Ecology and Evoluon 2016; 1–12 www.ecolevol.org
© 2016 The Authors. Ecology and Evoluon
published by John Wiley & Sons Ltd.
Received: 22 April 2016
Revised: 1 September 2016
Accepted: 4 September 2016
The relaonship between biodiversity and ecosystem funcon has increasingly been
debated as the cornerstone of the processes behind ecosystem services delivery.
Experimental and natural eld- based studies have come up with nonconsistent pat-
terns of biodiversity–ecosystem funcon, supporng either niche complementarity or
selecon eects hypothesis. Here, we used aboveground carbon (AGC) storage as
proxy for ecosystem funcon in a South African mistbelt forest, and analyzed its rela-
onship with species diversity, through funconal diversity and funconal dominance.
We hypothesized that (1) diversity inuences AGC through funconal diversity and
funconal dominance eects; and (2) eects of diversity on AGC would be greater for
funconal dominance than for funconal diversity. Community weight mean (CWM) of
funconal traits (wood density, specic leaf area, and maximum plant height) were
calculated to assess funconal dominance (selecon eects). As for funconal diver-
sity (complementarity eects), multrait funconal diversity indices were computed.
The rst hypothesis was tested using structural equaon modeling. For the second
hypothesis, eects of environmental variables such as slope and altude were tested
rst, and separate linear mixed- eects models were ed aerward for funconal di-
versity, funconal dominance, and both. Results showed that AGC varied signicantly
along the slope gradient, with lower values at steeper sites. Species diversity (richness)
had posive relaonship with AGC, even when slope eects were considered. As pre-
dicted, diversity eects on AGC were mediated through funconal diversity and func-
onal dominance, suggesng that both the niche complementarity and the selecon
eects are not exclusively aecng carbon storage. However, the eects were greater
for funconal diversity than for funconal dominance. Furthermore, funconal domi-
nance eects were strongly transmied by CWM of maximum plant height, reecng
the importance of forest vercal stracaon for diversity–carbon relaonship. We
therefore argue for stronger complementary eects that would be induced also by
complementary light- use eciency of tree and species growing in the understory layer.
carbon stock, community weight mean, funconal richness, maximum plant height, niche
complementarity, structural equaon modeling
1Department of Forest and Wood
Science, Stellenbosch University, Maeland,
2Laboratoire de Biomathémaques et
d’Esmaons Foresères, Université
d’Abomey-Calavi, Cotonou, Bénin
3South African Naonal Biodiversity
Instute, Kirstenbosch Research Centre,
Claremont, South Africa
4Department of Conservaon Ecology
and Entomology, Stellenbosch University,
Maeland, South Africa
Appliquée, Université d’Abomey-Calavi,
Sylvanus Mensah, Department of Forest
and Wood Science, Stellenbosch University,
Maeland, South Africa.
SHARE INTRA-ACP project; Naonal
Research Foundaon of South Africa;
African Forest Forum
Tree species diversity promotes aboveground carbon storage
through funconal diversity and funconal dominance
Sylvanus Mensah1,2 | Ruan Veldtman3,4 | Achille E. Assogbadjo5 |
Romain Glèlè Kakaï2 | Thomas Seifert1
This is an open access arcle under the terms of the Creave Commons Aribuon License, which permits use, distribuon and reproducon in any medium,
provided the original work is properly cited.
MENSAH ET AL.
1 | INTRODUCTION
The relaonship between biodiversity and carbon storage is being
debated as one of the current ecological topics (Cavanaugh et al., 2014;
Day, Baldauf, Rushauser, & Sunderland, 2014; Ruiz- Benito et al., 2014;
Ruiz- Jaen & Potvin, 2011), and some aspects of climate- related eects
have been well invesgated (Durán, Sánchez- Azofeifa, Rios, & Gianoli,
2015; Wu et al., 2015). Because biomass is an important component
of forest stand producvity, the relaonship between biomass carbon
and biodiversity can also be assimilated to the one of biodiversity and
ecosystem funcon (Lasky et al., 2014). Basically, two well- debated
mechanisms are commonly used to explain the role of plant diversity
in ecosystem resource dynamics, ecosystem processes, and funcons:
niche complementarity eects and selecon eects (Dı́az & Cabido,
2001; Tilman et al., 1997); the niche complementary eects hypoth-
esis assumes increasing diversity would promote greater variety of
funconal traits and provide opportunies to species to eciently use
the available resources, thereby increasing ecosystem funcon; the
selecon eects hypothesis suggests that in ecosystem with higher
diversity, there would be a higher probability of occurrence of domi-
nant species or traits that inuence ecosystem funconing. Currently,
great research eorts are made to elucidate how diversity components
(taxonomic diversity, funconal diversity, and funconal dominance)
drive biomass and carbon stocks, and the extent to which the ndings
support niche complementarity and selecon eects hypotheses.
Taxonomic diversity, expressed by species richness and alpha-
diversity indices, has been commonly used as a simple measure of
biodiversity (Mayeld et al., 2010; Tilman et al., 1997) and has been
shown to correlate posively with carbon stocks. However, because a
new species—with dierent funconal traits—added to an ecosystem
would likely contribute to the physiological processes, the eects of
taxonomic diversity on carbon storage could be treated as dierent
eects of funconal diversity (accounng for niche complementarity)
or/and funconal dominance (comprising selecon eects). The func-
onal diversity is known as “the value and range of funconal traits
of the organisms present in a given ecosystem”(Dı́az & Cabido, 2001,
pp 654) and therefore might be the starng point of elucidang the
mechanisms underlying the relaon between biodiversity and carbon
(Cadoe, Carscadden, & Mirotchnick, 2011; Naeem, 2002). Yet, some
recent reviews showed controversy in the relaonship between tax-
onomic and funconal diversity (Mayeld et al., 2010; Naeem, 2002;
Song, Wang, Li, & Zhou, 2014). On the one hand, following Tilman
et al. (1997) and Mouchet, Villéger, Mason, and Mouillot (2010), func-
onal diversity was posively correlated with species richness, and in
this case, taxonomic diversity can simply be used to replace funconal
diversity. On the other hand, it was pointed out that land use, the local
species pool, etc. could also inuence the relaonship between func-
onal and taxonomic diversity (Cadoe et al., 2011; Mayeld et al.,
2010). Consequently, whether diversity (species richness) eects on
ecosystem funcon are fully mediated by funconal diversity or code-
termined by selecon eects (dominance paerns) is sll well debated.
In tropical natural forests, where several species cohabit and fulll the
major ecosystem funcons, it is common to observe the abundance
and dominance of highly producve tree species, thus increasing the
chances that diversity–carbon relaonships are mediated by selec-
on eects. This was partly conrmed by our previous observaons
in South African mistbelt forests, especially the greater inuence of
the most dominant species on biomass stocks (Mensah, Veldtman,
du Toit, Glèlè Kakaï, & Seifert, 2016). More and more, research tends
to show how funconal diversity and/or funconal dominance play
a major role in ecosystem funconing (Baraloto et al., 2012; Clark,
Flynn, Buereld, & Reich, 2012; Ruiz- Jaen & Potvin, 2011; Song
et al., 2014). Understanding whether diversity eects on ecosystem
funcon are more likely mediated through funconal diversity than
funconal dominance, or vice versa, will bring substanal insights into
which mechanism is more relevant.
Very few studies have addressed the relaonships between diver-
sity and ecosystem funcon in natural mulspecies tropical forests.
Using aboveground tree carbon data in a northern mistbelt forest in
South Africa, we examined the relaonship between diversity and car-
bon stocks through the eects of funconal diversity and funconal
dominance. We hypothesized that (1) diversity inuences tree carbon
storage through both funconal diversity and funconal dominance
eects. However, there are insights that diversity and carbon relaon-
ships can be caused by covarying environmental factors (Cavanaugh
et al., 2014; Ouyang et al., 2016). Therefore, we considered altude
and slope as the most physical gradients in these forests, and tested
their eects on tree carbon storage. In addion, while accounng for
signicant environmental gradient eects, we also hypothesized that
(2) eects of diversity on carbon storage would be greater for func-
onal dominance than for funconal diversity.
2 | MATERIALS AND METHOD
2.1 | Study area
This study was carried out in the northern mistbelt forests in the
Limpopo Province, South Africa. These forests are generally found as
large patches on steep eastern slopes in the province (Geldenhuys,
1997, 2002). The site selected for this study was the Woodbush-De
Hoek nave forest complex (23°50′S, 30°03′E) near Tzaneen. The
area is characterized by an altudinal gradient from 1,050 to 1,800 m
above mean sea level and an annual rainfall ranging from 600 mm to
1,800 mm (Geldenhuys, 2002). Pine plantaons are established in the
surrounding environment by the State Department of Water Aairs
and Forestry and transferred to forest companies for commercial m-
ber producon. The main sectors for the management policy in the
landscape are mber producon, nature conservaon, and recrea-
on (hiking). The vegetaon in the Woodbush-De Hoek nave forest
is dominated by canopy and above canopy species such as Xymalos
monospora, Podocarpus lafolius, Syzygium gerrardii, and Cryptocarya
transvaalensis (Mensah, Glèlè Kakaï, & Seifert, 2016). The understory
vegetaon is represented by species such as Oxyanthus speciosus,
Peddiea africana, and Kraussia oribunda (Geldenhuys, 1997).
MENSAH ET AL.
2.2 | Forest sampling and aboveground carbon data
Stand data (species, canopy layer, tree density, basal area) were
obtained by means of a straed random sampling design set in a
707.612 ha (hectare) forest block in the Woodbush-De Hoek for-
est. The sampling design consisted of 30 replicates of 250 m2 circular
subplots, each inside a 500 m2 circular larger plot. These plots were
laid out in straed compartments obtained by subdividing the re-
search area on the basis of three classes of slope: at (1.5%–15.3%),
gentle (15.3%–29.19%), and steep (29.19%–43.1%); four classes of
aspect (North, South, West, and East); and three classes of eleva-
on: low (1,174–1,332 m a.s.l.), medium (1,332–1,490 m), and high
(1,490–1,648 m). Inside 250 m2 plots, species names, diameter at
breast height (dbh), and height of trees belonging to 5–10 cm dbh
were recorded, while only individuals having more than 10 cm dbh
were tagged and measured within the larger plots.
We used the mulspecies allometric biomass equaon developed
for the northern mistbelt forests (Mensah, Veldtman, du Toit, Glèlè
Kakaï, & Seifert, 2016) to calculate the aboveground biomass (AGB)
for all individual trees present in the plots. The allometric equaon
provided more accurate esmated biomass values, compared with the
exisng pantropical biomass equaon (Chave et al., 2005; Mensah,
Veldtman, & Seifert, 2016). The formula for the allometric biomass
equaon is as follows:
where AGB stands for the aboveground tree biomass in kilograms,
SWD the specic wood density (g/cm3), DBH the diameter at breast
height (cm), and H the total height (m). AGB was computed for each
individual tree, upscaled to plot and stand level for each diameter class
(i.e., for 5–10 cm dbh in the 30 smaller plots and for ≥10 cm dbh in the
30 larger plots), and summed up to obtain the values for dbh > 5cm.
Carbon values were determined aerward, by mulplying the abo-
veground biomass by a factor of 0.5 (Lung & Espira, 2015).
2.3 | Diversity and dominance metrics
Diversity was measured using taxonomic diversity, at each plot.
We used species richness to characterize the taxonomic diversity
(Magurran, 1988). Species richness at plot level is simply dened as
the number of disnct species enumerated inside each plot. To assess
funconal diversity, we considered the funconal traits that are rel-
evant to the ecosystem funcon of interest (i.e., biomass and carbon
storage). Because carbon storage is strongly dependent on wood and
foliage structures, we used traits such as specic wood density (WD),
specic leaf area (SLA), and maximum plant height (PHm). Data on
specic wood density were obtained from the Global Wood Density
Database (Zanne et al., 2009). In case mulple values were available
for a single species, the average wood density was used. When a par-
cular species was missing, we used the average genus wood density.
SLA and PHm were extracted from the TRY database (Kage et al.,
2011). As funconal diversity metrics, we esmated funconal rich-
ness (Fric), funconal evenness (Feve), funconal divergence (Fdiv),
funconal dispersion (Fdis), and Rao quadrac entropy (RaoQ) at each
plot (Baraloto et al., 2012; Cavanaugh et al., 2014; Finegan et al.,
2015; Villéger, Mason, & Mouillot, 2008), using the values of the
funconal traits with the “FD” package in R (Laliberté, Legendre, &
Shipley, 2015). These diversity indices are multrait funconal diver-
sity metrics that combine both the relave weight of each species and
the pairwise funconal dierence between species. A review of these
measures can be found in study by Mouchet et al. (2010).
Funconal dominance was assessed by esmang the plot- level com-
munity weight mean (CWM) for each funconal trait. CWM is the mean
of each species trait value weighted by the relave abundance (stem
number) of that species (Cavanaugh et al., 2014). The per- plot CWM was
esmated for WD, SLA, and PHm, again using the “FD” package in R.
2.4 | Data analyses
Here, we tested two hypotheses: (1) diversity eects on carbon stor-
age are mediated through both funconal diversity and funconal
dominance eects; and (2) eects of diversity on carbon storage are
greater for funconal dominance than for funconal diversity. The
rst hypothesis was tested using structural equaon models (SEM),
while the second hypothesis was tested using linear mixed- eects
models. For both SEM and linear mixed- eects models, AGC data
were log- transformed to meet the normality assumpon (Shapiro–
Wilk stasc = 0.97, p- value = .628).
2.4.1 | Structural equaon modeling
SEM oers the possibility to test hypothesized paerns of direct and
indirect relaonships among the measured variables. This is parcu-
larly important, as we assumed that the diversity eects would be
transmied through both funconal diversity and funconal domi-
nance. Therefore, we examined the indirect and direct eects of di-
versity (species richness) on aboveground carbon. We constructed
two separate structural equaon models represenng (1) full media-
on: Diversity eects are fully transmied by funconal diversity and
dominance metrics; and (2) paral mediaon: There are both direct
and indirect diversity eects through funconal diversity and func-
onal dominance metrics. Due to the presence of mulple measures
for funconal diversity, we used stepwise selecon techniques to
select the most relevant funconal diversity metrics for the above-
ground carbon data. As a result, only funconal richness (Fric) and
funconal evenness (Feve) were selected (p- value <.05). We did not
deny the potenal eects of environmental variables on the species
diversity and aboveground tree carbon. Nevertheless, we believe that
such eects could be beer assessed in mixed modeling approach (ad-
dressed in the next paragraph), thus simplifying the outputs of the
SEMs. The overall t of the SEMs was assessed using χ2 – square test
(a p- value >.05 would indicate an absence of signicant deviaons be-
tween data and model), the comparave t index (CFI), and the Akaike
informaon criterion (AIC) (Grace & Bollen, 2005). We used the stand-
ardized coecients to allow direct comparisons across paths (Grace &
Bollen, 2005). SEMs were ed in the R stascal soware package
MENSAH ET AL.
(R Development Core Team 2015), using the “sem” funcons in the
“lavaan” package version 0.5–19 (Rosseel, 2012).
2.4.2 | Linear mixed- eects models
Prior to the mixed- eects modeling, we tested for potenal environ-
mental variables and species richness eects on aboveground carbon
storage. Environmental variables are expected to have eects on plant
structures, growth, and survival (Mensah, Houehanou, Sogbohossou,
Assogbadjo, & Glèlè Kakaï, 2014; Wang, Fang, Tang, & Zhu, 2006)
and hence on standing aboveground biomass and carbon stocks.
Here, we focused on the variables that are determinant and quan-
able in the area, that is, topography (slope and altude) (Geldenhuys,
1997, 2002). Other environmental variables such as temperature and
precipitaon have also been proved to have much inuence on pro-
ducvity, biomass, and carbon stocks (Cavanaugh et al., 2014; Durán
et al., 2015), but were not explored here mainly because of their una-
vailability at the small scale in this study. Topography was character-
ized by classifying the slope and elevaon in three categorical levels.
Slope was categorized as at (low slope), gentle (intermediate slope),
and steep (high slope). As for elevaon, low, medium, and high cat-
egories were considered. Simple linear models were performed to test
for slope and elevaon eects on aboveground carbon storage. As a
result, only the slope showed slightly signicant impact on the car-
bon stock and therefore was considered for further analyses. Mulple
linear regressions were also ed on slope and species richness to
test their eects on aboveground carbon storage. For both simple
and mulple linear models, Shapiro–Wilk tests were used to check
for the normality of the log- transformed AGC data and of the residu-
als. Addionally, Breusch–Pagan tests and Durbin–Watson stascs
were used to test for homoscedascity and autocorrelaon between
We next examined the relaonship of each diversity component
(i.e., funconal diversity and funconal dominance) with carbon stor-
age, by ng separate linear mixed- eects models (Zuur, Ieno, Walker,
Saveliev, & Smith, 2009). We considered species richness and slope
as random factors, and each measure of funconal diversity (i.e., Fric,
Feve, Fdiv, Fdis, and RaoQ) and of funconal dominance (i.e., CWM
of WD, SLA, and PHm) as xed eects. The mixed- eects models
were ed to assess (1) individual eect of each funconal diversity
and funconal dominance metric; (2) combined eects of funconal
diversity metrics; (3) combined eects of funconal dominance met-
rics; and (4) combined eects of funconal diversity and funconal
dominance metrics. The best models were selected by performing a
backward eliminaon of nonsignicant eects (p- value >.05). The lin-
ear mixed- eects models were performed using the “lmer” funcon of
the “lmerTest” package (Kuznetsova, Brockho, & Christensen, 2016)
of the R stascal soware. The p- values reported were calculated
from the F test based on Saerthwaite approximaons to the de-
grees of freedom, in the package “lmerTest” (Kuznetsova et al., 2016).
The signicance of the random eects was assessed using likelihood
rao (LR) test, again in the package “lmerTest”. The performance of
ed models was assessed based on the t stascs such as Akaike
informaon criterion (AIC) and the marginal R square, which indicates
the proporon of variance explained by xed eects (Nakagawa &
3 | RESULTS
A total of 50 plant species were enumerated, belonging to 46 genera
and 33 families. The dominant families were Rutaceae (ve species),
Rubiaceae (four species), Slbaceae (three species), and Celastraceae
(three species). The number of species per plot, for trees ≥5 cm dbh,
ranged from ve species to 18 species, with an average species rich-
ness of 11 species per plot. Tree number varied from 19 to 67 stems,
with an average value of 42 stems per plot. The amount of above-
ground carbon was highly variable across all plots, and ranged from
49.1 MgC/ha to 476.1 MgC/ha, with an esmated average value of
3.1 | Diversity eects mediated through funconal
diversity and funconal dominance
The outputs of the structural equaon models ed to assess the me-
diated eects of diversity (species richness) on AGC, through func-
onal diversity and funconal dominance, are summarized in Table 1
and Figure 1. The rst model “full mediaon” had chi- square value of
11.59 (df = 7; p = .115), indicang good t and absence of signicant
deviaons between data and model.
In the “full mediaon” model, species richness showed a signi-
cant posive direct eect on funconal richness (R2 = 0.47; β = 0.69;
p < .001; Table 1), which also showed posive and signicant eect
on AGC (β = 0.47; p = .002; Table 1). Therefore, species richness,
through funconal richness, had a posive signicant eect on AGC
(β = 0.69*0.47 = 0.32). There was a nonsignicant eect of spe-
cies richness on funconal evenness (β = 0.02; p = .926; Table 1);
the laer, however, exhibited a signicant negave eect on AGC.
In addion, we found no signicant correlaon between funconal
richness and funconal evenness (β = 0.29; p = .090; Table 1), which
would suggest that the mediated eects of species richness were
transmied by funconal richness only. Among the funconal dom-
inance metrics, the CWM of maximum plant height did not retain any
signicant path. Only the CWM of wood density showed signicant
responses to species richness (R2 = 0.15; β = 0.38; p = .028), but did
not signicantly inuence the AGC (p = .275). In contrast, the CWM
of SLA had a negave signicant eect (β = −0.37; p = .039; Table 1)
on AGC, although not signicantly inuenced by species richness
(p = .324). The signicant residual correlaon between CWM of wood
density and CWM of SLA (β = 0.45; p = .003; Table 1) suggests that
the mediated eects of species richness are also transmied by these
Note that the “paral mediaon” model was ed by only adding a
direct path from species richness to AGC to the “full mediaon” model.
The chi- square value for the “paral mediaon” model was 7.57 with
6 degrees of freedom and a p value of .272, also indicang good t.
MENSAH ET AL.
There are similaries between the two models in terms of signicant
and nonsignicant paths (Table 1), but the “paral mediaon” model
exhibited slightly beer ts (CFI = 0.932; R2 = 0.52; AIC = 304.2) than
the “full mediaon” model (CFI = 0.977; R2 = 0.45; AIC = 306.2). The
added causal path from species richness to AGC was slightly signif-
icant at 0.05, suggesng an exisng true direct eect of diversity
on AGC. Both models suggest that species richness eects on abo-
veground carbon are mediated through funconal diversity and func-
3.2 | Eects of environmental variables, funconal
diversity, and funconal dominance on carbon storage
Not surprisingly, there were signicant eects of the environmental
variables, especially the slope which explained 14.05% of the varia-
on of the aboveground carbon (Table 2). Low slope showed regres-
sion coecient which was 0.53 signicantly higher than the baseline
(higher slope), whereas intermediate slope was not. This indicates that
carbon stock was signicantly higher at low slope sites than high slope
sites. Unlike slope, altude did not have any signicant inuence on
the aboveground tree carbon (F- stasc = 1.381; p = .268; Table 2).
Furthermore, while accounng for the eects of the slope, we also
found that species richness was signicant and showed a posive re-
laonship with AGC (β = 0.06; p = .016; Table 2).
The results of the separate linear mixed- eects models tesng the
individual eects of funconal diversity metrics revealed that only Feve
was signicant, and had a negave eect on AGC (β = −1.6; p = .037;
Table 3). Fdis, Fdiv, and RaoQ showed high values of probability (from
0.359 to 0.528), while Fric had a slightly signicant and posive eect
on AGC (p = .079; Table 3). While assessing the combined eects of
funconal diversity metrics, we found that Fdis, Fdiv, and RaoQ were
le out aer backward selecon for the nal model (Table 3). The ef-
fects of funconal diversity on AGC were thus shown by a signicant
posive eect of funconal richness (β = 135.6; p = .013; Table 3)
and a signicant negave eect of funconal evenness (β = −2.03;
p = .006; Table 3). Both funconal richness and evenness explained
27% of the variance of AGC.
All the three funconal dominance metrics used in this study
showed signicant eects on the aboveground carbon (Table 4).
Both CWM of SLA and CWM of WD showed negave eects, while
CWM of maximum plant height exhibited a posive eect (Table 4).
However, when assessing their combined eects on AGC, CWM of
SLA was not retained by the nal model, and the eects of funconal
dominance were only shown by posive and signicant eects of
TABLE1 Results of the structural equaon modeling carried out to test the eects of species richness on carbon stocks (AGC) via funconal
diversity and funconal dominance
Est.std SE Zp- value Est.std SE Zp- value
Full mediaon Paral mediaon
Path from species richness to Fric 0.69 0.14 5.02 <.001 0.69 0.14 5.02 <.001
Path from species richness to Feve 0.02 0.19 0.09 .926 0.02 0.19 0.09 .926
Path from species richness to CWM (PHm) 0.06 0.19 0.32 .750 0.06 0.19 0.32 .750
Path from species richness to CWM (SLA) −0.18 0.19 −0.99 .324 −0.18 0.19 −0.99 .324
Path from species richness to CWM (WD) 0.38 0.18 2.20 .028 0.38 0.18 2.20 .028
Path from Fric to AGC 0.47 0.16 3.04 .002 0.24 0.19 1.27 .203
Path from Feve to AGC −0.39 0.14 −2.70 .007 −0.38 0.14 −2.75 .006
Path from CWM (PHm) to AGC −0.10 0.22 −0.46 .642 −0.16 0.21 −0.77 .440
Path from CWM (SLA) to AGC −0.37 0.18 −2.06 .039 −0.30 0.17 −1.74 .081
Path from CWM (WD) to AGC −0.21 0.19 −1.09 .275 −0.33 0.20 −1.66 .096
Path from species richness to AGC 0.41 0.20 2.00 .046
Path from CWM (WD) to CWM (SLA) 0.45 0.15 3.02 .003 0.45 0.15 3.02 .003
Path from CWM (WD) to CWM (PHm) −0.71 0.09 −7.50 <.001 −0.71 0.09 −7.50 <.001
Path from CWM (SLA) to CWM (PHm) −0.63 0.11 −5.54 <.001 −0.63 0.11 −5.54 <.001
Path from Feve to Fric 0.29 0.17 1.69 .090 0.29 0.17 1.69 .090
Model t stascs
AIC 306.2 304.2
p- value (chi- square) .115 .275
Est.std, path standardized coecients; SE, standard error; Fric, funconal richness; Feve, funconal evenness; CWM, community weight mean; PHm, maxi-
mum plant height; SLA, specic leaf area; WD, wood density.
MENSAH ET AL.
CWM of maximum plant height and CWM of wood density, with 21%
explained variance (Table 4).
Examinaon of separate mixed- eects models for funconal di-
versity and funconal dominance revealed that the marginal R square
(variance explained by xed factors) in the diversity–AGC relaonship
was greater for funconal diversity (27%) than for funconal domi-
nance (21%). When considering funconal diversity and funconal
dominance measures in a same model, we found that 34% of the
variaons of AGC were explained by signicant eects of funconal
richness, funconal evenness, and CWM of maximum plant height
FIGURE1 Summary of the path model relating species diversity (species richness), and measures of functional diversity and of functional
dominance to the aboveground carbon (AGC); a: full mediation; b: partial mediation. CWM: community weight mean; PHm: maximum plant
height; SLA: specific leaf area; WD: wood density. The figures with parentheses are the coefficients of determination (R2), shown for dependent
variables. The figures without parentheses are the standardized path coefficients. The single- pointed arrows represent causal paths, while the
double- pointed arrows represent the residual correlations. The blue lines indicate the positive effects, while the red lines show negative effects;
Chisq, Chi- square statistic; DF, degree of freedom indicating the number of paths omitted from the model; Prob, probability of the data given
the model; Prob >.05 indicates the absence of significant discrepancy between the data and the model. CFI, comparative fit index; AIC, Akaike
information criterion. The significance of each path is given in Table 1
TABLE2 Results of simple and mulple linear models tesng the eects of elevaon, slope, and richness on aboveground carbon stock
Est. SE t value Pr (>|t|) SW BP DW
Elevaon (Intercept) 12.15 0.19 63.48 <0.001 0.849 0.240 1.68
Low −0.36 0.24 −1.48 0.152
Medium −0.09 0.23 −0.40 0.691
Adjusted R2 (%) 2.56
Slope (Intercept) 11.67 0.20 59.24 <0.001 0.927 0.211 1.69
Low 0.53 0.23 2.32 0.028
Medium 0.19 0.24 0.84 0.409
Adjusted R2 (%) 14.05
Slope + Species richness (Intercept) 10.98 0.32 34.19 <0.001 0.821 0.263 1.93
Low 0.51 0.21 2.45 0.021
Medium 0.16 0.22 0.72 0.479
Species richness 0.06 0.03 2.56 0.017
Adjusted R2 (%) 28.71
Est., esmates of regression coecients; SE, standard errors; SW, p- values for Shapiro–Wilk normality tests; BP, p- values for Breusch–Pagan tests; DW,
Durbin–Watson autocorrelaon stasc.
MENSAH ET AL.
(Table 5). For all the selected models, species richness as random fac-
tor had much less variability than slope. The nonsignicant variabil-
ity due to species richness in the mixed- eects models suggests that
much of its inuence on AGC has been considered by funconal diver-
sity and funconal dominance, as conrmed by the SEM.
4 | DISCUSSION
Our study explored the paerns of diversity–carbon stock relaonship
in mistbelt forests in South Africa, nding that carbon stocks varied
greatly as responses to environmental gradients, taxonomic diversity,
funconal diversity, and funconal dominance. Specically, the study
revealed that (1) slope gradient signicantly inuenced aboveground
carbon, with lower carbon stock found at steeper sites; (2) increasing
species diversity (species richness) increased tree carbon stock; (3) di-
versity eects on tree carbon stock were mediated through funconal
diversity and funconal dominance; (4) funconal diversity eects on
tree carbon stock were greater than those of funconal dominance;
and (5) the specic eects of funconal diversity and funconal domi-
nance on carbon stock varied with metrics and funconal traits.
4.1 | Eects of environmental variables on tree
We did not detect any signicant eect of altude on tree carbon stock,
according to Cavanaugh et al. (2014) who also reported in a global
scale study, a lack of signicant relaonship between forest carbon and
TABLE3 Results of linear mixed- eects models tesng the eects of funconal diversity on aboveground carbon stock
Fixed eects Random eects (variance)
Est. SE df tPr (>|t|) Sp.rich. Slope Rsd. Marg. R2AIC
(Intercept) 11.76 0.16 2.98 71.90 <0.001 0.00 0.05 0.15 0.09 30.74
Fric 103.06 56.38 24.19 1.83 0.079
(Intercept) 12.92 0.48 25.97 27.11 <0.001 0.00 0.03 0.15 0.13 37.96
Feve −1.66 0.75 24.58 −2.21 0.037
(Intercept) 11.75 0.27 8.16 43.48 <0.001 0.00 0.05 0.17 0.01 40.77
Fdis 1.00 1.57 25.82 0.64 0.528
(Intercept) 12.30 0.446 22.51 27.577 <0.001 0.01 0.02 0.16 0.03 41.95
Fdiv −0.64 0.686 25.47 −0.935 0.359
(Intercept) 11.77 0.22 4.18 53.14 <0.001 0.00 0.06 0.17 0.02 38.38
RaoQ 3.82 4.66 25.80 0.82 0.42
(Intercept) 12.97 0.43 24.83 30.08 <0.001 0.00 0.04 0.12 0.27 23.83
Fric 135.59 50.64 23.15 2.68 0.013
Feve −2.03 0.68 23.32 −2.97 0.006
Est., coecient esmates; SE, standard errors; Sp.rich., species richness; Rsd., residual variance; Marg. R2, marginal R square; Fric, funconal richness; Feve,
funconal evenness; Fdis, funconal dispersion; Fdiv, funconal divergence; RaoQ, Rao quadrac entropy.
TABLE4 Results of linear mixed- eects models tesng the eects of funconal dominance on aboveground carbon stock
Fixed eects Random eects (variance)
Est. SE df tPr (>|t|) Sp.rich. Slope Rsd. Marg. R2AIC
(Intercept) 13.92 0.66 18.99 21.15 <0.001 0.03 0.03 0.10 0.20 44.18
CWM (SLA) −0.02 0.01 17.55 −3.14 0.006
(Intercept) 10.21 0.51 20.18 20.14 <0.001 0.08 0.11** 0.07 0.17 41.29
CWM (PHm) 0.07 0.02 18.45 3.66 0.002
(Intercept) 14.85 1.19 16.42 12.46 <0.001 0.15 0.05 0.09 0.10 38.39
CWM (WD) −4.86 1.94 15.37 −2.50 0.024
(Intercept) 6.06 2.06 24.64 2.95 0.007 0.00 0.16** 0.11 0.21 38.03
CWM (PHm) 0.11 0.03 24.35 3.63 0.001
CWM (WD) 5.35 2.44 23.96 2.19 0.038
**Signicant at 0.01.
Est., coecient esmates; SE, standard errors; Sp.rich., species richness; Rsd., residual variance; Marg. R2, marginal R square; CWM (SLA), community
weight mean of specic leaf area; CWM (WD), community weight mean of wood density; CWM (PHm), community weight mean of maximum plant height.
MENSAH ET AL.
altude. Yet, this nding runs contrary to many previous studies that
examined the relaonships between altude and biomass or carbon
storage (de Caslho et al., 2006; Ensslin et al., 2015; Sharma, Baduni,
Gairola, Ghildiyal, & Suyal, 2010). On the one hand, some authors re-
ported that biomass and carbon stocks can decline with increasing al-
tude (de Caslho et al., 2006; Moser, Hertel, & Leuschner, 2007). On the
other hand, studies found posive correlaon between increasing tree
carbon and increasing altude (Gairola, Sharma, Ghildiyal, & Suyal, 2011;
Zhu et al., 2010). Furthermore, biomass and carbon stocks were found
to increase up to a certain altudinal limit (3,000 m a.s.l.) and aerward
decline sharply with higher altudinal values (Ensslin et al., 2015; Singh,
Adhikari, & Zobel, 1994). This lack of clarity on the relaonship between
altude and forest biomass may be partly due to the variaon in the al-
tudinal range among studies. For instance, most of the abovemenoned
studies that reported signicant eects of altude have covered greater
altudinal ranges well above 2,500 m a.s.l; the relaonship between al-
tude and carbon stocks in our study might have been hidden due to
the smaller altudinal range covered (1,000–1,800 mm), which might
have not been considerable enough to detect substanal variaon in
growth condions and hence biomass and carbon stock.
Unlike altude, slope showed signicant inuence, and accounted
for 14% of carbon variance, evidencing that dierences in carbon
stocks can result from topological constraints, parcularly dierence in
slope. Consistent with our results, slope has been idened as a poten-
al environmental variable that aects tree carbon (de Caslho et al.,
2006; Chave et al., 2003). Because aboveground carbon is intrinsically
related to wood and biomass producon, the inuence of slope can be
seen as prior impacts of environment on availability of resources (de
Caslho et al., 2006; Luizao et al., 2004), which in turn aect forest dy-
namics. For example, steeper slope will speed up nutrients and water
runo and constrain trees and will also favor highly water and nutri-
ent ecient species against others. Taking this into account, it follows
that tree growth and biomass producon can be potenally reduced
at higher slope sites, as results of moisture and nutrient stress (Clark,
Clark, & Oberbauer, 2010; Durán et al., 2015), whereas at and gen-
tle slope sites would allow for more water availability, to which plant
would likely respond posively. The signicant eect of slope supports
the fact that ecosystem funcons in general and biomass and carbon
storage in parcular are environment- structured (Wu et al., 2015).
4.2 | Increasing species diversity promotes tree
We found signicant and posive eects of species richness on
aboveground carbon, even when the eects of environmental fac-
tors (i.e., slope) were accounted for. While this nding accords with
some recent studies that controlled for the eects of environmen-
tal variables (Ouyang et al., 2016; Wu et al., 2015), it also supports
the commonly described paern in highly diverse natural forests;
that is, biomass and carbon stocks increase with increasing diversity.
Indeed, several local and global studies on forest ecosystems have
shown posive relaonship between species richness and forest bio-
mass or carbon (Cavanaugh et al., 2014; Con et al., 2013; Day et al.,
2014; Ruiz- Benito et al., 2014; Sharma et al., 2010; Wu et al., 2015).
In addion, studies in boreal (Paquee & Messier, 2011), temper-
ate (Paquee & Messier, 2011; Vilà et al., 2007), and tropical forests
(Barrufol et al., 2013) have also reported increases in producvity
with increasing diversity.
One can expect that increasing species diversity would increase
carbon storage because higher taxonomic diversity would lead to
higher stem density and forest producvity (Ruiz- Benito et al., 2014).
The posive eect of species diversity can also be explained through
the benets of plant–plant interacons such as facilitaon, where by
some species could enhance soil ferlity (by xing nitrogen) for the
producvity of other species. This fact is even oen used to support
the reason why mixed species communies of plantaons are far
more producve than monospecic stands. But it might also be well
possible that increasing species richness increases the chances of in-
clusion of highly producve and naturally favored dominant species
(Ruiz- Benito et al., 2014), as shown by our previous results on the in-
uence of most dominant species on carbon stocks in mistbelt forests
(Mensah, Veldtman, du Toit, Glèlè Kakaï, & Seifert, 2016).
TABLE5 Results of linear mixed- eects models tesng the combined eects of funconal diversity and funconal dominance on aboveground
carbon (AGC) stock
Model descripon Est. SE df tPr (>|t|)
+ Funconal dominance
Fixed eects (Intercept) 11.39 0.63 23.82 18.03 <0.001
Fric 143.50 42.65 21.99 3.37 0.003
Feve −1.72 0.58 22.15 −2.95 0.008
CWM (PHm) 0.06 0.02 22.80 3.10 0.005
Species richness 0.00
**Signicant at 0.01.
Est., coecient esmates; SE, standard errors; Fric, funconal richness; Feve, funconal evenness; CWM (PHm), community weight mean of maximum
MENSAH ET AL.
While our dataset in the mistbelt forests supports the posive spe-
cies richness–carbon relaonship, it must be noted that evidence of the
inverse eect also exists. For instance, studies by Ruiz- Jaen and Potvin
(2011) in natural forest of Barro Colorado Island in Central Panama and
Szwagrzyk and Gazda (2007) in natural forests of central Europe revealed
negave relaonship of species diversity with biomass and carbon
stocks. Furthermore, others studies found such relaonships nonsignif-
icant (see Gairola et al., 2011). These controversial outcomes suggest
that the eects of diversity on forest carbon may vary with other factors
such as the types and the successional stages of the forests (Lasky et al.,
2014; Wu et al., 2015), and also the specic dimension of the diversity
measure used (Con et al., 2013; Lasky et al., 2014; Ouyang et al., 2016).
4.3 | Diversity eects mediated through funconal
diversity and funconal dominance
The use of mulple diversity measures to provide addional insights
into the mechanisms behind diversity–producvity has gained in-
creasing interest in recent years (Cadoe et al., 2011; Con & Díaz,
2013; Finegan et al., 2015; Lasky et al., 2014; Ruiz- Benito et al., 2014;
Vance- Chalcra, Willig, Cox, Lugo, & Scatena, 2010; Ziter, Benne,
& Gonzalez, 2013). Accordingly, funconal diversity and dominance
metrics were also examined in this study. While most of these studies
tended to compare the relave eects of species richness and other
diversity measures, we have provided here an addional example of
exploring diversity eects on carbon stocks, by assuming that these
eects were mediated through funconal diversity and funconal
dominance. Our results on the structural equaon modeling conrm
this assumpon and therefore support the need to explore beyond
species richness to beer elucidate the mechanisms that govern di-
versity–producvity relaonship. The results further support the idea
that both complementarity and selecon eects are not exclusively
aecng carbon storage (Ruiz- Benito et al., 2014; Wu et al., 2015).
Diversity (species richness) promotes carbon stock through eects of
both funconal diversity and funconal dominance, partly because
these diversity components are based on specic funconal traits,
which would reect funconal dierences among the species (Dı ́az &
Cabido, 2001; Song et al., 2014). This nding can also be due to the
fact that increased species richness indirectly accounted for dier-
ences among species, in terms of ecological niche and resource use.
4.4 | Funconal diversity eects on tree
Of the ve funconal diversity indices used in this study, only funconal
richness and funconal evenness were found to explain variaon in car-
bon stock. There is a variety of evidence for funconal diversity eects
on biomass and carbon. A study by Finegan et al. (2015) in tropical rain
forests of Bolivia, Brazil, and Costa Rica found only funconal rich-
ness—among other funconal diversity indices—as signicant predictor
for biomass variaon. Yet, a study in unmanaged forest fragments in
Quebec revealed signicant and posive relaonships between func-
onal dispersion and AGC (Ziter et al., 2013). Similarly, Ouyang et al.
(2016) found signicant but negave eects of the Rao quadrac en-
tropy on stand biomass in subtropical forests in China. While we be-
lieve that these funconal diversity indices have their specic biological
meaning, in this study, the posive eect of funconal richness on the
AGC could be due to funconal richness being posively correlated
with species richness (SEM results; Villéger et al., 2008).
The funconal richness measures the amount of trait or niche
space lled by the species within a community (Clark et al., 2012;
Mason, Mouillot, Lee, & Wilson, 2005). It would increase carbon stor-
age because species with various traits would dier in resource use,
and would more eciently use the resources available within the com-
munity for higher growth and producvity, thus reecng the niche
complementarity eects (Finegan et al., 2015). Unlike funconal rich-
ness, funconal evenness did not show any relaonship with species
diversity; however, it did exhibit negave inuence on AGC. Following
Mason et al. (2005), the funconal evenness measures the evenness
of abundance distribuon in the lled niche space. Therefore, both
funconal richness and funconal evenness relate to the niche space
or secons of niche space, and funconal diversity as measured here
could reect some form of “niche dierences” (Carroll, Cardinale, &
Nisbet, 2011). Greater funconal diversity, that is, greater value and
range of funconal traits, would reect not only the magnitude of
“niche dierences”, but also the dierences in resource ulizaon by
species, thus promong diversity eects on ecosystem funconing.
This is in line with Carroll et al. (2011) who showed that increasing
niche dierence contributes to species coexistence and posive diver-
sity eects on biomass yield.
The unexpected lack of strong individual eect of funconal rich-
ness on aboveground carbon in this study might be due to the num-
ber of funconal traits used. In fact, only three funconal traits were
considered; although these traits were found to be crucial to explain
biomass allocaon paerns (Chave et al., 2009; Mensah, Glèlè Kakaï,
et al., 2016), they might not be as important as we thought for com-
plementary resource allocaon. Similarly, these funconal traits might
not be sucient enough to catch the enre variability needed to ex-
plain carbon variaon. Adding other funconal traits such as plant hy-
draulic conducvity, leaf mass per area, and nitrogen xing potenal
could have well captured the funconal variability.
4.5 | Funconal dominance eects on tree
The use of CWM values of funconal trait to predict funconal domi-
nance eects is supported by the understanding that CWM metric
reects dominance of traits and species within a given community,
and also in line with the fact that dominant species would induce
funconal shis in mean trait values (Ricoa & More, 2011). CWM
as funconal dominance metric could be used to elaborate on com-
peve dominance of species (Ricoa & More, 2011). Therefore,
funconal dominance could indicate some aspect of “relave tness
dierences” between competors (Carroll & Nisbet, 2015; Carroll
et al., 2011). Moreover, the nding that funconal dominance signi-
cantly inuenced tree carbon storage is consistent with the previous
MENSAH ET AL.
report that the magnitude of “relave tness dierences” strength-
ens the inuence of diversity on biomass yield (Carroll et al., 2011).
The funconal dominance eects, as measured in this study, varied
with the funconal trait. Specically, CWM of wood density revealed
negave and signicant eect on carbon stocks. It is not surprising
given that wood density is a potenal predictor of tree biomass, which
highly correlates with the carbon stock. There are some insights that
CWM of wood density is negavely related to the biomass incre-
ment, as being good predictor of individual tree diameter increments
(Finegan et al., 2015). However, aer examining biomass stocks in
tropical forests, Stegen, Swenson, Valencia, Enquist, and Thompson
(2009) pointed out that increasing wood density can decrease or in-
crease the carbon stock, regardless of whether trees have high or low
mean wood density. The authors therefore came to the conclusion
that no general relaonship exists between forest biomass and wood
density. The present nding about CWM of wood density means that
low wood density species grow faster and tend to store more bio-
mass; thus, it suggests that conserving and planng low wood density
species would likely help to increase the carbon stock.
Similarly, CWM of specic leaf area exhibited negave and signi-
cant eect on carbon stocks. This is consistent with other studies that
found negave relaonship between specic leaf area and plant biomass
(Finegan et al., 2015). Leaf area is important for the amount of radiant
energy intercepted by the plant. It is also generally known to facilitate the
transfer of CO2 and water between foliage and atmosphere. Therefore,
the signicant inuence of CWM of specic leaf area in this study sup-
ports the idea that leaf area captures a strategy of the plant for resource
consumpon, especially light (Mensah, Glèlè Kakaï, et al., 2016).
Community weight mean of maximum plant height showed
posive relaonship with carbon storage, as also reported in re-
cent studies (Con & Díaz, 2013; Finegan et al., 2015; Ruiz- Jaen &
Potvin, 2011). This is probably because tree height is a key variable
for species- specic or mulspecies biomass regressions. In addion,
maximal tree height is a potenal species trait, as it denes the lim-
its of compeon for light and thus for light consumpon (Poorter,
Bongers, & Bongers, 2006; Poorter, Bongers, Sterck, & Woll, 2005).
Examinaon of combined eects of funconal dominance metrics
revealed that only CWM of wood density and of maximum plant
height were retained in the nal model, with maximum plant height
being the most signicant predictor. Furthermore, only maximum
plant height was also retained among funconal dominance metrics
when we assessed the combined eects of funconal dominance and
funconal diversity. Tree height being closely related to tree diameter,
the posive and signicant relaonship between CWM of maximum
plant height and carbon stocks reects the potenal importance of
characteriscs of dominant and adult trees for ecosystem funcon-
ing and producvity, thus supporng the selecon eects hypothesis.
The important contribuon of dominant stems to forest biomass has
well been evidenced in some recent studies (Chave et al., 2003; Lung
& Espira, 2015). The study by Lung and Espira (2015) revealed that
tree stems larger than 50 cm have the greatest impact on forest bio-
mass, and <16% of the species pool accounted for over 62% of the
4.6 | Funconal diversity eects greater than
those of funconal dominance
When examining the percentage of variance explained, we found that
funconal diversity explained more variance than funconal domi-
nance (Tables 2 and 4). This rejects our second hypothesis, and sug-
gests that complementarity eects seem to be more important than
selecon eects. This nding contradicts Finegan et al.’s (2015) and
Ruiz- Jaen and Potvin’s (2011) results that selecon eects were more
important for the aboveground biomass and carbon stock in tropi-
cal forests. For this study, funconal dominance metrics (community
weight mean of funconal traits) were calculated using species rela-
ve abundance, while Ruiz- Jaen and Potvin (2011) and Finegan et al.
(2015) used species relave basal area and species relave biomass,
respecvely, as weighng variable. The strength of relaonship be-
tween community weight mean of traits and the ecosystem funcon
of interest could depend on the weighng variable. Biomass- or basal
area- weighted communies mean values would likely show stronger
relaon with biomass and carbon than abundance- based communies
mean values. Further studies should elaborate on this and show the
extent to which weighng variable can inuence our understanding of
weighted mean values’ eects on ecosystem funcons.
All being considered, it is important to menon that our result
actually supports the idea that these two hypotheses (complemen-
tarity and selecon eects) are not exclusive, and can contribute to
ecosystem funconing. Previous evidence of both complementarity
and selecon eects on ecosystem funcon suggests they can also
contribute at dierent proporons at dierent mes of ecosystem
development (Fargione et al., 2007). Both complementarity and se-
lecon eects mutually promote species coexistence. As pointed
out by Carroll et al. (2011), these two hypotheses could even be the
outcome of interacons of the “relave tness dierences” and the
“niche dierences”, whereby some species’ populaons could be sup-
pressed by dominant competors, to allow eecve ulizaon of the
available resources. The selecon eects reported here are strongly
transmied through specic maximum plant height, which reects the
inuence of dominant species and suggests a possible compeve ex-
clusion in terms of ulizaon of resources (e.g., light). In mulspecies,
mulstory natural forests, chances are high to observe dominant and
taller species that increase stand producvity, probably by achieving
higher absorpvity of photosynthecally acve radiaon, thus re-
ducing (through compeve dominance) the level of photosynthec
photon ux density available for understory species. However, it must
be noted that, even for these dominant species, interacons within
ontogenic stages (for example, compeon for light between seed-
lings, juveniles, and adults) could dene an ecient complementary
use of light for greater producvity. Furthermore, an ecient use by
understory species (limited to the subcanopy layer) of the available
photosynthec photon ux density, and also of decomposed lier
(from canopy and dominant trees leaves) may likely reect some com-
plementary eects on stand producvity. Therefore, selecon eects
(dominant traits and species) on ecosystem funcon would be appar-
ent in natural forests as we predicted, but complementary eects and
MENSAH ET AL.
ecient use of limited resources, especially by coexisng and under-
story species, could promote greater ecosystem funcon.
5 | CONCLUSION
This study examined the diversity–carbon stock relaonship in
mistbelt forests in South Africa and revealed that taxonomic diver-
sity (species richness) promotes carbon storage through funconal
diversity and funconal dominance. The study further highlighted that
both the niche complementarity and selecon hypotheses are impor-
tant for carbon storage. However, the eects of funconal diversity
(niche complementarity eects) were greater than funconal domi-
nance eects (selecon eects). Moreover, the eects of funconal
dominance were strongly transmied through the CWM of maximum
plant height, reecng the importance of forest vercal stracaon
for diversity–carbon relaonship. Therefore, complementary eects
would be induced also by complementary light- use eciency of spe-
cies and trees growing in the understory layer. We suggest that future
research on the relaon between diversity and forest carbon be ori-
ented toward a perspecve of forest canopy (or dominant species vs.
other species), to contribute addional insights into our understand-
ing of biodiversity–ecosystem funcon relaonship.
We sincerely acknowledge the anonymous reviewers and the editor
for their cricism and their construcve comments on the early ver-
sion of this arcle. This study was nancially supported by the SHARE
INTRA- ACP project, the Naonal Research Foundaon of South
Africa through the project “Catchman Letaba” in the RTF funding
scheme, and the African Forest Forum.
CONFLICT OF INTEREST
Baraloto, C., Hérault, B., Paine, C. E. T., Massot, H., Blanc, L., Bonal, D., …
Sabaer, D. (2012). Contrasng taxonomic and funconal responses
of a tropical tree community to selecve logging. Journal of Applied
Ecology, 49, 861–870.
Barrufol, M., Schmid, B., Bruelheide, H., Chi, X., Hector, A., Ma, K., …
Niklaus, P. A. (2013). Biodiversity promotes tree growth during succes-
sion in subtropical forest. PLoS ONE, 8, 1–9.
Cadoe, M. W., Carscadden, K., & Mirotchnick, N. (2011). Beyond species:
Funconal diversity and the maintenance of ecological processes and
services. Journal of Applied Ecology, 48, 1079–1087.
Carroll, I. T., Cardinale, B. J., & Nisbet, R. M. (2011). Niche and tness dif-
ferences relate the maintenance of diversity to ecosystem funcon.
Ecology, 92, 1157–1165.
Carroll, I. T., & Nisbet, R. M. (2015). Departures from neutrality induced
by niche and relave tness dierences. Theorecal Ecology, 8,
de Caslho, C. V., Magnusson, W. E., de Araújo, R. N. O., Luizão, R. C. C.,
Luizão, F. J., Lima, A. P., & Higuchi, N. (2006). Variaon in aboveground
tree live biomass in a central Amazonian Forest: Eects of soil and to-
pography. Forest Ecology and Management, 234, 85–96.
Cavanaugh, K. C., Gosnell, J. S., Davis, S. L., Ahumada, J., Boundja, P., Clark,
D. B., … Andelman, S. (2014). Carbon storage in tropical forests cor-
relates with taxonomic diversity and funconal dominance on a global
scale. Global Ecology and Biogeography, 23, 563–573.
Chave, J., Andalo, C., Brown, S., Cairns, M. A., Chambers, J. Q., Eamus, D., …
Yamakura, T. (2005). Tree allometry and improved esmaon of carbon
stocks and balance in tropical forests. Oecologia, 145, 87–99.
Chave, J., Condit, R., Lao, S., Caspersen, J. P., Foster, R. B., & Hubbell, S. P. (2003).
Spaal and temporal variaon of biomass in a tropical forest: Results from
a large census plot in Panama. Journal of Ecology, 91, 240–252.
Chave, J., Coomes, D., Jansen, S., Lewis, S. L., Swenson, N. G., & Zanne,
A. E. (2009). Towards a worldwide wood economics spectrum. Ecology
Leers, 12, 351–366.
Clark, D. B., Clark, D. A., & Oberbauer, S. F. (2010). Annual wood producon
in a tropical rain forest in NE Costa Rica linked to climac variaon but
not to increasing CO2. Global Change Biology, 16, 747–759.
Clark, C. M., Flynn, D. F. B., Buereld, B. J., & Reich, P. B. (2012). Tesng
the link between funconal diversity and ecosystem funconing in a
minnesota grassland experiment. PLoS ONE, 7, e52821.
Con, T. V., Thang, N. T., Ha, D. T. T., Khiem, C. C., Quy, T. H., Lam, V. T., …
Sato, T. (2013). Relaonship between aboveground biomass and mea-
sures of structure and species diversity in tropical forests of Vietnam.
Forest Ecology and Management, 310, 213–218.
Con, G., & Díaz, S. (2013). Plant funconal diversity and carbon storage - an
empirical test in semi- arid forest ecosystems. Journal of Ecology, 101, 18–28.
Day, M., Baldauf, C., Rushauser, E., & Sunderland, T. C. H. (2014).
Relaonships between tree species diversity and above- ground
biomass in Central African rainforests: Implicaons for REDD.
Environmental Conservaon, 41, 64–72.
Dı́az, S., & Cabido, M. (2001). Vive la diérence: Plant funconal diversity mat-
ters to ecosystem processes. Trends in Ecology & Evoluon, 16, 646–655.
Durán, S. M., Sánchez-Azofeifa, G. A., Rios, R. S., & Gianoli, E. (2015). The rela-
ve importance of climate, stand variables and liana abundance for carbon
storage in tropical forests. Global Ecology and Biogeography, 24, 939–949.
Ensslin, A., Ruen, G., Pommer, U., Zimmermann, R., Hemp, A., & Fischer,
M. (2015). Eects of elevaon and land use on the biomass of trees,
shrubs and herbs at Mount Kilimanjaro. Ecosphere, 6, 45.
Fargione, J., Tilman, D., Dybzinski, R., Lambers, J. H. R., Clark, C., Harpole,
W. S., … Loreau, M. (2007). From selecon to complementarity: Shis in
the causes of biodiversity- producvity relaonships in a long- term bio-
diversity experiment. Proceedings of the Royal Society B, 274, 871–876.
Finegan, B., Peña-Claros, M., de Oliveira, A., Ascarrunz, N., Bret-Harte, M. S.,
Carreño-Rocabado, G., … Poorter, L. (2015). Does funconal trait diver-
sity predict above- ground biomass and producvity of tropical forests?
Tesng three alternave hypotheses. Journal of Ecology, 103, 191–201.
Gairola, S., Sharma, C. M., Ghildiyal, S. K., & Suyal, S. (2011). Live tree bio-
mass and carbon variaon along an altudinal gradient in moist tem-
perate valley slopes of the Garhwal Himalaya (India). Current Science,
Geldenhuys, C. J. (1997). Nave forest regeneraon in pine and eucalypt
plantaons in Northern Province, South Africa. Forest Ecology and
Management, 99, 101–115.
Geldenhuys, C. J. (2002). Tropical Secondary Forest Management in Africa:
Reality and Perspecves; Food and Agriculture Organizaon: Rome, Italy.
Grace, J. B., & Bollen, K. A. (2005). Interpreng the results from mulple
regression and structural equaon models. Bullen of the Ecological
Society of America, 86, 283–295.
Kage, J., Diaz, S., Lavorel, S., Prence, I. C., Leadley, P., Bönisch, G., …
Wirth, C. (2011). TRY - a global database of plant traits. Global Change
Biology, 17, 2905–2935.
Kuznetsova, A., Brockho, P. B., & Christensen, R. H. B. (2016). lmerT-
est: Tests in Linear Mixed Eects Models. R package version 2.0-32.
Retrieved from hps://CRAN.R-project.org/package=lmerTest.
MENSAH ET AL.
Laliberté, E., Legendre, P., & Shipley, B. (2015). Measuring funconal diver-
sity (FD) from mulple traits, and other tools for funconal ecology. R
Package Version 1.0–12.
Lasky, J. R., Uriarte, M., Boukili, V. K., Erickson, D. L., John Kress, W., & Chazdon,
R. L. (2014). The relaonship between tree biodiversity and biomass dynam-
ics changes with tropical forest succession. Ecology Leers, 17, 1158–1167.
Luizao, R. C. C., Luizao, F. J., Paiva, R. Q., Monteiro, T. F., Sousa, L. S., &
Kruijt, B. (2004). Variaon of carbon and nitrogen cycling processes
along a topographic gradient in a central Amazonian forest. Global
Change Biology, 10, 592–600.
Lung, M., & Espira, A. (2015). The inuence of stand variables and human use
on biomass and carbon stocks of a transional African forest: Implicaons
for forest carbon projects. Forest Ecology and Management, 351, 36–46.
Magurran, A. E. (1988). Ecological diversity and its measurement. Dordrecht,
the Netherlands: Springer.
Mason, N. W. H., Mouillot, D., Lee, W. G., & Wilson, J. B. (2005). Funconal
richness, funconal evenness and funconal divergence: The primary
components of funconal diversity. Oikos, 111, 112–118.
Mayeld, M. M., Bonser, S. P., Morgan, J. W., Aubin, I., McNamara, S., &
Vesk, P. A. (2010). What does species richness tell us about funconal
trait diversity? Predicons and evidence for responses of species
and funconal trait diversity to land- use change. Global Ecology and
Biogeography, 19, 423–431.
Mensah, S., Glèlè Kakaï, R., & Seifert, T. (2016). Paerns of biomass alloca-
on between foliage and woody structure : The eects of tree size and
specic funconal traits. Annals of Forest Research, 59, 49–60.
Mensah, S., Houehanou, T. D., Sogbohossou, E. A., Assogbadjo, A. E., &
Glèlè Kakaï, R. (2014). Eect of human disturbance and climac vari-
ability on the populaon structure of Afzelia africana Sm. ex pers.
(Fabaceae–Caesalpinioideae) at country broad- scale (Bénin, West
Africa). South African Journal of Botany, 95, 165–173.
Mensah, S., Veldtman, R., du Toit, B., Glèlè Kakaï, R., & Seifert, T. (2016).
Aboveground biomass and carbon in a South African mistbelt forest
and the relaonships with tree species diversity and forest structures.
Forests, 79, 1–17.
Mensah, S., Veldtman, R., & Seifert, T. (2016). Allometric models for height
and above ground biomass of dominant tree species in South African
mistbelt forests, Southern Forests: a Journal of Forest Science, DOI:10.29
Moser, G., Hertel, D., & Leuschner, C. (2007). Altudinal change in LAI
and stand leaf biomass in tropical montane forests: A transect study in
ecuador and a pan- tropical meta- analysis. Ecosystems, 10, 924–935.
Mouchet, M. A., Villéger, S., Mason, N. W. H., & Mouillot, D. (2010). Funconal
diversity measures: An overview of their redundancy and their ability to
discriminate community assembly rules. Funconal Ecology, 24, 867–876.
Naeem, S. (2002). Disentangling the impacts of diversity on ecosystem
funconing in combinatorial experiments. Ecology, 83, 2925–2935.
Nakagawa, S., & Schielzeth, H. (2013). A general and simple method for
obtaining R2 from generalized linear mixed- eects models. Methods in
Ecology and Evoluon, 4, 133–142.
Ouyang, S., Xiang, W., Wang, X., Zeng, Y., Lei, P., Deng, X., & Peng, C.
(2016). Signicant eects of biodiversity on forest biomass during
the succession of subtropical forest in south China. Forest Ecology and
Management, 372, 291–302.
Paquee, A., & Messier, C. (2011). The eect of biodiversity on tree
producvity: From temperate to boreal forests. Global Ecology and
Biogeography, 20, 170–180.
Poorter, L., Bongers, L., & Bongers, F. (2006). Architecture of 54 moist- forest tree
species: Traits, trade- os, and funconal groups. Ecology, 87, 1289–1301.
Poorter, L., Bongers, F., Sterck, F., & Woll, H. (2005). Beyond the regener-
aon phase: Dierenaon of height- light trajectories among tropical
tree species. Journal of Ecology, 93, 256–267.
R Core Team (2015). R: A language and environment for stascal compung.
Vienna, Austria: R Foundaon for Stascal Compung. URL hps://
Ricoa, C., & More, M. (2011). CWM and Rao’s quadrac diversity: A
unied framework for funconal ecology. Oecologia, 167, 181–188.
Rosseel, Y. (2012). lavaan: An R package for structural equaon modeling.
Journal of Stascal Soware, 48, 1–36.
Ruiz-Benito, P., Gómez-Aparicio, L., Paquee, A., Messier, C., Kage, J., &
Zavala, M. A. (2014). Diversity increases carbon storage and tree produc-
vity in Spanish forests. Global Ecology and Biogeography, 23, 311–322.
Ruiz-Jaen, M. C., & Potvin, C. (2011). Can we predict carbon stocks in tropical
ecosystems from tree diversity? Comparing species and funconal diver-
sity in a plantaon and a natural forest. New Phytologist, 189, 978–987.
Sharma, C. M., Baduni, N. P., Gairola, S., Ghildiyal, S. K., & Suyal, S. (2010).
Tree diversity and carbon stocks of some major forest types of Garhwal
Himalaya, India. Forest Ecology and Management, 260, 2170–2179.
Singh, S. P., Adhikari, B. S., & Zobel, D. B. (1994). Biomass, Producvity,
Leaf Longevity, and Forest Structure in the Central Himalaya. Ecological
Monographs, 64, 401–421.
Song, Y., Wang, P., Li, G., & Zhou, D. (2014). Relaonships between func-
onal diversity and ecosystem funconing: A review. Acta Ecologica
Sinica, 34, 85–91.
Stegen, J. C., Swenson, N. G., Valencia, R., Enquist, B. J., & Thompson, J.
(2009). Above- ground forest biomass is not consistently related to wood
density in tropical forests. Global Ecology and Biogeography, 18, 617–625.
Szwagrzyk, J., & Gazda, A. (2007). Above- ground standing biomass and
tree species diversity in natural stands of Central Europe. Journal of
Vegetaon Science, 18, 555–562.
Tilman, D., Knops, J., Wedin, D., Reich, P., Ritchie, M., & Siemann, E. (1997).
The inuence of funconal diversity and composion on ecosystem
processes. Science, 277, 1300–1302.
Vance-Chalcra, H. D., Willig, M. R., Cox, S. B., Lugo, A. E., & Scatena, F. N.
(2010). Relaonship between aboveground biomass and mulple mea-
sures of biodiversity in subtropical forest of Puerto Rico. Biotropica, 42,
Vilà, M., Vayreda, J., Comas, L., Ibánez, J. J., Mata, T., & Obón, B. (2007).
Species richness and wood producon: A posive associaon in
Mediterranean forests. Ecology Leers, 10, 241–250.
Villéger, S., Mason, N. W. H., & Mouillot, D. (2008). New muldimensional
funconal diversity indices for a mulfaceted framework in funconal
ecology. Ecology, 89, 2290–2301.
Wang, X., Fang, J., Tang, Z., & Zhu, B. (2006). Climac control of primary
forest structure and DBH- height allometry in Northeast China. Forest
Ecology and Management, 234, 264–274.
Wu, X., Wang, X., Tang, Z., Shen, Z., Zheng, C., Xia, X., & Fang, J. (2015).
The relaonship between species richness and biomass changes from
boreal to subtropical forests in China. Ecography, 38, 602–613.
Zanne, A. E., Lopez-Gonzalez, G., Coomes, D. A., Ilic, J., Jansen, S., Lewis,
S. L., … Chave, J. (2009). Data from: Towards a worldwide wood eco-
nomics spectrum. Dryad Digital Repository, hp://dx.doi.org/10.5061/
Zhu, B., Wang, X., Fang, J., Piao, S., Shen, H., Zhao, S., & Peng, C. (2010).
Altudinal changes in carbon storage of temperate forests on Mt
Changbai, Northeast China. Journal of Plant Research, 123, 439–452.
Ziter, C., Benne, E. M., & Gonzalez, A. (2013). Funconal diversity and
management mediate aboveground carbon stocks in small forest frag-
ments. Ecosphere, 4, 1–21.
Zuur, A. F., Ieno, E. N., Walker, N., Saveliev, A. A., & Smith, G. M. (2009).
Mixed eects models and extensions in ecology with R. New York, NY:
Springer New York.
How to cite this arcle: Mensah, S., Veldtman, R., Assogbadjo, A. E.,
Glèlè Kakaï, R. and Seifert, T. (2016), Tree species diversity promotes
aboveground carbon storage through funconal diversity and
funconal dominance. Ecology and Evoluon, 00: 1–12. doi: