Reforestation Opportunities in Indonesia: Mitigating Climate ...
An economic and ecological multi-criteria evaluation of reforestation methods to recover burned...
Transcript of An economic and ecological multi-criteria evaluation of reforestation methods to recover burned...
An economic and ecological multi-criteria evaluationof reforestation methods to recover burned
Pinus nigra forests in NE Spain
Josep Maria Espelta*, Javier Retana, Abdessamad HabroukCentre de Recerca Ecologica i Aplicacions Forestals (CREAF) i Unitat d’Ecologia, Facultat de Ciencies,
Universitat Autonoma de Barcelona, 08193 Bellaterra, Barcelona, Spain
Received 12 March 2002; received in revised form 9 August 2002; accepted 1 November 2002
Abstract
Recurrence of recent large wildfires is threatening the permanence of Pinus nigra Arnold (Black pine) forests in central
Catalonia (NE Spain), due to the almost nil post-fire regeneration of this non-serotinous pine. Potential practices to carry out
extensive reforestation programs with P. nigra may differ widely in terms of their final success, economic cost and undesired
ecological impact. In this framework, we have analysed different types of vegetation clearing (mechanical, controlled burning or
grazing), soil preparation (ripping or planting holes) and reforestation methods (broadcast seeding, spot seeding and planting) to
restore P. nigra forests. We have compared these practices in terms of seedling establishment, but also in the light of their
economic costs and ecological impact, through a new multi-criteria analysis. Seedling establishment after sowing was very poor
and not influenced by vegetation clearing. In plantations, seedling survival was higher in the ripper treatment than in planting
holes for all vegetation clearing treatments except the control one. Nevertheless, the higher economic cost of the planting holes
treatment and the negative impact of mechanical clearing and burning on the small mammals community made the stated
differences in seedling survival irrelevant. Thus, the multi-criteria analysis revealed that the two most preferred options were
planting in uncleared or lightly grazed areas with soil preparation through ripping. This study gives some valuable insights about
the use of new decision-support tools in restoration programs and provides practical guidelines concerning the restoration of
extensive burned P. nigra forests.
# 2002 Elsevier Science B.V. All rights reserved.
Keywords: Post-fire restoration; Decision-support tools; Pinus nigra; Planting; Seeding
1. Introduction
Tree regeneration failure after large wildfires is
becoming one of the major factors threatening the
conservation of many forest ecosystems in the west
Mediterranean basin (Ferran and Vallejo, 1998).
Extensive and massive crown fires result in homo-
geneous burned landscapes where the absence of
surviving trees and the distance to the edge can
severely hinder the arrival of propagules and the
chances of a successful regeneration process (Retana
et al., 2002). In Catalonia (NE Spain), large wildfires
(>1000 ha) have accounted for 0.4% of the total
number of fires, but 75% of the total burned surface
from 1975 to 1999 (Salvador et al., 2000). Moreover,
since 1990, large wildfires have destroyed more than
Forest Ecology and Management 180 (2003) 185–198
* Corresponding author. Tel.: þ34-93-581-2028;
fax: þ34-93-581-1312.
E-mail address: [email protected] (J.M. Espelta).
0378-1127/02/$ – see front matter # 2002 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0378-1127(02)00599-6
25% of the total area occupied by Pinus nigra Arnold
(Black pine) forests (Gracia et al., 2000). P. nigra does
not have serotinous cones and, due to the early release
of seeds at the beginning of spring, does not maintain a
seed bank when summer wildfires occur (Habrouk
et al., 1999). For these reasons, this species has very
low or nil post-fire regeneration (Trabaud and Camp-
mant, 1991; Retana et al., 2002).
P. nigra forests have both economic importance for
wood production and ecological significance (e.g. P.
nigra forests are included in the EU endangered
habitats directive), and the lack of successful regen-
eration has given rise to major concern about the best
alternatives to recover them through suitable artificial
reforestation programs (Espelta, 1999). In the west
Mediterranean basin, water stress has been envisaged
as the most limiting factor for ecosystem reconstruc-
tion, with soil fertility achieving secondary impor-
tance (Vallejo et al., 2000). Moreover, recently
established seedlings may be particularly prone to
water stress (Baeza et al., 1991), because they have
not yet developed a root system capable of replacing
water lost through transpiration. In that framework,
site preparation before reforestation, including both
vegetation clearing and soil preparation, may be of
paramount importance to enhance the outcome of
planted seedlings (Graham et al., 1989; Flemming
et al., 1996). Vegetation clearing has been reported
to diminish competition for water and nutrients, thus
ameliorating the survival rates of seedlings (Harvey
et al., 1996; Ashby, 1997). On the other hand, soil
preparation by ploughing may reduce runoff losses,
increase the total water stored in the profile and
decrease the penetration resistance of the soil (Ser-
rada, 1990; Querejeta et al., 2001).
Methods to clear vegetation and soil preparation
techniques may differ widely in their economic cost,
the effects on seedling survival and growth, and even
the promotion of undesired indirect ecological
impacts on the local fauna and flora (Peman and
Navarro, 1998). Although the benefits of different site
preparation techniques in reforestation programs have
been thoroughly documented elsewhere, very few
studies have at the same time taken into account their
effects on plantation performance, their economic
costs and their ecological consequences. Notwith-
standing, a general consensus has been growing up
about the need for insights from ecology and economy,
as well as social awareness about management of
natural resources, to be brought together if restoration
efforts are to succeed on a large-scale (Edwards and
Abivardi, 1997; Holl and Howarth, 2000). Elucidation
of the best way to increase plantation performance
with the minimum economic cost and ecological
impact, may benefit from advanced multi-criteria
decision-support tools. These methods facilitate the
comparison of different alternatives, and can contain
information arising from this complex system where
each criterion (plantation performance, economic cost
and ecological impact) may be expressed in a different
way (e.g. quantitative or non-quantitative) or sub-
jected to different types of uncertainty (e.g. stochastic
or fuzzy) (see, among others, Munda, 1995; Ducey
and Larson, 1999).
In the framework of assessing the best alternatives
to reforest extensive burned forests in Mediterranean
areas, the main objectives of this study are: (i) to
evaluate the combined effects of three different meth-
ods of vegetation clearing (controlled burning, live-
stock grazing and mechanical clearing), two types of
reforestation strategies (seeding, planting) and two
soil preparation techniques (ripping, planting holes)
in the establishment of P. nigra seedlings in burned
areas in Catalonia (NE Spain) and (ii) to compare this
array of contrasting alternatives in light of their eco-
nomic cost, plantation performance and ecological
impact, using a multi-criteria decision-support tool
(NAIADE: new approach to imprecise assessment and
decision environment; Munda, 1995). This compar-
ison includes two different analyses, one considering
financial cost and survival and growth of P. nigra
seedlings in the different reforestation practices, and
the second including other ecological criteria such as
the impact of those practices in the community of
small mammals.
2. Material and methods
2.1. Study area
This study was carried out in the regions of Bages
and Bergueda (418450–42860N; 18380–2810E, Catalo-
nia, NE Spain). The climatic conditions correspond to
the subhumid Mediterranean climate (according to the
Thornwaite index), with mean annual temperature of
186 J.M. Espelta et al. / Forest Ecology and Management 180 (2003) 185–198
12 8C and mean annual precipitation of 600 mm.
These regions were affected by the largest historically
recorded wildfire in NE Spain, which burned ca.
24,300 forested hectare in July 1994. According to
the data provided by the Forest Ecological Inventory
of Catalonia (IEFC), the area was covered extensively
by forests (71%). The main forest tree species before
the fire was P. nigra (75%) while other tree species
present in the area were Pinus halepensis Mill., Pinus
sylvestris L., Quercus ilex L. and Quercus cerrioides
Wk. et Costa (Gracia et al., 2000). After the fire event,
and due to the failure of P. nigra, 39% of the surface
previously covered by P. nigra forests became grass-
lands or open shrublands.
2.2. Experimental design and data analysis
The study was carried out from 1998 to 2000 in
three zones in the burned area: Can Armengol, Vila-
dases and Soler de Jaumas (Fig. 1A), the three located
in limestones. In each experimental zone, 14 plots of
1 ha each were established to test the combination of
different treatments of: (i) vegetation clearing, (ii)
reforestation and (iii) soil preparation. Treatments
were randomly assigned and combined in the 14
experimental plots in each zone (see Fig. 1B).
Before the start of the experimental treatments,
cover and height of the vegetation was determined
in each plot (sampling 1). To do so, 10 transects of
25 m each were laid per plot. In each transect, species
composition and vegetation height were measured
every metre. At this time, no significant differences
were found among experimental treatments for vege-
tation cover (range of mean values per plot: 83–86%)
and height (range 38–43 cm), while the three zones
differed in vegetation cover (ANOVA, F ¼ 15:9,
P < 0:05, mean � standard error: 91 � 1, 89 � 1
and 77 � 1%, for Viladasses, Can Armengol and Soler
de Jaumas, respectively), but not in vegetation height
(F ¼ 0:9, P ¼ 0:42, Viladasses: 48 � 2 cm; Can
Armengol: 41 � 2 cm; Soler de Jaumas: 33 � 2 cm).
In each study zone, four different treatments of
vegetation clearing were carried out:
� Livestock grazing (hereafter G). Four plots per site
were grazed by sheep and cattle during the spring
and fall of 1998. According to the observed levels
of tips and leaves consumption, as well as the light
browsing of low palatable shrubs, the grazing
pressure could be classified as moderate (following
Etienne et al., 1996).
� Controlled burning (hereafter B). The vegetation of
four plots per site was burned in December 1998 by
a controlled fire.
� Mechanical clearing (hereafter M). The vegetation
of three plots per site was cleared by rolling chains
pulled behind a tractor, leaving the debris scattered
about the site.
� Control (hereafter C). The vegetation of three plots
per site was kept intact.
To evaluate the effect of these clearing treatments,
the same 10 transects per plot of Sampling 1 were
monitored again after vegetation clearing, in March
1999 (Sampling 2), and 1 year later, in March 2000
(Sampling 3). The change in vegetation cover and
height per transect between Samplings 1 and 2 were
the variables used to analyse the ability of the different
clearing treatments to eliminate vegetation, while the
difference between Samplings 1 and 3 allowed us to
evaluate their ability to suppress vegetation re-growth
during a longer period. The effects of vegetation
elimination treatment and zone in these variables were
analysed by ANOVA models.
In the three experimental zones, the two reforesta-
tion methodologies contrasted consisted in planting
(1 þ 0) P. nigra seedlings or sowing P. nigra seeds. In
each zone P. nigra seedlings were planted in eight
plots combining randomly the four vegetation clearing
treatments and the two soil preparation methodologies
(Fig. 1B). The two soil preparation methods consisted
of:
� Ripping (hereafter R). In four plots, each one with a
different treatment of vegetation clearing, plough-
ing of planting lines where seedlings were subse-
quently planted, was carried out using a double
teeth subsoiler mounted at the rear of a bulldozer,
which opened two parallel rips of 70–90 cm depth.
� Planting holes (hereafter P). In four plots, each one
with a different treatment of vegetation clearing,
seedlings were planted in holes (40 cm � 40 cm) up
to 50 cm depth. These holes were carried out with
an excavator fitted with an articulated arm.
Initial seedling density planted in the different plots
was 1400 seedlings/ha regularly distributed. Survival
J.M. Espelta et al. / Forest Ecology and Management 180 (2003) 185–198 187
of seedlings was checked 1 and 2 years after planta-
tion, in six subplots of 25 seedlings (total 150 seed-
lings) per plot. Height growth of five seedlings per
subplot was also monitored in the six subplots per plot.
Height growth of seedlings was established as the
difference between their initial height (measured just
after plantation) and their height at the end of the first
and second year after the start of the experiment.
Fig. 1. (A) Geographical location of the area of Bages and Bergueda affected by the wildfire of 1994 in Catalonia (NE Spain) and the three
zones where the restoration practices were carried out. V: Viladases, SJ: Soler de Jaumas, CA: Can Armengol. (B) Experimental design of the
combined treatments of vegetation clearing, soil preparation and reforestation methods in each study zone.
188 J.M. Espelta et al. / Forest Ecology and Management 180 (2003) 185–198
The effects of the vegetation clearing treatment, soil
preparation, zone and time after plantation on seedling
survival and seedling height growth were analysed by
repeated measures ANOVA models. Inspection of
residuals was carried out to check for normality and
homoscedasticity. The sequential Bonferroni method
was employed to control the group-wide type I error
rate (Rice, 1989). The individual values of the differ-
ent levels of each variable were compared with a post-
hoc test (Fisher’s protected least significant differ-
ence).
Reforestation by seeding involved two different
methodologies to sow P. nigra seeds: broadcast dis-
tribution and spot distribution. In each experimental
zone, broadcast seeding was carried out in four plots
with the different vegetation clearing treatments pre-
viously described (Fig. 1B). On the other hand, spot
distribution was only applied in two plots per zone,
one grazed by livestock and another burned. We
specifically applied this type of seedling under these
two vegetation clearing methods, because they were
those which left the least debris above the soil, in
contrast with the mechanical clearing and the un-
cleared (control) plots. The aim was to compare the
possible benefits of a spot seed distribution to protect
seeds from predation and desiccation in contrast to a
broadcast seedling. The two seeding methodologies
consisted of:
� Broadcast distribution treatment. A total of 0.5 kg
of P. nigra seeds (cf. 25,000 seeds/ha) were broad-
cast over the whole surface of each 1 ha plot. This
treatment was applied in four plots per site, one per
treatment of vegetation elimination.
� Spot distribution treatment. In this case, 0.25 kg of
P. nigra seeds were used per 1 ha plot. A total
number of 1400–1500 small spots per ha were
regularly treated by hand in each plot, where 6–8
seeds were sown.
Prior to seeding, seeds were embedded in water
(12 h) and lightly dusted with two commercial and
widely used repellents (Morkit1 and Mesurol1) to
avoid or minimise predation by birds and insects after
sowing.
Since plots for all combinations of vegetation elim-
ination treatment � seeding methodologies were not
present in the field design, we were forced to carry
out two different complementary ANOVA analyses
instead of the complete design: a one-way ANOVA of
the effect of the four vegetation elimination treatments
on seedling establishment for the broadcast distribu-
tion treatment, and a two-way ANOVA of the effect of
the two seed distribution treatments and two of the
vegetation elimination treatments (grazed and burned
plots) on seedling establishment. To achieve normality
and homoscedasticity of residuals, seedling establish-
ment (density of seedlings per ha) was log-trans-
formed.
To assess a possible negative ecological impact of
the treatments applied, the dynamics of the small
mammals community were evaluated. We selected
this animal group as it can be considered highly
sensitive to the treatments used to clear vegetation
and prepare soil for planting (Barrett and Peles,
1999). Five surveys were carried out from November
1998 (before the experimental treatments) to January
2000. Two tramping transects were placed in each
plot. Each transect was composed of seven Sherman
traps at a distance of 10 m from one another. Captures
included two consecutive nights checked every 12 h.
An animal abundance index (number of animals
caught per trap and night) was obtained in each plot.
Two variables were used for describing the commu-
nity of small mammals in each plot: population
density (i.e., the total number of animals captured
per plot and survey), and diversity (computed as the
Shannon diversity index). More information on the
methodology used in this survey is described in
Comas et al. (2001).
2.3. Multi-criteria analysis of the different
reforestation practices
The evaluation of the different reforestation alter-
natives was carried out through a multi-criteria ana-
lysis. This analysis was done with the NAIADE
method (Munda, 1995). NAIADE is a multi-criteria
evaluation method which performs the comparison of
alternatives on the basis of a set of criteria. It allows
the use of information affected by different types and
degrees of uncertainty, thus the values assigned to the
criteria for each alternative may be expressed in the
form of either crisp, stochastic, fuzzy numbers or
linguistic expressions. This is a discrete method
(the set of alternatives is finite) which does not use
traditional weighting criteria and which generates a
J.M. Espelta et al. / Forest Ecology and Management 180 (2003) 185–198 189
ranking of alternatives using a pairwise comparison
technique (JRC-Ispra, 1996). NAIADE multi-criteria
analysis is based on a comparison algorithm made up
by the following steps:
(1) Completion of the criteria/alternatives matrix.
Values associated to each criteria for each
alternative are assigned in the form of pure
numbers or using a quantitative definition
affected by different levels and types of uncer-
tainty (either fuzzy or stochastic).
(2) Pairwise comparison of alternatives using pre-
ference relations. Comparison of criteria scores
(values) of each pair of alternatives is carried out
by means of their distance. In the case of numeric
evaluation, the distance is defined as the
difference between the two numbers. In the case
of fuzzy or stochastic evaluation, the semantic
distance is used, which measures the distance
between two functions, taking into account the
position and the shape of the two functions (see
for further details about this calculation JRC-
Ispra (1996)). The comparison by means of the
semantic distance is based on preference rela-
tions expressed by the user. These preference
relations are defined by means of six functions
that allow to express (depending on the distance
between alternatives), for each criterion, an index
of credibility of the statements that an alternative
is much better, better, approximately equal,
worse or much worse than another. The cred-
ibility index goes from 0 (definitely non-credible)
to 1 (definitely credible) increasing monotoni-
cally within this range.
(3) Criteria aggregation. Through an aggregation
algorithm of the credibility indexes, NAIADE
calculates a preference intensity index of one
alternative with respect to another, using the aparameter (see below) to express the minimum
requirements on the credibility indexes. Only
those criteria whose indexes are above the athreshold will be counted in the analysis.
(4) Ranking of alternatives. This ranking is based on
the preference intensity indexes. The final rank-
ing comes from the intersection of two separate
rankings. The first one (fþ) is based on the better
and much better preference relations, while the
second one (f�) is based on the worse and much
worse preference relations. Both rankings range
from 0 to 1 to express the intensity of the
relations (see for further details Munda (1995);
JRC-Ispra (1996)).
In this multi-criteria analysis, we only tested the
planting alternatives. We excluded the seeding treat-
ments; because their almost complete failure (see
Section 3) means that they were not worth comparing.
To evaluate each alternative, we used five criteria:
1. Financial cost. Monetary cost of the different
combinations of vegetation clearing and soil
preparation methods conducted.
2. Survival of P. nigra seedlings. Density of P. nigra
seedlings surviving at the end of the experiment.
3. Height growth of P. nigra seedlings. Absolute
mean height growth of P. nigra seedlings between
the start and the end of the experiment.
4. Density of small mammals. Computed as the
difference between the abundance of small
mammals captured per plot at the beginning and
at the end of the 2 years study.
5. Diversity of small mammals. Computed as the
difference in the Shannon diversity index of the
community of small mammals at the beginning
and at the end of the 2 years study.
This set of criteria had a different form. Financial
cost was a numerical criterion. P. nigra survival and P.
nigra growth were used as fuzzy criteria, because
there existed interactions on these variables among
vegetation clearing, soil preparation and zone (see
Section 3). For this reason we assigned to each cell
the mean observed value in the corresponding alter-
native, but instead of using a single value, we used a
fuzzy function shaped by the mean value and the
maximum range of variability observed among the
three zones. Thus, in the different simulations ran with
the model we could better integrate the spatial varia-
bility of survival and growth. For each criterion, four
thresholds delimiting indifference, weak indifference,
preference and strong preference were defined, after
discussion with landowners and governmental tech-
nical services in charge of restoration programs in the
area (Departament de Medi Ambient, Generalitat de
Catalunya). In NAIADE, the inclusion of such thresh-
olds is necessary to define the preference relations
that will allow the pairwise comparison of alternatives
190 J.M. Espelta et al. / Forest Ecology and Management 180 (2003) 185–198
(see above) taking into account the point of view of
stakeholders (see Munda, 1995). As it was impossible
to define realistically such thresholds for the density
and diversity of small mammals, we used a linguistic
expression of 9 degrees (ranging from perfect to extre-
mely bad) to define its situation in comparison with the
initial situation (before vegetation clearing and soil
preparation). In the analysis of alternatives, financial
cost was clearly considered a tendency to minimise,
while the remaining criteria were considered tendencies
to maximise. Table 1 summarises the form, thresholds
and tendencies of the different criteria used.
We ran two different analyses, the first one only
considered financial cost and survival and growth of
P. nigra seedlings, while the second also considered the
density and diversity of small mammals. Each analysis
included 1000 simulations to cover the variation arising
from fuzzy variables, such as survival and height of
seedlings, which are defined by a function. The mini-
mum requirement for the credibility indexes was
a ¼ 0:4. This means that in the aggregation process
only those criteria whose comparison resulted in an
index of credibility above the a threshold were included.
3. Results
3.1. Vegetation clearing
The different treatments to clear vegetation showed
significant differences in the reduction of vegetation
cover and height (Table 2). The highest reduction in
vegetation cover was obtained with the controlled
burning treatment, while the highest reduction in
height was obtained with the mechanical clearing
(Fig. 2). The reduction in vegetation cover (but not
height) varied with zone (41–33%). The interaction
between vegetation clearing treatment and zone was
also significant for vegetation height, because differ-
ences among zones were higher for the mechanical
clearing treatment.
One year after vegetation clearing, there were sig-
nificant differences in vegetation recovery among
experimental treatments (Table 3). The grazing treat-
Table 1
Expression, trend, units and thresholds of indifference (m ¼¼), slight indifference (m ¼), preference/rejection (m h&mi) and strong preference/
rejection (m hh&mii) for the criteria used in the multi-criteria analysisa
Criteria Expression Trend Unit m ¼¼ m ¼ m h&mi m hh&mii
Financial cost Numeric Minimise Euro 60 90 105 120
Survival of P. nigra Fuzzy Maximise Number 35 70 105 140
Height growth of P. nigra Fuzzy Maximise cm per year 2 4 6 8
Diversity of small mammals Linguistic Maximise – – – – –
Density of small mammals Linguistic Maximise – – – – –
a Note that preference or rejection will depend upon the trend of the criteria (respectively, maximise or minimise). Number, number of
seedlings/ha.
Table 2
F values from ANOVA tests of effects of vegetation clearing
treatment, zone and plot (nested within plot) on the decrease of
vegetation cover and height per transect just after the elimination
treatmenta
Factor d.f. Vegetation
cover
Vegetation
height
Vegetation clearing (VC) 2 119.3 52.7
Zone (Z) 2 8.3 1.6
VC � Z 4 2.1 3.8
Plot (zone) 3 7.1 1.6
a Significant coefficients (at a ¼ 0:05 after applying the
sequential Bonferroni method) are indicated in bold.
Table 3
F values from ANOVA tests of effects of vegetation clearing
treatment (grazing, fire, mechanical clearing and control), zone and
plot (nested within plot) on the decrease of vegetation cover and
height per transect 1 year after the elimination of vegetationa
Factor d.f. Vegetation
cover
Vegetation
height
Vegetation clearing (VC) 2 7.5 9.9
Zone (Z) 2 11.0 1.1
VC � Z 4 3.3 0.9
Plot (zone) 3 8.6 0.2
a Significant coefficients (at a ¼ 0:05 after applying the
sequential Bonferroni method) are indicated in bold. Data of
vegetation height were log-transformed.
J.M. Espelta et al. / Forest Ecology and Management 180 (2003) 185–198 191
ment showed the lowest reduction in vegetation cover,
and even an increase in vegetation height, in both cases
higher than control plots, while plots where the
mechanical clearing and burning treatments were
applied showed the lower recovery (Fig. 3). Zone
and the interaction between zone and clearing treat-
ment were significant for vegetation cover, because
vegetation recovery for the burning treatment was
higher in Can Armengol than in the other two zones.
3.2. Seedling survival and height growth in the
plantation plots
All factors and most of their interactions affected
seedling survival (Table 4). Percentage of seedling
survival was higher in mechanical cleared plots
(mean � standard error: 57:0 � 1:5%) and in control
plots (55:7 � 1:4%), than in grazed (52:1 � 1:2%) or
burned plots (50:0 � 1:3%). Moreover, seedling sur-
vival was higher in plots where ripping was conducted
(57:0 � 0:9%) than in those with planting holes
(50:3 � 1:0%), and was also different in the three
zones included in the study. Most of the interactions
between these factors were also significant. Thus,
according to the interaction vegetation clearing�soil preparation (Table 4) seedling survival was higher
in the ripper treatment than in planting holes for all
vegetation clearing treatments except the control
(Fig. 4). In the control treatment there were also
different patterns in the three zones of study. The
three zones also showed significant differences for
the two soil preparation treatments. Thus, seedling
survival decreased from 63:6 � 0:9% in 1999 to
43:7 � 1:0% in 2000, although this factor also showed
Fig. 2. Reduction of (A) vegetation cover and (B) vegetation
height after the three treatments used to eliminate vegetation.
Vertical bars extend over þ1S.E. of the mean. Different letters
indicate significant differences among treatments according to the
Fisher PSLD post-hoc test.
Fig. 3. Increment (or reduction) of (A) vegetation cover and (B)
vegetation height 1 year after the elimination of vegetation. Vertical
bars extend over þ1S.E. of the mean. Different letters indicate
significant differences among treatments according to the Fisher
PSLD post-hoc test.
192 J.M. Espelta et al. / Forest Ecology and Management 180 (2003) 185–198
significant interactions with the other three. Thus, the
highest survival was found in the mechanical clearing
treatment in the first year, and in the control treatment
in the second. The significant interactions of higher
order were of difficult interpretation.
Seedling height growth also varied for all the main
factors considered (Table 4). Seedling height growth was
lower in plots where vegetation had been cleared by
controlled burning (6:8 � 0:2 cm) than in the other three
types (M: 7:5 � 0:2 cm; G: 8:1 � 0:2 cm and C:
8:1 � 0:2 cm); it was also lower in plots where soil
preparation was carried out through ripping (7:2�0:1 cm) than in those which used planting holes
(8:1 � 0:2 cm), and was also different in the three zones.
There were also significant differences between years
(year 1999: 7:4 � 0:1 cm; year 2000: 7:8 � 10:2 cm).
3.3. Seedling density in the sowing plots
No significant differences were found among plots
with different type of vegetation clearing (two-way
ANOVA, P ¼ 0:41), among zones (P ¼ 0:15) or for
Table 4
F values from ANOVA tests of effects of vegetation clearing
treatment, soil preparation treatment, zone and year after
plantation (1999 and 2000) on seedling survival and seedling
height growtha
Factor d.f. Seedling
survival
Seedling height
growth
Vegetation clearing (VC) 3 7.7 6.0
Soil preparation (SP) 1 32.4 9.0
Zone (Z) 2 228.2 6.3
Year after plantation (Y) 1 1101.1 8.2
VC � SP 3 33.7 0.1
VC � Z 6 16.3 2.8
SP � Z 2 15.9 1.7
VC � Y 3 12.2 0.6
SP � Y 1 1.9 4.7
Z � Y 2 52.4 3.4
VC � SP � Z 6 21.8 1.3
VC � SP � Y 3 16.3 3.8
VC � Z � Y 6 7.7 0.9
SP � Z � Y 2 11.1 0.1
VC � SP � Z � Y 6 2.1 0.4
a Significant coefficients (at a ¼ 0:05 after applying the
sequential Bonferroni method) are indicated in bold.
Fig. 4. Variations in seedling survival in plots where vegetation had been eliminated by the different treatments according to the soil
preparation methodology used. Vertical bars extend over þ1S.E. of the mean.
J.M. Espelta et al. / Forest Ecology and Management 180 (2003) 185–198 193
the interaction between both factors (P ¼ 0:92) for
the broadcast sowing treatment. In these plots, seed-
ling establishment was very low (range: 0–200 seed-
lings/ha). Similar results were found for the
comparison between the two types of seed distribution
in plots with different type of vegetation clearing: no
differences were found between seed distribution
types (three-way ANOVA, P ¼ 0:10; broadcast dis-
tribution: 7 � 4 seedlings/ha; spot distribution: 80�36 seedlings/ha), vegetation clearing treatments (P ¼0:18) or zones (P ¼ 0:12). None of the interactions
between these variables was significant (P > 0:50 in
all cases).
3.4. Multi-criteria analysis of the different
reforestation practices
Table 5 shows the alternative � criteria matrix for
the NAIADE comparison of the different reforestation
methods. The analysis including only financial costs,
seedling survival and seedling growth (Fig. 5A), indi-
cated soil ripping and preserving vegetation (control)
(C þ R) or mechanical clearing (M þ R) as the best
alternatives, while mechanical clearing and planting
holes (M þ P) obtained the worst score. In spite of
differences on seedling survival, this ranking of alter-
natives was mainly driven by the economic differences
between treatments of ripping (cheaper) and planting
holes (much more expensive) (Table 5). Thus, all
alternatives including ripping as the soil preparation
method were placed at the top of the ranking list, while
those including planting holes were at the bottom.
When the impact of treatments on the density and
diversity of small mammals was included, some
important changes occurred in the ranking list
(Fig. 5B): (i) C þ R appeared as the most preferred
alternative, (ii) B þ R and M þ R went down in the
ranking list, due to their negative impact on small
mammals, (iii) C þ P raised to the third position in
spite of its high financial cost, and (iv) B þ P and
M þ P obtained again the worst scores, due to their
negative ecological impact and high costs.
4. Discussion
In most Mediterranean countries, reforestation is
the most widespread large-scale restoration practice
aimed at reducing fire effects (Vallejo and Alloza,
1998). Reforestation through seeding was the com-
monest way to restore pine forests until the mid-
seventies, when increased possibilities of mechanical
site preparation as well as improved seedling produc-
tion led to a shift towards plantation practices (Castell
and Castello, 1996). Notwithstanding, in recent years
the need to reforest extensive burned areas has
renewed interest in comparing different seeding and
planting strategies (Moreno and Vallejo, 1999).
In our study, the final establishment of P. nigra
seedlings 2 years after the experiment onset ranged
drastically from 7 � 4 seedlings/ha in the broadcast
seeding treatments to 610 � 40 seedlings/ha in the
plantations. Establishment after the seeding treatment
was much lower than the only comparable experience
we know in Catalonia: results obtained by Castell and
Castello (1996) in an aerial seeding experience with P.
halepensis immediately after the Garraf (Barcelona,
NE Spain) 1994 wildfire that gave densities of 830–
7200 seedlings/ha. The discrepancy observed may be
attributed to the fact we seeded 4 years after the fire,
Table 5
Alternative � criteria matrix used to compare the different reforestation methods essayed in the studya
G þ R G þ P B þ R B þ P M þ R M þ P C þ R C þ P
Financial costb 1020 1500 1200 1680 1320 1800 1020 1500
Survival of P. nigra a560 a630 a644 a490 a798 a434 a602 a728
Height growth of P. nigra a7 a8 a6 a7 A7 a9 a7 a8
Diversity of small mammals �Good �Good �Bad �Bad Bad Bad �Good �Good
Density of small mammals Good Good �Bad �Bad Moderate Moderate Good Good
a Numbers preceded by a symbol (a: approximately) indicate a fuzzy variable. G, livestock grazing; B, controlled burning; M, mechanical
clearing; C, control; R, ripping and P, planting holes.b Financial cost includes the costs of clearing vegetation (burning or mechanically), soil preparation (ripping or holes), cost of seedlings
and planting.
194 J.M. Espelta et al. / Forest Ecology and Management 180 (2003) 185–198
Fig. 5. Ranking of alternatives according to: (A) financial cost, survival and height growth of P. nigra seedlings and (B) adding the impact on the diversity and density of the
community of small mammals. Left values (fþ), ranking based on the better and much better relations (values range from 0 to 1, indicating how an alternative is better than the
rest). Right numbers (f�), ranking based on the worse and much worse relations (values range from 0 to 1, indicating how an alternative is worse than the rest).
J.M.
Esp
eltaet
al./F
orest
Eco
log
ya
nd
Ma
na
gem
ent
18
0(2
00
3)
18
5–
19
81
95
once ground vegetation had achieved a cover of
77–91%. Moreover, although seeds of P. nigra were
dusted with an insect repellent, we observed a great
amount of predation by granivorous ants (Messor
spp.), which had already massively colonised the area
(pers. observation). The failure of the broadcast seed-
ing assay points out that, although it may be recom-
mended to wait a precautionary time to observe
whether natural regeneration occurs after a wildfire
(Espelta, 1999), the fast recovery of ground vegeta-
tion, as well as some animal groups, severely threatens
the success of broadcast seeding treatments.
The establishment of seedlings of P. nigra obtained
in the plantation experiences is low in comparison to
the success of plantations reported in temperate and
boreal forests (Graham et al., 1989; Gemmel et al.,
1996; Archibold et al., 2000). However, they are in the
range of other experiences carried out in Mediterra-
nean environments (Vallejo and Alloza, 1998), where
the transplant-shock to a very harsh environment and
the following chronic water stress usually cause low
survival rates. Mortality linked to the early post-
transplant period is especially high in Mediterranean
environments, because of the poor ability of the root
system of the newly established seedlings to replace
water lost through transpiration (Baeza et al., 1991). In
addition, survival rates usually further decrease after
the first summer drought (Vallejo et al., 2000). For
these reasons, although global survival of P. nigra
accounted roughly for 50% of the initial density
(610 seedlings/ha), this result allows us to predict that
development of P. nigra forests in the area will be
possible at a medium term scale (see Retana et al.,
2002).
In our study, vegetation clearing prior to planting
showed a rather low impact, as survival after 2 years
was quite similar in all plots, with even slightly higher
values in control than in treated plots. This lack of
major effects of vegetation clearing could be linked to
two factors. First, the fast re-growth of the cleared
vegetation through resprouting led all plots to present
after 2 years a rather similar and high cover (more than
75%), the initial differences among treatments vanish-
ing (between 40% in the mechanical clearing and
100% in the control). Second, it is important to note
that in a harsh environment such as the Mediterranean
climate, rather than competing, vegetation could act as
a shelter for young seedlings, protecting them from
excessive solar radiation and water stress (see exam-
ples of this nursery effect in Jobidon et al. (1998)).
Notwithstanding this general trend, the type of vege-
tation clearing interacted with the soil preparation
method. Linear ripping has been reported as an excel-
lent way to ameliorate water availability and nutrient
release for seedlings by retaining runoff, increasing
the total water stored in the profile and enhancing the
possibilities of the root systems to penetrate and obtain
water and nutrients (Varelides and Kritikos, 1995;
Querejeta et al., 2001). However, to capture run off
losses it will be important to break soil crusts and
obtain regular and deep plowing rips in all the surface
(Peman and Navarro, 1998). In our study areas, rip-
ping was much more easily and regularly conducted in
plots where vegetation was previously burned or
mechanically cleared, in comparison to the grazed
or control areas, where shrubs obstructed ripping. In
contrast, in those areas the punctual planting holes
carried out by the excavator produced deep and regular
planting holes. We hypothesise that these differences
in soil preparation are responsible for the differences
in seedling survival.
The previous stated differences among reforestation
alternatives give us a first insight about the best way to
tackle recovery of extensively burned P. nigra forests.
However, possible emerging recommendations turn
over when considering differences in economic cost
and ecological impact throughout the NAIADE multi-
criteria analysis. Thus, combination of a low economic
cost and low ecological impact makes control plots
with soil preparation through ripping the top ranked
alternative, while burned and mechanical cleared plots
with planting holes would be the less recommended
ones. The two analysis run allowed us to note that
burning/mechanical clearing and ripping were top
ranked when only financial costs and survival and
growth of P. nigra seedlings were considered, but fell
down when the negative impact on small mammals
was included: in control plots, density of small mam-
mals increased 100% 2 years after the experiment
onset, but this increase only accounted for 30% in the
mechanically cleared plots and even decreased a 5% in
the burned plots (Comas et al., 2001). Moreover, in
these cleared plots, the community of small mammals
was dominated by a single species (Apodemus sylva-
ticus), while in control plots five different species
were found (Comas et al., 2001). These results allow
196 J.M. Espelta et al. / Forest Ecology and Management 180 (2003) 185–198
us to draw a major conclusion: vegetation clearing,
although a common sense practice extensively repor-
ted in reforestation projects, would be meaningless
in those sub-Mediterranean forests because of its
relative low importance in terms of seedling survival
but high economic cost and impact on the small
mammals community. Economic costs and ecological
values (e.g. protection of the community of small
mammals) have been usually placed at opposite
extremes (see Jackson et al., 1995). However, in our
case their optimisation involves similar reforestation
alternatives.
Restoration represents a long term dedication of
natural resources and a substantial financial commit-
ment (Ehrenfeld, 2000). Practitioners need to know
how to allocate this limited funding, labour and time
for maximum effect in the design of restoration pro-
grams of large-scale projects (Clewell and Rieger,
1997). In that framework, 4 years after the fire,
plantation rather than seeding arises as the most
suitable reforestation practice (with the present tech-
nological limitations of seeding). On the other hand,
vegetation clearing does not match the expected ben-
efits in the performance of young seedlings. Finally,
the multi-criteria analysis carried out provides a valu-
able example that although restoration activities are
inextricably embedded within an economic frame-
work (Edwards and Abivardi, 1997), it is possible
to reconcile such scope with ecological criteria (see
Holl and Howarth, 2000) when choosing among dif-
ferent alternatives.
Acknowledgements
Thanks are given to Giussepe Munda for helpful
comments and assistance in running the NAIADE
program, to Lluıs Comas for relevant information
about the dynamics of the small mammals community
in the study area and to Jose Luıs Ordonez for field
assistance. We are grateful to Anselm Rodrigo and
Ramon Vallejo and two anonymous reviewers for their
very useful comments on the manuscript. This
research was partly funded by INTERREG II (EU)
project Reconstruction of forest landscapes affected
by large wildfires events, the Department of Environ-
ment (Generalitat de Catalunya) and the INIA project
SC98-070.
References
Archibold, O.W., Acton, C., Ripley, E.A., 2000. Effect of site
preparation on soil properties and vegetation cover, and the
growth and survival of white spruce (Picea glauca) seedlings in
Saskatchewan. For. Ecol. Manage. 131, 127–141.
Ashby, W.C., 1997. Soil ripping and herbicides enhance tree and
shrub restoration on stripmines. Restor. Ecol. 5, 169–177.
Baeza, J.M., Pastor, A., Martın, J., Ibanez, M., 1991. Mortalidad
postimplantacion en repoblaciones de Pinus halepensis,
Quercus ilex, Ceratonia siliqua y Tetraclinis articulata en la
provincia de Alicante. Studia Oecolog. 8, 139–146.
Barrett, G.W., Peles, J.D., 1999. Landscape ecology of small
mammals. Springer, Heidelberg.
Castell, C., Castello, J.I., 1996. Metodologıa y resultados de la
siembra aerea efectuada en el Parque Natural del Garraf.
Montes 46, 51–57.
Clewell, A., Rieger, J.P., 1997. What practitioners need from
restoration ecologists. Restor. Ecol. 5, 350–354.
Comas, Ll., Retana, J., Espelta, J.M., 2001. Efecto de los
tratamientos silvıcolas para la reforestacion de zonas afectadas
por grandes incendios sobre las poblaciones de microma-
mıferos. V Jornadas de la Sociedad Espanola de Conservacion
y Estudio de Mamıferos, Vitoria-Gasteiz, Spain, December
6–10.
Ducey, M.J., Larson, B.C., 1999. A fuzzy set approach to the
problem of sustainability. For. Ecol. Manage. 115, 29–40.
Edwards, P.J., Abivardi, C., 1997. Ecological engineering and
sustainable development. In: Urbanska, K.M., Webb, N.R.,
Edwards, P.J. (Eds.), Restoration Ecology and Sustainable
Development. Cambridge University Press, New York, pp.
325–352.
Ehrenfeld, J.G., 2000. Defining the limits of restoration: the need
for realistic goals. Restor. Ecol. 8, 2–9.
Espelta, J.M., 1999. La reconstruccio de paisatges forestals afectats
per grans incendis. Projecte pilot a l’incendi del Bages-
Bergueda del 1994. Silvicultura 27, 6–9.
Etienne, M., Derzko, M., Rigolot, E., 1996. Browse impact in
silvopastoral systems participating in fire prevention in the
French Mediterranean region. In: Etienne, M. (Ed.), Western
European Silvopastoral Systems. INRA Editions, Paris, pp. 93–
102.
Ferran, A., Vallejo, V.R., 1998. Long-term plant regeneration after
wildfires in Mediterranean ecosystems of NE Spain. In:
Trabaud, L. (Ed.), Fire Management and Landscape Ecology.
International Association of Wildland Fire, pp. 155–165.
Flemming, R.L., Black, T.A., Adams, R.S., 1996. Site preparation
effects on Douglas-fir and lodgepole pine water relations
following planting in a pine-grass dominated clearcut. For.
Ecol. Manage. 83, 47–60.
Gemmel, P., Nilsson, U., Welander, T., 1996. Development of oak
and beech seedlings planted under varying shelterwood
densities and with different site preparation methods in
Sweden. New For. 12, 141–161.
Gracia, C., Burriel, J.A., Ibanez, J.J., Mata, T., Vayreda, J., 2000.
Inventari ecologic i forestal de Catalunya. Regio forestal IV.
Centre de Recerca Ecologica i Aplicacions Forestals, Barcelona.
J.M. Espelta et al. / Forest Ecology and Management 180 (2003) 185–198 197
Graham, R.T., Harvey, A.E., Jurgensen, M.F., 1989. Effects of site
preparation on survival and growth of Douglas-fir (Pseudotsuga
menziessi Mirb. Franco.) seedlings. New For. 3, 89–98.
Habrouk, A., Retana, J., Espelta, J.M., 1999. Role of heat tolerance
and cone protection of seeds in the response of three pine
species to wildfires. Plant Ecol. 145, 91–99.
Harvey, A.E., Page-Dumroese, D.S., Jurgensen, M.F., Graham,
R.T., Tonn, J.R., 1996. Site preparation alters biomass, root and
ectomycorrhizal development of outplanted western white pine
and Douglas-fir. New For. 11, 255–270.
Holl, K.D., Howarth, R.B., 2000. Paying for restoration. Restor.
Ecol. 8, 260–267.
Jackson, L., Lopoukhine, N., Hillyard, D., 1995. Ecological
restoration: a definition and options. Restor. Ecol. 3, 71–75.
Jobidon, R., Charette, L., Bernier, P.Y., 1998. Initial size and
competing vegetation effects on water stress and growth of
Picea mariana (Mill.) BSP seedlings planted in three different
environments. For. Ecol. Manage. 103, 293–305.
JRC-Ispra, 1996. NAIADE v. 1.0. Institute for Systems, Informatics
and Safety, Ispra.
Moreno, J.M., Vallejo, V.R., 1999. Fire impacts on the ecosystem
and restoration: summary of the main findings from the delfi-
fire database. In: Volume of Proceedings of Forest Fires: Needs
and Innovation. International Symposium. Athens, Greece,
pp. 239–261.
Munda, G., 1995. Multicriteria Evaluation in a Fuzzy Environment:
Theory and Applications in Ecological Economicals. Physica-
Verlag, Heidelberg.
Peman, J., Navarro, R., 1998. Repoblaciones forestales. Publ.
Universitat de Lleida, Lleida.
Querejeta, J.I., Roldan, A., Albadalejo, J., Castillo, V., 2001. Soil
water availability improved by site preparation in a Pinus
halepensis afforestation under semiarid climate. For. Ecol.
Manage. 149, 115–128.
Retana, J., Espelta, J.M., Habrouk, A., Ordonez, J.L., Sola-
Morales, F., 2002. Regeneration patterns of three Mediterra-
nean pines and forest changes after a large wildfire in
northeastern Spain. Ecoscience 9, 89–97.
Rice, W.R., 1989. Analyzing tables of statistical tests. Evolution
43, 223–225.
Salvador, R., Valeriano, J., Pons, X., Dıaz-Delgado, R., 2000. A
semi-automatic methodology to detect fire scars in shrubs and
evergreen forests with Landsat MSS time series. Int. J. Remote
Sens. 4, 655–671.
Serrada, R., 1990. Consideraciones sobre el impacto de la
repoblacion forestal en el suelo. Ecologıa 1, 435–462.
Trabaud, L., Campmant, C., 1991. Difficulte de recolonisation
naturelle du pin de Salzmann Pinus nigra ssp. salzmannii
(Dunal) Franco Apres incendie. Biol. Cons. 58, 329–343.
Vallejo, V.R., Alloza, J.A., 1998. The restoration of burned lands:
the case of eastern Spain. In: Moreno, J.M. (Ed.), Large Forest
Fires. Backhuys Publishers, Lieden, pp. 91–108.
Vallejo, V.R., Bautista, S., Cortina, J., 2000. Restoration for soil
protection after disturbances. In: Trabaud, L. (Ed.), Life and
Environment in the Mediterranean. WIT Press, Boston, pp. 301–
343.
Varelides, C., Kritikos, T., 1995. Effect of site preparation intensity
and fertilization on Pinus pinaster survival and height
growth on three sites in northern Greece. For. Ecol. Manage.
73, 111–115.
198 J.M. Espelta et al. / Forest Ecology and Management 180 (2003) 185–198