EV-1
Coordination of aboveground and belowground responses to  
local-scale soil fertility differences between two contrasting 
Jamaican rain forest types
David A. Wardle, Peter J. Bellingham, Paul Kardol, Reiner Giesler and Edmund V. J. Tanner 
D. A. Wardle (david.wardle@slu.se) and P. Kardol, Dept of Forest Ecology and Management, Swedish Univ. of Agricultural Sciences, SE- 901 
83 Umeå, Sweden. – P. J. Bellingham, Landcare Research, PO Box 69040, Lincoln 7640, New Zealand. – R. Giesler, Climate Impacts 
Research Centre, Dept of Ecology and Environmental Sciences, Umeå Univ., Abisko, Sweden. – E. V. J. Tanner, Dept of Plant Sciences,  
Univ. of Cambridge, Cambridge, CB2 3EA, UK. 
There is growing interest in understanding how declining soil fertility in the prolonged absence of major disturbance drives 
ecological processes, or ‘ecosystem retrogression’. However, there are few well characterized study systems for exploring this 
phenomenon in the tropics, despite tropics occupying over 40% of the Earth’s terrestrial surface. We studied two types 
of montane rain forest in the Blue Mountains of Jamaica that represent distinct stages in ecosystem development, i.e. 
an earlier stage with shallow organic matter and a late stage with deep organic matter (hereafter ‘mull’ and ‘mor’ stages).  
We characterized responses of soil fertility and plant, soil microbial and nematode communities to the transition from mull 
to mor and whether these responses were coupled. For soil abiotic properties, we found this transition led to lower amounts 
of both nitrogen (N) and phosphorus (P) and an enhanced N to P ratio. This led to shorter-statured and less diverse  
forest, and convergence of tree species composition among plots. At the whole community (but not individual species) 
level foliar and litter N and P diminished from mull to mor, while foliar N to P and resorption efficiency of P relative to N 
increased, indicating increasing P relative to N limitation. We also found impairment of soil microbes (but not nematodes) 
and an increasing role of fungi relative to bacteria during the transition. Our results show that retrogression phenomena 
involving increasing nutrient (notably P) limitation can be important drivers in tropical systems, and are likely to involve 
aboveground–belowground feedbacks whereby plants produce litter of diminishing quality, impairing soil microbial  
processes and thus reducing the supply of nutrients from the soil for plant growth. Such feedbacks between plants and  
the soil, mediated by plant litter and organic matter quality, may serve as major though often overlooked drivers of long 
term environmental change.
At local spatial scales (tens to hundreds of meters), vari-
ability in soil fertility, notably the availability and supply 
of nutrients, is a fundamental driver of both aboveground 
and belowground community and ecosystem properties 
(Vitousek 2004). It has frequently been proposed that the 
aboveground and belowground ecosystem components show 
coordinated responses to soil fertility, because fertility governs 
the quality of resources that plants return to the belowground 
subsystem, and regulates the composition and activity of soil 
organisms that govern the supply of nutrients from the soil 
for plant growth (Berendse 1998, Wardle et al. 2004a, van 
der Putten et al. 2013). However, the evidence for coordina-
tion between the two components is mixed. For example, 
some studies have shown that nitrogen (N) addition can 
impede plant litter decomposition when added at levels 
known to enhance plant growth (Fog 1988, Carreiro et al. 
2000), and plant production and litter decomposition can 
show contrasting responses to both N and phosphorus (P) 
fertilization (Hobbie and Vitousek 2000). Further, while 
some studies have found plant and microbial communities to 
show a coupled response to nutrient addition (Suding et al. 
2008), others have found plants and microbes to respond to 
nutrient addition independently of each other (Wardle et al. 
2013) or to respond to nutrient addition in similar ways in 
only some environmental settings (Sundqvist et al. 2014). 
Hence, the view that aboveground and belowground subsys-
tems display simple and coordinated responses to variation 
in nutrient availability (Wardle et al. 2004b) appears to have 
mixed support.
The evaluation of how aboveground and belowground 
components of ecosystems respond to soil fertility in natural 
ecosystems is arguably done most effectively by comparing 
sites that are close to one another and which differ in nutrient 
availability but which have the same climate, parent mate-
rial and topography. Such comparisons are scarce because 
of problems with potentially confounding factors. Some of 
the most effective comparisons have involved the use of 
retrogressive chronosequences (Walker and Syers 1976, 
Vitousek 2004, Wardle et al. 2004b, Peltzer et al. 2010). 
These chronosequences consist of sites that differ in their 
© 2014 The Authors. Oikos © 2014 Nordic Society Oikos
Subject Editor: Wim van der Putten. Accepted 3 June 2014
Oikos 000: 001–013, 2014 
doi: 10.1111/oik.01584
EV-2
stages of ecosystem development, and for which limitation 
by nutrients (and notably the limitation of P relative to 
N) increases over the time frame of several centuries to 
millennia (Peltzer et al. 2010). Only a handful of retro-
gressive chronosequences have been well characterized, 
and these include examples in temperate (Richardson et al. 
2004), Mediterranean-type (Laliberté et al. 2012), boreal 
(Wardle et al. 2012) and montane tropical (Vitousek 2004) 
ecosystems. Few studies have explicitly considered both above-
ground and belowground community responses to changes in 
soil fertility during retrogression; some have found coordi-
nated responses (Williamson et al. 2005, Doblas-Miranda 
et al. 2008) while others have not (Wardle et al. 2012). 
Further, despite the tropics occupying 40% of the Earth’s 
land surface, these studies have been performed mostly in 
boreal and temperate ecosystems, with almost nothing 
known about whether similar responses occur for tropical 
systems outside of Hawaii (Peltzer et al. 2010).
In the present study we considered replicated forest plots 
that have a comparable climate, topography and parent 
material in a montane tropical rainforested area in the Blue 
Mountains of Jamaica (Grubb and Tanner 1976, Tanner 
1977). This area consists of several patches of each of two 
types of forest on ridge tops that are on ‘mull’ and ‘mor’ soil 
(sensu Muller 1884), i.e. ‘mull’ and ‘mor’ forest (sensu Tanner 
1977), often adjacent or within 100 m of each other. Here, 
the mor soil consists of a thick uppermost layer of humus 
(usually  20 cm deep) while mull soil consists of organic 
material mixed into the soil and without a thick humus layer 
(Tanner 1977). The vegetation composition differs greatly 
between the two forest types, and the mor forest is shorter 
statured and with a much deeper and more acidic humus 
layer than is the mull forest (Tanner 1977, 1980). Soils in 
the mor forest also contain lower levels of available nutrients 
(on a per area basis) and support slower rates of those soil 
processes that enhance N availability when compared with 
the mull forest (Tanner 1977, Brearley 2013); further, litter 
bag studies show that mass loss from decomposition of tree 
litters placed in mor forest plots occurs more slowly (Tanner 
1981). The mull forest represents an earlier stage of ecosys-
tem development than the mor forest, and when mor forest 
vegetation and humus is removed (Sugden et al. 1985) or 
when fresh mineral soil in mor forest is exposed by uproot-
ing of canopy trees following hurricane damage (Bellingham 
et al. 1995), the newly exposed surface is colonized by tree 
species that are characteristic of the mull forest but absent 
in the mor forest. As such, the mor forest patches represent 
the stage that the mull forest patches are likely to reach if 
soil profiles were not disturbed, and ecosystem retrogression 
was therefore able to proceed (Grubb and Tanner 1976). The 
patchiness of the forest may also be reinforced by chance 
colonization of ridge-crest sites by tree species that can form 
mor humus and alter their local environment (through 
production of acidic, nutrient-poor litter), in turn favoring 
mor-forming species and disfavoring many species that domi-
nate on mull soils (Grubb and Tanner 1976, Tanner 1977).
We explored differences between six mull and six mor 
plots for several aboveground and belowground properties, 
to test the following three hypotheses which are based on 
our understanding of ecosystem retrogression (Peltzer et al. 
2010). 1) Soil abiotic properties: soils in the mull plots will 
have higher amounts of the most biologically available forms 
of N and P, and a greater ratio of N to P, relative to the 
mor plots. This is based on the expectation that retrogres-
sion is characterized by decreasing plant-available nutrients 
and increasing limitation of P relative to N (Wardle et al. 
2004b, Peltzer et al. 2010). 2) Aboveground properties: in 
addition to having a shorter stature, trees in mor plots will 
produce leaves and litter with lower N and P concentrations 
and greater N to P ratios than those in mull plots. Further, 
resorption of both foliar N and P prior to leaf senescence will 
be greater in mor plots, and P will be resorbed more than 
N in these plots. These differences are expected both at the 
individual species and whole plot levels, and are expected 
to reflect the differences in soil fertility between the mull 
and mor plots. 3) Soil biological properties: soils in the mor 
plots will support lower densities of soil primary decompos-
ers (bacteria and fungi) and nematodes that feed on bacteria, 
fungi and each other, relative to the mull plots. Further, we 
expect that the ratio of fungi to bacteria and fungal-feeding to 
bacterial-feeding nematodes to be greatest in the mor plots, 
given that infertile ecosystems promote the fungal-based 
energy channel that promotes more conservative nutrient 
cycling (Coleman et al. 1983, Wardle et al. 2004a). Answer-
ing these questions in combination will inform on how local 
scale variability in soil fertility in tropical forests affects key 
aspects of both the aboveground and belowground compo-
nents of ecosystems, the degree to which these components 
show coordinated responses to soil fertility and may feed 
back to one another, and ultimately the mechanistic basis 
by which forested ecosystems may respond to environmental 
change over time scales of centuries to millenia.
Methods
Study system, plots and soil sampling
The study was conducted on ridge forest in the western Blue 
Mountains of Jamaica (18°05′N, 79°39′W), between Belle 
Vue Peak (1822 m) and John Crow Peak (1762 m). The 
annual rainfall averages 2500–3000 mm year1 (Kapos and 
Tanner 1985) and the mean monthly maximum tempera-
tures range from 18.5 to 20.5°C, and minimum temperatures 
from 11 to 12°C (Tanner 1980). The region is in the Atlantic 
hurricane belt, and hurricane eyes pass over or within 15 
km of the Blue Mountains every 25 years (Tanner and 
Bellingham 2006). The ridge includes several patches of both 
‘mull’ and ‘mor’ forest. Soils in mull forest are character-
ized by having a thin O-horizon ( 2 cm thick) and a well- 
developed A-horizon with a high organic matter content 
mixed into it (i.e. with  30% soil organic matter) (Plice 
1946, Grubb and Tanner 1976). Soils in mor forest are 
characterized by a well-developed O-horizon dominated by 
organic material. The underlying mineral soil does show fea-
tures of an A-horizon, but visible signs of E-horizon forma-
tion can be seen in the surface mineral soil (FAO/UNESCO 
1988). Previous measurements in this study system reveal 
that for the mor forest soil carbon (C) concentration is 47% 
in the O-horizon (0–30 cm depth) and between 18% for 30–
45 cm depth and 5% for 45–60 cm (Tanner 1977). Mean-
while C concentration for the mull forest soil is 29% for 0–8 
EV-3
cm depth, 18% for 8–11 cm and 1% for 11–68 cm (Tanner 
1977). Mineral soils are derived from granodiorite, andesites 
and sedimentary rock, and are loams (Hafkenscheid 2000). 
The forests, through being on a ridge, are free draining and 
very rarely and only locally become waterlogged (Grubb and 
Tanner 1976, Tanner 1977).
For the present study, we selected six patches of mull 
forest and six of mor forest along the ridge at altitudes between 
ca 1500–1625 m, based on known vegetation composition 
of the two forest types (Tanner 1977). Within each patch 
we set up a 10  10 m plot between 16 and 23 September 
2011, and all measurements and sampling were performed 
at that time. One each of these mull and mor plots are the 
same as those used for previous work in this system (Tanner 
1977, 1980, 1981). All plots were at least 60 m from the 
next nearest plot, and the total area containing the plots was 
ca 1600  100 m (16 ha). The parent material underlying 
the plots is mainly derived from metamorphosed Cretaceous 
sediments and granodiorite. Parent material and topography 
are comparable between the two forest types.
Within each plot one composite soil sample (i.e. humus 
and/or mineral soil) was collected to 10 cm depth at the 
time of plot measurement. This was done by selecting four 
points (one in the center of each quadrant of the plot) and 
for each point using a trowel to collect soil to that depth; 
all soil within the plot was bulked. The depth of the humus 
layer was also measured at each point using a ruler. All soil 
samples were packaged at field moisture levels in double 
plastic bags, tightly positioned together in a single insulated 
box with minimal extra space, and sent by courier to Sweden 
where they were kept at 4°C before analyses were performed. 
There was a two week delay between collection in Jamaica 
and analysis in Sweden for reasons beyond our control, and 
as is characteristic for studies that have collected soils from 
remote locations and shipped them to laboratories elsewhere, 
we assume that any temperature changes in the samples 
during transportation would have been equivalent for all 
samples and would not have introduced treatment bias.
Abiotic soil properties
A subsample of each soil sample was sieved to 4 mm for 
measurement of abiotic properties. These included pH in 
a KCl extract, loss on ignition as a measure of soil organic 
matter (SOM) using a muffle furnace (360°C, 24 h), con-
centrations of total C, nitrogen (N) and stable N isotopes 
determined by combustion (with an elemental analyser 
connected with a mass spectrometer), phosphorus (P) deter-
mined by colorimetry on an auto-analyzer III after Kjeldahl 
digestion, and ammonium, nitrate and phosphate (following 
extraction with 1 M KCl) by colorimetry as above. Here, 
natural abundances of 15N were expressed in per mil (‰) 
deviation from the international standard, i.e. atmospheric 
N. Ratios of C to N, C to P and N to P were calculated from 
the total C, N and P values. Lower natural abundance 
values of 15N are typically associated with more conserva-
tive nutrient cycling and greater nutrient limitation (Menge 
et al. 2011, Brearley 2013). To characterize soil P com-
position, we performed a five-step sequential extraction 
for each soil sample as described by Hedley et al. (1982) 
and modified by Binkley et al. (2000) and Giesler et al. 
(2004); this provides a direct estimate of the five different 
operationally-defined P pools of contrasting lability. These 
include membrane-extractable inorganic P (Pi), NaHCO3-
extractable Pi and organic P (Po), and NaOH-extractable 
Pi and Po and HCl-extractable P. Membrane-extractable Pi 
and NaHCO3-extractable Pi and Po are considered to be 
more labile than NaOH and HCl extractable P (Giesler 
et al. 2004).
All concentrations of total soil C, N and P, and for all 
forms of both N and P, were expressed on both a per unit 
soil weight basis and a per area basis to 10 cm soil depth. To 
determine the latter requires a measure of soil bulk density; 
we calculated bulk density (g cm3) for each soil sample from 
the following equation based on the total C concentration of 
soils previously collected from this study system (Tanner and 
Bellingham, 2006), i.e.
bulk density  0.998  EXP(–0.054  soil C); R 2  0.939.
Plant variables
For each plot, the diameter at breast height (DBH) of each 
stem with a DBH greater than 5 cm was measured and 
identified to species. The diameter of each stem was then 
converted to aboveground biomass by using allometric 
relationships developed in this study system by Tanner 
(1980). This biomass data was used to determine stand-
ing biomass of each woody species in the plot, as well as 
total biomass and the Shannon–Wiener diversity index. 
The total species richness of the measured stems was also 
determined.
For each of the dominant and subordinate but abundant 
woody plant species in each plot, foliar and litter samples were 
collected; a mean of 8.2 species were sampled per plot which 
on average comprised a total of 89.4% of the total woody 
plant biomass. For each species at least 25 fully expanded live 
leaves were collected from the canopy using a long handled 
pruner; we sampled only the most recently produced fully 
expanded leaves on each branchlet. At least 25 freshly fallen 
dead leaves were collected for the same species from the litter 
layer. For each species, the mean oven-dry weight per live 
leaf and per dead leaf was determined (for use in nutrient 
resorption calculations) and the total N and P concentra-
tions measured by micro-Kjeldahl analysis; leaf and litter N 
to P ratios were determined from these values. For each spe-
cies in each plot, we also determined resorption efficiency of 
both N and P as the percentage resorption during leaf senes-
cence (Killingbeck 1996). Because substantial leaf mass loss 
can occur during senescence which needs to be corrected for 
when values of resorption efficiency are calculated, we used 
the measurements of mean live leaf mass and mean dead 
leaf mass for each sample to provide this correction (van 
Heerwaarden et al. 2003). As such, percent resorption 
efficiency for both N and P was calculated as:
percent efficiency  100  ((MLM  CL) – (MDM  CD))/ 
               (MLM  CL)
where MLM is mean live leaf mass, MDM is mean dead leaf 
mass, CL is the concentration of N or P in live leaves, and 
CD is the concentration of N or P in dead leaves.
EV-4
were heat-killed, and fixed using 35% formaldehyde diluted 
to 4%. Subsequently approximately 200 nematodes per 
sample were identified to family level, and placed into five 
functional feeding groups following Yeates et al. (1993), i.e. 
bacterial feeders, fungal feeders, plant feeders, omnivores 
and predators. For each soil sample nematode densities were 
expressed per unit soil weight, per unit SOM, and per unit 
area to 10 cm depth as described above. For each sample 
we also used this data to determine the richness of nema-
tode families, nematode diversity (Shannon–Wiener index), 
and the ratio of fungal feeding nematodes to fungal plus 
bacterial-feeding nematodes.
Data analysis
All aboveground and belowground response variables were 
analysed using one-way ANOVA (using STATISTIX ver. 
10.0) with each plot as a replicate to test for differences 
between mull and mor forests; data were transformed 
if needed to conform to ANOVA assumptions. Further, 
differences between mor and mull plots for the commu-
nity composition for trees (species biomass data), PLFAs 
and nematode families were each analyzed using principal 
component analysis (PCA), using CANOCO ver. 4.5 (ter 
Braak and Šmilauer 2002). To determine whether com-
munity composition for each of these three groups was sig-
nificantly different between mor and mull plots, we used 
redundancy analysis (RDA) with Monte Carlo permutation 
tests (999 unrestricted permutations), using CANOCO ver. 
4.5. All analyses for PCA and RDA analyses were run using 
proportional biomass or abundance data. As a measure of 
b-diversity for each of the tree, PLFA, and nematode com-
munity data sets, we calculated the inverse of the percentage 
similarity index:
dissimilarity  min( )x yi i
i
N
∑
where xi and yi is the proportional abundance density 
(Ni / N, N  (Ni)) of the i-th taxon in the two communities 
being compared (Lepš et al. 2001). We calculated ‘overall’ 
b-diversity for mull plots as the mean percentage dissimilar-
ity of each mull plot with all other mull plots, and for mor 
plots as the mean percentage dissimilarity of each mor plot 
with all other mull plots.
Results
Soil abiotic variables
Humus depth was shallower in the mull than mor plots 
(mean  SE: mull: 1.7  0.5 cm; mor: 21.3  2.9 cm; 
F1,10  43.8, p  0.001). The pH in the top 10 cm of soil 
was greatest in the mull plots (mull: 4.46  0.10; mor: 
3.58  0.06; F1,10  58.3, p  0.001), as was soil bulk den-
sity (mull: 0.463  0.032 g cm3; mor: 0.064  0.001 g 
cm3; F1,10  148.6, p  0.001). Levels of SOM were signifi-
cantly greater in the mor than mull plots on a concentration 
basis (mull: 35.6  3.0%; mor: 94.5  0.9%; F1,10  344.9, 
p  0.001) and significantly greater for the mull than 
mor plots on a per unit area basis to 10 cm depth (mull: 
For each plant nutrient measure in each plot (i.e. foliar 
and litter N, P and N to P ratio, and N and P resorption) we 
calculated a plot-level abundance weighted measure by using 
the following equation according to Garnier et al. (2007):
Plot-level measure  
i
n
=
∑
1
 pi  nutrienti
where pi is the biomass of species i as a proportion of the 
total biomass for all species collected in that plot, and nutri-
enti is the value of the plant nutrient measure for species i. 
We also repeated these calculations but without abundance-
weighting (so that all species were weighted equally) to 
provide a plot-level unweighted measure for each variable. 
Comparison of weighted and non-weighted plot-level nutri-
ent measures enables assessment of the role of dominant 
versus subordinate species in determining these measures 
(Mason et al. 2012, Kichenin et al. 2013).
The height of the forest canopy was measured at four 
positions within each plot, by measuring the length of the 
pruner pole needed to sample canopy foliage.
Soil biological variables
A 4 mm sieved subsample of the soil sample collected 
from each plot was used for determination of soil micro-
bial variables. We measured substrate-induced respiration 
(SIR; a relative measure of active microbial biomass) as 
described by Anderson and Domsch (1978) and modified 
by Wardle (1993). Briefly, a 3 g (dry weight) subsample of 
soil was amended to 150% moisture content (dry weight 
basis), amended with 30 mg glucose, placed in a 130 ml 
airtight vessel, and incubated at 22°C. Evolution of CO2 
between 1 h and 4 h was then determined by injecting 1 ml 
subsamples of headspace gas into a gas analyzer, and used 
as a measure of SIR. We also measured the composition of 
microbial phospholipid fatty acids (PLFAs) for assessing the 
microbial community by using the method of Bligh and 
Dyer (1959) as modified by White et al. (1979), and used by 
Sundqvist et al. (2014); different PLFAs represent different 
subsets of the soil microbial community. For each soil, the 
abundance of each fatty acid extracted was expressed as rela-
tive nmoles per g of dry soil using standard nomenclature. 
The PLFAs i15:0, a15:0, i16:0, 16:1w9, 16:1w7t, 16:1w7c, 
i17:0, a17:0, cy17:0, 18:1w7 and cy19:0 were used to indi-
cate relative bacterial biomass, the PLFA 18:2w6 was used to 
indicate relative fungal biomass and the PLFAs 10Me17:0 
and 10Me18:0 were used to indicate relative actinomycete 
biomass. The PLFAs used for calculation of total branched 
PLFAs (i.e. gram positive bacteria) were i15:0, a15:0, i16:0 
and i17:0, while those used to determine total cyclic PLFAs 
were cyc17:0 and cyc19:0. For each soil sample we expressed 
SIR and PLFA concentrations per unit soil weight, per unit 
SOM, and per unit area to 10 cm depth using the soil bulk 
density measures described above. Further, for each sample 
we calculated the Shannon–Wiener diversity index to repre-
sent the diversity of all PLFAs, and the ratio of fungal PLFAs 
to the sum of fungal plus bacterial PLFAs to represent the 
fungal to bacterial ratio.
An unsieved 100–150 g (wet weight) subsample of each 
soil sample was used to characterize the soil nematode com-
munity, by using a sugar flotation method. The nematodes 
EV-5
per unit soil weight and per unit area. Amongst the Hedley 
fractions, concentrations of NaHCO3-extractable Pi and Po 
were significantly higher in the mor than mull plots whereas 
NaOH-extractable Po was higher in the mull than mor plots, 
when expressed per unit soil weight (Table 1). Meanwhile, 
all Hedley fractions except membrane-extractable Pi were 
considerably higher in the mull than in the mor plots on a 
per unit soil weight basis (Table 1).
Plant variables
Total tree aboveground biomass did not differ significantly 
between the mull and mor plots, but canopy height and tree 
species richness and diversity (Shannon–Wiener index) were 
all significantly less in the mor plots (Table 2). Further, of 
the 12 tree species that occurred most frequently in the mull 
16052  447 g m2; mor: 6105  88 g m2; F1,10  475.6, 
p  0.001).
The concentration of total C and N was greatest in 
mor plot soil, while that for P was greatest in mull plot soil 
(Fig. 1). On a per area basis to 10 cm depth, total C, N and 
P were all substantially greater in the mull than mor plots 
(Fig. 1). Meanwhile, the ratios of total C to N, C to P and 
N to P were all considerably greater in the mor than mull 
plots (Fig. 1). The d15N values of soil were greatest in the 
mull plots (mull: 2.38  0.41‰; mor: –1.73  0.24‰; 
F1,10  74.6, p  0.001).
Mull plots had substantially more soil nitrate both on a 
per soil weight and per unit area basis than mor plots, while 
mor plots had substantially more ammonium per soil weight 
(Table 1). Soil phosphate was greater (but marginally not 
significantly so at p  0.05) in the mor than mull plots both 
Figure 1. Soil carbon (C), nitrogen (N) and phosphorus (P) expressed both as a concentration and per unit area, and ratios of soil C to N, 
C to P and N to P. Values are means and standard errors. Note that the vertical axis scale varies among different variables. F-values are from 
one-way ANOVAs comparing mull and mor plots following ln-transformation of data; degrees of freedom are 1,10. *, **, ***indicate that 
the means for mull and mor plots differ at 0.05, 0.01 and 0.001, respectively.
EV-6
Table 1. Chemical forms of soil nitrogen and phosphorus expressed both per unit soil weight and per unit area to 10 cm depth in mull and 
mor plots. Values are means  standard errors (n  6). For the Hedley fractionation data, HCl–extractable Pi was below detection limit for all 
samples and is thus not shown. Bold indicates statistically significant differences between mull and mor at p  0.05.
Per unit soil weight Per unit area
Measure Mull mor F- and p- valuea mull mor F- and p- valuea
(mg g1 soil) (mg m2)
Ammonium 1.69  1.38 7.38  1.09  10.5 (0.009) 82.6  67.4 47.6  7.0 0.3 (0.617)
Nitrate 20.19  2.75 0.03  0.03  53.8 ( 0.001) 900.7  87.5 0.2  0.2 105.8 ( 0.001)
Phosphate 0.007  0.004 0.485  0.223 4.7 (0.056) 0.311  0.211 3.097  1.406 3.8 (0.079)
Hedley fractionation datab (mg g1 soil) (g m2)
Membrane-extractable Pi 0.7  0.3 31.2  15.2 4.0 (0.072) 0.03  0.01 0.23  0.11 3.1 (0.104)
NaHCO3-extractable Pi 19.8  3.3 39.0  2.5 21.2 (0.001) 0.91  0.16 0.29  0.02 15.4 (0.003)
NaHCO3-extractable Po 33.3  8.6 56.5  3.9 6.0 (0.034) 1.47  0.30 0.42  0.03 11.8 (0.006)
NaOH-extractable Pi 117.0  2.7 105.5  2.1 0.7 (0.432) 5.32  0.55 0.79  0.05 66.7 ( 0.001)
NaOH-extractable Po 582.1  68.3 380.5  27.5 7.5 (0.021) 26.3  2.97 2.89  0.26 62.1 ( 0.001)
 aF-values are from one way ANOVA with 1,10 degrees of freedom
bPi  inorganic phosphorus, Po  organic phosphorus
Table 2. Data on tree biomass, canopy height, species richness 
and diversity (Shannon–Wiener index), and biomass of species 
occurring most frequently in mull and/or mor plots. Values are 
means  standard errors (n  6). Bold indicates statistically signifi-
cant differences between mull and mor at p  0.05.
Response variable Mull plots Mor plots
Test statistic 
and p-valuea
Total biomass (kg 
100 m2)
2420  295 1847  153 3.0 (0.116)
Canopy height (m) 13.0  0.8 8.0  0.5 23.6 ( 0.001)
Species richness 
per 100 m2
18.0  1.3 10.0  0.6 30.0 ( 0.001)
Shannon–Wiener 
index
2.11  0.09 1.75  0.08 8.6 (0.015)
Species biomass 
(kg 100 m2)
Alchornea 
latifolia
315  141 333  90 0.1 (0.715)
Chaetocarpus 
globosus
0  0 189  172 10.8 ( 0.001)
Clethra 
occidentalis
425  125 121  28 4.3 (0.034)
Clusia have-
tioides
0  0 46  37 8.9 ( 0.001)
Cyrilla 
racemiflora
32  79 347  13 9.6 ( 0.001)
Eugenia 
virgultosa
100  40 1  1 7.1 ( 0.001)
Lyonia octandra 0  0 612  103 10.8 ( 0.001)
Pittosporum 
undulatum
46  25 13  6 1.0 (0.318)
Podocarpus 
urbanii
374  212 3  2 3.3 (0.074)
Sideroxylon 
montanum
284  231 0  0 8.9 ( 0.001)
Solanum 
punctulatum
213  128 0  0 8.9 ( 0.001)
Vaccinium 
meridionale
45  63 101  84 2.4 (0.130)
 aKruskall–Wallis test statistic for species biomasses, F1,10 from 
ANOVA for all other variables
and/or mor plots, four had significantly higher biomass in the 
mull plots, and four had significantly higher biomass in the 
mor plots (Table 2). The PCA revealed large compositional 
differences between mull and mor plots (Fig. 2a), and the 
redundancy analysis revealed these differences to be highly 
statistically significant (F-ratio  10.1, p  0.002, propor-
tion of total variation explained  50.3%). Values for b-di-
versity were greater in the mull than mor plots (0.61  0.02 
and 0.31  0.01, respectively).
Community level measures of foliar and litter N and P 
were always greater in the mull than mor plots regardless 
of whether or not they were abundance-weighted, and 
these differences were significant at p  0.05 with only 
one exception (Fig. 3). Community-level foliar N to P 
ratios were greater in the mull than mor plots but this 
was significant at p  0.05 only when measures were not 
abundance-weighted. In contrast, community level lit-
ter N to P ratios were greater in the mor than mull plots 
but this was significant at p  0.05 only when measures 
were abundance-weighted. Community measures of N 
resorption did not differ significantly between mull and 
mor plots but community measures of P resorption were 
greater in the mor plots; these were significant at p  0.05 
when measures were abundance-weighted but marginally 
non-significant when unweighted (Fig. 3).
At the individual species level, foliar N and P both varied 
at least two-fold among species in both the mull and mor 
plots (Supplementary material Appendix 1 Table A1). Of 
the four most widespread species, foliar N was significantly 
greater in the mull plots at p  0.05 for Clethra occidentalis 
and marginally non-significantly greater for Alchornea lati-
folia. Foliar P did not differ significantly between mull and 
mor plots for any species. All four species had a higher foliar 
N to P ratio in the mull plots, and significantly so at p  0.05 
for three of them. Litter N varied at least three-fold and P 
varied at least seven-fold among species in both the mull and 
mor plots (Supplementary material Appendix 1 Table A2). 
Litter N, P and N to P ratios for the four most widespread 
species never differed significantly between mull and mor at 
p  0.05, but there were marginally non-significant greater 
values in mull plots for litter N for two species and for litter 
N to P for one species. Resorption of N and P varied sub-
stantially among species in both mull and mor plots, with 
values for N resorption ranging from below zero to over 
70%, and those for P resorption ranging from below zero to 
over 85% (Supplementary material Appendix 1 Table A3). 
Resorption of N did not differ between mull and mor plots 
for any of the four most widespread species, but P resorp-
tion was significantly greater at p  0.05 in the mor plots for 
EV-7
Figure 2. Species–sample bi-plots resulting from principal compo-
nent analysis of the community composition of tree species (A), soil 
microbes (phospholipid fatty acids or PLFAs) (B) and soil nema-
tode families (C). Percentages along the axes correspond to 
the amount of explained variation in community composition. 
Each point indicates one mor or mull plot. For clarity, only the best 
fitting tree and nematode taxa are plotted. Abbreviations for tree 
species: Cha glo  Chaetocarpus globosus, Cin mon  Cinnamomum 
montanum, Cle occ  Clethra occidentalis, Clu hav  Clusia 
havetioides, Cyr rac  Cyrilla racemiflora, Den pen  Dendropanax 
pendulus, Eug mar  Eugenia marshiana, Gua gla  Guarea glabra, 
Hed arb  Hedyosmum arborescens, Ile obc  Ilex obcordata, Lyo 
oct  Lyonia octandra, May jam  Maytenus jamaicensis, Oco 
pat  Ocotea patens, Pod urb  Podocarpus urbanii, Psy slo  Psycho-
tria sloanei, Sch sci  Schefflera sciodaphyllum, Sid mon  Sideroxylon 
montanum, Sol pun  Solanum punctulatum, Tur occ  Turpinia 
occidentalis, Urb cri  Urbananthus critoniformis.
A. latifolia and marginally non-significantly greater for 
Pittosporum undulatum.
Soil biotic variables
Measures of SIR (i.e. active microbial biomass) were signifi-
cantly greater at p  0.05 in mor than mull soil on a per 
unit soil weight basis and greater in mull than mor soil 
on a per unit area basis; SIR was also marginally non-
significantly greater in mull than mor soil on a per unit 
weight SOM basis (Fig. 4). For PLFA data expressed on a 
per unit soil weight basis, fungal and branched bacterial 
PLFAs were significantly greater in mor than mull soil and 
no other PLFA group differed between the two soil types. 
All PLFA groups significantly differed between the mull 
and mor soil when expressed either on a per unit SOM 
basis or per unit area (to 10 cm depth) basis; in all but one 
case (i.e. fungal PLFAs per unit SOM) the highest values 
occurred for the mull plots (Fig. 4). The ratio of fungal to 
bacterial PLFAs and diversity of PLFAs (Shannon–Wiener 
index) were both greatest for the mor soils (Table 3). The 
PCA revealed large compositional differences in PLFA 
composition between mull and mor plots (Fig. 2b), and 
the redundancy analysis revealed these differences to be 
highly statistically significant (F-ratio  41.3, p  0.002, 
proportion of total variation explained  80.5%). Values 
for b-diversity were higher in the mull than mor plots 
(0.09  0.01 and 0.05  0.01 respectively).
For all nematode feeding groups, numbers per unit soil 
weight were greater for the mor than mull plots, and these 
differences were significant at p  0.05 for all groups 
except the predators (Fig. 5). The same trend occurred 
when nematodes were expressed per unit OM, but here 
differences between mull and mor plots were significant 
only for bacterial-feeders, with differences for fungal-
feeders and omnivores being marginally non-significant. 
There were no significant differences between mull and 
mor plots for any feeding group when densities were 
expressed per unit area, although densities of predators 
were marginally non-significantly greater for mull soils 
(Fig. 5). The ratio of fungal-feeding to bacterial-feeding 
nematodes and nematode diversity (Shannon–Wiener 
index) did not differ between mull and mor plots, but mor 
soils supported a greater richness of nematode families 
than did mull soils (Table 3). The PCA did not reveal any 
compositional differences in nematodes between mull and 
mor plots (Fig. 2c), and this was supported by the redun-
dancy analysis (F-ratio  0.3, p  0.630, proportion of 
total variation explained  2.5%). Values for b-diversity 
were lower in the mor than mull plots (0.31  0.03 and 
0.38  0.04 respectively).
EV-8
Figure 3. Abundance-weighted and unweighted whole plot level measures of foliar and litter nitrogen (N), phosphorus (P) and N to P 
ratios, and N and P resorption during leaf senescence. Values are means and standard errors. F-values are from one-way ANOVAs compar-
ing mull and mor plots; degrees of freedom are 1,10. †, *, **, ***indicate that the means for mull and mor plots differ at 0.10, 0.05, 0.01 
and 0.001, respectively.
EV-9
Figure 4. Measures of soil microbes measured as phospholipid fatty acids (PLFAs) and substrate-induced respiration (SIR), expressed per 
unit soil weight, per unit soil organic matter (SOM), and per unit area to 10 cm depth. AMF  arbuscular mycorrhizal fungi; ACTIN  actin-
omycetes. Values are means and standard errors. Note that the vertical axis scale varies among different variables. F-values are from one-way 
ANOVAs comparing mull and mor plots; degrees of freedom are 1,10. †, *, **, ***indicate that the means for mull and mor plots differ at 
0.10, 0.05, 0.01 and 0.001, respectively.
Table 3. Community–level variables for the soil microflora using phospholipid fatty acid (PLFA) data, and nematode fauna, in mull and mor 
plots. Values are means  standard errors (n  6). Bold indicates statistically significant differences between mull and mor at p  0.05.
Response variable Mull plots Mor plots F- and p-valuea
Ratio of fungal to fungal bacterial PLFAs 0.054  0.002 0.169  0.019 37.2 ( 0.001)
Ratio of FF to FF BF nematodesb 0.086  0.031 0.050  0.013 1.2 (0.299)
Diversity (Shannon–Wiener index) for PLFAs 2.28  0.01 2.36  0.01 24.1 ( 0.001)
Diversity (Shannon–Wiener index) for nematode families 1.20  0.23 1.36  0.37 0.1 (0.728)
Richness of families for nematodes 6.00  0.63 8.66  0.76 7.2 (0.024)
 aDerived from one-way ANOVA; degrees of freedom are 1,10
bFF  fungal-feeders; BF  bacterial-feeders
Discussion
We found that the transition from mull to mor stands 
involves changes both belowground and aboveground that 
are characteristic of retrogression or soil aging. These include 
diminished soil fertility, reduced forest stature and diversity, 
reduced nutrients in foliage and litter at the whole commu-
nity (though usually not the individual species) level, and 
lower densities of microbes (though not nematodes) per unit 
soil C and per unit area. We now discuss these results and 
their significance for understanding how local scale variabil-
ity of soil fertility in tropical forests affects key aspects of 
both the aboveground and belowground subsystems.
We found several changes in soil abiotic properties 
from mull to mor that are mostly consistent with changes 
expected during retrogression. These are in line with mineral 
nutrient data reported from this system by Tanner (1977), 
and with our first hypothesis. First, we showed that mor 
soils have a higher N to P ratio than mull soils, in line with 
what occurs during retrogression in other systems (Wardle 
et al. 2004b, Peltzer et al. 2010). This is likely because N 
(which can be fixed biologically) can be replenished while 
P (which is derived from parent material) cannot (Vitousek 
2004). In our study system, N2-fixing trees are very scarce, 
but it is known for other montane tropical forests that free 
living N2-fixing bacteria, such as those associated with dead 
leaves, can fix appreciable amounts of N (Crews et al. 2000). 
Second, we showed a marked increase in SOM depth, 
which is likely to stem from less decomposer activity (Fig. 4) 
and lower quality litter entering the decomposer subsystem 
(Fig. 3). This is indicative of a shift of C storage from vegeta-
tion to longer term organic matter pools in the soil (Peltzer 
et al. 2010). Third, we found that concentrations of nitrate 
and total P diminished from mull to mor. Conversely, the 
concentrations of total N, ammonium, phosphate and of 
two of the three most labile forms of P from the Hedley 
EV-10
Figure 5. Densities of major functional groupings of soil nematodes, expressed per unit soil weight, per unit soil organic matter (SOM), 
and per unit area to 10 cm depth. AMF  arbuscular mycorrhizal fungi; ACTIN  actinomycetes. Values are means and standard errors. 
Note that the vertical axis scale varies among different variables. F-values are from one-way ANOVAs comparing mull and mor plots fol-
lowing ln-transformation of data; degrees of freedom are 1,10. †, *, **, ***indicate that the means for mull and mor plots differ at 0.10, 
0.05, 0.01 and 0.001, respectively.
fractionation increased, but this was mostly due to the greater 
amount of organic matter in the mor soil, and when data are 
expressed per unit area to adjust for this, most of these mea-
sures decreased. Further d15N values were lower in the mor 
plots, in line with what has been shown for both this system 
(Brearley 2013) and other retrogressive sequences (Selmants 
and Hart 2008, Menge et al. 2011), and is consistent with 
more closed N cycling through diminished N availability 
in the mor sites. In total, our results provide evidence that 
as mull plots transition to mor plots over time, there are 
distinct decreases in N and P availability in the top 10 cm of 
the soil layer (where most plant rooting occurs) and reduced 
total P relative to total N.
The changes that occurred in abiotic soil properties 
from the mull to mor transition had important effects on 
forest stand characteristics. Although mor plots did not have 
significantly less biomass than mull plots they were signifi-
cantly shorter, as expected for forests undergoing retrogres-
sion (Crews et al. 1995, Richardson et al. 2004). Further, 
there were large compositional differences, with both less 
a-diversity and b-diversity in mull than in mor plots. This 
arises because the mull plots have a much larger potential 
pool of tree species (Tanner 1977), with different subsets 
of this pool establishing on different plots, but with these 
plots supporting fewer species from a smaller species pool as 
they transition to mor plots over time. The fewer species that 
can establish on mor plots leads to convergence of species 
composition and thus less variation of composition among 
mor than among mull plots (Walker and del Moral 2003, 
Walker et al. 2010). Our findings are consistent with work 
from some other chronosequences that show a smaller poten-
tial species pool and thus reduced tree diversity as retrogres-
sion proceeds (Wardle et al. 2008), despite total a-diversity 
(including understory species) frequently increasing (Laliberté 
et al. 2013).
We found our hypotheses regarding changes in foliar 
and litter N and P characteristics to be more consistently 
supported at the whole community level than at the indi-
vidual species level. At the community level, we found both 
foliar and litter N and P to be less on the mor than the mull 
plots, in line with the reduced soil fertility in these plots. 
We also found two lines of evidence for greater limitation by 
P than by N in the mor plots. First, the litter N to P ratio 
was greater in the mor plots though only for abundance-
weighted measures (and with a weak but significant trend in 
the opposing direction for the foliar N to P ratio when abun-
dance weighting was not performed). Second, community-
level foliar resorption values for P but not for N were greater 
in the mor than mull plots. These findings are broadly in 
line with other studies showing community-level measures 
to reflect increasing P relative to N limitation as retrogres-
sion proceeds (Richardson et al. 2005, Wardle et al. 2009, 
Hayes et al. 2014). Further, with the exception of the lit-
ter N to P ratio, we found comparable differences in N and 
P characteristics between the mull and mor plots regardless 
of whether or not the community-level measure was abun-
dance weighted, meaning that dominant and minor species 
contribute similarly to these differences (Mason et al. 2012, 
EV-11
indicative of poorer resource quality in the mor plots, which 
may arise both from diminishing soil fertility exerting direct 
effects and indirect effects via poorer quality of plant litter 
entering the soil. This will in turn lead to reduced micro-
bial activity, reduced mineralization of nutrients for plant 
growth, and impaired decomposition leading to build up of 
soil organic matter.
We note that while there has been a growing interest in 
identifying and interpreting retrogressive chronosequences 
from around the world (Peltzer et al. 2010), with one excep-
tion all described examples are from the temperate or boreal 
zones. Our work, alongside that from Hawaii (Vitousek 
2004), suggests that the type of retrogressive phenomena 
identified elsewhere also occurs in tropical systems at least in 
areas relatively unimpacted by humans. This is consistent with 
suggestions that controls of N relative to P availability, and 
the ecological consequences of this, are governed by similar 
biogeochemical processes in vastly differing climatic regions, 
even if the time scale involved may vary (Wardle et al. 2004b, 
Peltzer et al. 2010). While we are unable to identify the pre-
cise time frame over which this retrogression occurs (attempts 
to use 14C dating on soil organic matter were unsuccessful, 
presumably because of soil bioturbation), the size and struc-
ture of the trees suggest that this process occurs at least over 
the timeframe of centuries, probably longer.
Our results provide evidence for the type of feedback 
mechanism between the aboveground and belowground 
subsystems proposed for this study system by Grubb and 
Tanner (1976), and which operates in a comparable man-
ner to that previously shown for a fire-driven retrogres-
sive chronosequence in boreal forest (Wardle et al. 2012). 
In such a feedback, a decreased availability of nutrients 
(notably P) over time would lead to diminished soil fertil-
ity, and replacement of plant species that produce high qual-
ity litter with those that produce poorer litter. The poorer 
quality litter entering the decomposer subsystem would in 
turn impair microbial activity and contribute to reduced 
mineralization of nutrients required for plant growth and 
consequently a build-up of organic matter. Over a time 
scale from centuries to millennia, this reduction in nutrient 
supply would lead to retrogressive ecosystems characterized 
by reduced vegetation stature and productivity. Most studies 
on plant–soil feedbacks have considered only plant associa-
tions with soil organisms that interact with their roots in the 
short term, rather than with the decomposer subsystem over 
much larger spatial and temporal scales (Wardle et al. 2012, 
van der Putten et al. 2013). As such, study systems such as 
the one investigated here can be used to inform on likely 
mechanisms through which feedbacks between the plant 
community and the belowground subsystem, mediated by 
plant litter and organic matter quality, can contribute to eco-
system succession and (over sufficient timescales) ecosystem 
retrogression. More broadly, greater consideration of these 
sorts of feedbacks delivers the potential to better understand 
drivers of ecosystem change in forested ecosystems over the 
time scale of centuries to millennia, including those drivers 
resulting from human-driven global change phenomena. 
Acknowledgements – We thank Kelley Gundale for technical help, 
Carl-Magnus Mörth and Heike Siegmund at the Stockhom Isotope 
Lab (Stockholm Univ.) for analytical help and Adit Sharma for 
Kichenin et al. 2013). At the community level, our results 
show that the greater limitation of both N and P during 
retrogression, and the greater limitation of P relative to N, 
leads to the return to the decomposer subsystem of litter that 
has poorer quality and with an elevated N to P ratio.
However, when foliar and litter N and P characteristics 
for each of the four species that occurred commonly on both 
mull and mor plots was considered, the trends identified at 
the whole community level for were mostly unsupported. 
Consistent with Tanner (1977), we found foliar N and P 
levels within species to be rather unresponsive to the tran-
sition from mull to mor, suggesting a rather low plastic-
ity at the individual species level. This is in line with other 
work along retrogressive chronosequences (Lagerström et al. 
2013) and other gradients of soil fertility (Kichenin et al. 
2013) in revealing that traits of individual species can show 
contrasting responses both to each other and to whole com-
munity measures. These various responses emerge because 
of the complex nature of interactions and consequences for 
nutrient acquisition among coexisting species (Kichenin 
et al. 2013). Our results also contrast with some studies 
that have pointed to a strong role of intraspecific variation 
in determining trait values at the whole community level 
(Albert et al. 2010, Violle et al. 2012). The fact that we find 
clear responses of foliar and litter characteristics at the whole 
community but not individual species level emerges because 
differences between mull and mor plots appear to be driven 
by genera that occur largely only in the mull plots (e.g. 
Solanum, Eugenia) or only in the mor plots (e.g. Clusia, 
Lyonia, Vaccinium, Chaetocarpus), and therefore by turnover 
of species between the mull and mor plots (Wardle et al. 
2009).
We found that while microbial measures as assessed by 
the SIR and PLFA techniques were enhanced in the mor 
plots when expressed per unit soil mass, this arose simply 
through the mor soils containing more SOM; when data 
was expressed per unit SOM (or per unit area) the micro-
flora was impaired, in line with our third hypothesis. This is 
reflective of SOM being of lower quality in the mor than the 
mull plots (Insam and Domsch 1988). Further, we found 
the ratio of fungal to bacterial biomass to increase from mull 
to mor, in line with what has been shown during retrogres-
sion for some other chronosequences (Williamson et al. 
2005, Wardle et al. 2012). This is also consistent with our 
third hypothesis, and is indicative of lower resource qual-
ity and more conservative nutrient cycling in the mor soil 
(Coleman et al. 1983, Wardle et al. 2004a). We did not, 
however, find nematode groups to be impaired in mor 
relative to mull soils, and indeed the reverse was sometimes 
true even when densities were expressed per unit SOM. 
This is likely to be due to greater soil moisture availabil-
ity resulting from higher organic matter levels in the mor 
soil creating a microclimate that is more conducive for 
nematodes (which require moisture for motility and feed-
ing; Bakonyi et al. 2007), and with this effect counteract-
ing any effect of lower fertility in the mor soil. The deeper 
organic matter layer in the mor plots, and probably therefore 
a greater diversity of microhabitats, could also help explain 
the higher diversity of both microbial PLFAs and nematodes 
in the mor plots despite plant diversity being less. In total, 
the data for the primary decomposers (microbes) is strongly 
EV-12
Hedley, M. J. et al. 1982. Changes in inorganic and organic 
phosphorus fractions induced by cultivation practices and by 
laboratory incubations. – Soil Sci. Soc. Am. J. 46: 970–976.
Hobbie, S. E. and Vitousek, P. M. 2000. Nutrient limitation of 
decomposition in Hawaiian forests. – Ecology 81: 1867–1877.
Insam, H. and Domsch, K. H. 1988. Relationship between soil 
organic carbon and microbial biomass on chronosequences of 
reclamation sites. – Microb. Ecol. 15: 177–188.
Kapos, V. and Tanner, E. V. J. 1985. Water relations of Jamaican 
upper montane rain forest trees. – Ecology 66: 241–250.
Kichenin, E. et al. 2013. Contrasting effects of plant inter- and 
intraspecific variation on community-level trait measures along 
an environmental gradient. – Funct. Ecol. 27: 1254–1261.
Killingbeck, K. T. 1996. Nutrients in senesced leaves: keys to 
the search for potential resorption and resorption proficiency. 
– Ecology 77:1716–1727.
Lagerström, A. et al. 2013. Decoupled responses of tree and shrub 
leaf and litter trait values to ecosystem retrogression across an 
island area gradient. – Plant Soil 367: 183–197.
Laliberté, E. et al. 2012. Experimental assessment of nutrient 
limitation along a 2-million year dune chronosequence in the 
south-western Australia biodiversity hotspot. – J. Ecol. 100: 
631–642.
Laliberté, E. et al. 2013. How does pedogenesis drive plant diver-
sity? – Trends Ecol. Evol. 28: 331–340.
Lepš, J. et al. 2001. Habitat and successional status of plants in 
relation to the communities of their leaf-chewing herbivores 
in Papua New Guinea. – J. Ecol. 89: 186–199.
Mason, N. W. H. et al. 2012. Changes in co-existence mechanisms 
along a long-term soil chronosequence revealed by functional 
trait diversity. – J. Ecol. 100: 678–689.
Menge, D. N. L. et al. 2011. Declining foliar and litter d15N 
diverge from soil, epiphyte, and input d15N along a 120 000 
year temperate rainforest chronosequence. – New Phytol. 190: 
941–952.
Muller, P. E. 1884. Studier over skovjord, som bidrag til skovdyr-
kningens theori. II. Om muld og mor i egeskove og paa heder. 
– Tidsskr. Skovbrug 7: 1–232.
Peltzer, D. A. et al. 2010. Understanding ecosystem retrogression. 
– Ecol. Monogr. 80: 509–529.
Plice, M. J. 1946. What is mull soil? – Soil Sci. Soc. Am. J. 10: 
410–413.
Richardson, S. J. et al. 2004. Rapid development of phosphorus 
limitation in temperate rainforest along the Franz Josef soil 
chronosequence. – Oecologia 139: 267–276.
Richardson, S. J. et al. 2005. Resorption proficiency along a chron-
osequence: responses among communities and within species. 
– Ecology 86:20–25.
Selmants, P. C. and Hart, S. C. 2008. Substrate age and tree islands 
influence carbon and nitrogen dynamics across a retrogressive 
semiarid chronosequence. — Global Biogeochem. Cy. 22: 
GB1021.
Suding, K. N. et al. 2008. Plant and microbe contribution to com-
munity resilience in a directionally changing environment. 
— Ecol. Monog. 78: 313–329.
Sugden, A. M. et al. 1985. Regeneration following clearing in a 
Jamaican montane forest: results of a ten year study. – J. Trop. 
Ecol. 1: 329–351.
Sundqvist, M. K. et al. 2014. Plant and microbial responses to 
nitrogen and phosphorus addition across an elevational gradi-
ent in subarctic tundra. – Ecology 85: in press.
Tanner, E. V. J. 1977. Four montane rain forests of Jamaica: 
a quantitative characterization of the floristics, the soils and 
the foliar nutrient levels, and a discussion of the interrelations. 
– J. Ecol. 65: 883–918.
Tanner, E. V. J. 1980. Studies on the biomass and productivity 
in a series of rainforest plots in Jamaica. – J. Ecol. 68: 
573–588.
ensuring safe shipment of the samples on short notice in the face of 
unforeseen obstacles. Drs J. Cohen and K. McLaren at the Univ. of 
the West Indies also provided invaluable help. This work was sup-
ported by a Wallenberg Scholars award to DAW.
References
Albert, C. H. et al. 2010. Intraspecific functional variability: extent, 
structure and sources of variation. – J. Ecol. 98: 604–613.
Anderson, J. P. E. and Domsch, K. H. 1978. A physiologically 
active method for the quantification of microbial biomass in 
soil. – Soil Biol. Biochem. 10: 215–221.
Bakonyi, G. et al. 2007. Soil nematode community structure as 
affected by temperature and moisture in a temperate semiarid 
shrubland. – Appl. Soil Ecol. 37: 31–40.
Bellingham, P. J. et al. 1995. Damage and responsiveness of 
Jamaican montane tree species after disturbance by a hurri-
cane. – Ecology 76: 2562–2580.
Berendse, F. 1998. Effects of dominant plant species on soils during 
succession in nutrient poor ecosystems. – Biogeochem. 42: 
73–88.
Binkley, D. et al. 2000. Soil phosphorus pools and supply under 
the influence of Eucalyptus saligna and nitrogen-fixing Albizia 
facaltaria. – For. Ecol. Manage. 128: 241–247.
Bligh, E. G. and Dyer, W. G. 1959. A rapid method of total lipid 
extraction and purification. – Can. J. Biochem. Physiol. 37: 
911–917.
Brearley, F. Q. 2013. Nitrogen stable isotopes indicate differences 
in nitrogen cycling between two contrasting Jamaican 
montane forests. – Plant Soil 367: 465–476.
Carreiro, M. M. et al. 2000. Microbial enzyme shifts explain litter 
decay responses to simulated nitrogen deposition. – Ecology 
81: 2359–2365.
Coleman, D. C. et al. 1983. Biological strategies of nutrient cycling 
in soil systems. – Adv. Ecol. Res. 13: 1–51.
Crews, T. E. et al. 1995. Changes in soil phosphorus fractions and 
ecosystem dynamics across a long chronosequence in Hawai‘i. 
– Ecology 76: 1407–1424.
Crews, T. E. et al. 2000. Changes in asymbiotic, heterotrophic 
nitrogen fixation on leaf litter of Metrosideros polymorpha with 
long—term ecosystem development in Hawaii. – Ecosystems 
3: 386–395.
Doblas-Miranda, E. et al. 2008. Changes in the community 
structure and diversity of soil invertebrates across the Franz 
Josef Glacier chronosequence. – Soil Biol. Biochem. 40: 
1069–1081.
FAO/UNESCO 1988. FAO/UNESCO soil map of the 
world–revised legend. World Resources Report No. 60. 
 – FAO, Rome, Italy.
Fog, K. 1988. The effect of added nitrogen on the rate of decom-
position of added matter. – Biol. Rev. 63: 433–462.
Garnier, E. et al. 2007. Assessing the effects of land—use change 
on plant traits, communities and ecosystem functioning in 
grasslands: a standardized methodology and lessons from an 
application to 11 European sites. – Ann. Bot. 99: 967–985.
Giesler, R. et al. 2004. Microbially available phosphorus in boreal 
forests: effects of aluminum and iron accumulation in the 
humus layer. – Ecosystems 7: 208–217.
Grubb, P. J. and Tanner, E. V. J. 1976. The montane forests and soils 
of Jamaica: a reassessment. – J. Arnold Arb. 57: 313–368.
Hafkenscheid, R. 2000. Hydrology and biogeochemistry of tropi-
cal montane rain forests of contrasting stature in the Jamaican 
Blue Mountains. – PhD thesis, Vrije Univ., Amsterdam.
Hayes, P. et al. 2014. Foliar nutrient concentrations and resorption 
efficiency in plants of contrasting nutrient-acquisition strate-
gies along a 2-million year dune chronosequence. – J. Ecol. 
102: 396–410.
EV-13
New Zealand forest and scrubland ecosystems. – Funct. Ecol. 
7: 346–355.
Wardle, D. A. et al. 2004a. Ecological linkages between above-
ground and belowground biota. – Science 304: 1629–1633.
Wardle, D. A. et al. 2004b. Ecosystem properties and forest decline 
in contrasting long-term chronosequences. – Science 305: 
509–513.
Wardle, D. A. et al. 2008. The response of plant diversity to eco-
system retrogression: evidence from contrasting long-term 
chronosequences. – Oikos 117: 93–103.
Wardle, D. A. et al. 2009. Among- and within-species variation in 
plant litter decomposition in contrasting long term chronose-
quences. – Funct. Ecol. 23: 442–453.
Wardle, D. A. et al. 2012. Linking vegetation change, carbon 
sequestration and biodiversity: insights from island ecosystems 
in a long term natural experiment. – J. Ecol. 100: 16–30.
Wardle, D. A. et al. 2013. Decoupled long term effects of nutrient 
enrichment on aboveground and belowground properties in 
subalpine tundra. – Ecology 94: 904–919.
White, D. C. et al. 1979. Determination of the sedimentary micro-
bial biomass by extractable lipid phosphate. – Oecologia 40: 
51–62.
Williamson, W. M. et al. 2005. Changes in soil microbial 
and nematode communities during ecosystem retrogression 
across a long term chronosequence. – Soil Biol. Biochem. 37: 
1289–1301.
Yeates, G. W. et al. 1993. Feeding habits in soil nematode families 
and genera – an outline for soil ecologists. – J. Nematol. 25: 
315–331.
Tanner, E. V. J. 1981. The decomposition of leaf litter in Jamaican 
montane rain forests. – J. Ecol. 69: 263–275.
Tanner, E. V. J. and Bellingham, P. J. 2006. Less diverse forest 
is more resistant to hurricane disturbance: evidence from mon-
tane rain forests in Jamaica. – J. Ecol. 94: 1003–1010.
ter Braak C. J. F. and Šmilauer, P. 2002. CANOCO Reference 
manual and CanoDraw for Windows user’s guide: software for 
canonical community ordination (ver. 4.5). — Microcom-
puter Power. 
van der Putten, W. H. et al. 2013. Plant–soil feedback: the past, 
the present and future challenges. – J. Ecol. 101: 265–276.
van Heerwaarden, L. M. et al. 2003. Current measures of nutrient 
resorption efficiency lead to a substantial underestimation of 
real resorption efficiency: facts and solutions. – Oikos 101: 
664–669.
Violle, C. et al. 2012. The return of the variance: intraspecific 
variability in community ecology. – Trends Ecol. Evol. 27: 
244–252
Vitousek, P. M. 2004. Nutrient cycling and limitation: Hawai’i as 
a model system. – Princeton Univ. Press.
Walker, L. R. and del Moral, R. 2003. Primary succession and 
ecosystem rehabilitation. – Cambridge Univ. Press.
Walker, L. R. et al. 2010. The use of chronosequences in studies 
of ecological succession and soil development. – J. Ecol. 98: 
725–736.
Walker, T. W. and Syers, J. K. 1976. The fate of phosphorus during 
pedogenesis. – Geoderma 15: 1–19.
Wardle, D. A. 1993. Changes in the microbial biomass and 
metabolic quotient during leaf litter succession in some 
Supplementary material (available online as Appendix 
oik.01584 at www.oikosjournal.org/readers/appendix). 
Appendix 1.