Annals of Botany 109: 299 –307, 2012
doi:10.1093/aob/mcr260, available online at www.aob.oxfordjournals.org
Soil seed bank recovery occurs more rapidly than expected in semi-arid
Mediterranean gypsum vegetation
J. M. Olano1,*, I. Caballero1 and A. Escudero2
1
Área de Botánica, Departamento de Ciencias Agroforestales, Escuela de Ingenierı́as Agrarias, Univ. de Valladolid, Los
Pajaritos s/n, ES-42003 Soria, Spain and 2Área de Biodiversidad y Conservación, Departamento de Biologı́a & Geologı́a,
ESCET, Univ. Rey Juan Carlos, C/ Tulipán s/n, ES-28933 Móstoles, Spain
* For correspondence. E-mail [email protected]
Received: 21 June 2011 Returned for revision: 25 July 2011 Accepted: 5 September 2011 Published electronically: 16 October 2011
† Background and Aims Seed banks are critical in arid ecosystems and ensure the persistence of species. Despite
the importance of seed banks, knowledge about their formation and the extent to which a seed bank can recover
after severe perturbation remains scarce. If undisturbed, soil seed banks reflect a long vegetation history; therefore, we would expect that new soil seed banks and those of undisturbed soils require long periods to become
similar with respect to both density and composition. In contrast, if soil seed banks are only a short- to midterm reservoir in which long-term accumulation constitutes only a tiny fraction, they will recover rapidly from
the vegetation. To shed light on this question, we evaluated seed bank formation in a semi-arid gypsum
community.
† Methods Soils from 300 plots were replaced with sterilized soil in an undisturbed semi-arid Mediterranean community. Seasonal changes in seed bank density and composition were monitored for 3 years by comparing paired
sterilized and control soil samples at each plot.
† Key Results Differences in seed bank density between sterilized and control soil disappeared after 18 months.
The composition of sterilized seed banks was correlated with that of the control plots from the first sampling date,
and both were highly correlated with vegetation. Nearly 24 % of the seed bank density could be attributed to
secondary dispersal. Most seeds died before emergence (66.41 –71.33 %), whereas the rest either emerged
(14.08–15.48 %) or persisted in the soil (14.59 –18.11 %).
† Conclusions Seed banks can recover very rapidly even under the limiting and stressful conditions of semi-arid
environments. This recovery is based mainly on the seed rain at small scales together with secondary dispersal
from intact seed banks in the vicinity. These results emphasize the relevance of processes occurring on short
spatial scales in determining community structure.
Key words: Gypsum vegetation, secondary dispersal, seed bank formation, seed fate, semi-arid environment.
IN T RO DU C T IO N
Given that seeds can persist in the soil for a long time before
germination (Laskurain et al., 2004), soil seed banks can
promote community stability by reducing environmental stochasticity in highly unpredictable ecosystems (Fenner, 1995)
and lead to the coexistence of species through the well-known
storage effect (Chesson, 2000). The prevailing paradigm is that
persistent seed banks guarantee the persistence of plant species
(Kalisz and McPeek, 1992) even when environmental conditions become extreme enough to kill all standing individuals.
Consequently, seed banks are well developed in ecosystems
where recurrent drastic perturbations such as wildfire
(Valbuena et al., 2000) or floods (Hölzel and Otte, 2004) are
responsible for the high mortality of standing individuals or
in stressful environments where recruitment conditions are
unpredictable (Kemp, 1989). These seed banks would constitute the memory of the vegetation history (Templeton and
Levin, 1979) because they accumulate seeds produced over
extended periods (Milberg, 1995; Bekker et al., 1999) and
also serve as a genetic reservoir of past selective events
(Morris et al., 2002) that can slow the rate of evolutionary
change (Nunney, 2002).
A soil seed bank is a highly dynamic community compartment (Fenner, 1995) structured at various spatial (Lortie et al.,
2005) and temporal scales (Bertiller, 1992; Caballero et al.,
2008a). Seed bank dynamics are the complex result of processes determining losses such as seed predation, seed decay
and death, deep seed burial to layers from where emergence
onto the soil surface is improbable and germination, and
gains through plant seed production, together with primary
and secondary dispersal. The relative importance of, and interactions between, these processes may determine changes in
seed density and composition over time, which under
extreme circumstances may become dramatic (Gutiérrez
et al., 2000). The existence of abrupt changes in density and
composition, even on very small temporal scales, apparently
contradicts the basic role of the seed bank to ensure local
population persistence. However, the seed bank and standing
vegetation maintain a close relationship on contrasting
spatial and time scales, suggesting that seed banks contribute
to population persistence despite rapid fluctuations in density
and composition (Olano et al., 2005; Caballero et al., 2008a).
Despite previous findings regarding the spatial and temporal
patterns of soil seed banks, knowledge about seed bank formation and the extent to which a seed bank can recover after
# The Author 2011. Published by Oxford University Press on behalf of the Annals of Botany Company. All rights reserved.
For Permissions, please email: [email protected]
300
Olano et al. — Soil seed bank recovery in semi-arid vegetation
severe perturbation remains incomplete (Hyatt and Casper,
2000; Luzuriaga et al., 2005). If soil seed banks reflect a
long vegetation history, we would expect that new soil seed
banks require long periods to become similar to those of undisturbed soils with respect to both density and composition,
especially in stressful environments where plant performance
and community dynamics are known to be slow (Miriti
et al., 2007). In contrast, if soil seed banks are mainly
shaped by annual seed rain and are only a short- to mid-term
reservoir in which long-term accumulation constitutes only a
tiny fraction, both transient and persistent seed banks will
recover rapidly from vegetation seed rain.
To test these two opposing hypotheses, we removed soil
seed banks from 300 locations with minimum perturbation
and replaced the soil after sterilization. We did not aim to
evaluate the role and genesis of soil seed banks after large disturbances as such an experimental set-up would constitute a
typical scenario for primary succession research. Our aim
here was to evaluate seed bank recovery after a small perturbation with surrounding areas being intact to gain insight
into soil seed bank formation and dynamics. A systematic
comparison of the new seed banks of these soils with paired
control soil seed banks was conducted over a 3-year period.
In addition, we monitored the surrounding vegetation over
small spatial scales over this 3-year period and monitored all
the seedlings emerging from the sterilized soil. This sampling
schema allowed the evaluation of seed gains according to
primary and secondary dispersal as well as losses and their
causes in order to determine their relative importance in the
formation of the new seed bank. Our model ecosystem was a
gypsum semi-arid community, which has a well-known and
diverse seed bank with shifts in composition and density at
contrasting spatial and temporal scales from yearly fluctuations
to shifts along secondary succession (Caballero et al., 2003,
2005, 2008a, b; Olano et al., 2005; Martı́nez-Duro et al.,
2009a). Specifically, we aimed to clarify how new soil seed
banks are formed, their relationship to vegetation, and
whether the composition similarity between vegetation and
seed bank increases with time.
M E T H O DS
Study site
Our field experiment was conducted at Belinchón
(Cuenca province), 80 km south-east of Madrid, in central
Spain (4083′ 33.15′′ N, 383′ 3.67′′ W, 720 m a.s.l.). The climate
is upper semi-arid meso-mediterranean (Rivas-Martinez and
Loidi, 1997) with an annual rainfall of 441 mm but with an
extreme summer drought (only a 5.6 % of the annual rainfall
occurs during July and August). Mean daily maximum and
minimum temperatures are, respectively, 10.6 and – 0.8 8C in
January and 33.5 and 14.5 8C in July. The year 2004 had a
very wet rainy season (449.8 mm of rainfall from previous
October to June), whereas 2005 and 2006 were dry years
(annual rainfall: 176.8 and 212.2 mm, respectively). The
soils were classified as calcic gypsisols that developed over
gypsum parental rocks (Monturiol and Alcalá del Olmo,
1990). Gypsum outcrops are characterized by a succession of
small hills with steep slopes. Our working area was situated
in a gently sloping area of private land with a shrub community dominated by genuine gypsophytes such as
Helianthemum squamatum and Lepidium subulatum; in
addition, there were some gypso-nitrophilous perennials,
such as Frankenia thymifolia, which are characteristic of the
plant communities in the lower part ( piedmont) of gypsum
slopes (Rubio and Escudero, 2000). Shrubs were interspersed
within open areas with a rich annual plant community with
perennial ground cover projection rarely surpassing 15 %.
The area was exposed to light, occasional sheep grazing.
Sampling design
In September 2003, we selected a representative rectangular
34 × 36-m area of gypsum community on a gentle slope. We
established a 17 × 18-m grid perpendicular to the maximum
slope, providing 306 grid cells (2.25 m2). In the centre of
300 cells, we established a 0.25-m2 (0.5 × 0.5-m) sampling
plot. We placed a cylinder (6.3 cm diameter, 15 cm deep) in
the middle of each sampling plot. The soil was then
removed and discarded. The cylinder was later refilled with
soil from the vicinity that had been sterilized in an autoclave
at 120 8C for 70 min to kill all seeds and propagules. The
cylinders had a 0.1-mm mesh in the bottom to separate the
sterilized soil from the untreated soil below whilst still maintaining contact between the two and allowing free seasonal
vertical water movement. As our goal was to evaluate seed
bank recovery on a small scale by maintaining ecosystem function and dynamics, we selected a relatively small cylinder size.
Soil seed bank monitoring started in April 2004, 6 months
after the onset of the experiment. Cylinders with sterilized
soil were extracted from 50 randomly selected plots and kept
in a paper bag. Intact soil seed banks were evaluated in the
same plots by extracting three soil cores (2.85 cm diameter)
and 5 cm depth at random around each cylinder, merging
them, and keeping them in another paper bag. As we expected
a much lower seed bank density in sterilized soil, the sampled
area was smaller for soil cores than for cylinders. Differences
in sample depth were not relevant because seed bank in arid
environments is in the first 3 cm of soil (Childs and Goodall,
1973). In addition, we selected three small soil cores surrounding the cylinders instead of a large core in order to better
capture the seed bank composition surrounding the sterilized
cylinders. Subsequent samplings were performed in
September and April. The last sampling took place in
September 2006. As three cylinders were lost in April 2006
and two in September 2006, the sampling sizes on those
dates were reduced to 47 and 48, respectively.
During the flowering peak of annuals in April, we also evaluated the percentage cover of all species present in the standing vegetation in the sampling plot around each cylinder. We
counted and identified at the species level all seedlings that
emerged within each cylinder to estimate seed bank losses
due to emergence.
Vegetation and seedling surveys were only performed
where cylinders had not been extracted at previous sampling
dates. Thus, we sampled 300 plots in April 2004, 200 in
April 2005 and 95 in April 2006. Our vegetation model
system exhibited strong seasonality with emergences occurring from late autumn to spring, peaking in early spring
Olano et al. — Soil seed bank recovery in semi-arid vegetation
(Escudero et al., 1999, 2000); seed rain occurred almost
exclusively in late spring and summer (more than 95 % of
released seeds; Caballero, 2006). Accordingly, we sampled
the soil seed banks at two dates: in September once the
seed rain had ended, before the start of the autumn rains
and subsequent emergence, and in late April just before
the seed rain and after the emergence period. As a consequence, the September sampling corresponded to transient
and persistent seed bank components, whereas the April
sampling was an estimate of the persistent seed bank.
Hereafter, the April and September seed banks are referred
to as the persistent and total seed banks, respectively. We
are aware that some of the seeds detected in the April monitoring correspond to seeds recently shed by a reduced group
of early-flowering plants such as Erophila verna, Clypeola
jonthlaspi and Hornungia petraea. However, from a previous seed rain experiment in this system, this fraction
falls below 1 % (see Caballero, 2006); consequently, we
think this seed bank can be considered an adequate reflection of the persistent bank.
We obtained data on the viable seed bank by monitoring
seedling emergence in a greenhouse. Soil samples were kept
at 4 8C for 2 months and were subsequently washed and
sieved through a 0.5-cm mesh to retain the coarse fraction.
The resulting material was sieved again through a 0.1-mm
mesh to reduce the fine material volume and scarify the
seeds, which favours germination. The resulting soil samples
were placed in trays lying on 8 × 8 -cm perlite-filled plastic
pots in a greenhouse. To improve the water-carrying capacity
of soil samples, a sterile substratum of vermiculite and peat
(2 : 1) was added and thoroughly mixed with the sampled
soil until a 1-cm depth was reached. Ten pots were filled
with the sterile substrate only and placed in the greenhouse
to detect contamination. Ten additional pots were filled with
the sterilized soil to detect the validity of the sterilization treatment; their positions were regularly changed to prevent any
localization effects. Each seedling was identified and
removed as soon as it emerged. When species-level identification was not feasible, seedlings were transplanted into individual pots and allowed to grow until identification was
possible. After 13 months, the soil in the pots was crumbled
to enhance germination, and emergence was monitored for
another 3 months. Finally, the pots were watered with a gibberellic acid (GA3) solution (1000 p.p.m.) to stimulate the emergence of seeds with physiological dormancy. Emergence was
monitored for another 2 months. This procedure is based on
that described by ter Heerdt et al. (1996), but we surveyed
the samples for 18 months and included the GA3 treatment
to guarantee the best possible estimation of the viable seed
bank (Olano et al., 2002). We are also aware that no seed
bank analysis is able to provide an exact value of the viable
seed bank as some species may exhibit emergence cues that
are not fulfilled in greenhouses (Baskin and Baskin, 1998).
Nevertheless, our approach combining several dormancybreaking tools, such as the mechanical abrasion linked to
sieve manipulation and the use of phytohormones to break secondary dormancy, uses the best available methods to fulfil
these requirements. No emergence was detected in the
control pots with sterilized soil, thus confirming the effectiveness of the sterilization treatment.
301
Statistical analysis
Linear models were built to compare seed bank density
between cylinders (sterilized soil) and intact field soil
samples (hereafter referred to as controls) at different sampling
dates. As seed density fit normality assumptions after being log
transformed, we used identity as link function and assumed a
normal distribution of errors.
Mantel tests were used to evaluate whether control and
cylinder seed bank compositions were similar (converged) or
not over time. This procedure compares two similarity
matrices computed for the same objects (Legendre and
Legendre, 1998) by evaluating the sum of the cross-products
between both matrices. The obtained standardized Mantel statistic (Mantel R or RM) can be interpreted as parametric Pearson
correlations among similarity indices. As cells in a similarity
matrix are not independent of each other, significance was
tested by the permutations of the n objects that originally
created the n × n matrix. The significance of the observed correlation is the proportion of such permutations that leads to a
higher correlation coefficient. The individual seed counts and
vegetation cover were not normally distributed; thus, the original values were log transformed before the matrices were constructed on the basis of Euclidean distance. Mantel statistical
significance was tested using a randomization approach with
999 permutations. We compared each cylinder with its adjacent intact control at each of the sampling dates. Thus, the
Mantel test examined whether sterilized cylinders were more
similar to their neighbouring seed banks than expected by
chance as well as the intensity of such relationship.
Unfortunately, we had a labelling problem at the last sampling
date (September 2006) that prevented us from comparing seed
bank composition between both compartments for that date.
In addition, we performed Mantel tests to evaluate the
relationship between the composition of vegetation and both
cylinder and control seed banks. The April vegetation distance
matrix was built using log-transformed cover data and
Euclidean distance. Separate tests were also performed to
compare cylinders and controls in 2004 and 2005.
We used ordinations to evaluate the trajectory of seed bank
composition over time as well as the relative importance of
season, time since the start of the experiment and sterilization
treatment. To select an appropriate ordination technique, the
seed matrix was subjected to a detrended correspondence
analysis (DCA; ter Braak, 1986), with detrending by segments
and non-linear rescaling of the axes. We followed the recommendations of ter Braak (1986) and conducted a canonical
correspondence analysis (CCA) because the lengths of the
extracted gradients of the seed bank dataset were greater
than three standard deviation units. The constraining matrix
included season (April or September) and sterilization treatment as dummy variables and time as a continuous variable.
We also included an interaction term among time and treatment to evaluate whether temporal sequence differed among
sterilized conditions. The total variation explained (TVE)
was calculated as the relationship between the trace and sum
of all canonical extracted axes. A Monte Carlo permutation
test was performed to determine the accuracy of the relationship (999 randomizations) between the two datasets. The
sum of all canonical eigenvalues or traces was used to
302
Olano et al. — Soil seed bank recovery in semi-arid vegetation
calculate the F-ratio statistic (ter Braak, 1986). The relationship between the two data sets was considered significant
only when P , 0.05, adjusted for multiple comparisons by
the Bonferroni correction. If the corresponding CCA model
was significant, a forward stepwise procedure was carried out
to select a reduced model including only the significant variables of the constraining matrix. Explanatory variables were
incorporated one at a time and step by step in the order of
decreasing eigenvalues after removing the variation accounted
for by the variables already included. The process was concluded when the new variable was not significant (P .
0.05). Improvement of the reduced model with each new
selected variable was determined by a Monte Carlo permutation test with 999 randomizations. The term l indicates the
eigenvalue and thus the canonical correlations between the
optimally scaled categories. All analyses were conducted
using CANOCO for Windows v. 4.0 (ter Braak and
Smilauer, 1997).
Seed bank dynamics
Our sampling schema can estimate the dynamics of the
different components of seed banks. Seed rain occurring
between April and September was incorporated into transient
seed bank. Once the seeds were in the soil, they could
emerge, remain in the seed bank or die. This can be described
by the following equation:
Total seed bankt−1 = Emergencet + Persistent seed bankt
+ Dead seedst
(1)
Emergence losses could be estimated from the seedling counts
in the cylinders; the April seed bank mostly corresponded to
the persistent seed bank, and the previous September seed
bank corresponded to the sum of the persistent and transient
seed banks. Thus, seed losses could be calculated as a subtraction between the two seed banks. Seed losses were the sum of
emergences based on field counts and pre-emergence mortality. Therefore, we were able to calculate seed fate for the
years 2005 and 2006.
60
***
*
April 2004
Sept. 2004
Although this model does not incorporate seed losses or
gains due to secondary seed dispersal, given the gentle slope
of the study area, secondary dispersal outside the grid was
probably low (Garcı́a Fayos et al., 2010). At the cylinder
level, as soon as the seed bank density of sterilized cylinders
equalled that of the controls (April 2005), we expected that
seed bank gains and losses by dispersal were the same.
However, the April 2004 seed bank and emergences in the
cylinders could be attributed mostly to secondary dispersal
since our experiment started in September 2003 after the
seed rain peak.
R E S U LT S
Density
A total of 24 472 seedlings were identified in the greenhouse
emerging from 590 soil samples: 10 585 seedlings emerged
from the 295 control samples and 13 887 emerged from the
295 cylinders belonging to at least 66 and 68 species, respectively. A total of 6335 seedlings from 61 species emerged
within the cylinders in the field during the study period. The
number of plant species reached 84 in the 595 0.25-m2 vegetation samples (see Supplementary Data, available online,
for a detailed list). The density in the intact control soils
ranged from 48 301 seeds m22 in September 2004 after an
exceptionally wet growing season to 4138 seeds m22 in
April 2004 immediately after the onset of our experiment,
thus reflecting strong seasonality in seed bank density
(Fig. 1). Seed bank density in the controls showed a clear
decreasing trend after a maximum in September 2004.
Sterilized cylinders exhibited a similar pattern but with
lower densities for the first two sampling dates. Differences
among cylinders and control seed bank densities were
evident during 2004 (April: F ¼ 31.107, P , 0.001;
September: F ¼ 6.074, P ¼ 0.015) but disappeared as early
as April 2005 and on subsequent sampling dates.
Composition
The composition of sterilized and control seed banks exhibited a significant Mantel correlation (Fig. 2). This correlation
Control
Treatment
Seeds (m–2 × 1000)
50
40
30
20
10
0
April 2005
Sept. 2005
April 2006
Sept. 2006
F I G . 1. Seed bank density over the study period in controls and cylinders (+ s.e.). Significant differences between control and cylinders for each sampling date
based on ANOVA are shown: ***P , 0.001, *P , 0.05.
Olano et al. — Soil seed bank recovery in semi-arid vegetation
303
0·8
0·7
Mantel r value
0·6
0·5
0·4
0·3
0·2
0·1
0·0
April 2004
Sept. 2004
April 2005
Sept. 2005
April 2006
F I G . 2. Mantel test comparing seed bank composition between controls and cylinders over time. All correlations are significant at P ¼ 0.001, except that of April
2004 (P ¼ 0.03). Error bars are Mantel r confidence intervals.
1·0
Control
Cylinder
0·8
Apr 2005
Apr 2004
CA second axis
0·6
Apr 2005
0·4
Apr 2004
Apr 2006
0·2
Sep 2004
0
–0·2
Sep 2006
–0·4
–0·4
–0·2
Apr 2006
Sep 2005
Sep 2004
Sep 2005
0
Sep 2006
0·2
0·4
0·6
0·8
1·0
1·2
CA first axis
F I G . 3. Correspondence analysis biplot showing the centroid position and error bars for control and cylinder seed bank compositions, as indicated.
was significant as early as the first sampling date in April 2004.
The Mantel correlation was higher for September (total) seed
banks than for April ( persistent) seed banks and reached a
maximum in September 2005.
The September seed bank composition was correlated with
previous April vegetation in cylinders (2004: rm ¼ 0.368,
P ¼ 0.003; 2005: rm ¼ 0.425, P ¼ 0.001) and in controls
(2004: rm ¼ 0.359, P ¼ 0.001; 2005: rm ¼ 0.406, P ¼
0.002). Both seed banks showed similar values for Mantel correlation with vegetation but with strong inter-annual variation,
with higher values in 2005.
CA biplots clearly show the differences between September
and April seed banks as well as the relatively close match
between cylinders and control seed banks (Fig. 3). Seed
bank composition was related mainly to season (l ¼ 0.10;
F ¼ 19.01, P , 0.001) followed by time since the start of
the experiment (l ¼ 0.05; F ¼ 9.69, P , 0.001); sterilization
treatment played a significant but much smaller role (l ¼
0.02; F ¼ 2.68, P , 0.001). Moreover, no effect of time ×
sterilization treatment was detected (l , 0.01; F ¼ 1.18,
P ¼ 0.201), indicating no differences in the trajectories
between sterilized and undisturbed seed banks. The three significant predictors accounted for a relatively low (5.5 %) but
highly significant (P , 0.001) proportion of the total variance,
which suggests that other determinants may control the variation in the data set. Nevertheless, a low variance is expected
when applying CCA to data sets with a large number of
samples and species.
Seed bank dynamics
At the first sampling date (April 2004), the cylinders had
2165 emerged seedlings m22 and the seed bank density was
988 seeds m22. As no seeds were produced between
304
Olano et al. — Soil seed bank recovery in semi-arid vegetation
September 2003 and April 2004, their presence in the cylinders could be attributed exclusively to secondary dispersal.
This is an underestimation because secondary dispersal
accounts for an undetermined but probably large number of
seeds that died before the sampling date without being
detected, as mortality accounted for approx. 2 – 2.5 times
more than the sum of emerged seedlings in the April ( persistent) seed banks in 2005 and 2006. Nevertheless, the ratio
between April 2004 cylinders and control seed banks (23.88
%) can serve as a rough estimate of the relative magnitude
of secondary dispersal.
We evaluated the different components of cylinder seed
fates in 2005 and 2006 using our additive model (Fig. 4). A
significant number of seeds present in the September seed
bank emerged the following April (2005: 14.08 %; 2006:
15.48 %), and a slightly larger quantity persisted in the seed
bank (2005: 14.59 %; 2006: 18.11 %). Nevertheless, the
majority of the seeds that presented in the September seed
bank died before emergence between September and April
(2005: 71.33 %; 2006: 66.41 %).
Seed rain could be estimated as the difference between the
current year’s September and April seed banks. The seed
rain in 2004 was much larger in the controls (44 017 seeds
m22) than in the cylinders (33 574 seeds m22). However,
almost no differences were observed between cylinders and
controls in 2005 (cylinders: 22 806 seeds m22, control:
22 248 seeds m22) and 2006 (cylinders: 10 409 seeds m22,
control: 10 717 seeds m22).
DISCUSSION
Contrary to our expectations, the results show that soil seed
bank recovery under the harsh conditions of the Iberian
gypsum steppes is very rapid. We found no differences in
seed bank density 18 months after the onset of the experiment.
Furthermore, differences in the density and composition
between control and sterilized cylinders diminished rapidly.
Arid and semi-arid environments are characterized by low
rainfall in addition to high temporal variability in precipitation
(Sher et al., 2004), with plant performance being strongly
dependent on water availability. As a consequence, the pace
of seed bank recovery may be strongly dependent on the particular sequence of dry and wet years occurring after the
destruction of the soil seed bank. In our case, the velocity of
this process may have been at least a partial consequence of
a very humid spring at the onset of the experiment in 2004.
Nevertheless, seed bank recovery and similarity to the
control seed bank were maintained in subsequent dry years.
Limited seed dispersal appears to be the standard in arid
environments (Ellner and Shmida, 1981; Gutterman and
Shem-Tov, 1996; Venable et al., 2008), suggesting that the
seed bank is mainly formed by direct seed rain from vegetation
at small spatial scales. This idea is concordant with our previous observational studies in gypsum environments in
which we repeatedly found a close relationship between both
community compartments, seed banks and vegetation at
small scales (Olano et al., 2005; Caballero et al., 2008a).
Secondary dispersal, particularly via run-off, is considered
to be a key factor in seed bank spatial patterning and composition (Garcı́a-Fayos et al., 2010). Soil seed bank heterogeneity
in arid environments is attributed to the accumulation of seeds
in favourable microenvironments (Caballero et al., 2008b; Li,
2008) with relevant effects in the subsequent plant life stage
community structure (Thompson and Katul, 2009). Although
it is necessary to disentangle the relative contributions of secondary dispersal and direct seed rain to understand seed bank
formation, this is rarely done (Martı́nez-Duro et al., 2009b).
Our experiment allows an estimate of secondary dispersal to
be obtained and indicates that nearly 25 % of the seed bank
comes from lateral movements. Furthermore, the significant
correlation regarding seed bank composition between control
and sterilized cylinders in April 2004 indicates that run-off
processes do not involve long-distance movements but rather
only small-scale horizontal displacements from the vicinity.
We detected dramatic seasonal density losses affecting
82– 85 % of the transient seed bank, which appears to be
common in gypsum environments (Caballero et al., 2005,
2008a) and arid ecosystems in general (Kemp, 1989;
Gutiérrez and Meserve, 2003). A significant fraction of
these seeds was accounted for by emergence (14 – 15 %
4865 ± 355
2165 ± 212
4312 ± 502
(14·1 %)
33 574 ± 4016
Emergence
Emergence
Seed rain
Emergence
Seed rain
5042 ± 713
994 ± 153
Survival
(14·6 %)
5610 ± 1123
Survival
Mortality
2004
(15·5 %)
22 806 ± 3222
2005
(18·1 %)
Mortality
24 654 ± 4091
18 144 ± 4091
(71·3 %)
(66·4 %)
2006
F I G . 4. Flowchart of seed bank dynamics in experimental cylinders. The April seed bank (boxed) is considered persistent. Entrances (seed rain and previous
April persistent seed bank) are shown with grey arrows (dotted for April seed bank). Exits (emergence and deaths) are depicted by black arrows (dotted for
emergence). Values are seed number m22 + s.e.; percentages are calculated with respect to values for entrances (seed rain + previous April seed bank).
Olano et al. — Soil seed bank recovery in semi-arid vegetation
from the September seed bank), which is higher than the
magnitudes reported for perennials in this system
(Quintana-Ascencio et al., 2009; Martı́nez-Duro et al.,
2009b). Nevertheless, the contribution of emergence to
seed losses seems to be modest at best. We are aware of
the possibility of underestimating real germination because
of our inability to detect emergence failure after germination
(Baskin and Baskin, 1998). Although this fact could
increase the relative importance of emergence, most of the
seeds in the September seed bank (66 – 71 %) were lost
due to processes related to pre-emergence mortality.
Pre-emergence mortality in semi-arid environments is attributed to seed predation due to vertebrates (Kelt et al., 2004)
or invertebrates (Martı́nez-Duro et al., 2010) and fungal
attacks (Meyer et al., 2010).
Our results strongly suggest that the average persistence
period in the studied seed bank is rather low, a finding that
contradicts the consensus regarding the existence of a longterm seed bank in semi-arid environments. This widely
accepted idea is based on the existence of large persistent
seed banks and the results of burial experiments that indicate
that the potential longevity of individual seeds is large
(Saatkamp et al., 2009). However, our unexpected results
seem to be corroborated by two additional pieces of evidence:
(1) the low ratio between persistent and total seed banks at
community and individual species levels (see Caballero
et al., 2005, 2008a and Supplementary Data, available
online), indicating that average seed life is shorter than
expected; and (2) the only study where seed bank age in arid
environments has been effectively evaluated using 14C
(Moriuchi et al., 2000) found a seed bank age structure dominated by seeds from the preceding season.
In this sense, the seed bank age structure in many arid ecosystems may be younger than that in forested ecosystems
where long-term seed banks are common (Olano et al.,
2002). This disparity may reflect the existence of frequent
windows of opportunity for effective recruitment in many
arid environments (Moriuchi et al., 2000). In fact, although
rainfall variability is fairly high in our locality (CV of spring
rainfall ¼ 45 %, climate data from Arganda, Madrid, n ¼ 40
years), rainfall during the winter – spring period guarantees
the reproduction of annuals during most years. Thus, even in
a spring as dry as 2005 (the driest in the last 40 years),
annuals were still able to produce seeds owing to their reproductive plasticity (Caballero et al., 2008a). This, together
with the high seed bank mortality, suggests that investing in
a long-term seed bank may not be as critical for many
species in the community as previously thought. We propose
that persistence in these environments may occur as a combination of short- to mid-term (a few years) seed banks and
regular seed production from vegetation even under extreme
conditions (Aragón et al., 2007). The rapid equalization
between seed bank and vegetation is concordant with this
hypothesis. Furthermore, as in other arid environments, there
is a strong relationship between the seed bank and vegetation
(Olano et al., 2005; Hopfensberger, 2007) with temporally
shifting community compositions as a consequence of differential emergence and seed production due to annual climatic
conditions (see our spiral model; Caballero et al., 2008a).
This general trend does not contradict the fact that some
305
species may possess longer-term seed persistence in the soil,
as may be the case of some gypsophyte annual specialists.
Campanula fastigiata probably constitutes the best example
of a persistent (April) seed bank, which accounts for 45– 61
% of the September seed bank (see Supplementary Data).
Distributing emergence over time reduces the risk of a population crash due to a massive reproductive failure during an
unfavourable year (Venable, 1989). Thus, the seed bank provides a mechanism to reduce the impact of environmental stochasticity on population persistence (Venable and Brown,
1988). Nevertheless, delayed germination indicates a potential
reduction of population growth as a result of a longer mean
time to reproduction (Harper, 1977; Kalisz and McPeek,
1992) and enhanced seed mortality due to longer persistence
in soil. Thus, seed persistence in soil and consequently seed
bank formation should be a trade-off between minimizing
the risk of pre-reproductive failure and maximizing population
growth. The equilibrium point depends on the characteristics
of the species and habitats in question. Therefore, pioneer
species in successional environments in which windows of
opportunity occur seldom maintain long-term seed banks that
allow them to persist for decades until the next perturbation
(Olano et al., 2002). Most annuals in arid ecosystems may
exhibit ample variation in their longevity depending on the frequency of favourable conditions and as a consequence of individual responses to these alternative evolutionary strategies.
In conclusion, our results show that soil seed bank genesis
and recovery can be extremely rapid processes even under
the limiting and stressful conditions of semi-arid gypsum
environments. This recovery is based mainly on the seed
rain on a small scale together with secondary dispersal from
intact seed banks in the vicinity. These results also emphasize
the relevance of processes occurring over a short spatial scale
in determining community structure (Siewert and Tielborger,
2010). Obviously, extrapolating these results to larger spatial
scales, especially with respect to disturbance, must be done
very cautiously. Our results do not underestimate the strong
role of seed banks in ensuring community persistence in semiarid gypsum environments. Instead, our results and hypothesis
suggest that this function can be maintained without a longlived seed bank through a highly dynamic process based on
seed replenishment and small-scale horizontal movements
from seed sources. As a consequence, the relationship
between vegetation and seed bank composition is tight and
more dynamic than previously thought as seed turnover is
high (Caballero et al., 2008a).
S U P P L E M E N TA RY D ATA
Supplementary data are available online at www.aob.oxfordjournals.org and consists of full details of the species that
emerged within the cylinders.
ACK NOW LED GE MENTS
We thank the Pajarón family for kindly allowing us to work on
their properties in Belinchón. Greenhouse space was provided
by the Plant Physiology Unit of the Basque Country
University. We are grateful to Alison Powell and two anonymous reviewers for detailed and constructive criticisms of
306
Olano et al. — Soil seed bank recovery in semi-arid vegetation
an earlier version of the manuscript. This work was supported
by the Spanish Ministerio de Ciencia e Innovación project
ISLAS CGL2009-13190-C03-03 and C03-01, Junta de
Castilla y León project VA006A10-2 and REMEDINAL2
(P2009/AMB-1783).
L I T E R AT U R E CI T E D
Aragón CF, Albert MJ, Giménez-Benavides L, Luzuriaga AL, Escudero
A. 2007. Environmental scales on the reproduction of a gypsophyte: a
hierarchical approach. Annals of Botany 99: 519– 527.
Baskin CC, Baskin JM. 1998. Seeds. Ecology, biogeography, and evolution
of dormancy and germination, 1st edn. San Diego: Academic Press.
Bekker RM, Lammerts EJ, Schutter A, Grootjans AP. 1999. Vegetation
development in dune slacks: the role of persistent seed banks. Journal
of Vegetation Science 10: 745–754.
Bertiller MB. 1992. Seasonal variation in the seed bank of a Patagonian grassland in relation to grazing and topography. Journal of Vegetation Science
3: 47– 54.
ter Braak CJF. 1986. Canonical correspondence analysis, a new eigenvector
technique for multivariate direct gradient analysis. Ecology 67:
1167–1179.
ter Braak CJF, Smilauer P. 1997. Canoco for Windows version 4.0.
Wageningen: Centre for Biometry.
Caballero I. 2006. Estructura espacio-temporal de un banco de semillas: Las
comunidades gipsı́colas del centro de la Penı́nsula Ibérica. PhD Thesis,
Universidad del Paı́s Vasco, Bilbao, Spain.
Caballero I, Olano JM, Loidi J, Escudero A. 2003. Seed bank structure
along a semi-arid gypsum gradient in central Spain. Journal of Arid
Environments 55: 287– 299.
Caballero I, Olano JM, Luzuriaga AL, Escudero A. 2005. Spatial coherence between seasonal seed banks in a semi-arid gypsum community,
density changes but structure does not. Seed Science Research 15:
153–160.
Caballero I, Olano J.M., Loidi J, Escudero A. 2008a. A model for
small-scale seed bank and standing vegetation connection along time.
Oikos 117: 1788–1795.
Caballero I, Olano J.M., Loidi J, Escudero A. 2008b. Seed bank spatial
structure in semi-arid environments: beyond the patch-bare area dichotomy. Plant Ecology 195: 215– 223.
Chesson P. 2000. Mechanisms of maintenance of species diversity. Annual
Review of Ecology and Systematics 31: 343– 366.
Childs S, Goodall DW. 1973. Seed reserves of desert soils. Research
Memorandum, RM 73-75. Logan, UT: US International Biological
Program, Desert Biome. Available online at http://digitalcommons.usu.
edu/dbiome_memo/.
Ellner S, Shmida A. 1981. Why area adaptations for long-range seed dispersal
rare in desert plants. Oecologia 51: 133–144.
Escudero A, Somolinos RC, Olano JM, Rubio A. 1999. Factors controlling
the establishment of Helianthemum squamatum (L.) Dum., an endemic
gypsophite of semi-arid Spain. Journal of Ecology 87: 290– 302.
Escudero A, Iriondo JM, Olano JM, Rubio A, Somolinos R. 2000. Factors
affecting establishment of a gypsophyte: the case of Lepidium subulatum
(Brassicaceae). American Journal of Botany 87: 861–871.
Fenner M. 1995. Ecology of seed banks. In: Kigel J, Galili G, eds. Seed development and germination. New York: Marcel Dekker, 507–528.
Garcı́a-Fayos P, Bochet E, Cerda A. 2010. Seed removal susceptibility
through soil erosion shapes vegetation composition. Plant and Soil 334:
289–297.
Gutiérrez JR, Meserve PL. 2003. El Niño effects on soil seed bank dynamics
in north-central Chile. Oecologia 134: 511–517.
Gutiérrez JR, Arancio G, Jaksic FM. 2000. Variation in vegetation and seed
bank in a Chilean semi-arid community affected by ENSO 1997. Journal
of Vegetation Science 11: 641–648.
Gutterman Y, Shem-Tov S. 1996. Structure and function of the mucilaginous
seed coats of Plantago coronopus inhabiting the Negev Desert of Israel.
Israel Journal of Plant Sciences 44: 125– 134.
Harper JL. 1977. Population biology of plants, 1st edn. London: Academic
Press.
ter Heerdt GNJ, Verweij GL, Bekker RM, Bakker JP. 1996. An improved
method for seed bank analysis: seedling emergence after removing the
soil by sieving. Functional Ecology 10: 144 –151.
Hölzel N, Otte A. 2004. Inter-annual variation in the soil seed bank of floodmeadows over two years with different flooding patterns. Plant Ecology
174: 279–291.
Hopfensberger K. 2007. A review of similarity between seed bank and standing vegetation across ecosystems. Oikos 116: 1438– 1448.
Hyatt LA, Casper BB. 2000. Seed bank formation during early secondary succession in a temperate deciduous forest. Journal of Ecology 88: 516–527.
Kalisz S, McPeek MA. 1992. Demography of an age structured annual:
resampled projection matrices, elasticity analyses and seed bank effects.
Ecology 73: 1082–1093.
Kelt DA, Meserve PL, Gutiérrez JR. 2004. Seed removal by small
mammals, birds and ants in semi-arid Chile, and comparison with other
systems. Journal of Biogeography 31: 931– 942.
Kemp PR. 1989. Seed banks and vegetation processes in deserts. In Leck MA,
Parker VT, Simpson RL, eds. Ecology of soil seed banks. San Diego:
Academic Press, 257 –281.
Laskurain NA, Escudero A, Olano JM, Loidi J. 2004. Seedling dynamics of
shrubs in a fully closed temperate forest: greater than expected.
Ecography 27: 650–658.
Li F-R. 2008. Presence of shrubs influences the spatial pattern of soil seed
banks in desert herbaceous vegetation. Journal of Vegetation Science
19: 537 –548.
Legendre P, Legendre L. 1998. Numerical ecology, 1st edn. Amsterdam:
Elsevier Science.
Lortie CJ, Ellis E, Novoplansky A, Turkington R. 2005. Implications of
spatial pattern and local density on community-level interactions. Oikos
109: 495–502.
Luzuriaga AL, Escudero A, Olano JM, Loidi J. 2005. Regenerative role of
seed banks following an intense soil disturbance. Acta Oecologica 27:
57–66.
Martı́nez-Duro E, Ferrandis P, Escudero A, Luzuriaga AL, Herranz JM.
2009a. Secondary old-field succession in an ecosystem with restrictive
soils: does time from abandonment matter? Applied Vegetation Science
13: 234 –248.
Martı́nez-Duro E, Ferrandis P, Herranz JM. 2009b. Factors controlling the
regenerative cycle of Thymus funkii subsp funkii in a semi-arid gypsum
steppe: a seed bank dynamics perspective. Journal of Arid
Environments 73: 252– 259.
Martı́nez-Duro E, Ferrandis P, Herranz JM, Copete MA. 2010. Do seed
harvesting ants threaten the viability of a critically endangered nonmyrmecochorous perennial plant population? A complex interaction.
Population Ecology 52: 397–405.
Meyer SE, Stewart TE, Clement S. 2010. The quick and the deadly: growth
vs virulence in a seed bank pathogen. New Phytologist 187: 209–216.
Milberg P. 1995. Soil seed bank after eighteen years of succession from grassland to forest. Oikos 72: 3 –13.
Miriti MN, Rodrı́guez-Buriticá S, Wright SJ, Howe HF. 2007. Episodic
death across species of desert shrubs. Ecology 88: 32–36.
Monturiol F, Alcalá del Olmo L. 1990. Mapa de asociaciones de suelos de la
Comunidad de Madrid. Escala 1:200.000. Madrid: Consejo Superior de
Investigaciones Cientı́ficas.
Moriuchi KS, Venable DL, Pake CE. 2000. Direct measurement of the seed bank
age structure of a Sonoran Desert annual plant. Ecology 81: 1133–1138.
Morris AB, Baucom RS, Cruzan MB. 2002. Stratified analysis of the soil
seed bank in the cedar glade endemic Astragalus bibullatus: evidence
for historical changes in genetic structure. American Journal of Botany
89: 29– 36.
Nunney L. 2002. The effective size of annual plant populations: the interaction of a seed bank with fluctuating population size in maintaining
genetic variation. American Naturalist 160: 195– 204.
Olano JM, Caballero I, Laskurain NA, Loidi J, Escudero A. 2002. Seed
bank spatial pattern in a temperate secondary forest. Journal of
Vegetation Science 13: 775 –784.
Olano JM, Caballero I, Loidi J, Escudero A. 2005. Prediction of plant cover
from seed bank analysis in a semi-arid plant community on gypsum.
Journal of Vegetation Science 16: 215 –222.
Quintana-Ascencio PF, Caballero I, Olano JM, Escudero A, Albert MJ.
2009. Does habitat structure matter? Spatially explicit population
modelling of an Iberian gypsum endemic. Population Ecology 51:
317– 328.
Olano et al. — Soil seed bank recovery in semi-arid vegetation
Rivas-Martinez S, Loidi J. 1997. Bioclimatology of the Iberian Peninsula.
Itinera Geobotánica 13: 41–47.
Rubio A, Escudero A. 2000. Small-scale spatial soil–plant relationship in
semi-arid gypsum environments. Plant and Soil 220: 139–150.
Saatkamp A, Affre L, Dutoit T, Poschlod P. 2009. The seed bank longevity
index revisited: limited reliability evident from a burial experiment and
database analyses. Annals of Botany 104: 715–724.
Sher AA, Goldberg DE, Novoplansky A. 2004. The effect of mean and variance in resource supply on survival of annuals from Mediterranean and
desert environments. Oecologia 144: 353–362.
Siewert W, Tielborger K. 2010. Dispersal–dormancy relationships in annual
plants: putting model predictions to the test. American Naturalist 176:
490–500.
Templeton AR, Levin DA. 1979. Evolutionary consequences of seed pools.
American Naturalist 114: 232–249.
307
Thompson S, Katul G. 2009. Secondary seed dispersal and its role in landscape organization. Geophysical Research Letters 36: L02402,
doi:10.1029/2008GL036044.
Valbuena L, Tárrega R, Luis-Calabuig E. 2000. Seed banks of Erica australis and Calluna vulgaris in a heathland subjected to experimental fire.
Journal of Vegetation Science 11: 161– 166.
Venable DL. 1989. Modeling the evolutionary ecology of seed banks. In: Leck
MA, Parker VT, Simpson RL, eds. Ecology of soil seed banks. San Diego:
Academic Press, 67–87.
Venable DL, Brown JS. 1988. The selective interactions of dispersal, dormancy, and seed size as adaptations for reducing risk in variable environments. American Naturalist 131: 360 –384.
Venable DL, Flores-Martinez A, Muller-Landau HC, Barron-Gafford G,
Becerra JX. 2008. Seed dispersal of desert annuals. Ecology 89:
2218– 2227.