Content uploaded by Montserrat Vilà
Author content
All content in this area was uploaded by Montserrat Vilà on Mar 12, 2014
Content may be subject to copyright.
Responses of Mediterranean Plant Species to different fire frequencies in
Garraf Natural Park (Catalonia, Spain): field observations and modelling
predictions.
Francisco Lloret
1,
*, Juli G. Pausas
2
and Montserrat Vilà
1
1
Centre de Recerca Ecològica i Aplicacions Forestals (CREAF), Universitat Autònoma Barcelona, Bellaterra,
08193 Barcelona, Spain;
2
Centro de Estudios Ambientales del Mediterráneo (CEAM), Parc Tecnològic, C.
Charles Darwin 14, 46980 Paterna, València, Spain; *Author for correspondence
Received 26 June 2000; accepted in revised form 25 January 2002
Key words: Ampelodesmos mauritanica, Fire recurrence, Model, Resprouting, Simulation
Abstract
Dynamics of the coexisting Mediterranean species Pinus halepensis,Quercus coccifera,Erica multiflora,Ros-
marinus offıcinalis,Cistus albidus, C. salviifolius and Ampelodesmos mauritanica, with contrasted life history
traits have been studied under different fire scenarios, following two approaches: a) field survey in areas with
three different fire histories (unburned for the last 31 years, once burned in 1982, and twice burned in 1982 and
1994), and b) simulations with different fire recurrence using the FATE vegetation model. We compared ob-
served abundance in the field survey to simulation outputs obtained from fire scenarios that mimicked field fire
histories. Substantial mismatching did not occur between field survey and simulations. Higher fire recurrences
were associated with an increase in the resprouting Ampelodesmos grass, together with a decrease in Pinus abun-
dance. Resprouting shrubs did not show contrasting changes, but trends of increase in Quercus and decrease in
Erica were observed. The seeders Rosmarinus and Cistus achieved maximum abundance at intermediate fire re-
currence. We also performed ten 200 year simulations of increasing fire recurrence with average times between
fires of 100, 40, 20, 10, and 5 years. A scenario without fire was also simulated. High fire recurrence produces an
increase in Ampelodesmos, a grass which is becoming dominant in the area, and a small increase in Erica, but
Quercus abundance decreases and Pinus disappears. Rosmarinus and Cistus abundance peaks at intermediate fire
frequencies. When comparing these simulations to those in which Ampelodesmos was excluded, we found that
the absence of the grass only increased Cistus occurrence in the community, this effect being more important at
frequent fire recurrence. The study suggests that simple models based on life history traits may be useful in
interpreting plant community dynamics in Mediterranean ecosystems that are greatly influenced by differences in
the fire regime.
Introduction
Vegetation dynamics in Mediterranean-type ecosys-
tems are largely determined by the fire regime (Hanes
1977; Gill 1981; Kruger and Bigalke 1984; Trabaud
1994; Fuentes et al. 1994). Several regeneration
mechanisms involved in species response to fire (i.e.
germination, resprouting) may produce a fast post-fire
recovery of the compositional and structural attributes
of the vegetation (Hanes 1977; Pausas 1999a). Vari-
ation in fire regime related to increasing fire recur-
rence, intensity or timing (Sousa 1984; Johnson and
Gutsell 1994), however, may also change vegetation
(Bond et al. 1984; Pausas (1999b, 2001)). For exam-
ple, fires of high intensity may increase mortality of
resprouters (Lloret and López-Soria 1993; Moreno
and Oechel 1993). High fire recurrence may prevent
seeders from replenishing seed banks or may deplete
bud banks of resprouters (Zedler et al. 1983), but
might favor species that combine high resprouting
ability and fast reproduction and seedling establish-
ment after fire (Vilà et al. 2001). Furthermore, long
223
Plant Ecology 167: 223–235, 2003.
© 2003 Kluwer Academic Publishers. Printed in the Netherlands.
inter-fire periods may reduce abundance of short-
lived opportunistic species whose populations in-
crease shortly after fires, but decline over time (Tra-
baud and Lepart 1980; Haidinger and Keeley 1993).
Therefore, the vegetation response to changes in fire
regimes of Mediterranean-type ecosystems should be
strongly influenced by the life-history traits of plant
species.
The number, size and frequency of fires have in-
creased in the last several decades in the northern
Mediterranean Basin. Changes in fire regimes have
resulted from changes in land use (Moreno et al.
1998; Pausas and Vallejo 1999) and increases in cli-
matic fire hazard (Piñol et al. 1998a). A shift toward
communities dominated by low structured, herba-
ceous vegetation has been proposed for the Mediter-
ranean Basin under scenarios of high disturbance fre-
quency (Bolòs 1962; Naveh 1974; Trabaud 1991).
For example, dominance of large tussock grasses
rather than shrubs has been locally reported in areas
with frequent fires, and the existence of a positive
feedback between fire and grass has been proposed
(D’Antonio and Vitousek 1992). Frequent fires thus
may promote a shift in Mediterranean-type vegetation
from shrublands to savannas (Vilàand Lloret 2000).
Although long-term consequences of this shift are not
completely understood, a change in dominant life
forms should be expected.
Models of vegetation dynamics are useful tools for
investigating the long-term consequences of different
scenarios involving climate (Solomon 1986; Bug-
mann 1996), harvesting (Pausas and Austin 1998), or
fire regime (Malanson 1985; Pausas (1998, 1999b)).
They are especially important for studying the conse-
quences of interval-dependent processes (Bond and
van Wilgen 1996), in which the experimental ap-
proach is difficult to apply. Interval-dependent pro-
cesses such as plant establishment, maturation and
dormancy are key factors for predicting long-term
consequences of alternative fire scenarios.
The FATE model (Funtional Attributes in Terres-
trial Ecosystems; Moore and Noble (1990)) repre-
sents a good compromise between minimal data re-
quirement and realistic description of the vegetation
dynamics in fire-prone systems (Pausas 1999b). FATE
is a deterministic qualitative vegetation model based
on the vital attributes approach (Noble and Slatyer
1980) and on the assumption that the best way to un-
derstand the dynamics of plant communities is to
know how individual plants perform in their environ-
ment.
In this study, we explored whether increasing fire
recurrence in Mediterranean-type ecosystems would
favour a shift from pine forests and evergreen shru-
blands to more grass-dominated communities. Focus-
sing on shrubland-pine forest formations of Garraf
Natural Park (Catalonia, Spain), we applied the FATE
model to predict variation in the abundance of domi-
nant species under different historical fire regimes,
and we compared model results to field observations.
We also ran the model under scenarios of increasing
fire recurrence to evaluate long-term effects of a large,
perennial grass (Ampelodesmos mauritanica) on com-
munity dynamics.
Material and methods
Study area
The Garraf Natural Park (Garraf hereafter) is located
about 30 Km south of Barcelona (NE Spain, 41°15⬘
N, 2°0⬘E). The area (almost 10,000 ha) is a karstic
massif ranging from sea level to 600 m altitude. Soils
are Jurassic and Cretaceous limestones and marls
with a high presence of rock outcrops and abandoned
terraces (old fields). We stratified the study area into
3 classes based on the fire history over the past three
decades, during which we have precise, spatial record
of fire occurrence. These classes were: 1) small, scat-
tered areas that have remained unburned for at least
the last 31 years (hereafter, unburned areas), 2) a
7000 ha area that was burned in July 1982 (hereafter,
once-burned areas), and 3) a 4800 ha area within the
7000 ha area that was burned again in April 1994
(hereafter, twice-burned areas).
The climate is typically Mediterranean, with mild
and moderately moist winters, and warm and dry
summers. Mean annual rainfall at the closest climate
station is 548 mm (Vilanova i la Geltrú, Barcelona),
with a pronounced summer drought (around 100 mm
of rainfall from June to August). Mean annual tem-
perature is 16.7 °C. Mean maximum and minimum
temperatures are reached in August (30.6 °C) and Jan-
uary (0.5 °C), respectively.
The vegetation is dominated by evergreen, sclero-
phyllous shrublands 1.5 m high and open Pinus
halepensis forests (Bolòs 1962). One main souce of
variability in vegetation is land use history, especially
abandoned terraces and uncultivated slopes that tend
to correspond to deeper and stony soils, respectively.
A second source is a coastal to inland gradient that
224
determines climatic variation; moister conditions oc-
cur inland, where higher altitudes enhance rainfall.
Given the large extension of two fires (in 1982 and
1994), the different land use histories and the climatic
gradient are well represented in all three types of fire
histories (Riera and Castell 1997).
Through many centuries, grazing by domestic live-
stock, mainly sheep and goats, has been important in
all the area comprising Garraf. Cores from ancient
lakes have provided charcoal dated to the Middle
Ages, which has been attributed to burning of wood-
lands to increase pastures (Riera-Mora and Esteban-
Amat 1994). However, agricultural practices have not
been historically intensive in the area because of the
dominance of stony soils with superficial limestone
bedrocks. Since the end of the 18th century and dur-
ing the 19th, vinyeard expansion increased (Ferrer
1998), but in the 1890’s phyloxera arrived in Garraf
(Giralt 1990) leading to the abandonment of most of
these areas, which have not been re-cultivated. These
areas have been colonized by shrublands and Pinus
halepensis forests. This type of ecosystem, with a
long history of human influence, is dominant in the
Mediterranean Basin, where pristine, natural wild-
lands are the exception (Naveh 1974).
Field survey
An extensive survey was conducted in Garraf from
January to March 1996. This survey occurred 31
years at least since the last fire in the unburned area,
14 years after the last fire in the once-burned area, and
two years after the most recent fire in the twice-
burned area. We randomly selected a subset of 92
quadrats from a 500 ×500 m grid map covering the
park. We chose 30 quadrats in the unburned and the
twice burned areas, and 32 quadrats in the once
burned areas. Within each quadrat we established one
10 ×10 m stand. Within each fire history area, stand
site selection was stratified to balance the different
combinations of aspect (north or south), topographi-
cal location (steep or flat areas), and soil type (pres-
ence of rock outcrops in more or less 30% of the total
soil surface). Vegetation cover of perennial species in
each stand was estimated by the point intercept sam-
pling procedure. We recorded the presence of the spe-
cies every 0.5 m along the four sides of the stand
quadrat and in a 10 m transect located in the middle
of the stand. Thus, in each stand 94 points were used
to estimate cover of perennial species. Annuals were
not included because their relative cover is low in this
type of woodlands (Folch 1981). This procedure is
appropriate to quantify the plant cover of dominant
species, six of which were selected for this study (see
below). Comparisons between the three types of fire
history were performed by one-factor ANOVA and
the Fisher’s PLSD test was used for post-hoc pairwise
comparisons (SuperAnova procedure, Abacus Con-
cepts (1989)). Data on cover percentage of each spe-
cies were ln(x+1) transformed before analysis to nor-
malise residuals.
Model and species description
All simulations were performed using the FATE
model (Moore and Noble 1990). FATE is a general
model of vegetation dynamics, which is based on the
performance of individual plants in a stand. It predicts
vegetation dynamics at a qualitative level and from
simple parameters describing life history traits; these
include maturation time, lifespan, resprouting and
germination ability after fire, seed ability to colonize
a new site (hereafter, seed arrival), seed dormancy,
and shade tolerance (Table 1). The model is determin-
istic and simulates cohorts of plants that pass through
a series of discrete stages: seeds, seedlings, immature
and mature (adult) plants. The model is not spatially
explicit, but species interactions are included in the
model by considering groups of coexisting species.
Then the model estimates the performance of the dif-
ferent species from the response of the different
stages to decreasing light levels caused by the pres-
ence of neighbours. The model runs at annual time
steps, and the outputs are qualitative descriptions of
the abundance of each stage, measured on a scale of
absent, low, medium, and high. Fire events, and the
respective post-fire vegetation recovery may be in-
cluded stochastically or at given times, allowing the
simulation of different fire recurrences. Fire intensity
is not considered by the model. A detailed description
of the model is given by Moore and Noble (1990).
The model was set to simulate a community com-
posed of six native species. These species typically
grow along the Mediterranean coast of NE Spain
(Bolòs 1962), and are representative of the different
life-history types occurring in these communities.
Ampelodesmos mauritanica (hereafter Ampelodes-
mos) is a perennial, large resprouting tussock grass.
This species is distributed through the Mediterranean
Basin from Spain to Western Greece, and from East-
ern Morocco to Tunisia. In the NE Iberian Peninsula,
it occurs in only two areas: in Garraf, and around
225
Tarragona, about 100 km southwest of Barcelona
(ORCA 1985). In both areas wildfires have been com-
mon in the last decades. The causes of the absence of
this species in the other sites along the coast is yet
unknown. In Garraf, Ampelodesmos may attain local
dominance, with increasing populations in areas that
have been more frequently burned (Vilàet al. 2001).
Pinus halepensis (hereafter Pinus) is a non-resprouter
needle-leaved evergreen pine with serotinous cones;
Quercus coccifera (hereafter Quercus) is a broadle-
aved evergreen resprouter shrub. Erica multiflora
(hereafter Erica) is a ericoid-leaved resprouter shrub;
Rosmarinus offıcinalis (hereafter, Rosmarinus)isa
narrow-leaved evergreen non-resprouter shrub. Fi-
nally, Cistus spp., including two species that were
pooled (Cistus salvifoliius and C. albidus, hereafter
Cistus), are broadleaved non-resprouter shrubs. The
main species attributes considered in the model are
given in Table 1. Time of first reproduction and
lifespan are approximate, and have been obtained
from regular visits to burned and unburned areas, and
from published information (Cucó1987). Ability to
resprout has been obtained from descriptions of other
authors (Cucó1987; Papió1994), from our own vis-
its to burned areas, and in the case of Erica from ex-
perimental burnings (Lloret and López-Soria 1993).
Fire effects on seeds have been obtained from heat-
ing experiments (Salvador and Lloret 1995; Habrouk
et al. 1999; Vilàet al. 2001) and field observations
(Lloret 1998). Dispersal ability is used in FATE to
estimate the rate of seed input from outside the simu-
lated stand. We assumed it to be similar (wide dis-
persal) for all species except Quercus; its recruitment
from seeds is very low in the area as a consequence
of acorn predation by rodents and the short time of
seed viability (Lloret, pers. obs.). Seed dormancy, and
shade effect on germination and survival have been
estimated from field observations and published in-
formation (Papió1994; Lloret 1998).
Simulation scenarios
The first step in the application of vegetation dynam-
ics models was to compare the predictions of the
model behaviour to field observations. Therefore, a
first set of scenarios utilized the observed fire recur-
rence patterns for the three fire history areas in Gar-
raf. The second set of simulations aimed to predict
possible long-term responses to changes in fire recur-
rence.
The first set of the three scenarios used a period of
31 years, starting at 1965. Before this date, we do not
have reliable information on fire distribution in the
area. This period mimicked the three types of fire
history identified in the field survey. Thus, the three
scenarios were: a) no-fire (hereafter, unburned scenar-
io), b) a fire 17 years after the setup (hereafter, once-
burned), c) a fire 17 years after the initiation and a
second fire 12 years later (hereafter, twice-burned).
We do not know the real initial abundances for the
different species. Therefore, we assumed a commu-
nity with the same abundance of the different species.
The default condition used in the model, which con-
sidered several coexisting species, was low levels of
adults for each species. This condition may influence
simulation results, for example, by understimating the
real abundance of long-lived species, such as trees.
After a single run, the model output and the field sur-
vey were compared by contrasting the qualitative out-
put for each species with the respective percentile
Table 1. Qualitative life history-traits of the six Mediterranean species considered in the FATE model. Data were obtained from direct field
observations in the study area and from the literature (Cucó1987; Lloret and López-Soria 1993; Papió1994; Salvador and Lloret 1995;
Lloret 1998; Habrouk et al. 1999; Vilàand Lloret 2000). (All, Most, Half, Few, None: qualitative description of the proportion of individuals)
Ampelodesmos Pinus Quercus Erica Rosmarinus Cistus
Time to first reproduction (yrs) 5 10 10 7 4 3
Lifespan (yrs) 25 120 250 40 25 15
Fraction of individuals able to resprout All None All Most None None
Seeds killed by fire All Few All Most Half None
Seed arrival Yes Yes No Yes Yes Yes
Seed innate dormancy (yrs) No Yes (5) No No No Yes (15)
Fraction of seeds with dormancy broken by fire None All None None None Most
Germination under shade Medium Very low Low Low Low Very low
Survival under shade Low Low Medium Very low Low Very low
226
distribution of percent cover, after ln(x+1) transfor-
mation.
The second set of scenarios considered periods of
200 years, in which the probability of a fire in a given
year was 0.01, 0.025, 0.05, 0.1, and 0.2. These sce-
narios simulated fire regimes with average time be-
tween fire (fire return) equivalent to 100, 40, 20, 10,
and 5 years, respectively (hereafter F100, F40, F20,
F10 and F5). Piñol et al. (1998b) have estimated that
fire return intervals in the region range approximately
from 25 to 130 years, though higher recurrence at a
single site is also common (Trabaud et al. 1993).
Given that in the absence of fire the model output is
always the same, a single simulation of a scenario
without fires was also performed. For all species, ten
simulations for each fire scenario were obtained. In
these simulations, fire occurred stochastically follow-
ing the respective probabilities (see above). Then, we
calculated the mean percentage of years during the
200 year period in which each species and stage were
present. As in the first set of scenarios, initial abun-
dances in all simulations were equal, low levels of
adults of each species.
For each species pairwise comparisons between
fire scenarios (F100, F40, F20, F10, and F5) were
performed by post-hoc Fisher’PLSD tests, after one-
factor ANOVA in which the main factor was the fire
scenario. In this analysis, the dependent variable was
the number of years along the 200 year simulation
period in which adult plants (including low, medium
and high abundances) were present. This estimation
avoided the specific effect of the last fire, which af-
fected final abundances. Visual screenings of the out-
puts along the 200 years simulations did not show
great discrepancies between the final abundance and
the pattern of abundance along the 200 year period.
For our purpose (comparisons between scenarios), the
percentage of years of the different abundance classes
was considered a good summary of this abundance
pattern. Successful establishment in the community
was considered to occur when the mature (adult)
stage was obtained. Comparisons between each fire
scenario and the no-fire scenario were performed for
the same variable by two-tailed t-tests for significant
differences between a population mean (10 simula-
tions for each fire scenario) and a constant (1 simu-
lation for the no-fire scenario).
Simulations of fire scenarios and no fire scenario
using the same set of species, excluding Ampelodes-
mos, were used to compare changes in presence or
absence of this grass on the community. We analyzed
the results of these simulations using a two-factor
ANOVA in which the main factors were fire scenario
and the presence of Ampelodesmos in the simulation.
As in the previous analysis, the dependent variable
was the number of years during the 200 year period
in which adult plants were present.
Results
Field observations
Cover of Ampelodesmos was more than 1.5 times as
high in the twice burned areas than in the once burned
areas and more than twice that in the unburned areas;
these cover values were significantly different (Ta-
ble 2). The post-fire Ampelodesmos cover increase
has been related to the pronounced ability of this spe-
cies to resprout and to recruit after fire (Vilàet al.
2001). The plant cover of Quercus was also twice as
high in burned areas than in the unburned areas, be-
ing the values of once and twice burned areas nearly
identical. Pinus cover in once burned areas was
nearly tenfold lower than in unburned areas, while
Erica and Rosmarinus did not show significant dif-
ferences between unburned and once burned areas.
Plant cover of Pinus,Erica,Rosmarinus, and Cistus
was lower in the twice burned area, although this
trend was not significant in the case of the seeding
Cistus. The short time since the last fire in the twice
burned area, explain the low cover values of the small
resprouter Erica and the seeders Pinus,Rosmarinus
and Cistus.
Comparisons between field observations and
short-term simulations
The observed Ampelodesmos increase, Pinus de-
crease, and the lack of change in Erica with increas-
ing fire recurrence were all predicted by the single run
model which simulated the respective fire histories
(Figure 1). However, the observed increase of Quer-
cus abundance is not well predicted by the model,
which indicated outputs with little variation in this
species in the three fire history scenarios. This dis-
crepancy may result from competitive effects of Pi-
nus trees on Quercus coccifera shrubs in the under-
storey of old-unburned forests, an effect which is not
well developed in the model.
The model also predicts a decrease in Rosmarinus
abundance following fires. Such decreases were ob-
227
served in the field only in the twice-burned areas.
Rosmarinus has high growth of seedlings after fire
(Lloret 1998), which may not be considered suffi-
ciently in the model. Finally, the increase of Cistus
observed in burned areas is also predicted by the
model, despite low abundances in the field surveys.
If we compare the six species through their respec-
tive FATE results, field and model observations match
well in the unburned scenario, except for the pre-
dicted medium value of Pinus. In the once-burned
scenario, the most important discrepancy is found in
Pinus and Rosmarinus. These species are predicted by
the model to be absent, but they were present in the
field survey. In the twice-burned scenario, the quali-
tative output of the model is medium abundance for
Cistus and Erica, but the mean cover percentage of
these species in the field was only 0.4 and 1.6, respec-
tively. In the case of Ampelodesmos, the predicted
model output is medium abundance, while field sur-
vey obtained a mean cover of 35.8% for this species
in twice burned areas. The large grass Ampelodesmos
is able to achieve high values of plant cover only two
years after fire, while the seeder Cistus and the small
resprouter Erica need more time. We believe that this
discrepancy may arise because the model does not
consider plant growth rates.
Long-term simulations
The six types of long-term simulations produced a
gradient of fire return intervals from an absence of fire
during a 200 year period to 5-year fire returns (Ta-
ble 3). Figure 2 shows the different trends for each
species.
All species were predicted to show significant
changes in abundances with different annual proba-
bilities of fire (Table 4). Ampelodesmos abundance,
expressed as the number of years in the 200 year
simulation period with presence of adults, was pre-
dicted to increase with annual probabilities of fire <
0.05. Pinus abundance was predicted to peak at low
to intermediate annual probabilities of fire. Adults of
this species were predicted to decrease in the F10
scenario and almost disappear in the F5 scenario.
Quercus was predicted to have the highest abundance
values in the absence of fire or at the F100 scenario,
decreasing at the higher annual probabilities of fire.
The long lifespan of this species and its ability to es-
tablish in the shrubland understorey could explain this
dominance at long fire return times. Erica abundance
was predicted to be low in the no fire and in the F100
scenario, with a maximum abundance at the F10 and
F5 scenarios. Rosmarinus is predicted by simulations
to be more abundant at the F100, F40 and F20 sce-
narios, significantly decreasing at the higher annual
probabilities of fire recurrence. Cistus is predicted to
have the lowest abundance values, with a peak at in-
termediate annual probabilities of fire (F10 and F20
scenarios).
Under a scenario without fire, the model predicts a
community dominated by the resprouter shrub Quer-
cus and a mixture of seeders (Rosmarinus,Pinus) and
resprouters with shorter life span (Erica,Ampelodes-
mos). Low annual probabilñity of fire (F100) does not
change this pattern very much, except for an increase
of Pinus, which reaches a maximum at about annual
probability of fire of 0.025 (F40), as a consequence
of its long lifespan and its ability to establish in open
sites after the death of shrubs. However, high annual
probability of fire is predicted to cause important
changes in the vegetation: Pinus tends to disappear,
while the resprouting, perennial grass Ampelodesmos
becomes dominant. Quercus cover also decreases
probably because the short time between fires does
Table 2. Mean (+ SE) percent cover of each of the six species in the field survey. F and p values were obtained from one-factor ANOVA
comparing percent cover in different fire regimes, after ln(x+1) transformation of data (see Figure 1 for percentile distribution). Values fol-
lowed by different letters indicate significant differences (P < 0.05) between stands with different fire history obtained from Fisher pairwise
tests.
Unburned Once-burned Twice-burned F
2,89
p
Ampelodesmos 15.6 (3.5) a 22.9 (4.0) b 35.8 (3.3) c 11.01 0.0001
Pinus 42.1 (5.3) a 5.9 (1.9) b 0.5 (0.2) c 77.06 0.0001
Quercus 11.6 (3.0) a 23.8 (4.1) b 23.9 (3.4) b 4.93 0.0093
Erica 7.7 (2.0) a 4.3 (1.0) ab 1.6 (0.4) b 4.15 0.0188
Rosmarinus 6.6 (1.4) a 9.8 (2.1) a 0.6 (0.2) b 15.34 0.0001
Cistus 1.1 (0.9) a 0.8 (0.4) a 0.4 (0.2) a 0.22 0.8045
228
not allow this long-lived species to reach high values
of plant cover. As a consequence of the loss of domi-
nance of Quercus, another species such as Erica
would be favoured. The seeder Cistus shows a peak
at intermediate annual probabilities of fire, when the
time between fires is long enough to ensure reproduc-
tion.
When we compare the outputs of simulations with
and without Ampelodesmos, we conclude that the
presence or absence of this grass in the community
Figure 1. Percentile distribution of the observed cover abundance (%) after ln(x+1) transformation of each species in the areas with different
fire history: unburned for at least the last twenty years (unburned areas), burned in July 1982 (once burned areas), and burned in July 1982
and again in April 1994 (twice burned areas). The bottom and the top of each box represent 25% and 75% of the data, the middle line
represents the median value of the data (50%), the bars represent 10% and 90% of the data, and the open circles 5% and 95%, respectively
(see Table 2 for mean and SE values). Simulation outputs for each scenario are written at the top of each box: Abs = absent (or presence of
seeds only), Low = low abundance, Med = medium abundance.
229
does not significantly change the results for most of
the species considered, even with increasing annual
probability of fire. The absence of Ampelodesmos is
associated with a significant increase of the adult
presence in Cistus simulations (Figure 3). This
change was particularly important at high annual
probabilities of fire (F20, F10 and F5 scenarios). All
other four species simulations did not change signifi-
cantly by the Ampelodesmos presence. Fire and Fire
×Ampelodesmos interaction had a non significant ef-
fect on species abundance (Table 3).
Discussion
Field observations
The field survey revealed differences in community
structure that were related to fire history. Life-history
traits may explain the variability of species abun-
dance in relation to fire recurrence. The post-fire re-
sprouting ability of perennial grasses has been
broadly studied in grasslands (Vogl 1974; Silva et al.
1991; Gitay et al. 1992; Masters et al. 1992), but this
has been less considered in Mediterranean-type eco-
systems, in spite of their local abundance, and their
relevance in fire-related processes (Bond and van
Wilgen 1996). Although Pinus halepensis has been
reported to regenerate well after fire (Trabaud 2000),
success may be diminished by frequent fires that limit
seed storage (Thanos and Daskalakou 2000), by the
presence of shading neighbours (Espelta 1996), such
as the resprouting Quercus, and by the existence of
shallow, stony soils in Garraf (Riera and Castell
1997). Lower values of Quercus in unburned areas
may be related to its limited ability to colonize un-
burned, old fields.
In addition to fire recurrence, event-dependent pro-
cesses, such as fire season or climatic conditions after
fire, are important in determining vegetation recovery
after fire (Bond et al. 1984; Le Maitre 1988; Trabaud
1991). In our case, fall and spring rainfall ensured
post-fire germination and resprouting in 1982 and
1994, respectively. Fire season may particularly influ-
ence the ability of the seed bank to restore popula-
tions (van Wilgen et al. 1992). In 1982, the fire was
in the summer, when the seed bank had been recently
filled. In 1994 the fire was in the spring when there
was a reduced seed bank of Pinus and Rosmarinus
(Cistus has a more permanent seed bank). Data on
Rosmarinus indicate, however that the levels of ger-
mination in 1995 were similar to the observed before
the fire (Lloret 1998). Pinus regeneration may have
been more influenced by the season of this fire.
The fire histories considered in our study include
the number of fire events in the last two decades and
time since last fire. The unburned and once burned
areas are likely to be comparable because of fast re-
covery of the structure and composition of Mediter-
ranean-type communities after disturbance (Keeley
1986; Malanson and Trabaud 1987). Comparisons in-
cluding the more recently burned areas (two years
old) should be considered with caution because the
time since the last fire was short. For instance, seeder
species, such as Pinus,Rosmarinus or Cistus are
likely to show low abundances in twice-burned areas
because of this effect.
Comparisons between field observations and
short-term simulations
Given the large variation in the field data the model
predicts qualitative changes in abundances of species.
There are several reasons for the inaccuracies of some
Table 3. Mean, standard deviation, minimum and maximum (n = 10), of the mean fire return intervals obtained in each 200 year simulation
period under the different fires scenarios.
Scenario name Annual fire probability Fire-return interval (years)
Mean SD Min Max
No Fire 0
F100 0.010 145.8 58.9 66.7 –
*
F40 0.025 40.2 22.3 25.0 100.0
F20 0.050 23.9 8.4 14.3 40.0
F10 0.100 10.0 2.1 5.7 13.3
F5 0.200 5.5 0.6 4.5 6.3
*
No fire during 200 years
230
Figure 2. Abundance of five stages of six species predicted for each of six fire scenarios described in Table 3. Abundance was estimated as
the mean percentage of years of the 200 year simulation period in which each of five stages was present. (Bars with different letters at the top
indicate significant differences in presence of adults (Low, intermediate and high abundance) between scenarios (P < 0.05, Fisher⬘s PLSD test
after ANOVA, except for comparisons between fire and no fire scenarios, in which two-tailed t-tests were used).
231
predictions. First, community history cannot be ex-
actly reproduced by the model. For example, the pres-
ence of large pine stumps suggests quite a long period
without extensive fires before 1982. The presence of
the slow growing shrub Juniperus phoenicea in un-
burned areas also suggests a relatively long period
without fire (Riera and Castell 1997). In the simulated
scenarios, the initial conditions are considered to be
the same (low abundance of adults) for all scenarios,
but we do not know the real abundance of each spe-
cies three decades ago. In unburned areas, higher
abundances of Pinus in the field than in the model
outputs may be explained by this inaccuracy: Pinus
could have been more abundant at the beginning of
the considered period in these areas (Riera and Cas-
tell 1997).
Second, the FATE model is not spatially explicit
and does not consider environmental heterogeneity
resulting from historical processes, including fire and
date of agricultural abandonment. Since the model
only produces a single qualitative estimation of plant
abundance, the comparison between the distribution
of species abundance in the surveyed plots and the
model output allows only roughly estimate the adjust-
ment of the model to the field variability.
Finally, discrepancies may also be due to species
interactions or population processes which are not
considered with enough detail by FATE. Despite these
inconsistencies, the pattern emerging from both ap-
proaches shows that fire is associated with an increase
of the resprouting Ampelodesmos grass together with
a decrease of pine forested areas. Resprouting shrubs
do not show a significant shift, but a trend of Quer-
cus to increase and of Erica to decrease is observed
from field observations. For the seeder species, the
Rosmarinus decrease with fire and the Cistus increase
in burned areas are predicted by the model.
Long-term simulations
Long term simulations allow prediction of possible
vegetation responses to different fire regimes. When
annual probability of fire increases, a dominance of
resprouters, and particularly Ampelodesmos arises.
Table 4. Results of the ANOVA performed for each species in which the dependent variable was the number of years in the 200 year simu-
lations (n = 10) with adults present. Five fire scenarios (simulations with annual proabbailties of 0.01, 0.025, 0.05, 0.1 and 0.2) and the
presence or absence of Ampelodesmos were the main effects.
Fire scenarios (F) Ampelodesmos (A) F ×A Error
df MS F df MS F df MS F df MS
Pinus 4 52139.8 23.61
**
1 5431.7 2.46 ns 4 1375 0.62 ns 90 22.1
Quercus 4 14283.8 410.02
**
1 53.3 1.53 ns 4 77.9 2.34 ns 90 34.8
Erica 4 19571.2 35.34
**
1 1.4 0.01 ns 4 187.9 0.34 ns 90 553.7
Rosmarinus 4 26890.6 64.22
**
1 510.8 1.22 ns 4 329.3 0.79 ns 90 418.7
Cistus 4 3580.4 14.01
**
1 5882.9 23.00
**
4 916.6 3.58
**
90 255.8
ns: not significant,
**
: p < 0.001
Figure 3. Cistus abundance in the different fire scenarios without Ampelodesmos.Cistus abundance was estimated as the mean percentage
of years of the 200 year simulation period in which each of five stages was present. Bars with different letters at the top indicate significant
differences between scenarios (P < 0.05, Fisher⬘s PLSD test after ANOVA, except for comparisons between fire and no fire scenarios, in
which two-tailed t-tests were used).
232
Seeder species peak at intermediate fire recurrences.
This peak seems dependent on the species lifespan.
The long lived resprouter Quercus is dominant when
fires become rare. The adjustment of life-history strat-
egies to fire recurrence is important for understand-
ing Mediterranean-type communities. For example,
fire-persister species have been found to be prevalent
over seeder, fire-recruiter species in Californian chap-
arral that remained unburned for as much as a cen-
tury (Keeley 1992). In contrast, seeder Proteaceae
from South Africa need a narrow window of fire in-
tervals for survival (van Wilgen et al. 1992).
Ampelodesmos removal experiments performed in
shrublands of Garraf suggest that this large grass does
not competitively suppress other resprouting shrubs
or seedlings of seeder species (Vilàand Lloret 2000).
Although the dynamics of the community seem more
influenced by the fire regime than by the direct effect
of this grass, a positive feedback between fire and
Ampelodesmos abundance may occur (D’Antonio and
Vitousek 1992): fire recurrence may increase in Am-
pelodesmos dominated communities because this
grass increases fine fuel loads (Vilàet al. 2001). The
model suggests that Cistus may become more fre-
quent in the community if Ampelodesmos is not
present. This trend increases at higher annual prob-
ability of fire. Therefore, under high annual probabil-
ity of fire, the success of this seeder, short-lived
shrub, which has often been considered to be favored
by fire (Trabaud 1987), would be mitigated by the
existence of the fast growing, resprouter grass Am-
pelodesmos.
The effect of fire intensity, extent and season on
post-fire regeneration has been widely explored
(Whelan 1995). Fire recurrence has been less studied,
probably because of difficulties in obtaining field
data. The importance of a shift in fire recurrence in
some ecosystems is a matter of concern for land man-
agers and policy makers (Lavorel et al. 1998). In spite
of the recognized ability of Mediterranean-type veg-
etation to maintain its composition after a fire (Tra-
baud 1994), this study shows that high annual prob-
ability of fire may change the relative abundance of
species with different life history traits. Zedler et al.
(1983) showed how short intervals between fires pro-
duced a shift from shrub to grass vegetation in the
Mediterranean California. Similarly, Naveh (1999)
has shown how the combined effect of recurrent fires
and heavy grazing pressure may lead to degraded
scrublands after the depletion of the herbaceous seed
bank. The current wide spectrum of life history types
found in Mediterranean-type ecosystems provides the
basis for alternative vegetation pathways under differ-
ent disturbance regimes. Models based on life history
traits are appropriate tools for testing the conse-
quences of alternative disturbance scenarios on
changes in vegetation structure.
Acknowledgements
We thank I. Noble for providing the FATE model and
P.H. Zedler, D. Goldberg, B. Platt and two anony-
mous reviewers for their useful comments. We also
thank the field assistance provided by several Biology
and Environmental Science students from the Univer-
sitat Autònoma de Barcelona, and particularly by C.
Casanovas, I. Gimeno, U. Gamper, A. Ballés, and C.
Vil à. Thanks are also due to the Servei de Parcs
(Diputacióde Barcelona) for providing support in the
research performed in the Garraf Natural Park. This
study has been funded by the EC project LUCIFER
(NV4-CT96-0320), and by the Spanish Government
(CICYT postdoc contract to JGP, and AGF-97-0533
project). CEAM is supported by Generalitat Valenci-
ana and BANCAIXA.
References
Abacus Concepts 1989. Superanova. User’s guide for MacIntosh.
5th edn. Abacus Concepts, Berkeley.
Bolòs O. 1962. El Paisaje Vegetal Barcelonés. Universitat de Bar-
celona, Barcelona.
Bolòs O., Vigo J., Masalles R.M. and Ninot J.M. 1990. Flora Man-
ual dels Països Catalans. Pòrtic, Barcelona.
Bond W.J. and van Wilgen B.W. 1996. Fire and Plants. Chapman
and Hall, London.
Bond W.J., Vlok J. and Viviers M. 1984. Variation in seedling re-
cruitment of Cape Proteaceae after fire. J. Ecol. 72: 209–221.
Bugmann H.K.M. 1996. Functional types of trees in temperate and
boreal forests: classification and testing. J. Veg. Sci. 7: 359–
370.
CucóM.L. 1987. Mecanismes de regeneració. In: Terradas J. (ed.),
Ecosistemes Terrestres. La Resposta Als IncendisiaD⬘altres
Pertorbacions. Diputacióde barcelona, Barcelona, pp. 45–62.
D’Antonio C.M. and Vitousek P.M. 1992. Biological invasions by
exotic grasses, the grass-fire cycle, and global change. Ann.
Rev. Ecol. Syst. 23: 63–87.
Espelta J. 1996. La regeneracióde boscos d⬘alzina (Quercus ilex
L.) i pi blanc (Pinus halepensis Mill.): Estudi experimental de
la resposta de les plàntules a la intensitat de llumialadis-
ponibilitat d⬘aigua. PhD Dissertation, Universitat Autònoma
Barcelona, Bellaterra, Spain.
233
Ferrer L. 1998. Expansióagrícola en el segle XVIII. In: Mestre J.
and Hurtado V. (eds), Atles d’História de Catalunya. 62th edn.,
Barcelona, pp. 172–173.
Folch R. 1981. La Vegetaciódels Païssos Catalans. Ketres, Barce-
lona.
Fuentes E.R., Segura A.M. and Holmgren M. 1994. Are the re-
sponses of matorral shrubs different from those in an ecosys-
tem with reputed fire history? In: Moreno J.M. and Oechel W.C.
(eds), The Role of Fire in Mediterranean-Type Ecosystems.
Springer-Verlag, New York.
Gill A.M. 1981. Coping with fire. In: Pate J.S. and McComb A.J.
(eds), The Biology of Australian Plants. University Western
Australia Press, pp. 65–87.
Gitay H., Lee W.G., Allen R.B. and Wilson J.B. 1992. Recovery of
Chionochloa rigida tussocks from fires in South Island, New
Zealand. J. Environmental Management 35: 249–259.
Giralt E. 1990. L’agricultura. In: Història Econòmica de la Catalu-
nya Contemporànea, Poblaciói Agricultura. Vol. 2. Enciclopè-
dia Catalana, Barcelona, pp. 121–309.
Haidinger T.L. and Keeley J.E. 1993. Role of high fire frequency
in destruction of mixed chaparral. Madroño 40: 141–147.
Hanes T. 1977. California chaparral. In: Barbour M.G. and Major
J. (eds), Terrestrial Vegetation of California. Wiley, New York,
pp. 417–470.
Habrouk A., Retana J. and Espelta J.M. 1999. Role of heat toler-
ance and cone protection of seeds in the response of three pine
species to wildfires. Plant Ecology 145: 91–99.
Johnson E.A. and Gutsell S.L. 1994. Fire frequency models, meth-
ods and interpretations. Adv. Ecol. Res. 28: 239–287.
Keeley J.E. 1986. Resilience of Mediterranean shrub communities
to fires. In: Dell B., Hopkins A.J.M. and Lamont B.B. (eds),
Resilience in Mediterranean-type Ecosystems. Dr W. Junk Pub-
lishers, Dordrecht, The Netherlands, pp. 95–112.
Keeley J.E. 1992. Recruitment of seedlings and vegetative sprouts
in unburned chaparral. Ecology 73: 1194–1208.
Kruger F.J. and Bigalke R.C. 1984. Fire in fynbos. In: Booysen P.V.
and Tainton N.M. (eds), Ecological Effects of Fire in South Af-
rican Ecosystems. Springer, Berlin, pp. 69–114.
Lavorel S., Canadell J., Rambal S. and Terradas J. 1998. Mediter-
ranean terrestrial ecosystems: research priorities on global
change effects. Global Ecology and Biogeography Letters 7:
157–166.
Le Maitre D.C. 1988. Effects of season of burn on the regeneration
of two Proteaceae with soil-stored seed. S. Afr. J. Bot. 54:
575–580.
Lloret F. and López-Soria L. 1993. Resprouting of Erica multiflora
after experimental fire treatments. J. Veg. Sci. 4: 367–374.
Lloret F. 1998. Fire, canopy cover and seedling dynamics in Medi-
terranean shrublands of Northeastern Spain. J. Veg. Sci. 9: 417–
430.
Lloret F. and López-Soria L. 1993. Resprouting of Erica multiflora
after experimental fire treatments. J. Veg. Sci. 4: 367–374.
Malanson G.P. 1985. Simulation of competition between alterna-
tive shrub life history strategies through recurrent fires. Ecol.
Model. 27: 271.
Malanson G.P. and Trabaud L. 1987. Ordination analysis of com-
ponents of resilience of Quercus coccifera garrigue. Ecology
68: 463–472.
Masters R.A., Vogel K.P. and Mitchell R.B. 1992. Response of
Central Plains tallgrass prairies to fire, fertilizer, and atrazine.
J. Range Management 45: 291–29.
Moore A.D. and Noble I.R. 1990. An individualistic model of veg-
etation stand dynamics. J. Environ. Manage. 31: 61–81.
Moreno J.M. and Oechel W.C. 1993. Demography of Adenostoma
fasciculatum in southern California chaparral. Oecologia 90:
50–60.
Moreno J.M., Vázquez A. and Vélez R. 1998. Recent history of
forest fires in Spain. In: Moreno J.M. (ed.), Large Forest Fires.
Backhuys Publishers, Leiden, The Netherlands, pp. 159–185.
Naveh Z. 1974. Effects of fire in the Mediterranean Region. In:
Kozlowski T.T. and Ahlgren C.E. (eds), Fire and Ecosystems.
Academic Press, New York, pp. 401–434.
Naveh Z. 1999. The role of fire as an evolutionary and ecological
factor on the landscapes and vegetation of Mt. Carmel. J. Med-
iterranean Ecology 1: 11–26.
Noble I.R. and Slatyer R.O. 1980. The use of vital attributes to
predict successional changes in plant communities subject to
recurrent disturbance. Vegetatio 43: 5–21.
ORCA 1985. Atlas Corològic de la Flora Vascular dels Països Cat-
alans. Vol. I. Institut d⬘Estudis Catalans, Barcelona.
PapióC. 1994. Ecologia del foc i regeneracióen garrigues i pinedes
mediterrànies. Institut d⬘Estudis Catalans, Barcelona.
Pausas J.G. 1998. Modelling fire-prone vegetation dynamics. In:
Trabaud L. (ed.), Fire Management and Landscape Ecology.
Int. Ass. Wildland Fire, Fairfield, WA, USA, pp. 327–334.
Pausas J.G. 1999a. Mediterranean vegetation dynamics: modelling
problems and functional types. Plant Ecology 140: 27–39.
Pausas J.G. 1999b. The response of plant functional types to
changes in fire regime in Mediterranean ecosystems. A simula-
tion approach. J. Veg. Sci. 10: 717–722.
Pausas J.G. 2001. Resprouting vs seeding-a Mediterranean per-
spective. Oikos 94: 193–194.
Pausas J.G. and Austin M.P. 1998. Potential impact of harvesting
for the long-term conservation of arboreal marsupials. Land-
scape Ecology 13: 103–109.
Pausas J.G. and Vallejo R.V. 1999. The role of fire in European
Mediterranean ecosystems. In: Chuvieco E. (ed.), Remote
Sensing of Large Wildfires in the European Mediterranean Ba-
sin. Springer-Verlag, Berlin, pp. 3–16.
Piñol J., Terradas J. and Lloret F. 1998a. Climate warming, wild-
fire hazard, and wildfire occurrence in coastal eastern Spain.
Climatic Change 38: 345–357.
Piñol J., Terradas J. and Lloret F. 1998b. Risk factors in wildfires
along the Mediterranean coast of the Iberian peninsula. In: Tra-
baud L. (ed.), Fire Management and Landscape Ecology. Int.
Ass. Wildland Fire, Fairfield, WA, USA, pp. 297–303.
Riera-Mora S. and Esteban-Amat A. 1994. Vegetation history and
human activity during the last 6000 years on the central Cata-
lan coast (northeastern Iberian Peninsula). Veget. Hist. Archae-
obot. 3: 7–23.
Riera J. and Castell C. 1997. Efectes dels incendis forestals recur-
rents sobre la distribucióde dues espècies del Parc Natural del
Garraf: el pi blanc (Pinus halepensis) i la savina (Juniperus
phoenicea). ButlletíInstitucióCatalana Història Natural 65:
105–116.
Solomon A.M. 1986. Transient response of forest to CO
2
-induced
climate change: simulation modeling experiments in eastern
North America. Oecologia 68: 567–579.
Salvador R. and Lloret F. 1995. Germinación en el laboratorio de
varias especies arbustivas mediterráneas: efecto de la temper-
atura. Orsis 10: 25–34.
234
Silva J.F., Raventos J., Caswell H. and Trevisan M.C. 1991. Popu-
lation response to fire in a tropical savanna grass, Andropogon
semiberbis: a matrix model approach. J. Ecology 79: 345–356.
Sousa W.P. 1984. The role of disturbance in natural communities.
Ann. Rev. Ecol. Syst. 15: 353–391.
Thanos C.A. and Daskalakou E.N. 2000. Reproduction in Pinus
hapensis and P. brutia. In: Ne⬘eman G. and Trabaud L. (eds),
Ecology, Biogeography and Management of Pinus halepensis
and Pinus brutia Forest Ecosystems in the Mediterranean Ba-
sin. Backhuys, Leiden, The Netherlands, pp. 79–90.
Trabaud L. 1987. Fire and survival traits of plants. In: Trabaud L.
(ed.), The Role of Fire in Ecological Systems. SPB Academic
Publishing, The Hague, pp. 65–89.
Trabaud L. 1991. Fire regimes and phytomass growth dynamics in
aQuercus coccifera garrigue. J. Veg. Sci 2: 307–314.
Trabaud L. 1994. Postfire plant community dynamics in the Medi-
terranean Basin. In: Moreno J.M. and Oechel W.C. (eds), The
Role of Fire in Mediterranean-type Ecosystems. Springer-Ver-
lag, New York, pp. 1–15.
Trabaud L. 2000. Post-fire regeneration of Pinus halepensis forests
in the West Mediterranean. In: Ne⬘eman G. and Trabau L. (eds),
Ecology, Biogeography and Management of Pinus halepensis
and Pinus brutia Forest Ecosystems in the Mediterranean Ba-
sin. Backhuys, Leiden, The Netherlands, pp. 257–268.
Trabaud L. and Lepart J. 1980. Diversity and stability in garrigue
ecosystems after fire. Vegetatio 43: 49–57.
Trabaud L., Christensen N.L. and Gill A.M. 1993. Historical Bio-
geography of Fire in Temperate and Mediterranean ecosystems.
In: Crutzen P.J. and Goldammer J.G. (eds), The Ecological, At-
mospheric, and Climatic Importance of Vegetation Fires. John
Wiley & Sons, New York, pp. 277–295.
van Wilgen B.W., Bond W.J. and Richardson D.M. 1992. Ecosys-
tem management. In: Cowling R. (ed.), The Ecology of Fyn-
boss. Nutrients, Fire and Diversity. Oxford University Press,
Cape Town, South Africa, pp. 345–371.
Vil àM. and Lloret F. 2000. Woody species tolerance to the expand-
ing perennial tussock-grass Ampelodesmos mauritanica after
fire. J. Veg. Sci. 11: 597–606.
Vil àM., Lloret F., Ogheri E. and Terradas J. 2001. Positive fire-
grass feedback in Mediterranean basin shrublands. Forest Ecol.
Manage. 147: 3–14.
Vogl R.J. 1974. Effects of fire on grasslands. In: Koslowski T.T.
and Ahlgren C.E. (eds), Fire and Ecosystems. Academic Press,
New York, pp. 139–194.
Whelan R.J. 1995. The Ecology of Fire. Cambridge University
Press, Cambridge.
Zedler P.H., Gautier C.R., and McMaster G.S. 1983. Vegetation
change in response to extreme events: the effect of a short in-
terval between fires in California chaparral and coastal scrub.
Ecology 64: 809–818.
235