 354, 1825—1829.
N a rd o to , G .B., O m etto, }., Ehleringer, J.R ., H iguchi, N ., B ustam ante, M .M .D . &
M artinehi, L.A. (2008). U nderstanding the influences o f spatial patterns o n N
availabihty w ithin the B razhian A m azon forest. Ecosystems, 11, 1234—1246.
O kin, G.S., M ahow ald, N ., Chadw ick, O .A . & A rtaxo, P. (2004). Im p a c t o f desert
du st o n the biogeochem istry o f p h o sp h o ru s in terrestrial ecosystem s. Global
Biogeochem. Cy., 18, (GB2005).
o f species density an d prim ary productivity in herbaceous p la n t com m unities.
O lander, L.P. & V itousek, P.M . (2000). R egulation o f soil phosphatase and chitinase
Oikos, 89, 428M 39.
G race,}., Lloyd,}., M cIntyre,}., M iranda, A.C., M eir, P., M iranda, H.S. et a l (1995).
O stertag, R. (2001). Effects o f nitrogen and p h o sp h o ro u s availabihty o n fine-root
C arb o n dioxide uptake by an un d istu rb ed tropical rain fo rest in Southw est
A m azonia, 1992—1993. Science, 270, 778—780.
H edin, L .O ., B rookshire, E.N.}., M enge, D .N .L . & B arron, A.R. (2009). T h e
nitrogen p aradox in tropical forest ecosystem s. A nnu. Rev. Ecol E vo l Syst., 40,
613-635.
H obble, S.E. & V itousek, P.M . (2000). N u trie n t hm itatio n o f d e co m p osition in
H aw aiian forests. Ecology, 81, 1867—1877.
H oldridge, L.R., G renke, W .C., H athew ay, W .H ., Liang, T. & Tosi, }.A. (1971).
Forest Environments in Tropical Life Zones: A Pilot Study. Perg am o n Press, O xford.
H olm , S. (1979). A sim ple sequentiahy rejective m ultiple te st procedure. Scand. f.
Stat., 6, 65-70.
H o u lto n , B., Sigman, D . & H edin, L. (2006). Iso to p ic evidence for large gaseous
nitrogen losses fro m tropical rainforests. Proc. N a tl Acad. Sci. U SA, 103, 8745—
8750.
H o u lto n , B.Z., W ang, Y .P., V itousek, P.M . & Field, C.B. (2008). A unifying
fram ew ork fo r dinitrogen fixation in th e terrestrial biosphere. Nature, 454, 327U34.
H ungate, B.A., D ukes, }.S., Shaw, R., L uo, Y. & Field, C.B. (2003). N itro g en and
chm ate change. Science, 302, 1512—1513.
© 2011 Blackw eh P ubhshing L td /C N R S
activity by N and P availabihty. Biogeochemistty, 49, 175—190.
dynam ics in H aw ahan M ontane Forests. Ecology, 82, 485—499.
Paoh, G .D . & C urran, L.M. (2007). Soil nutrients hm it fine htter p ro d u c tio n and
tree grow th in m ature low land forest o f Southw estern B orneo. Ecosystems, 10,
503-518.
Paoh, G .D ., C urran, L.M. & Zak, D .R . (2005). P hosphorus efficiency o f B ornean
rain forest productivity: E vidence against the unim odal efficiency hypothesis.
Ecolo^, 86, 1548-1561.
P earson, H.L. & V itousek, P.M . (2001). Stand dynam ics, nitrogen accum ulation,
and sym biotice nitrogen fixation in regenerating stands o i Acacia koa. Ecol A p p l,
11, 1381-1394.
Pett-Ridge, J.C ., H en d erso n , G .M . & Chadw ick, O.A. (2009). Investigating soil
processes across the H aw aiian Islands using U -series nuchdes. Geochim. Cosmochim. Acta, 73, A 1022—A1022.
Phihips, O .L., M ahh, Y., H iguchi, N ., Laurence, W .F., N u n ez, P.V., V asquez, R.M.
et a l (1998). Changes in the c arb o n balance o f tropical forests: evidence from
long-term plots. Science, 282, 439—442.
P order, S. & Hilley, G .E . (2010). Linking chronosequences w ith the rest o f the
w orld —predicting soil p h o sp h o ru s c o n te n t in eroding landscapes. Biogeochemistty,
102, 153-166.
Nutrients, climate a n d tropical NPP 9
Review a nd Synthesis
P order, S., V itousek, P., Chadw ick, O ., C ham berlain, C. & HiUey, G . (2007). Uplift,
erosion and p h o sp h o ru s hm itatio n in terrestrial ecosystem s. Ecosystems, 10, 158—
170.
Pow ers, J.S., M ontgom ery, R.A., A dair, B.C., Brearley, F.Q ., D eW alt, S.J., C astanho,
C.T. et al. (2009). D eco m p o sitio n in tropical forests: a p an-tropical study o f the
effects o f h tter type, h tter p lacem en t an d m esofaunal exclusion across a p re ­
cipitation g ra d ie n t./. E col, 97, 801—811.
Q uesada, C.A., Lloyd, J., Schw artz, M., Patino, S., Baker, T., Czimczik, C. et al
(2009). Soils o f A m azonia w ith particular reference to the R A IN F O R sites.
Biogeosciences, 6, 3851—3921.
Reed, S.C., Cleveland, C.C. & T o w n sen d , A.R. (2007). C ontrols over free-hving
nitrogen fixation in a low land tropical rain forest. Biotropica, 39, 585—592.
R eich, P.B. & O leksyn, J. (2004). G lobal patterns o f p la n t leaf N and P in
relation to tem perature an d latitude. Proc. N a tl A.cad. Sci. USA., 101, 11001—
11006.
R eich, P.B., W alters, M.B. & F h sw o rth , D.S. (1997). F ro m tropics to tundra:
G lobal convergence in p lan t functioning. Proc. N a tl Acad. Sci., 94, 13730—
13734.
R eich, P.B., Tjoelker, M .G ., Pregitzer, K.S., W right, I.J., O leksyn, J. & M achado,
J.L. (2008). Scahng o f respiration to n itro g en in leaves, stem s and roots o f higher
land plants. Ecol Eett., 11, 793—801.
R eich, P.B., Oleksyn, J. & W right, I.J. (2009). L eaf p h o sp h o ru s influences the
photosyn th esis-n itro g en relation: a cross-biom e analysis o f 314 species. Oecologia,
160, 207-212.
R ichter, D .D ., A hen, H ., Li, J., M arkew itz, D . & Raikes, J. (2006). Bioavahabihty o f
T reseder, K .K . & V itousek, P.M. (2001). E ffects o fn u trie n t availabihty o n investm ent
in acquisition o f N and P in H aw aiian rain forests. Ecology, 82, 946—954.
V enables, W .N . & Ripley, B .D . (2002). Modem applied statistics with S. Springer, N ew
York.
V itousek, P.M. (1984). Litterfah, nutrien t cychng, and n u trien t hm itation in tropical
forests. Ecology, 65, 285—298.
V itousek, P. (2004). Nutrient Cycling and Eimitation: H a w a ii as a Model System.
P rinceton U niversity Press, Princeton.
V itousek, P.M. & Farrington, H . (1997). N u trie n t hm itation and soil developm ent:
E xperim ental te st o f a biogeochem ical theory. Biogeochemistty, 37, 63—75.
V itousek, P.M . & Sanford, R.L. (1986). N u trie n t cychng in m oist tropical forest.
A n n . Rev. E col Syst., 17, 137—167.
V itousek, P.M ., A sner, G .P., Chadw ick, O .A . & H otchkiss, S. (2009). Landscapelevel variation in forest structure and biogeochem istry across a substrate age
gradient in Hawah. Ecology, 90, 3074—3086.
V itousek, P., P order, S., H o u lto n , B. & C hadw ick, O . (2010). Terrestrial p h o s ­
pho ru s hm itation: m echanism s, im phcations, and nitrogen—p h o sp h o ru s interac­
tions. Ecol A p p l, 20, 5—15.
W alker, T.W . & Syers, J.K . (1976). T h e fate o f p h o sp o ru s during pedogenesis.
Geoderma, 15, 1—19.
W arren, C.R., M cG rath, J.F. & A dam s, M.A. (2005). D ifferential effects o f N , P and
K o n photosynthesis and partitioning o f N in Pinuspinaster needles. A n n . For. Sci.,
62, 1-8.
W ieder, W .R., C leveland, C.C. & T ow nsend, A.R. (2009). C ontrols ov er leaf htter
d ecom position in w et tropical forests. Ecology, 90, 3333—3341.
slowly cychng soil phosp h o ru s: m ajor restructuring o f soil P fractions over four
W right, S J. & van Schaik, C.P. (1994). L ight and the phenology o f tropical trees.
decades in an aggrading forest. Oecologia, 150, 259—271.
M®. N at., 143, 192-199.
W right, I.J., Reich, P.B., W estoby, M., Ackerly, D .D ., B aruch, Z., B ongers, F. et al
R obertson, G. (1984). N itrification and nitro g en m ineralization in a low land rain­
forest succession in C osta Rica, C entral Am erica. Oecologia, 61, 99—104.
(2004). T h e w orldw ide leaf econom ics spectrum . Nature, 428, 821—827.
R obertson, G . & Tiedje, J.M . (1988). D enitrification in a h u m id tropical rainforest.
W right, S.J., Y avitt, J.B ., W urzburger, N ., T urner, B.L., T anner, E.V.J., Sayers, E J .
Nature, 336, 756-759.
Saleska, S.R., Miller, S.D., M atross, D .M ., G oulden, M .L., W ofsy, S.C., da, R ocha
g ro w th and htter p ro d u c tio n in a low land tropical forest. Ecology, D O I: 10.1890/
et al. (2003). C arb o n in A m azo n forests: unexpected seasonal fluxes and distur­
bance-induced losses. Science, 302, 1554—1557.
Schuur, E.A . (2003). Productivity and global chm ate revisited: the sensitivity o f
et a l (2011). Potassium , p h o sp h o ru s o r nitrogen hm it ro o t ahocation, tree
10-1558.1.
Z hao, M. & R unning, S.W. (2010). D ro u ght-induced reduction in global terrestrial
n e t prim ary p ro d u c tio n from 2000—2009. Science, 329, 940—943.
tropical fo rest g ro w th to precipitation. Ecology, 84, 1165—1170.
Silver, W.L. (1994). Is n u trien t availabihty related to p la n t n u trien t use in hum id
tropical forests? Oecologia, 98, 336—343.
Silver, W .L., T h o m p so n , A.W ., Firestone, M .K ., Reich, A. & Fw el, J.J. (2005).
N itro g e n re te n tio n and loss in tropical plantations and old g ro w th forests. Ecol
A p p l, 15, 1604-1614.
Silver, W .L., T h o m p so n , A.W ., H erm an, D .J. & F irestone, M .K . (2011). Is there
evidence fo r hm itations to nitro g en m ineralization in u p p e r m o n tan e tropical
forests? In: Forests in the Mist: Science fo r Conserving and Managing Tropical Montane
Cloud Forest (eds Bruijnzeel, L.A., Juvik, J., Scatena, F.N ., H am ilton, L.S. & B ubb,
P.). U niversity o f H awaii Press, H o nolulu, pp. 418—427.
Stark, J.M . & H art, S.C. (1997). H ig h rates o f nitrification an d nitrate tu rn o v er in
undisturb ed coniferous forests. Nature, 385, 61—64.
Syers, J.K ., Jo h n sto n , A .E . & C urtin, D . (2008). Efficiency of Soil and Fertiliser Phos­
phorus Use: Reconciling Changing Concepts of Soil Phosphorus Behavior with Agronomic
Information. F A O Fertilige'f and Plant Nutrition Bulletin. FA O , Rome.
T anner, E.V.J., V itousek, P.M . & Cuevas, E. (1998). E x p erim en tal investigation o f
nutrien t h m itatio n o f forest g ro w th o n w et tropical m ountains. Ecolo^, 79, 10—22.
T em pler, P.M ., Silver, W .L., Pett-Ridge, J. & Firestone, M .K . (2008). Plant and
m icrobial controls o n nitro g en re te n tio n and loss in tropical fo rest soils. Ecology,
89, 3030-3040.
T h o rn to n , P.E., Lam arque, J.F ., R osenbloom , N .A . & M ahow ald, N .M . (2007).
Influence o f c a rb o n -n itro g en cycle co uphng o n land m o d el response to C O 2
fertilization and chm ate variabihty. Global Biogeochem. Cy., 21, (GB4018).
T ow nsend, A.R., Cleveland, C.C., A sner, G . & B ustam ante, M .M .C. (2007). C o n ­
trols o f fohar N :P ratios in tropical rain forests. Ecology, 88, 107—118.
T ow nsend, A.R., A sner, G .P. & Cleveland, C.C. (2008). T h e biogeochem ical h e t­
erogeneity o f tropical forests. Trends Ecol E vo l, 23, 424—431.
T ow nsend, A.R., C leveland, C., H o u lto n , B., A lden, C. & W hite, J.W .C. (2011).
M ulti-elem ent regulation o f th e tropical forest c arb o n cycle. Front. Ecol
Environ., 9, 9—17.
SUPPORTING INFORM ATIO N
A dditional Supporting Inform ation may be found in the online
version o f this article:
Figure SI Linear relationships betw een (a) HtterfaU and (b) biomass
increm ent and A N PP. Relationships were used to generate A N PP
values w hen either com ponent was unavaUable in the Uterature.
Figure S2 A N P P vs. M A T (a and b) and M AP (c and d) in the upland
and lowland forest sites in the database.
Table SI M aximum (Max.), m inim um (Min.) and m edian values for
the cUmate, carbon and nutrient cycUng variables used in the analysis.
T he ‘uplands + low land’ colum n represents calculated values for aU
sites in the database. T he ‘lowlands’ colum n represents values from
sites w ith M A T > 20.7
and the ‘uplands’ colum n represents
values from sites w ith M A T < 20.7 °C.
As a service to our authors and readers, this journal provides
supporting inform ation suppUed by the authors. Such materials are
peer-reviewed and may be re-organized for online deUvery, but are n o t
copy edited o r typeset. Technical support issues arising from
supporting inform ation (other than missing fUes) should be addressed
to the authors.
E ditor, D avid H ooper
M anuscript received 8 June 2011
M anuscript accepted 15 Ju n e 2011
'2 0 1 1 Blackwell Publishing L td /C N R S