Patterns of tree community composition along a coastal
dune chronosequence in lowland temperate rain forest
in New Zealand
Benjamin L. Turner ? Andrew Wells ?
Kelly M. Andersen ? Leo M. Condron
Received: 2 April 2012 / Accepted: 27 July 2012 / Published online: 26 September 2012
 Springer Science+Business Media B.V. (outside the USA) 2012
Abstract Soil chronosequences provide an opportu-
nity to examine the influence of long-term pedogenesis
on the biomass and composition of associated tree
communities. We assessed variation in the species
composition of trees, saplings, and seedlings, and the
basal area of adult trees, in lowland temperate rain forest
along the Haast chronosequence on the west coast of the
South Island of New Zealand. The sequence consists of
Holocene dune ridges formed following periodic earth-
quake disturbance and is characterized by rapid podzol
development, including a marked decline in phosphorus
concentrations, accumulation of a thick organic horizon,
and formation of a cemented iron pan. Tree basal area
increased for the first few hundred years and then
declined in parallel with the decline in total soil
phosphorus, consistent with the concept of forest
retrogression. There were also marked changes in the
composition of the tree community, from dominance by
conifers on young soils to a mixed conifer?angiosperm
forest on old soils. Although a variety of factors could
account for these changes, partial Mantel tests revealed
strong correlations between tree community composi-
tion and soil nutrients. The relationships differed among
life history stages, however, because the adult tree
community composition was correlated strongly with
nutrients in the mineral soil, whereas the seedling
community composition was correlated with nutrients
in the organic horizon, presumably reflecting differ-
ences in rooting depth. The changes in the tree
community at Haast are consistent with disturbance-
related succession in conifer?angiosperm forests in the
region, but the opposite of patterns along the nearby
Franz Josef post-glacial chronosequence, where coni-
fers are most abundant on old soils. The Haast
chronosequence is therefore an important additional
example of forest retrogression linked to long-term soil
phosphorus depletion, and provides evidence for the
role of soil nutrients in determining the distribution of
tree species during long-term succession in lowland
temperate rain forests in New Zealand.
Keywords Chronosequence  Holocene  Mantel
tests  New Zealand  Nitrogen  Nothofagus 
Phosphorus  Podocarpaceae  Podzol  Retrogression
Introduction
Long-term ecosystem development is characterized by a
slow decline in soil phosphorus availability to plants
(Walker and Syers 1976). This occurs through the
Electronic supplementary material The online version of
this article (doi:10.1007/s11258-012-0108-3) contains
supplementary material, which is available to authorized users.
B. L. Turner (&)  K. M. Andersen
Smithsonian Tropical Research Institute, Apartado
0843-03092, Balboa, Ancon, Republic of Panama
e-mail: TurnerBL@si.edu
A. Wells  L. M. Condron
Agriculture and Life Sciences, Lincoln University,
PO Box 84, Lincoln 7647, Canterbury, New Zealand
123
Plant Ecol (2012) 213:1525?1541
DOI 10.1007/s11258-012-0108-3
gradual loss of total phosphorus via leaching and erosion
(Hedin et al. 2003) as well as chemical transformations
of phosphorus compounds in the soil, including a
depletion of primary mineral phosphate and an accumu-
lation of phosphorus bound in secondary minerals and
organic compounds (Crews et al. 1995; Parfitt et al. 2005;
Turner et al. 2007). In the absence of rejuvenating
disturbance, the decline in phosphorus availability leads
to a corresponding decline in the biomass of the
vegetation, termed ?retrogression? (Wardle et al. 2004),
as well as changes in the diversity and composition of the
plant community (Mark et al. 1988; Richardson et al.
2005; Walker et al. 1981; Wardle et al. 2008). Soil
chronosequences therefore provide opportunities to
obtain information on long-term drivers of plant com-
munity assembly (Peltzer et al. 2010; Walkeret al. 2010).
One of the most well-studied soil chronosequences
occurs at Franz Josef, on the west coast of the South
Island of New Zealand. The sequence is formed in post-
glacial deposits up to 120,000 years old and exhibits a
clear decline in phosphorus availability (Parfitt et al.
2005; Walker and Syers 1976), with a corresponding
decline in forest biomass on the oldest surfaces (Wardle
et al. 2004). A similar pattern occurs along the Waitutu
chronosequence of marine uplift terraces, also on the
South Island of New Zealand (Mark et al. 1988; Parfitt
et al. 2005). There are marked changes in the forest
community during ecosystem development along these
two sequences, with dominance by evergreen angio-
sperms on young soils and an increasing abundance of
conifers in the family Podocarpaceae as soils age
(Coomes et al. 2005; Richardson et al. 2004).
The increasing abundance of conifers as ecosystem
development proceeds is assumed to indicate the
superior ability of the podocarps to survive on
phosphorus-poor soils, having the advantage of slow
growth, long-lived leaves, and other nutrient-conserv-
ing strategies (Coomes and Bellingham 2011; Rich-
ardson et al. 2005). The presence of tree ferns on
young soils and waterlogging on old soils also appear
to be important in structuring the podocarp community
along the sequences (Coomes et al. 2005; Gaxiola
et al. 2010). However, the trend toward increasing
conifer dominance differs from disturbance-related
successional patterns observed in the region, whereby
large earthquakes along the Alpine Fault promote
initial dominance by conifers, followed by slow
progression to a mixed conifer?angiosperm forest
(Ogden and Stewart 1995; Stewart et al. 1998; Wells
et al. 2001). This suggests factors other than pedo-
genesis-driven changes in nutrient status might be
more important in determining forest community
composition in New Zealand temperate rain forests.
Here we report the results of a survey of tree
communities along the Haast chronosequence, a prog-
radational dune sequence formed by periodic distur-
bance events linked to earthquakes on the Alpine fault
that runs through the spine of the South Island of New
Zealand (Turner et al. 2012; Wells and Goff 2006). The
Haast chronosequence provides a strong test of factors
influencing plant community composition during eco-
system development, because the dunes are formed from
the same parent material, have similar topography, occur
in a small area of uniform climate, support undisturbed
forest, and the dates of dune formation are reasonably
well constrained. Our specific aims were (i) to quantify
changes in the tree community along the sequence,
including basal area and the species composition of the
adult, sapling, and seedling communities, and (ii) assess
the extent to which plant community composition was
linked to dispersal limitation or changes in soil nutrients.
We were particularly interested in whether patterns in
the tree community would correspond to those observed
at the nearby Franz Josef chronosequence, a very
different sequence in terms of soil properties and
sequence age, but under similar rainfall and vegetation.
Materials and methods
The Haast chronosequence
The Haast dune system is located northeast of the town
of Haast, on the west coast of the South Island of New
Fig. 1 The Haast chronosequence, showing a an aerial image
of the entire sequence with the approximate transect line
indicated by the blue bar, with youngest dunes on the left close
to the road, and oldest dunes on the right; b, c a young dune
(Dune 4, 517 years BP), with abundant tree ferns (D. squarrosa
and C. smithii) under a canopy of large D. cupressinum, growing
on a weakly developed soil (Typic Udipsamment); d, e an
intermediate-aged dune (Dune 8, 1,826 years BP) with a mixed
conifer?angiosperm tree community on a moderately-developed
soil (Spodic Udipsamments) with a clear bleached eluvial
horizon; f, g an old dune (Dune 12, 3,903 years BP) with a mixed
conifer?angiosperm tree community on a well-developed
Spodosol (Typic Placorthod) with a bleached eluvial horizon,
a spodic B horizon, and a continuous cemented placic horizon
(iron pan). Photo credits: a Land Information New Zealand;
b?f B.L. Turner. (Color figure online)
c
1526 Plant Ecol (2012) 213:1525?1541
123
Plant Ecol (2012) 213:1525?1541 1527
123
Zealand (434302000S, 169403000E) (Fig. 1). The sys-
tem extends *10 km alongshore and 5 km inland,
with dunes 20?100 m wide rising up to 20 m above
the adjacent dune slacks. Dune building episodes are
thought to result from earthquakes along the Alpine
Fault, which cause widespread disturbance including
landslides, rock falls, floodplain aggradation, and tree
falls (Wells and Goff 2007). After each event, a pulse
of sediment is transported rapidly to the ocean via the
Haast River and then deposited as a linear dune either
side of the river mouth. The youngest dune is thus
located closest to the ocean, with dunes becoming
progressively older inland.
Dune building is a regional phenomenon and has
occurred following all known earthquakes since AD
1220, and presumably since sea level stabilized during
the mid-Holocene, ca. 7,000 years ago (Wells and
Goff 2007). This is supported by the consistent
earthquake record spanning the last 8,000 years (Ber-
ryman et al. 2012). The earliest dunes have been dated
precisely using tree rings and historical records (Wells
and Goff 2007), while older dune dates are estimated
based on a 14C date for the oldest dune (see Turner
et al. 2012). Parent material is assumed to be similar
across the sequence based on the mineralogy of
unweathered sand (Palmer et al. 1986).
Modern mean annual temperature is 11.3 C and
mean annual rainfall is 3,455 mm spread relatively
evenly throughout the year, with no month having
\200 mm (New Zealand Meteorological Service
1983). Climate at the site has varied considerably
since the onset of dune formation, with evidence for a
dry period between 3,000 and 4,000 years BP and a
substantial cold period around 3,500 years BP (Li et al.
2008). However, forests in the region have been mixed
conifer?broadleaf temperate rain forest since at least
7,700 years BP (Li et al. 2008).
Soil development and changes in nutrient
availability
Detailed information on soil development and changes
in soil properties are reported elsewhere (Eger et al.
2011; Palmer et al. 1986; Turner et al. 2012). In brief,
the soils develop rapidly to podzols under the perhu-
mid climate, from Typic Udipsamments (Entisols) on
young dunes (\800 years BP) to Typic Placorthods
(Spodosols) on older dunes (3,903?6,500 years BP)
(Fig. 1). The latter soils are characterized by a thick
organic horizon, a bleached eluvial horizon, a spodic B
horizon (i.e., subsoil accumulation of iron and organic
matter), and a cemented iron pan (placic horizon) that
forms at around 10?20 cm depth in the mineral soil
(Turner et al. 2012). The pan is likely to have
ecological significance, because it impedes vertical
water movement and root penetration, physically
preventing access to nutrients in deeper horizons.
The thickness of the organic horizon increases from
\1 cm on the youngest dune to 17?28 cm on the older
dunes and there is rapid acidification and depletion of
exchangeable base cations in the early stages of
pedogenesis. Soils are very acidic, with mineral soil
pH (measured in water) B4.2 for the majority of the
sequence and declining to *3.5 on the oldest dunes
(Turner et al. 2012; see supplementary material).
Patterns of total nitrogen and phosphorus along the
sequence are summarized in Fig. 2. Total nitrogen
concentrations increase from low values in the young
soils to maximum and stable values in older soils, with
a slight decline in the total nitrogen concentration with
age (Fig. 2a). In contrast, total phosphorus concentra-
tions in both organic and surface mineral soil horizons
decline rapidly in the first few hundred years of
pedogenesis, from high initial values in young soils to
the lowest values in old soils (Fig. 2b). The corre-
sponding N:P ratios increase with soil age (Fig. 2c).
The decline in total phosphorus in mineral soil in the
early stages of pedogenesis was accounted for by
primary mineral phosphate (i.e., acid-extractable
inorganic phosphate). Organic phosphorus accounted
for 72?84 % of the total phosphorus for all but the
youngest dune (*20 %) (Turner et al. 2012). Infor-
mation on soil properties for the five dunes studied for
vegetation is reported in supplementary material.
Vegetation measurements
We established three replicate 10 9 20 m vegetation
plots on five of the eleven dunes studied previously
(Turner et al. 2012): Dune 3 (312 years BP), Dune 4
(517 years BP), Dune 8 (1,826 years BP), Dune 13
(4,422 years BP), and Dune 17 (6,500 years BP). Plots
were established around the intersection of the sam-
pling transect across the dunes and were therefore
effectively located randomly. The first plot was
established at the intersection of the transect line with
the dune crest, with the second and third plots located
either side along the dune, beginning 30 m from the
1528 Plant Ecol (2012) 213:1525?1541
123
edge of the first plot. The plot size (200 m2 rather than
the standard 400 m2) and shape were necessitated by
the narrow shape of the dune ridges. Plots were
positioned 20 m along the dune crest and 10 m down
the inland side of the dune ridge; the coastal side was
not sampled because colonization of this portion of
dune ridge by trees is delayed until after the next
youngest dune has formed (Wells and Goff 2007). The
triplicate plots on each dune gave a total area of
600 m2 per dune.
All stems C10 cm diameter at breast height (dbh)
(1.4 m) were identified and measured for calculation
of basal area. Saplings (stems\10 cm dbh and taller
than 1.4 m) were identified and enumerated, but
diameters were not recorded. Seedlings (stems
between 30 cm and 1.4 m in height) were identified
and enumerated over the whole plot. Ground cover,
including bryophytes, litter, ferns, and bare ground,
was estimated over the entire plot using standard
procedures in New Zealand forest ecology (Hurst and
Allen 2007) (see supplementary material).
Statistical analysis
Multivariate analyses of tree communities, soil prop-
erties, and dune age were conducted using the
statistical software package ??vegan?? version 1.17.3
(Oksanen 2010) in the R programming environment
version 2.11.1 (R Development Core Team,
www.r-project.org). We used Bray?Curtis dissimilar-
ity indices to calculate between-dune differences in
tree, sapling, and seedling communities based on
mean species abundance data and absolute (m2 ha-1)
and proportion (%) of the total basal area for the tree
community. Floristic dissimilarity matrices were
based on data from Turner et al. (2012), which were
square root transformed and submitted to a Wisconsin
double standardization procedure, with species data
divided by their maxima and dunes standardized to
equal totals (Oksanen 2010). We used double stan-
dardization because it ranks species regardless of their
taxonomic status and adult stature. Analyses without
double standardization yielded qualitatively similar
results. The main difference was that the signal
between the seedling community and the organic
horizon was lost, while the sapling?mineral soil cor-
relation was strengthened (data not shown). We used
Euclidean distance to calculate between-dune differ-
ences in age, soil, and organic horizon. Dune age was
used as a proxy for geographic distance.
Mineral soil and organic horizon data were simpli-
fied into three multivariate variables based on separate
principal components analyses (PCA) prior to calcu-
lating distance matrices. For the mineral soil PCA, we
included pH, total carbon, total nitrogen, inorganic
nitrogen, readily-exchangeable phosphorus, C:N ratio,
C:P ratio, sand, silt, and clay. Mineral soil PCA axis 1
captured 70.5 % of the variation, with C:N and silt
loading positively and readily-exchangeable phos-
phate, pH, and sand loading negatively. Mineral soil
PCA axis 2 captured 15.5 % of the variation in soil
variables, with total carbon and nitrogen loading
positively. Mineral soil PCA axis 3 captured 11.0 % of
the variation in soil variables, with inorganic nitrogen
To
ta
l N
 in
 o
rg
an
ic 
ho
riz
on
8
12
16
20
Total N
 in m
ineral soil0
1
2
Organic horizon
Mineral soil
To
ta
l P
 in
 o
rg
an
ic 
ho
riz
on
0.4
0.6
0.8
1.0 Total P in m
ineral soil0.1
0.2
0.3
Organic horizon
Mineral soil
Dune age (years B.P.)
0 2000 4000 6000
N
:P
 o
rg
an
ic 
ho
riz
on
10
15
20
25
30
35
N
:P m
ineral soil
0
3
6
9
12
Organic horizon
Mineral soil
a
b
c
Fig. 2 Changes in a total nitrogen (g N kg-1), b total
phosphorus (g P kg-1), and c the N:P ratio in the organic horizon
and mineral soil (0?20 cm) along the Haast dune sequence,
Westland, New Zealand (data from Turner et al. 2012)
Plant Ecol (2012) 213:1525?1541 1529
123
and C:P ratio loading positively and clay loading
negatively. For the organic horizon PCA, we included
horizon depth, pH, total carbon, total nitrogen, total
phosphorus, and C:N ratio. Organic horizon PCA axis
1 captured 70.7 % of variation, with total nitrogen,
total carbon, and depth loading negatively, and pH
loading positively. Organic horizon PCA axis 2
captured 18.2 % of the variation, with C:N ratio
loading positively. Axis 3 captured 10.8 % of the
variation, with total phosphorus loading negatively.
Thus, for both mineral and organic soils, the three axes
accounted for 99.7 % of the variation and each axis
accounted for at least 10 % of the variation.
Mantel tests were performed to assess the correla-
tion between plant floristic distance matrices and the
mineral soil, organic horizon, and dune age distance
matrices. Partial Mantel tests were used to examine the
correlation between pair-wise differences in the plant
matrices and mineral soil and organic horizon matrices
controlling for dune age (??pure mineral?? and ??pure
organic??) and between pair-wise differences in the
plant matrices and dune age controlling for soil
variables (??pure age??). The standardized Mantel test
statistic (r) was calculated based on the Pearson?s
product-moment correlation coefficient and signifi-
cance was assessed using randomization tests with
1,000 permutations. In addition, we examined corre-
lations between soil properties and stem density,
relative abundance, and basal area of adult conifers,
woody angiosperms, and tree ferns.
Results
Plant diversity in forest census plots
There were three key stages of forest structure along the
chronosequence (Fig. 1). Fourteen species were present
along the sequence as individuals C10 cm dbh, includ-
ing five species in the Podocarpaceae, two tree ferns, and
seven woody angiosperms (Table 1). Note that we
include Phyllocladus within the Podocarpaceae,
Table 1 Basal area (m2 ha-1) of trees C10 cm dbh at five sites along the Haast dune sequence, New Zealand
Species Family Dune 3
(392 BP)
Dune 4
(517 BP)
Dune 8
(1,826 BP)
Dune 13
(4,422 BP)
Dune 17
(6,500 BP)
Basal area (m2 ha-1)
Conifers
Dacrydium cupressinum Podocarpaceae 108.2 ? 12.1 181.3 ? 12.0 36.8 ? 8.6 54.1 ? 38.1 0.0
Phyllocladus alpinus Podocarpaceae 0.0 0.0 0.0 1.1 ? 1.1 0.8 ? 0.6
Podocarpus hallii Podocarpaceae 0.0 0.0 1.2 ? 0.6 0.9 ? 0.9 1.7 ? 1.5
Podocarpus totara Podocarpaceae 7.5 ? 7.3 0.0 0.0 0.0 0.0
Prumnopitys ferruginea Podocarpaceae 8.0 ? 3.3 6.2 ? 3.3 11.2 ? 7.4 15.4 ? 7.5 10.0 ? 3.7
Total conifers 123.7 ? 2.8 187.5 ? 14.3 49.1 ? 2.4 71.4 ? 42.8 12.4 ? 4.5
Woody angiosperms
Elaeocarpus dentatus Elaeocarpaceae 0.0 0.0 0.0 0.2 ? 0.2 0.0
Elaeocarpus hookerianus Elaeocarpaceae 0.0 0.0 0.0 0.4 ? 0.4 0.4 ? 0.4
Hedycarya arborea Monimiaceae 0.0 0.8 ? 0.8 0.0 0.0 0.0
Metrosideros umbellata Myrtaceae 0.0 0.0 0.0 0.3 ? 0.3 0.0
Nothofagus menziesii Fagaceae 0.0 0.0 0.0 0.3 ? 0.3 15.9 ? 8.3
Pseudopanax crassifolius Araliaceae 0.0 0.0 0.0 0.0 0.1 ? 0.1
Weinmannia racemosa Cunoniaceae 0.6 ? 0.2 0.3 ? 0.3 25.0 ? 11.7 29.1 ? 4.6 31.7 ? 17.9
Total woody angiosperms 0.6 ? 0.2 1.1 ? 0.7 25.0 ? 11.7 30.3 ? 4.4 48.2 ? 20.6
Tree ferns
Cyathea smithii Cyatheaceae 19.1 ? 2.0 10.6 ? 1.6 0.0 0.0 0.0
Dicksonia squarrosa Dicksoniaceae 11.9 ? 6.4 6.6 ? 0.2 0.2 ? 0.2 0.0 0.0
Total tree ferns 31.0 ? 6.3 17.1 ? 1.6 0.2 ? 0.2 0.0 0.0
Values are mean ? standard error of three replicate 200 m2 plots located along the dune ridge at each site
1530 Plant Ecol (2012) 213:1525?1541
123
because the recent elevation of the genus to family level
(Phyllocladaceae) is not supported by phylogenetic
analyses (Biffin et al. 2011). One conifer, Dacrycar-
pus dacrydioides (Podocarpaceae), occurred as sap-
lings throughout the sequence, but not as tree-sized
individuals (see below). The youngest dune studied
supported a further species of woody angiosperm
(Pseudowintera colorata, Winteraceae) as a sapling-
sized plant, as well as seedlings of two other
angiosperms in the Araliaceae (Pseudopanax arboreus
and Schefflera digitata).
Basal area of trees C10 cm diameter at breast
height
Basal area for stems C10 cm dbh was initially high for
the two youngest dunes studied and then declined
(Table 1; Fig. 3). Maximum basal area (206 ?
13 m2 ha-1) was recorded on Dune 4 (517 years BP)
and the minimum basal area (60.6 ? 24.5 m2 ha-1)
was recorded on the oldest dune (6,500 years BP). An
increase in basal area on Dune 13 (4,422 years BP) was
due to a single large D. cupressinum individual of
161 cm dbh, reflected in the large standard error for
conifer basal area at this site (Table 1; Fig. 3).
Most of the basal area (C80 %) on the two youngest
dunes studied was accounted for by podocarps
(124?188 m2 ha-1, Table 1), although they declined
markedly as dunes aged, accounting for only
12.4 m2 ha-1 (23 % of total basal area) on the oldest
dune (Fig. 3). Tree ferns accounted for 31 and 17 % of
the total basal area on the two youngest dunes studied,
respectively. Both Cyathea smithii and Dicksonia
squarrosa were present along the sequence, with
C. smithii always in greatest abundance (Table 1,
supplementary material). Tree ferns were absent from
plots on the two oldest dunes studied (Table 1; Fig. 3),
although a few individuals were noted outside the
plots on these dunes.
Woody angiosperms accounted for only a small
proportion of the basal area on the youngest dunes
(\1 m2 ha-1,\1 % total basal area), but increased on
the older dunes to a maximum of 48.2 m2 ha-1 (77 %
of the total basal area) (Table 1; Fig. 3; supplementary
Table 1). The majority of the basal area of woody
angiosperms C10 cm dbh was W. racemosa, which
accounted for 31?49 % of the total basal area on the
oldest three dunes (Table 1; supplementary Table 1).
In addition, Nothofagus menziesii accounted for 27 %
of the total basal area on the oldest dune
(15.9 ? 8.3 m2 ha-1) (Table 1; Fig. 3; supplemen-
tary Table 1). This species is shown separately from
other woody angiosperms in Figs. 3 and 4, because it
represents the only ectomycorrhizal species along the
sequence. Other woody angiosperms present as trees
C10 cm dbh in the plots included Hedycarya arborea,
Metrosideros umbellata, Pseudopanax crassifolius,
and two species of Elaeocarpus (Table 1).
a
Ba
sa
l a
re
a 
(m
2  
ha
-
1 )
0
50
100
150
200
250
Total basal area
Conifers
Woody angiosperms
Tree ferns
Nothofagus menziesii
b
Dune system
3 4 8 13 17
Ba
sa
l a
re
a 
(%
 of
 to
tal
)
0
20
40
60
80
100
Fig. 3 Basal area of individuals C10 cm dbh at five sites of
contrasting age along the Haast dune sequence, Westland, New
Zealand. Values are the mean ? standard error of three
replicate 10 m 9 20 m plots located along the dune crest at
each site. N. menziesii is shown separately from the woody
angiosperms to illustrate the pattern for the only ectomycorrhi-
zal species that occurs along the sequence and is not included in
the woody angiosperms data
Plant Ecol (2012) 213:1525?1541 1531
123
Abundance of adult trees, saplings, and seedlings
along the chronosequence
The number of individuals C10 cm dbh ranged from
1,100 to 2,183 stems ha-1, being highest on the youn-
gest dunes and lowest on the three oldest dunes (Table 2;
Fig. 4). The majority of the stems C10 cm dbh on the
youngest two dunes were tree ferns ([1,500 stems ha-1
on Dune 3), with woody angiosperms constituting the
majority of stems on the oldest two dunes (Fig. 4).
These were dominated by W. racemosa (59 % of total
stems C10 cm dbh), with a smaller contribution of
N. menziesii (9 % of total stems C10 cm dbh) (supple-
mentary material). Podocarps accounted for a relatively
consistent number of stems C10 cm dbh across the five
dunes (317?583 stems ha-1) (Fig. 4).
The number of saplings was greatest on Dune 8
(3,117 ? 1,090 stems ha-1) and smallest on the old-
est dune (567 ? 203 stems ha-1) (Table 2; Fig. 4). In
contrast to individuals C10 cm dbh, the number of
woody angiosperm saplings declined, and podocarp
saplings increased, along the sequence, at least as a
proportion of the total saplings (Fig. 4). Saplings were
rarely [5 m tall and therefore did not reach the
canopy. Most of the saplings on the youngest three
dunes were woody angiosperms, mainly species in the
genus Coprosma, as well as a considerable number of
W. racemosa individuals. Five species of Coprosma
have been recorded along the sequence (Dickinson
and Mark 1994). In our plots, they accounted for
[1,000 stems ha-1 of sapling-sized individuals on the
three youngest dunes, including virtually all the
saplings (98 ? 2 %) on Dune 4 (392 years BP);
however, Coprosma spp. accounted for only around
10 % of the total saplings on the oldest two dunes
a
In
di
vid
ua
ls 
(st
em
s h
a-1
)
0
500
1000
1500
2000
2500
3000
Total individuals
Conifers 
Woody angiosperms
Tree ferns
Nothofagus menziesii
b
0
1000
2000
3000
4000
5000
c
Dune system
3 4 8 13 17
0
1000
2000
3000
4000
Fig. 4 Number of a trees (C10 cm dbh), b saplings (\10 cm
dbh and taller than 1.4 m), and c seedlings (0.3?1.4 m high) at
five sites along the Haast dune sequence, New Zealand. Values
are the mean ? standard error of three replicate 10 m 9 20 m
plots located along the dune crest at each site. N. menziesii is
shown separately from the woody angiosperms to illustrate the
pattern for the only ectomycorrhizal species that occurs along
the sequence and is not included in the woody angiosperms data;
saplings were only recorded for Dune 13 (33 individuals ha-1).
No conifer saplings were recorded on Dune 4, but this point is
included on the figure (other instances of zero adults, saplings,
or seedlings, are not shown)
b
1532 Plant Ecol (2012) 213:1525?1541
123
(see supplementary material). Coprosma saplings
undoubtedly increase the total basal area of woody
angiosperms on the young dunes, although we
estimate this to be \3 m2 ha-1 out of a total basal
area of[150 m2 ha-1. Nothofagus menziesii occurred
as saplings only on Dune 13 (33 stems ha-1), con-
tributing 1 % of the total saplings (Fig. 4). Podocarps
represented a considerable proportion of the total
saplings on the oldest three dunes studied (23?53 %),
being particularly abundant on Dunes 8 and 13
(717?1,033 stems ha-1, respectively), but were rare
on the two youngest dunes (B2 % of total saplings)
(Fig. 4). Thus, saplings on the youngest two dunes
were almost all woody angiosperms, while on the
oldest two dunes saplings were evenly distributed
between angiosperms and conifers.
Seedling abundance varied between 883 and 2,917
individuals per hectare along the sequence, with the
highest values recorded for Dune 8 and Dune 13
(Table 2; Fig. 4). Seedlings followed a similar pattern
to saplings, with young dunes dominated by woody
angiosperms and relatively similar numbers of angio-
sperms and podocarps on older dunes. In fact, the
majority of the seedlings on all dunes were woody
angiosperms (63?96 % of total seedlings). Most of
these were Coprosma spp., which accounted for
617?1,100 stems ha-1 and up to 93 % of the total
seedlings (see supplementary material). In addition,
W. racemosa (17?433 stems ha-1, 3?15 % of total seed-
lings) and P. crassifolius (up to 9 % of total seedlings)
were abundant along the entire sequence (see supple-
mentary material). Podocarp seedlings were abundant
only on Dune 8 and 13 (*1,000 stems ha-1), mainly
as two species (see below), but were rare on both the
youngest and oldest dunes (\100 stems ha-1) (Fig. 4).
Seedlings of N. menziesii occurred only on Dune 8 and
13; on the latter dune, there were 483 stems ha-1,
accounting for 31 % of the total seedlings at that site
(Fig. 4). It was notable, however, that no seedlings or
saplings of N. menziesii were present in plots on the
oldest dune, despite the presence of several adults
C10 cm dbh.
Distribution of the Podocarpaceae
along the chronosequence
Six species of podocarps were identified in plots on the
five dunes studied and their distribution varied mark-
edly along the sequence. On the three youngest dunes,
most of the basal area and total individuals C10 cm
dbh were accounted for by D. cupressinum, but this
species declined along the sequence and was present in
plots on the oldest dune only as saplings (Fig. 5,
supplementary material). Seedlings of this species
occurred only on the youngest dune in very low
numbers. Occasional large individuals of D. cupress-
inum are, however, present as scattered emergents
outside our plots on the oldest dune.
In contrast, Prumnopitys ferruginea increased
markedly along the sequence, eventually forming the
dominant podocarp in terms of both basal area and
number of individuals (as both trees and saplings) on
the oldest dune. Of the other podocarp species,
Podocarpus totara was present on Dune 3 but not
elsewhere, while P. hallii was absent on young dunes
but present on older dunes (Fig. 5). Both Podocarpus
Table 2 Summary of vegetation at five sites of contrasting age along the Haast dune sequence, Westland, New Zealand
Dune stage Dune age (years BP) Basal area Individuals
Treesa Treesa Saplingsb Seedlingsc
m2 ha-1 Stems ha-1
Dune 3 392 155.2 ? 4.6 2,183 ? 322 1,967 ? 377 1,100 ? 153
Dune 4 517 205.8 ? 13.4 1,583 ? 192 1,050 ? 260 900 ? 350
Dune 8 1,826 74.3 ? 10.1 1,100 ? 58 3,117 ? 1,090 2,917 ? 145
Dune 13 4,422 101.7 ? 41.2 1,133 ? 169 2,000 ? 889 2,583 ? 806
Dune 17 6,500 60.6 ? 24.5 1,183 ? 60 567 ? 203 883 ? 448
Values are the mean ? standard error of three replicate 10 m 9 20 m plots located along the dune crest at each site
a Stems C10 cm dbh
b Stems \10 cm dbh (1.4 m) and C1.4 m high
c Stems [30 cm and \1.4 m in height
Plant Ecol (2012) 213:1525?1541 1533
123
species made a relatively small contribution to the
basal area of trees C10 cm dbh along the sequence,
although there were a large number of saplings and
seedlings of P. hallii on Dunes 8 and 13 (but low
numbers on other dunes) and P. totara does occur on
the older dunes of other sequences in the region (A.
Wells, personal observation). Similarly, there were
large numbers of saplings and seedlings of Phyllocla-
dus alpinus on Dunes 8 and 13, but few elsewhere, and
this species made only a small contribution to the total
basal area, with adults present only on the two oldest
dunes studied (Fig. 5). Dacrycarpus dacrydioides was
the only podocarp not present as individuals C10 cm
dbh anywhere along the sequence, and saplings and
seedlings of this species were infrequent on all dunes
studied. Thus, Dacrydium cupressinum dominates on
young soils, Prumnopitys ferruginea dominates the
conifer community on old soils, but only Podocarpus
hallii and Phyllocladus alpinus (and to a lesser extent
Prumnopitys ferruginea) appear to be regenerating
successfully under the existing canopy.
Plant communities in relation to dune age, mineral
soil, and organic horizon
A number of significant correlations (p \ 0.05)
occurred between forest composition, soil properties,
and dune age. Notably, conifer basal area was
positively correlated with mineral soil nitrogen and
readily-exchangeable phosphate (both r = 0.94),
angiosperm basal area was correlated negatively with
soil pH (r = -0.98) and total phosphorus in mineral
soil (r = -0.96), while tree fern basal area was
correlated positively with soil pH (r = 0.89) and total
phosphorus in mineral soil (r = 0.91), and negatively
with the N:P ratio in the organic horizon (r = -0.92).
Dune age was correlated positively with the basal area
of angiosperms (r = 0.95), but not conifers or tree
ferns. Dune age was, however, correlated negatively
with soil pH (r = -0.96) and total phosphorus in
mineral soil (r = -0.94). A full correlation table is
presented in supplementary material, including corre-
lations for abundance and relative basal area.
Mantel tests showed significant associations
between the plant communities, soil variables (as
PCA axes), dune age, and/or the organic horizon
variables for at least one plant community variable
(Table 3; Fig. 6). Differences in mineral soil variables
were related to dune age (r = 0.72, p = 0.04),
c
Dune system
3 4 8 13 17
0
200
400
600
800
a
0
100
200
300
400
500
600
b
In
di
vid
ua
ls 
(st
em
s h
a-1
)
0
200
400
600
800
1000
Dacrydium cupressinum
Podocarpus hallii
Prumnopitys ferruginea
Phyllocladus alpinus
Dacrycarpus dacrydiodes
Podocarpus totara
Fig. 5 Number of individual a trees, b saplings, and c seedlings
of the five podocarp species at five sites of contrasting age along
the Haast dune sequence, New Zealand. Values are the
mean ? standard error of three replicate 10 m 9 20 m plots
located along the dune crest at each site
1534 Plant Ecol (2012) 213:1525?1541
123
whereas differences in organic horizon variables were
independent of dune age (r = 0.18, p [ 0.05).
For trees C10 cm dbh, community composition based
on abundance of individuals was correlated with mineral
soil properties (r = 0.95, p\0.05) and dune age
(r = 0.76?0.87, p\0.05), but not organic horizon
properties (r = 0.45, p[0.05) (Table 3; Fig. 6). How-
ever, partial Mantel tests also revealed that while the
??pure mineral soil?? effect was correlated significantly
with the tree community composition (r = 0.90,
p\0.05), the ??pure organic horizon?? and ??pure age??
effects were not correlated significantly (Table 3; Fig. 6).
For sapling community composition, there was a
significant relationship with mineral soil properties
(r = 0.42, p \ 0.05), but not with dune age or organic
horizon (Table 3; Fig. 6). When correcting for vari-
ation associated with dune age, neither ??pure mineral
soil?? nor ??pure organic horizon?? was correlated
significantly with the sapling community.
The seedling community composition was not
correlated significantly with either mineral soil prop-
erties or dune age, but was correlated significantly
with organic horizon properties (r = 0.55, p \ 0.01),
including after correction for variation related to age
(r = 0.58, p \ 0.05) (Table 3; Fig. 6). A separate
analysis of seedling community against either organic
horizon nutrients or organic horizon depth showed that
the seedling community was only correlated with the
nutrient matrix (results not shown).
Tree basal area, expressed both as absolute values
(m2 ha-1) and relative dominance (% of the total basal
area), was correlated significantly with mineral soil
properties (r = 0.94, p \ 0.05) and dune age (r =
0.85?0.87, p \ 0.05), but not organic horizon properties
(r = 0.43?0.47, p [ 0.05) (Table 3; Fig. 6). Similarly,
partial Mantel tests showed that tree basal area matrices
were correlated with ??pure mineral soil?? and ??pure
age,?? but not ??pure organic horizon?? (Table 3; Fig. 6).
Discussion
Changes in forest structure and composition
during ecosystem development
Clear changes in the forest community occurred along
the Haast chronosequence, including a marked decline
in basal area of trees C10 cm dbh and a shift in the
relative dominance of tree species. The decline in
basal area is consistent with the concept of forest
Table 3 Mantel and partial Mantel tests examining the relationship between floristic dissimilarity as a function of differences in soil
and organic horizon variables and age between dunes
Community Mantel tests Partial Mantel tests
Mineral
soil
Organic
horizon
Age ?Pure mineral?
(mineral soil | age)
?Pure organic?
(organic horizon | age)
?Pure age?
(age | soil)
Abundance
Trees (stems ha-1) 0.95* 0.45 0.78* 0.90* 0.52 0.28
Saplings
(stems ha-1)
0.42* 0.18 0.21 0.41 0.17 -0.17
Seedlings
(stems ha-1)
0.04 0.55** -0.11 0.12 0.58* -0.17
Basal area
Tree basal area
(m2 ha-1)
0.94* 0.47 0.87* 0.87* 0.66 0.72**
Relative dominance
(%)
0.94* 0.43 0.88* 0.87* 0.62 0.77**
Soil properties
Mineral soil
(0?20 cm)
? ? 0.74* ? ? ?
Organic horizon ? ? 0.17 ? ? ?
Values represent the Mantel test statistic (r) based on the Pearson?s product-moment correlation coefficient
* p \ 0.05, ** p \ 0.01
Plant Ecol (2012) 213:1525?1541 1535
123
??retrogression?? during long-term ecosystem develop-
ment (Wardle et al. 2004), although it also reflects in part
the transition from monodominant D. cupressinum
stands to a mixed conifer?angiosperm community.
Retrogression linked to a long-term reduction in phos-
phorus availability is a characteristic of chronose-
quences with a variety of climates and parent
materials and is distinct from shorter-term changes in
basal area linked to successional trends, such as age-
related declines in forest productivity, because it occurs
over multiple generations and is driven by long-term
pedological processes (Peltzer et al. 2010). Although
retrogression has been most commonly observed over
long time scales ([100,000 years), it can also occur over
relatively short time scales of a few thousand years
(Wardle et al. 1997) and is likely to have been driven at
Haast by the rapid rate of soil development and
associated phosphorus loss (Turner et al. 2012).
Basal areas reported here for the young dunes at Haast
are at the upper end of the range for D. cupressinum
stands in New Zealand, but are representative of young
dunes on similar sequences in the region (A. Wells,
personal observations). There was no bias in the
positioning of our plots, which were effectively located
randomly based on the intersection of the transect line
with the dune crests. However, it is possible that our
Fig. 6 Floristic similarity of tree, sapling, and seedling com-
munities (Y axes) for all pairs of dunes as a function of
differences in soil nutrients and age between the pairs. Floristic
similarity was based on species abundance per dune. The mineral
soil and organic horizon distance matrices were calculated using
the first three axes of PCA. * p\0.05, ** p\0.01
1536 Plant Ecol (2012) 213:1525?1541
123
transect passed through particularly dense stands of
D. cupressinum on the young dunes that are not
representative of the area as a whole. A previous study
of vegetation in a single 10 9 20 m plot on each of six
alternating dune?slack features along the Haast
sequence reported basal areas of stems[10 cm dbh up
to 78.5 m2 ha-1 and tree densities of 229?1,415
stems ha-1 at breast height (Dickinson and Mark
1994). We measured larger basal areas and stem
densities in our plots on young dunes dominated by emer-
gent D. cupressinum, although these were similar to
values reported previously for mixed conifer?angio-
sperm forest in the Westland region, which typically
contain 200?500 stems ha-1 of D. cupressinum[10 cm
dbh (Ogden and Stewart 1995). Basal area estimates of
D. cupressinum stands include up to 133 m2 ha-1 for
stems[5 cm dbh on moraines (Stewart et al. 1998).
Unlike the chronosequence on the Hawaiian Islands,
where a single tree species, Metrosideros polymorpha,
forms the dominant canopy tree across the entire
4.1 million year sequence and retrogression appears to
be clearly linked to nutrient status (Herbert and Fownes
1999), the changes in basal area along the Haast
sequence occurred in parallel with marked changes in
the tree community composition. Although this means
that the changes at Haast more closely represent how
most ecosystems respond to long-term soil development,
it complicates the assignment of the decline in basal area
to changes in nutrients alone, because other factors such
as population dynamics and dispersal limitation might
also be involved. For example, it has been suggested that
dispersal effects (distance from the seed source) play a
greater role than site age or soil development in shaping
species distribution along the early stages of the Glacier
Bay chronosequence (Fastie 1995). Given the relatively
limited age of the Haast sequence, the pronounced
changes in the forest community might be regulated by
the strong nutrient gradient, dispersal limitation, or
successional patterns unrelated to pedogenesis. Wider
patterns of forest succession and disturbance dynamics
over multiple generations, which may occur regardless
of changes in nutrient status, must also be considered
(discussed below).
Regional patterns of disturbance-related forest
succession
Disturbance is a key factor shaping forest communi-
ties in lowland conifer?angiosperm forests of New
Zealand (Duncan 1993; Ogden 1985; Veblen and
Stewart 1982; Wells et al. 2001). Large scale distur-
bances such as landslides, flooding, and volcanic
eruptions cause major soil disturbance and favor
extensive even-aged cohorts of either angiosperms or
conifers with high basal area and low seedling density
(Rose et al. 1992; Stewart and Rose 1990; Stewart
et al. 1998; Wells et al. 2001). Without further major
disturbance, the initial cohort collapses through
smaller scale (gap phase) canopy disturbances and is
replaced by a forest with a lower basal area but greater
compositional and structural diversity (Kershaw and
McGlone 1995; Ogden and Stewart 1995; Stewart and
Rose 1990; Veblen and Stewart 1982). This favors the
regeneration of fast growing angiosperms such as W.
racemosa and promotes shade tolerant conifer species
such as P. ferruginea, the most shade-tolerant of the
New Zealand podocarps (Ogden and Stewart 1995),
which regenerates in tree-fall gaps (Stewart et al.
1998) or even in the absence of gaps (Lusk and Smith
1998). Many of these successional changes are
directly related to the diversity and structural com-
plexity of micro-sites for seedling establishment,
reflected in part in the marked changes in ground
cover along the Haast sequence (see supplementary
material Fig. S1). Overall, the effect is that smaller
scale disturbances tend to lead to greater diversity in
site conditions and therefore a more diverse forest
community (e.g., Duncan 1993). This is reflected in
the composition of the tree community on old dunes at
Haast, which indicates that gap-phase disturbance has
been dominant for at least several centuries.
Dacrydium cupressinum appears to quickly colo-
nize freshly exposed sand dunes and then dominate the
basal area of the forest for hundreds of years (Dunes
1?6). This species is light demanding and presumably
well dispersed by birds (Norton et al. 1988). This is
followed by increasing dominance by angiosperms,
greater abundance of shade-tolerant podocarps, and a
much lower basal area (Dunes 8?17). There were very
few seedlings or saplings of D. cupressinum on any
dune, indicating the importance of open sites (perhaps
also associated with high nutrient availability) for this
species to become established. However, seedlings
and saplings of Phyllocladus alpinus, Podocarpus
hallii, and to a lesser extent Prumnopitys ferruginea,
were abundant on 1,800?3,900 year-old dunes. This
pattern may be linked to the presence of tree ferns,
which were abundant on young dunes, but virtually
Plant Ecol (2012) 213:1525?1541 1537
123
absent on older dunes. Tree ferns are relatively
nutrient demanding and are therefore common on
young fertile soils (Richardson et al. 2005), where they
appear to play a key role in suppressing podocarp
regeneration through intense shading (Coomes et al.
2005). Thus, changes in tree community composition
along the Haast chronosequence are broadly consistent
with the regional pattern of disturbance-related suc-
cession in lowland conifer?angiosperm forests.
Soil nutrients versus dispersal in the regulation
of tree community composition
Although the Haast chronosequence is characterized
by marked changes in soil nutrients with soil age
(Turner et al. 2012), the distribution of tree species
might also be influenced by dispersal limitation and
successional patterns unrelated to pedogenesis (Fastie
1995; Johnson and Miyanishi 2008). Given the
relatively short distances between dunes at Haast, it
might be expected that chance and dispersal limitation
are of minor significance in structuring the forest
community. However, dispersal is important in the
colonization of young sites, because well-dispersed
species (e.g. W. racemosa) are most likely to establish
quickly after the stabilization of a new dune. Even
species like N. menziesii, sometimes assumed to
disperse slowly (Wardle 1991), can colonize young
surfaces within a few years of stabilization (Wardle
1980). As N. menziesii has been present in forests
around the Haast chronosequence for the last several
thousand years (Li et al. 2008), it seems likely that it
had opportunity to colonize young dunes, suggesting
that factors other than dispersal have determined its
distribution along the chronosequence.
Fortunately, the close proximity of the Haast dunes,
despite marked differences in their ages, allows
examination of the relative importance of nutrients
and dispersal in determining the plant community
composition along the sequence, because the age
sequence can be used as a measure of geographic
distance. Dispersal limitation theory states that there
should be a log-linear relationship between distance
and floristic similarity (Condit et al. 2002), so at Haast
we would expect a similar relationship between dune
age and floristic similarity. The adult tree community
shows decreasing floristic similarity between dunes as
the age (i.e. distance) difference between them
increases. However, there was no relationship between
floristic similarity and difference in age of the dunes
for the sapling or seedling community, indicating that
processes other than dispersal limitation, such as soil
nutrients, may be important in filtering the plant
community.
The role of soil nutrients in structuring the plant
community at Haast is demonstrated clearly by the
results of the Mantel and partial Mantel tests, which
showed that tree species distribution is explained by
soil nutrients to a much greater extent than dispersal
(i.e. dune age). The significance of nutrients was
further emphasized by the finding that the abundance
and basal area of adult tree species (and to a lesser
extent sapling species abundance) along the sequence
were related strongly to nutrients in the mineral soil,
whereas the distribution of seedling species was
related strongly to nutrients in the organic horizon.
This presumably reflects differences in rooting depth:
adults are rooted in mineral soil throughout the
sequence, whereas seedlings are rooted predominantly
in the organic horizon, particularly on the older dunes
with thick organic horizons. Thus, seedlings obtain
nutrients from organic soil rather than the mineral soil,
especially during the early stages of their growth.
These results indicate strong environmental filtering
of plant communities at all life history stages based on
soil nutrient status.
Leaf litter is certainly known to exert species-
specific effects on seedlings regeneration in temperate
and tropical forests (Molofsky and Augspurger 1992;
Sydes and Grime 1981), but this is primarily related to
litter depth rather than nutrient content. Importantly,
there are few examples where adults and seedlings of
the same species show different habitat associations
based on nutrient availability. Webb and Peart (2000)
suggested that seedlings would be more strongly
related to a habitat than adults if seedling establishment
occurred in suboptimal habitats, which would indicate
the importance of negative density dependence
(Comita et al. 2010). However, seedlings and adults
at Haast are associated most strongly with nutrients in
the organic horizon and mineral soil, respectively,
corresponding to their likely patterns of rooting depth.
It therefore seems likely that seedlings are widely
dispersed, consistent with the relatively small distance
across the dune sequence, and that nutrient availability,
rather than density dependence, is the most likely
explanation for the observed patterns in plant commu-
nity composition along the chronosequence.
1538 Plant Ecol (2012) 213:1525?1541
123
Comparison of the Haast chronosequence
with the nearby Franz Josef chronosequence
Although the pattern of tree community composition
along the Haast chronosequence is consistent with that
following large-scale disturbance in Westland, it
differs markedly from the pattern of vegetation change
at the nearby Franz Josef chronosequence, despite a
similar overall species pool (although N. menziesii is
not present at Franz Josef). At Haast, podocarps
dominate the basal area on young dunes, with an
increasing proportion of the basal area formed by
woody angiosperms as soils age. At Franz Josef, in
contrast, young soils support only angiosperms,
D. cupressinum adults first appear on intermediate-
aged soils (12,000 years BP), and podocarps become a
major proportion of the tree community on only the
oldest soils (Richardson et al. 2004). Stevens (1968)
termed this pattern ??backward succession?? with
reference to the absence of podocarps on the high
glacial terraces in Westland, as opposed to ??forward
succession?? associated with decreasing fertility and
podzolisation. A similar pattern to Franz Josef occurs
along the Waitutu chronosequence (Coomes et al.
2005; Mark et al. 1988), although this is perhaps
linked to waterlogging on the older terraces, because
the podocarp species that occur there, including
Halocarpus biformis and Lepidothamnus intermedius,
appear to be flood-tolerant (Gaxiola et al. 2010). In
contrast, the dune soils at Haast are sandy and well
drained, even on older soils with an iron pan, so
waterlogging does not appear to be an important factor
in structuring the plant community along the sequence.
Soil nutrient status appears to play an important
role in structuring the vegetation communities along
New Zealand chronosequences. The modern distribu-
tion of podocarps in New Zealand is biased toward
harsh edaphic conditions (Leathwick 1995) and the
increasing abundance of podocarps as soils age along
the Franz Josef and Waitutu chronosequences has been
attributed to their superior ability to tolerate nutrient-
poor soils, being slow growing and having long-lived
leaves that retain nutrients for long periods (Coomes
et al. 2005; Richardson et al. 2005). Could differences
in nutrient status reconcile the differences in vegeta-
tion patterns between the Haast and Franz Josef
chronosequences? (Although Waitutu represents
another example of soil and forest change over a long
time scale, we exclude it here because the youngest
surfaces are *79,000 years old.) Phosphorus cer-
tainly seems to offer a potential explanation for the
presence of D. cupressinum along the two sequences.
The early stages of the Franz Josef chronosequence,
for example, contain much greater concentrations of
total phosphorus (*800 mg P kg-1) than early stages
of the Haast (*300 mg P kg-1) sequence (Parfitt
et al. 2005; Turner et al. 2012). This is a product of the
marked differences in parent material between the two
sequences: slightly weathered sand grains at Haast,
compared with unweathered coarse moraine at Franz
Josef. Along both sequences, however, D. cupressi-
num appears on soils with total phosphorus concen-
trations of *300 mg P kg-1; these are only a few
hundred years old at Haast, but *12,000 years old at
Franz Josef. In addition, soils with *100 mg P kg-1
of total phosphorus along both sequences support
canopies containing podocarps and woody angio-
sperms in approximately equal proportions.
We therefore conclude that pedogenic changes in
soil nutrients not only influence variation in tree
community composition along the Haast chronose-
quence, but might also help to reconcile differences in
community development among sequences. The link
between soil phosphorus and the distribution of tree
species warrants further investigation, perhaps through
studies on additional sequences in the region devel-
oped on contrasting parent materials.
Acknowledgments We thank Peter Bellingham and David
Coomes for helpful discussion. Kelly Andersen was supported
by a Smithsonian Institution Postdoctoral Fellowship. Funding
for travel and consumables was provided by Lincoln University.
References
Berryman KR, Cochran UA, Clark KJ, Biasi GP, Langridge RM,
Villamor P (2012) Major earthquakes occur regularly on an
isolated plate boundary fault. Science 336:1690?1693
Biffin E, Conran J, Lowe A (2011) Podocarp evolution: a
molecular phylogenetic perspective. In: Turner BL,
Cernusak LA (eds) Ecology of the Podocarpaeae in tropical
forests. Smithsonian Institution Scholarly Press, Wash-
ington, DC, pp 1?20
Comita LS, Muller-Landau HC, Aguilar S, Hubbell SP (2010)
Asymmetric density dependence shapes species abun-
dances in a tropical tree community. Science 329:330?332
Condit R, Pitman N, Leigh EG, Chave J, Terborgh J, Foster RB,
Nunez P, Aguilar S, Valencia R, Villa G, Muller-Landau
HC, Losos E, Hubbell SP (2002) Beta-diversity in tropical
forest trees. Science 295:666?669
Plant Ecol (2012) 213:1525?1541 1539
123
Coomes DA, Bellingham PJ (2011) Temperate and tropical
podocarps: how ecologically alike are they? In: Turner BL,
Cernusak LA (eds) Ecology of the Podocarpaceae in
tropical forests. Smithsonian Institution Scholarly Press,
Washington, DC, pp 119?140
Coomes DA, Allen RB, Bentley WA, Burrows LE, Canham CD,
Fagan L, Forsyth DM, Gaxiola-Alcantar A, Parfitt RL,
Ruscoe WA, Wardle DA, Wilson DJ, Wright EF (2005)
The hare, the tortoise and the crocodile: the ecology of
angiosperm dominance, conifer persistence and fern fil-
tering. J Ecol 93:918?935
Crews TE, Kitayama K, Fownes JH, Riley RH, Herbert DA,
Mueller-Dombois D, Vitousek PM (1995) Changes in soil
phosphorus fractions and ecosystem dynamics across a
long chronosequence in Hawaii. Ecology 76:1407?1424
Dickinson KJM, Mark AF (1994) Forest?wetland vegetation
patterns associated with a Holocene dune?slack sequence,
Haast Ecological District, south western New Zealand.
J Biogeogr 21:259?281
Duncan RP (1993) Flood disturbance and the coexistence of
species in a lowland podocarp forest, south Westland, New
Zealand. J Ecol 81:403?416
Eger A, Almond PC, Condron LM (2011) Pedogenesis, soil
mass balance, phosphorus dynamics and vegetation com-
munities across a Holocene soil chronosequence in a super-
humid climate, South Westland, New Zealand. Geoderma
163:185?196
Fastie CL (1995) Causes and ecosystem consequences of mul-
tiple pathways of primary succession at Glacier Bay,
Alaska. Ecology 76:1899?1916
Gaxiola A, McNeill SM, Coomes DA (2010) What drives ret-
rogressive succession? Plant strategies to tolerate infertile
and poorly drained soils. Funct Ecol 24:714?722
Hedin LO, Vitousek PM, Matson PA (2003) Nutrient losses over
four million years of tropical forest development. Ecology
84:2231?2255
Herbert DA, Fownes JH (1999) Forest productivity and effi-
ciency of resource use across a chronosequence of tropical
montane soils. Ecosystems 2:242?254
Hurst JM, Allen RB (2007) The RECCE method for describing
New Zealand vegetation?expanded manual. Version 4.
Landcare Research, Lincoln
Johnson EA, Miyanishi K (2008) Testing the assumptions of
chronosequences in succession. Ecol Lett 11:1?13
Kershaw AP, McGlone MS (1995) The quaternary history of the
southern conifers. In: Enright NJ, Hill RS (eds) Ecology of
the southern conifers. Smithsonian Institution Press,
Washington, DC, pp 30?63
Leathwick JR (1995) Climatic relationships of some New
Zealand forest tree species. J Veg Sci 6:237?248
Li X, Rapson GL, Flenley JR (2008) Holocene vegetational and
climatic history, Sponge Swamp, Haast, south-western
New Zealand. Quat Int 184:129?138
Lusk CH, Smith B (1998) Life history differences and tree
species coexistence in an old-growth New Zealand rain
forest. Ecology 79:795?806
Mark AF, Grealish G, Ward CM, Wilson JB (1988) Ecological
studies of a marine terrace sequence in the Waitutu Eco-
logical District of southern New Zealand. Part 1: The
vegetation and soil patterns. J R Soc N Z 18:29?58
Molofsky J, Augspurger CK (1992) The effect of leaf litter on
early seed establishment in a tropical forest. Ecology
73:68?77
New Zealand Meteorological Service (1983) Summaries of
climatological observations to 1980. New Zealand Mete-
orological Service Miscellaneous Publication 177
Norton DA, Herbert JW, Beveridge AE (1988) The ecology of
Dacrydium cupressinum: a review. N Z J Bot 26:37?62
Ogden J (1985) An introduction to plant demography with
special reference to New Zealand trees. N Z J Bot
23:751?772
Ogden J, Stewart GH (1995) Community dynamics of the New
Zealand conifers. In: Enright NJ, Hill RS (eds) Ecology of
the southern conifers. Smithsonian Institution Press,
Washington, DC, pp 81?119
Oksanen J (2010) Multivariate analysis of ecological commu-
nities in R: vegan tutorial
Palmer RWP, Doyle RB, Grealish GJ, Almond PC (1986) Soil
studies in south Westland, 1984/1985. Soil Bureau District
Office Report NP 2. Department of Scientific and Industrial
Research, New Zealand
Parfitt RL, Ross DJ, Coomes DA, Richardson SJ, Smale MC,
Dahlgren RA (2005) N and P in New Zealand soil chron-
osequences and relationships with foliar N and P. Bio-
geochemistry 75:305?328
Peltzer DA, Wardle DA, Allison VJ, Baisden WT, Bardgett RD,
Chadwick OA, Condron LM, Parfitt RL, Porder S, Rich-
ardson SJ, Turner BL, Vitousek PM, Walker J, Walker LR
(2010) Understanding ecosystem retrogression. Ecol Mo-
nogr 80:509?529
Richardson SJ, Peltzer DA, Allen RB, McGlone MS, Parfitt RL
(2004) Rapid development of phosphorus limitation in
temperate rainforest along the Franz Josef soil chronose-
quence. Oecologia 139:267?276
Richardson SJ, Peltzer DA, Allen RB, McGlone MS (2005)
Resorption proficiency along a chronosequence: responses
among communities and within species. Ecology 86:20?25
Rose AB, Pekelharing CJ, Platt KH (1992) Magnitude of canopy
dieback and implications for conservation of southern rata-
kamahi (Metrosideros umbellata?Weinmannia racemosa)
forests, central Westland, New Zealand. N Z J Ecol
16:23?32
Stevens PR (1968) A chronosequence of soils near the Franz
Josef Glacier. Lincoln College, University of Canterbury,
Canterbury
Stewart GH, Rose AB (1990) The significance of life history
strategies in the developmental history of mixed beech
(Nothofagus) forests, New Zealand. Vegetatio 87:101?114
Stewart GH, White JC, Duncan RP (1998) The structure and
dynamics of rimu-dominated forests on glacial moraines,
South Westland. N Z For 42:22?27
Sydes C, Grime JP (1981) Effects of tree leaf litter on herba-
ceous vegetation in deciduous woodland. J Ecol 69:
249?262
Turner BL, Condron LM, Richardson SJ, Peltzer DA, Allison VJ
(2007) Soil organic phosphorus transformations during
pedogenesis. Ecosystems 10:1166?1181
Turner BL, Condron LM, Wells A, Andersen KM (2012) Soil
nutrient dynamics during podzol development under low-
land temperate rain forest. Catena 97:50?62
1540 Plant Ecol (2012) 213:1525?1541
123
Veblen TT, Stewart GH (1982) On the conifer regeneration gap
in New Zealand: the dynamics of Libocedrus bidwilli
stands on South Island. J Ecol 70:413?436
Walker TW, Syers JK (1976) The fate of phosphorus during
pedogenesis. Geoderma 15:1?19
Walker J, Thompson CH, Fergus IF, Tunstall BR (1981) Plant
succession and soil development in coastal sand dunes of
subtropical eastern Australia. In: West DC, Shugart HH,
Botkin DB (eds) Forest succession: concepts and applica-
tion. Springer, New York, pp 107?131
Walker LR, Wardle DA, Bardgett RD, Clarkson BD (2010) The
use of chronosequences in studies of ecological succession
and soil development. J Ecol 98:725?736
Wardle P (1980) Primary succession in Westland National Park
and its vicinity, New Zealand. N Z J Bot 18:221?232
Wardle P (1991) Vegetation of New Zealand. Cambridge Uni-
versity Press, Cambridge
Wardle DA, Zackrisson O, Ho?rnberg G, Gallet C (1997) The
influence of island area on ecosystem properties. Science
277:1296?1299
Wardle DA, Walker LR, Bardgett RD (2004) Ecosystem prop-
erties and forest decline in contrasting long-term chrono-
sequences. Science 305:509?513
Wardle DA, Bardgett RD, Walker LR, Peltzer DA, Lagerstro?m
A (2008) The response of plant diversity to ecosystem
retrogression: evidence from contrasting long-term
chronosequences. Oikos 117:93?103
Webb OW, Peart DR (2000) Habitat associations of trees and
seedlings in a Bornean rain forest. J Ecol 88:464?478
Wells A, Goff J (2006) Coastal dune ridge systems as chrono-
logical markers of palaeoseismic activity: a 650-yr record
from southwest New Zealand. Holocene 16:543?550
Wells A, Goff J (2007) Coastal dunes in Westland, New Zea-
land, provide a record of paleoseismic activity on the
Alpine fault. Geology 35:731?734
Wells A, Duncan RP, Stewart GH (2001) Forest dynamics in
Westland, New Zealand: the importance of large, infrequent
earthquake-induced disturbance. J Ecol 89:1006?1018
Plant Ecol (2012) 213:1525?1541 1541
123