Content uploaded by Adrián Regos
Author content
All content in this area was uploaded by Adrián Regos on Jul 02, 2020
Content may be subject to copyright.
Content uploaded by Adrián Regos
Author content
All content in this area was uploaded by Adrián Regos on Jun 18, 2020
Content may be subject to copyright.
Contents lists available at ScienceDirect
Ecosystem Services
journal homepage: www.elsevier.com/locate/ecoser
Mountain farmland protection and fire-smart management jointly reduce
fire hazard and enhance biodiversity and carbon sequestration
Silvana Pais
a,b
, Núria Aquilué
c,d
, João Campos
a
, Ângelo Sil
a,f,g
, Bruno Marcos
a
,
Fernando Martínez-Freiría
a
, Jesús Domínguez
e
, Lluís Brotons
d,h
, João P. Honrado
a,i
,
Adrián Regos
a,e,⁎
a
InBIO/CIBIO – Centro de Investigação em Biodiversidade e Recursos Genéticos, Campus Agrário de Vairão, Rua Padre Armando Quintas, n° 7, 4485-661 Vairão, Portugal
b
proMetheus, Research Unit in Materials, Energy and Environment for Sustainability, Instituto Politécnico of Viana do Castelo, 4990-706 Ponte de Lima, Portugal
c
Centre d’Étude de la Forêt, Université du Québec à Montréal, Montréal, QC, Canada
d
InForest JRU (CTFC-CREAF), Crta. Antiga St Llorenç de Morunys km 2, 25280 Solsona, Catalonia, Spain
e
Departamento de Zooloxía, Xenética e Antropoloxía Física, Universidade de Santiago de Compostela, 15782 Santiago de Compostela, Spain
f
CITAB – Centro de Investigação e de Tecnologias Agro-Ambientais e Biológicas, Universidade de Trás-os-Montes e Alto Douro, Apartado 1013, 5001-801, Portugal
g
CIMO – Centro de Investigação de Montanha (CIMO), Instituto Politécnico de Bragança, Campus de Santa Apolónia, 5300-253 Bragança, Portugal
h
CSIC, 08193 Cerdanyola del Vallès, Spain
i
Faculdade de Ciências da Universidade do Porto, Rua do Campo Alegre, S/N, Edifício FC4, 4169-007 Porto, Portugal
ARTICLE INFO
Keywords:
Biomod2
Fire-smart landscape management
Fire suppression
InVEST model
Land abandonment
REMAINS model
Wildfires
ABSTRACT
The environmental and socio-economic impacts of wildfires are foreseen to increase across southern Europe over
the next decades regardless of increasing resources allocated for fire suppression. This study aims to identify fire-
smart management strategies that promote wildfire hazard reduction, climate regulation ecosystem service and
biodiversity conservation. Here we simulate fire-landscape dynamics, carbon sequestration and species dis-
tribution (116 vertebrates) in the Transboundary Biosphere Reserve Gerês-Xurés (NW Iberia). We envisage 11
scenarios resulting from different management strategies following four storylines: Business-as-usual (BAU),
expansion of High Nature Value farmlands (HNVf), Fire-Smart forest management, and HNVf plus Fire-Smart.
Fire-landscape simulations reveal an increase of up to 25% of annual burned area. HNVf areas may counter-
balance this increasing fire impact, especially when combined with fire-smart strategies (reductions of up to 50%
between 2031 and 2050). The Fire-Smart and BAU scenarios attain the highest estimates for total carbon se-
questered. A decrease in habitat suitability (around 18%) since 1990 is predicted for species of conservation
concern under the BAU scenario, while HNVf would support the best outcomes in terms of conservation. Our
study highlights the benefits of integrating fire hazard control, ecosystem service supply and biodiversity con-
servation to inform better decision-making in mountain landscapes of Southern Europe.
1. Introduction
Wildfires are a major component of disturbance regimes worldwide
(Keeley et al., 2012). Despite the increasing amount of resources in-
vested in fire suppression, the number of extreme fire events has largely
increased over the last decades in southern Europe, overriding current
fire-suppression systems (San-Miguel-Ayanz et al., 2013). In southern
Europe, Spain, Greece and Portugal are the countries most affected by
wildfires, both in terms of fire occurrence and total burned area (see
Gonçalves and Sousa, 2017). In 2017, more than one hundred people
died in Portugal due to large forest fires that overtook fire-fighting
capabilities (Tedim et al., 2018). Although the total burned area in
much of the Mediterranean region has decreased in recent years (Turco
et al., 2016), extreme fires (i.e. large fires burning at high intensities)
have become more frequent —which entails severe environmental and
socio-economic impacts (Tedim et al., 2013). This increase in the
number of extreme wildfires is due to increasing stand-level fuel ac-
cumulation and landscape-level fuel connectivity caused by long-
standing land abandonment processes (which favor vegetation en-
croachment and forest densification) (Moreira et al., 2011), and more
adverse climatic conditions (i.e., longer periods with hot and dry con-
ditions, and high-speed winds) (Tedim et al., 2013; Turco et al., 2019).
https://doi.org/10.1016/j.ecoser.2020.101143
Received 9 March 2020; Received in revised form 15 June 2020; Accepted 16 June 2020
⁎
Corresponding author at: InBIO/CIBIO – Centro de Investigação em Biodiversidade e Recursos Genéticos, Campus Agrário de Vairão, Rua Padre Armando Quintas,
n° 7, 4485-661 Vairão, Portugal (Adrián Regos).
E-mail address: adrian.regos@usc.es (A. Regos).
Ecosystem Services 44 (2020) 101143
2212-0416/ © 2020 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/BY-NC-ND/4.0/).
T
Agricultural abandonment has shaped rural mountain areas in many
parts of the Mediterranean Europe since the last century, owing to di-
verse socio-economic and biophysical constraints such as reduced job
opportunities, poor generational renewal, low accessibility and soil
productivity (Cerqueira et al., 2010; MacDonald et al., 2000;Lasanta
et al., 2016). The cessation of traditional livestock and agricultural
practices caused by rural exodus has favored more homogeneous and
flammable landscapes (Moreira et al., 2011) —with strong side-effects
on fire regime, ecosystem services and biodiversity (van der Zanden
et al., 2017). Rural communities traditionally used fire as a tool to
manage mountain landscapes (for pastoral activities among other mo-
tivations; Chas-Amil et al., 2015; Tedim et al., 2016), being common a
high frequency of low-intensity small-sized fires (Catry et al., 2009;
Chas-Amil et al., 2010). The increasing amount of available fuel (under
warmer and drier climatic conditions; Tedim et al., 2018) together with
a high number of fire ignitions have led to altered fire regimes
(Fernandes, 2013). In addition, fire-suppression policies characterized
by a rigid response to fire occurrence and fire exclusion (i.e. trying to
eliminate fires from the landscape) are still largely prevalent in
southern Europe, indirectly contributing to foster extreme wildfires (the
also-known ‘Fire paradox’) (Fernandes et al., 2016c). The fire-vegeta-
tion feedbacks and their complex interactions with fire-suppression
policies and land-use changes make landscape dynamics difficult to
predict, which challenges decision-making due to the large uncertainty
of alternative management scenarios.
The progressive loss of traditional low-intensity farming systems in
Southern Europe, sustaining what has been called ‘High Nature Value
farmlands’ (HNVf, defined as socio-ecological systems underlying the
maintenance of low-intensity farming systems supporting the occur-
rence of several species and habitats; Lomba et al., 2015), has been
widely associated with population declines of many wild species
(Ribeiro et al., 2014). Species associated with wet grasslands, pastures
and low-intensively managed agricultural lands are the most negatively
affected by land abandonment processes across Europe (e.g. grassland
waterbirds and farmland bird species; Franks et al., 2018; Lehikoinen
et al., 2018). On the contrary, wilderness and forest-dwelling species,
including emblematic animals such as wolfs, bears or eagles, strongly
benefit from land abandonment (Navarro and Pereira, 2012). This has
represented a conservation opportunity in areas that became no longer
viable or attractive from a socio-economic viewpoint (see ‘ecological
rewilding’ concept, Navarro and Pereira, 2012). In relation to eco-
system services, land abandonment has boosted climate regulation (e.g.
carbon storage and sequestration) (Sil et al., 2017), recreational (e.g.
birdwatching or ecotourism), wood provision and water regulation
services provided by forest ecosystems (Carvalho-Santos et al., 2015;
Cerqueira et al., 2015; Sil et al., 2016). However, agricultural aban-
donment also decreases the fire regulation capacity and the fire pro-
tection ecosystem service in mountain landscapes (Sil et al., 2019b),
among other ecosystem services such as food provision, pest and dis-
ease control (Renard et al., 2015).
Various fuel-treatment practices (such as prescribed burning, me-
chanical treatments such as forest thinning or mastication) have been
proposed over the last decades to reduce fuel quantity, fuel continuity,
and the associated risk of high-severity forest fires (Agee and Skinner,
2005; Omi, 2015). However, the environmental and economic sus-
tainability of these forest practices is not always considered as they are
only designed to cope with wildfires (Mclver et al., 2013). In fact, the
challenge for managers and policy makers is no longer simply how to
reduce wildfire impacts but how to reconcile socio-economic impacts of
fires with their ecological benefits (Pausas and Keeley, 2019; Sil et al.,
2019a). Fire-smart management (defined as “as an integrated approach
primarily based on fuel treatments through which the socio-economic im-
pacts of fire are minimized while its ecological benefits are maximized”;
Hirsch et al., 2001) has emerged as an promising option to incorporate
the role of fire as (socio-)ecological process into strategic planning to
achieve a more sustainable coexistence with wildfires (Fernandes,
2013; Hirsch et al., 2001; Tedim et al., 2016). Fire-smart management
would clearly enable a more balanced integration of positive (e.g. re-
ducing species competition, diseases and pests or fire intensity, and
increase fire protection in wildland-urban interfaces; Pausas and
Keeley, 2019) and negative contributions of fire to human well-being,
which would inform better decision making in fire management policy
and land-use planning (Sil et al., 2019a). In practice, fire-smart land-
scapes can be obtained by fuel-reduction treatments and by fuel type
conversion, rather than by fuel isolation (Fernandes, 2013). From this
perspective, proactive management should therefore focus on reshaping
vegetation (fuel) configuration to foster more fire-resistant and/or fire-
resilient landscapes (Fernandes, 2013) while simultaneously ensuring
the long-term supply of ecosystem services and the conservation of
biodiversity (Hirsch et al., 2001; Tedim et al., 2016). To the best of our
knowledge, and despite the advantages of integrated fire management
for decision-making, there are no studies assessing the effects of fire-
smart landscape management on ecosystem services and biodiversity
conservation.
The present study aims to identify ‘win-win’ situations to reduce the
impact of wildfires and maximize the provision of carbon storage and
sequestration and biodiversity conservation in fire-prone regions af-
fected by rural abandonment. In particular, we assessed the potential
trade-offs between wildfire mitigation (measured through total burned
and suppressed area), climate regulation ecosystem services (i.e.,
carbon storage and sequestration) and biodiversity conservation (i.e.,
habitat availability for 116 vertebrate species) under fire-smart man-
agement scenarios in a transboundary (Spain-Portugal) mountain re-
gion severely affected by rural abandonment and wildfires. This study
illustrates the potential of wildfire-landscape dynamic model simula-
tions to support more informed decisions for fire-suppression and
landscape planning.
2. Material and methods
2.1. Study area
The study was conducted in the Transboundary Biosphere Reserve
Gerês-Xurés (hereafter BR-GX) (ca. 276,000 ha, of which 71% in
Portugal and the remaining 29% in Spain), a representative mountain
landscape of NW Iberian Peninsula (Fig. 1). This mountainous area is
covered by a complex hydrographic network running on a rugged relief,
with an elevation ranging from 15 m to 1,545 m, composed of deep
valleys, plains and steep slopes (Regos et al., 2015). The region is lo-
cated at the transition between the Mediterranean and Eurosiberian
(Temperate) biogeographic zones, close to the Atlantic coast. The study
area includes the entire reserve, encompassing three EU Natura 2000
sites besides two nationally designated protected areas, the Peneda-
Gerês National Park in Portugal and the Baixa Limia - Serra do Xurés
Natural Park in Spain. Although our study is conducted in the entire
Biosphere Reserve, we intend to discern the management impacts both
within and outside protected areas (i.e. National Parks in Portugal and
Spain), given the differences between both areas in terms of socio-
economic values and protection measures, which would influence how
the different management strategies could be implemented.
Like in other mountain areas in Mediterranean Europe, population
has decreased in the BR-GX by 28% between 1990–2010, with declines
of up 50% in the Spanish side (www.ine.pt and www.ine.es). This de-
population together with the ageing of the remaining farmers has been
accompanied by the abandonment of traditional agricultural (with
losses of approximately 10,000 hectares between 1990 and 2010) and
livestock activities (including burning, grazing and extensive agri-
culture) in the Biosphere Reserve (Regos et al., 2015, and Appendix B)
—as it has been reported for other mountain landscapes of NW Iberian
Peninsula (Morán-Ordóñez et al., 2013). The landscape in the study
area is dominated by shrubs (broom, gorse and heath, c.a. 32% of the
study area) and sparsely vegetated areas (rocky areas with poor soils
S. Pais, et al. Ecosystem Services 44 (2020) 101143
2
and little vegetation, 25%), maintained by fire and extensive agropas-
toral activities; followed by a variety of fragmented forests, such as
deciduous woodlands (mostly represented by Quercus robur and Q.
pyrenaica; 18%) and coniferous plantations (dominated by Pinus syl-
vestris and P. pinaster; 11%) (Regos et al., 2015). Deciduous oaks (the
climax vegetation of the region) are less prone to fire than the more
flammable pine species (see Fernandes et al., 2016a and references
therein).
The study area is classified within the intermediate-cool-small
pyrome (characterized by intermediate fire return but fairly small fires,
see Archibald et al., 2013). Locally, the area is subjected to frequent
human-induced wildfires, linked to the profound socio-economic
changes suffered by these territories (e.g., vandalism, arson, revenge,
land use change attempts) (Calviño-Cancela et al., 2016; Chas-Amil
et al., 2015, 2010). Despite the large increase in fire-suppression re-
sources over the last 20 years, the fire regime in the study area is
characterized by large numbers of fire events and total burned area
(12,755 fires between 1983 and 2010, burning a total of 195,000 ha),
with areas burned up to 5–6 times (Regos et al., 2015). Between 1983
and 2010, a total of 54,041 hectares burned by around 9,512 fires in
Spain. In the Portuguese side, 3,243 fires were recorded, burning a total
of 141,038 hectares.
2.2. Modelling framework and management scenarios
2.2.1. Workflow
To quantify the potential impacts of alternative fire management
and land-use policies on fire regime, carbon storage and sequestration
and biodiversity, we coupled fire-landscape dynamic modelling with
species distribution and carbon sequestration models (Fig. 2). The fire-
landscape simulations allowed quantifying the impact of fire and land
management on fire regime (namely, burned and suppressed area)
under each management scenario, as well as the temporal dynamics of
the main land cover (LC) types. Historic land-use/cover maps and the
temporal projections obtained from the fire-landscape model were used
as inputs in species distribution and carbon models, calibrated for past
conditions (1990–2010) and projected to future landscape conditions
under each management scenario (2011–2050) (Fig. 2).
Fig. 1. Location of the Transboundary Biosphere Reserve Gerês-Xurés in the Iberian Peninsula, and the several protected areas over a digital elevation model of the
area. Coordinates are UTM, Zone 29 N.
S. Pais, et al. Ecosystem Services 44 (2020) 101143
3
2.2.2. The fire-landscape model
We used a spatially explicit process-based model (REMAINS) that
integrates the main factors driving fire-landscape dynamics in Southern
European mountain landscapes. The model allows investigating how
the spatiotemporal interactions between fire-vegetation dynamics, fire
management (i.e. fire-suppression strategies) and land-use changes af-
fect fire regime (and thereby landscape composition and dynamics) at
short- and medium-timescales. It also allows quantifying the effects of
fire management on total burned and suppressed area (calculated as the
difference between potential area to be burnt in a year and the final
burnt area). The REMAINS model reproduces fire-landscape dynamics
according to pre-designed scenario storylines (Table 1). In particular,
the model simulates wildfires (including fire ignition, spread, burning
and extinction), vegetation dynamics (i.e., natural succession and post-
fire regeneration), land-use changes (e.g. agriculture abandonment or
intensification) and forest management (e.g. increase of intensive
plantations for timber production) (details can be found in Appendix
A).
The model was implemented using the Spatially Explicit Landscape
Event Simulator (Fall and Fall, 2001), based on previous experiences
using similar approaches focusing on fire-vegetation dynamics (fire-
succession model MEDFIRE; Brotons et al., 2013; Duane et al., 2019)
and anthropogenic land-use changes (land-use/cover change model
MEDLUC (Aquilué et al., 2017). At each time step (1 year), the model
Fig. 2. Diagram of the modelling workflow including the fire-landscape, biodiversity and carbon modelling modules, the input data and the resulting outputs.
Table 1
Landscape and fire management storylines for the study area.
Name Storyline description
Business-as-usual – BAU It envisages a future landscape derived from the historical fire regime and land-use change trends reported between 1987 and 2010,
clearly dominated by land abandonment processes (Regos et al., 2015).
High Nature Value farmlands – HNVf Related to initiatives aimed at reverting farmland abandonment and mimicking EU environmental and rural policies on fire regime and
biodiversity conservation (Lomba et al., 2015; Moreira and Pe’er, 2018) in the BR-GX as a counterpoint of the current BAU scenario.
Fire-Smart It aims at controlling final burnt area by intervening on vegetation covers (e.g. promoting the gradual conversions of coniferous forests to
native oak woodlands) to foster more fire-resistant (less flammable) and/or fire-resilient landscapes (Fernandes, 2013). Assuming the
same amount of fire suppression resources applied nowadays, a more effective fire-suppression system would be expected due to lower
fire spread rates found in oaks than in coniferous forests (Fernandes, 2013).
HNVf + Fire Smart It envisages an integrated management policy that combines the promotion of more resistant and less flammable landscapes (‘Fire-smart’)
with policies aimed at gradually increasing agricultural areas (HNVf), as an opportunity for fire suppression and farmland/grassland
biodiversity conservation.
S. Pais, et al. Ecosystem Services 44 (2020) 101143
4
simulates fire ignition, spread and extinction until reaching a target
annual burnt area defined according to statistical data for each ad-
ministrative region of the study area (Portugal and Spain) between
1983 and 2010 (INCF, n.d.; MAPAMA, 2018). The target fire sizes are
also a model input, but the final fire size emerges from the spatial in-
teraction between the location of fire ignitions, landscape composition,
topography, and fire suppression. The probability of fire ignition is a
function of human-related and biophysical variables (see Appendix A
for details). The spread rate is formulated as a polynomial expression
with three factors (slope, aspect and fire-proneness of each LC type)
adapted from Duane et al. (2016). Two types of fire-suppression stra-
tegies are implemented: (1) ‘Active fire suppression’, in which sup-
pression of a fire front starts when the fire spread rate is below a spe-
cific threshold, mimicking the current capacity of fire brigades to
extinct low-intensity fires; and (2) ‘Passive fire suppression’, based on
opportunities derived from the presence of agricultural areas (set as
1 ha), which is assumed to be sufficient to interrupt the continuity of
highly flammable vegetation, thus mimicking the advantage that fire
brigades can take in heterogeneous landscape mosaics.
Land-use changes are modelled using a demand-allocation ap-
proach. In a demand-allocation framework the users set the demand (or
quantity of change) and the LULC change model uses a spatial proce-
dure to allocate the change (i.e. to select the cells to be transformed to
the target land-cover type) (Aquilué et al., 2017). The demand or
quantity-of-change by time step were based on a landscape change
analysis performed for the 1987–2010 period (details can be found in
Appendix B). Land-use/cover maps at 30-m resolution were derived
from satellite images of the Landsat archive by using supervised clas-
sification methods (see details in Regos et al., 2016). In particular, four
types of LC transitions are modelled per year: (1) market-oriented forest
plantations, the transition of scrublands to pine plantations; (2) ‘fire-
smart’ forest plantations, the transition of pine plantations to oak for-
ests; (3) rural abandonment, the conversion of crops and grasslands to
semi-natural vegetation areas identified as scrublands; and (4) agri-
cultural intensification, the conversion of scrublands to cultivated land.
Changes are simulated in locations with a higher likelihood to be
transformed to the target land-cover type. A transition-potential vari-
able accounts for this likelihood and is computed for each LC transition
adopting the neighbor factor approach introduced by Verburg et al.,
(2004) (see Appendix A).
2.2.3. Fire-smart landscape management scenarios
Framed within four major storylines (Table 1), we designed 11
scenarios that combine different land management strategies in fire-
prone landscapes (Table 2).
Scenario parameters regarding LC transitions were established
considering historic land-use/cover changes. The annual conversion
rate from cropland to scrubland (i.e. land abandonment rate) was set at
400 ha, according to the land-use change analysis between 1987 and
2010 (Table 2 and Appendix B). In ‘HNVf’ scenarios, the land aban-
donment rate was set to 0 and the annual conversion rate from scrub-
land to cropland was gradually increased from 400 to 1600 hectares
(Table 2). The natural succession rate (value of 1.6 in Table 2) was also
calculated from the land-use/cover change analysis, indicating the rate
at which scrubs are converted into deciduous woodlands. This para-
meter was progressively increased in ‘Fire-smart’ scenarios to favor the
conversion into oak woodlands. In ‘Fire-smart’ scenarios, the conver-
sion of coniferous into deciduous woodlands were also implemented to
convert half the total area or the total area of the coniferous forest area
(rate of 0.5 and 1, respectively, in Table 2) into deciduous over the 40-
year simulation (Table 2). One hundred replicates of each scenario were
simulated for a 40-year period (2010–2050) to deal with the un-
certainty associated with fire and LC transitions stochasticity.
2.2.4. Carbon sequestration modelling
We conducted a biophysical assessment of the climate regulation
ecosystem service (hereafter CRES) based on the carbon sequestration
ecosystem function, by applying the InVEST (Integrated Valuation of
Ecosystem Services and Tradeoffs) model to the BR-GX study area. We
evaluated the impact of fire and land-use management scenarios on this
ecosystem service over a period of 63 years (1987–2050). The total
carbon sequestered (Tg C; i.e. the carbon sequestered by all carbon
pools accumulated over time) and the total carbon sequestration rate
(Tg yr
−1
; i.e. the carbon sequestered per year by all carbon pools) were
assumed as proxies of CRES.
The carbon sequestration and storage module of the InVEST model
(Sharp et al., 2018) was used to perform the simulations. Uncertainty in
future carbon sequestration estimates was addressed by running the
InVEST carbon module for each management scenario and replicate.
The carbon module links the carbon stocks in four carbon pools: Above-
and belowground biomass (AGB and BGB, respectively), litter (DOM)
and soil organic carbon (SOC) to each LC class type available in the
study area. It returns the carbon stored in the landscape and estimates
the carbon sequestered over time by comparing levels of carbon based
on simulated LC spatial data. Land cover databases of the BR-GX for
years 1987, 2000 and 2010 (30-m spatial resolution) and the simulated
landscape scenarios (2011–2050) classified in five major LC classes
(i.e., cropland, shrubland, coniferous forest, native oak woodlands, and
sparsely vegetated areas) were used to feed spatial requirements of the
carbon storage and sequestration module of the InVEST model. To es-
timate the carbon stocks in each of these pools per LC class, carbon data
on AGB and BGB, DOM and SOC for each of the major LC classes was
collected from: data available in published scientific literature at local
or regional scale (Sil et al., 2017), and official statistics from the Por-
tuguese and Spanish national forest inventories (Appendix C).
The carbon stocks in AGB and BGB in forest cover classes were
computed based on the application of biomass allometric equations
(Montero et al., 2005) to estimate the biomass available for the
Table 2
Management scenarios, their related storylines and annual LC conversion rates. Conversions from cropland to scrubland and from scrubland to oak is a natural
succession process while the other two conversions are anthropogenic.
Acronym Related storylines Conversion from scrubland
to oak (%)
Conversion from cropland to
scrubland (ha)
Conversion from scrubland to
cropland (ha)
Conversion of coniferous forest to
deciduous woodlands (%)
BAU BAU 1.6 400 0 0
HNV_1 HNVf 1.6 0 400 0
HNV_2 HNVf 1.6 0 800 0
HNV_3 HNVf 1.6 0 1200 0
HNV_4 HNVf 1.6 0 1600 0
FireSmart_1 Fire-Smart 1.6 400 0 0.5
FireSmart_2 Fire-Smart 1.6 400 0 1
FireSmart_3 Fire-Smart 2.0 400 0 1
FireSmart_4 Fire-Smart 2.4 400 0 1
HNV_FireSmart_1 HNVf + Fire Smart 1.6 0 800 1
HNV_FireSmart_2 HNVf + Fire Smart 2.4 0 1600 1
S. Pais, et al. Ecosystem Services 44 (2020) 101143
5
dominant species occurring within the area, and then converted into
carbon through applying a carbon content factor (Montero et al., 2005)
as shown in Appendix C. In addition, data on carbon in AGB available
from the fifth Portuguese national forest inventory (http://www2.icnf.
pt/portal/florestas/ifn/ifn5) was directly used after applying a con-
version factor (from CO
2
equivalent to C: 12 kg C/44 kg
CO
2
= 0.2727). Carbon stocks in each carbon pool for all the LC classes
were maintained constant over time (assuming that carbon pools are in
a steady state), which means that the carbon sequestration or emission
only occurs when a pixel of a given LC type changes between dates,
whereas if the LC type is kept unchanged between dates, the carbon
sequestration/emission rate will be zero for that time period.
2.2.5. Species distribution modelling
We used correlative species distribution models (SDMs) to predict
the impacts of alternative fire and landscape management scenarios
(Table 2) on biodiversity. SDMs were calibrated using occurrence spe-
cies data available from local atlas between 1990 and 2010 (Table 3).
These atlases document the occurrence of breeding avifauna and her-
petofauna in the Peneda Gerês National Park (Pimenta and Santarém,
1996; Soares et al., 2005) and in the Baixa Limia Xurés National Park
(Domínguez et al., 2012, 2005). SDMs were performed at the spatial
resolution of 1 and 2 km, depending on the atlas data used, being
subsequently projected onto a 1-km grid covering the whole BR-GX for
past and future environmental conditions (1990–2050) simulated under
each management scenario at decanal resolution. SDMs were developed
only for bird species with more than 10 presences (fewer records were
available) and for herpetiles with more than 30 presences. We used a
higher presence threshold for amphibians and reptiles for guaranteeing
the quality of our models, as ectothermic phisiology makes modelling
the distribution of these taxonomic groups potentially more challenging
when using only habitat and topography as explanatory variables. Fi-
nally, we obtained suitable data for 116 vertebrate species (93, 15 and 8
species of birds, reptiles and amphibians, respectively).
The datasets from which environmental variables (topographic and
land-use/cover information) were computed to calibrate the SDMs were
selected according to their temporal proximity to the atlas surveys.
Topographic information (altitude, slope and aspect) was obtained from
the Advanced Spaceborne Thermal Emission and Reflection Radiometer
(ASTER) Global Digital Elevation Model (GDEM), at an initial spatial
resolution of 30 m (https://asterweb.jpl.nasa.gov/gdem.asp). Land-use/
cover information (percentage of coniferous forests, deciduous wood-
lands, agricultural areas, rocky areas with sparse vegetation, and scrub-
land) was obtained from (1) the abovementioned Landsat-derived maps
for past conditions (1987, 2000 and 2010), and (2) fire-landscape model
simulations between 2011 and 2050 under each management scenario
(30-m spatial resolution). Topographic and land-use/cover information
was aggregated into the spatial resolution of species data (1 km).
Habitat suitability projections were based on consensus prediction
from six widely used modelling techniques available in the R package
‘Biomod2′ (Thuiller et al., 2009): Generalized Linear Models, General-
ized Additive Models, Random Forests, Artificial Neural Networks,
Generalized Models of Boosted Regression, and Multivariate Adaptive
Regression Splines (Thuiller et al., 2009). The combination of different
modelling algorithms in the final consensus prediction (hereafter, en-
semble modelling approach) was performed using a weighted mean
considering the weights proportional to the selected evaluation scores
(i.e. the higher the area under the curve of the model, the greater the
importance in the ensemble modelling; see Thuiller et al., 2009). This
approach aims to control the uncertainty arising from individual model
predictions, and to provide more informative and ecologically robust
predictions (Martínez-Freiría et al., 2017, 2015; Thuiller et al., 2009).
We used a repeated (at least 10 times) split-sample approach to produce
predictions independent of the training data. Each model run was fitted
using 80% of the data and evaluated against the remaining 20% by
using the area under the curve (AUC) of a receiver operating char-
acteristics (ROC) (Fielding and Bell, 1997). The final projection was
obtained by computing a consensus of single-model projections using a
weighted average approach for each vertebrate species model (Mar-
mion et al., 2009). AUC values were considered as model weights, using
at least 10 model replicates for AUC values higher than 0.65 (see Ap-
pendix D for more details). Ensemble models were converted to a
binary classification of predicted presence/absence according to ROC
optimized thresholds (that maximize sensitivity and specificity) from
‘Biomod2′ package (Thuiller et al., 2009). The total areas of habitat
availability predicted by the models were summarized for groups of
species defined according to different conservation/interest criteria: (1)
protection under the Birds and Habitats European directives; (2) re-
gional IUCN conservation status in Portugal and Spain; and (3) endemic
species from Iberian Peninsula. For the IUCN criteria, species with
status of “Least Concern” and “Near threatened” were grouped as non-
threatened, while species with status of “Vulnerable”, “Endangered”
and “Critically Endangered” were grouped as threatened (see Appendix
D for conservation status at species level). For computational efficiency,
biodiversity and carbon models were only projected under scenarios
with the highest contrast and most extreme fire regime and landscape
trends (i.e. 'BAU', 'HNV_4', 'FireSmart_4' and 'HNV_FireSmart_2', see
Table 2).
3. Results
3.1. Future fire regime under management scenarios
According to our simulations, larger areas are expected to burn in
the period 2031–2050 than those of historical records (1990–2010),
with an annual increase of 2,000 ha, on average, under the business-as-
usual scenario (Fig. 3). In contrast, alternative management policies
aimed at increasing farmland areas (i.e. HNVf scenarios) would lead to
a gradual reduction of the burned area in relation to the reference va-
lues (1990–2010) (Fig. 3). Indeed, the area expected to be burned by
large fires (larger than 1,000 ha) between 2031 and 2050 could be
potentially reduced from 20,000 ha under the BAU scenario to
10,000 ha under the HNVf scenarios (i.e. reduction of 50%; see
'HNVf_4′ in Fig. 3, and Appendix E). ‘Fire-smart’ forest management
strategies were not predicted to significantly modify fire regime in the
short term (2011–2030), allowing only a significant reduction in
burned area when combined with policies focused on the valorization
of agricultural areas (see ‘HNVf_FireSmart_2′ in Fig. 3).
Regarding the effectiveness of the two fire suppression strategies,
active fire suppression did not differ significantly from the BAU sce-
nario between 2011 and 2030, even if suppressed area slightly in-
creased from 2030 onwards (Fig. 3). In contrast, taking advantage of
the opportunities created by the agricultural areas of at least 1 hectare
(i.e. passive fire suppression) produced significant differences across
scenarios (Fig. 3). The implementation of management policies aimed
at promoting agricultural activities will likely increase future fire-sup-
pression opportunities, leading to larger suppressed areas (from approx.
20,000 ha suppressed between 2031 and 2050 under the ‘BAU’ scenario
up to 60,000 ha under the ‘HNVf_4′ scenario; Fig. 3).
Table 3
Description of data used for calibrating the species distribution models.
Source Sampling
location
Spatial
resolution
Taxonomic
group
Temporal
resolution
N ° Units
Atlas PGPN 2 km Birds 1990–1995 238
Atlas BLXNP 1 km Birds 1998–2000 397
Atlas BLXNP 2 km Birds 2010 147
Atlas PGPN 1 km Amphibians
and reptiles
1998–2003 583
Atlas BLXNP 1 km Amphibians
and reptiles
2010 337
S. Pais, et al. Ecosystem Services 44 (2020) 101143
6
3.2. Carbon sequestration under management scenarios
The land-use/cover changes observed between 1987 and 2010 and
simulated under each management scenario largely affected the supply
of the climate regulation ecosystem service over time (measured
through total carbon sequestered and carbon sequestration rate;
Appendix C). Between 1987 and 2010, ecosystems in the BR-GX se-
questered a total of 2.87 Tg C at an average rate of 0.12 Tg C yr
−1
.
Under the future scenarios, the results indicate that the fire-smart and
business-as-usual scenarios would show the highest estimates for total
sequestered carbon (4.79 ± 0.23 and 3.63 ± 0.27 Tg C, respectively)
and for carbon sequestration rate (0.48 ± 0.02 and 0.36 ± 0.03 Tg C
yr
−1
, respectively), while management options aimed at promoting
agricultural activities would lead to the lowest total carbon sequestered
(0.27 ± 0.13 Tg C) and carbon sequestration rate (0.03 ± 0.01 Tg C
yr
−1
); still, if combined with a fire-smart strategy, higher levels of
carbon sequestered would be achieved (1.23 ± 0.17 Tg C and
0.12 ± 0.02 Tg C yr
−1
).
3.3. Biodiversity conservation under management scenarios
We obtained ensemble models with high predictive accuracy, both
overall (AUC = 0.925 ± 0.07; TSS = 0.737 ± 0.167) and when
Fig. 3. Burned area and area suppressed by active and passive fire suppression strategies under each scenario in the short- (2011–2030) and medium-term
(2031–2050) (see scenario acronyms in Table 2). Black line shows the area burned between 1990–2010. For all boxplots, lower and upper whiskers encompass the
95% interval, lower and upper hinges indicate the first and third quartiles, and the central line indicates the median value (solid dots are outliners). Intra-boxplot
variability is computed from the different values obtained for burned and suppressed area from each model simulation, and represents the uncertainty associated
with fire stochasticicy.
Table 4
Predictive accuracy metrics (Area under the curve – AUC; True Skill Statistics –
TSS) of species distribution models according to taxonomic groups.
AUC TSS
Birds 0.929 ± 0.06 0.746 ± 0.168
Amphibious 0.913 ± 0.07 0.71 ± 0.156
Reptiles 0.91 ± 0.07 0.694 ± 0.169
S. Pais, et al. Ecosystem Services 44 (2020) 101143
7
considering the different taxonomic groups separately (Table 4).
Our results suggest that the business-as-usual scenario and the im-
plementation of fire-smart strategies would lead to a steep decrease in
habitat suitability (Fig. 5), independently of the protection status of the
area and the species (with habitat losses varying between 10 and more
than 30%, see Fig. F1 in Appendix F). Despite some variability of ha-
bitat suitability predictions, policies aimed at increasing agricultural
areas (alone or in combination with fire smart strategies) are expected
to largely increase, or at least avoid losses, in habitat availability for
some species (Fig. 5). For species without legal protection and currently
non-threatened, our results indicate an overall increase of 20–30% of
suitable habitat in relation to 2010; see Fig. F.1 in Appendix F) over the
next 40 years under the ‘HNVf’ and ‘HNVf + FireSmart’ scenarios (see
Fig. 5). On the contrary, general decreases of habitat availability were
predicted for protected (under the European directives and according to
regional IUCN criteria) and vulnerable (endemic) species, both inside
and outside protected areas. Nonetheless, ‘HNVf’ and ‘HNVf + Fire-
Smart’ represent the best-case scenarios by allowing a potential habitat
suitability stabilization after 2010 (see Fig. 5 and Fig. F.1 in Appendix
F).
4. Discussion
This study shows the benefits of integrating proactive land-use po-
licies and fire-smart management strategies at the regional scale (and in
a transboundary context) to promote sustainable solutions to the
wildfire problem in abandoned mountain landscapes. Overall, our re-
sults highlight that land-use policies aimed at promoting farmland areas
would provide fire-suppression opportunities while simultaneously
ensuring biodiversity conservation within (and around) protected
areas. In addition, our results suggest that fire-smart management
strategies based on large-scale forest conversions would foster the cli-
mate regulation ecosystem service (through carbon sequestration). This
study illustrates how scenario planning supported by fire-landscape
model simulations can help to better inform policy and decision making
on fire and land-use management considering multiple societally re-
levant goals.
4.1. Sources of uncertainty and limitations
Our results confirm that the impact of wildfires in the BR-GX will be
even higher in the future than nowadays under the business-as-usual
scenario, due to the maintenance of rural abandonment processes
(Loepfe et al., 2010) and current policies focused on fire exclusion
(Moreira et al., 2019). However, these results can be relatively opti-
mistic since climate change was not explicitly included in these sce-
narios. Further model developments should take into account climate-
fire relationships and the complex interactions among climate, vege-
tation dynamics and fire management (Abatzoglou et al., 2018), given
the expected increase of drought and high temperature conditions in
Southern Europe (Fernandes et al., 2014; Moreira et al., 2011). It would
be also worth exploring other more informative measures of wildfire
impacts such as fire severity rather than total burned area, an issue
especially relevant under the current context of land abandonment and
climate warming.
Another potential limitation of this study is related to the assess-
ment of a single ecosystem service —the modelling of carbon storage
and sequestration. Although other ecosystem services might be in-
directly evaluated, such as fire regulation capacity and fire protection
services (Depietri and Orenstein, 2019; Sil et al., 2019b) or those re-
lated to biodiversity (e.g. pest control and seed dispersal; Whelan et al.,
2015), we acknowledge that future studies should focus on analyzing a
larger set of ecosystem services. We also recognize the potential lim-
itation of considering broad land cover types instead of particular
species or more detailed typologies. For instance, we assume that all
types of croplands provide the same fire suppression opportunities,
which might be an oversimplification (see e.g. HNVf typologies in
Lomba et al., 2015). In addition, different species within the same
shrubland and forest type could slightly differ in their post-fire response
(Calvo et al., 2003, 2002), which might affect post-fire regeneration
and carbon sequestration rates at local scales. Besides, while our models
can estimate carbon sequestration from natural succession, post-fire
recovery processes and land-use changes, direct carbon emissions from
wildfires are not currently considered —an issue that should be taken
into account in future model developments.
In terms of biodiversity, our species distribution models presented
high predictive accuracy metrics, which allow us to interpret our results
with confidence. Potential uncertainties could be associated with model
downscaling (from 2 km to 1 km), but this procedure has been proven
effective for capturing general environmental patterns and to predict
potential distributions at finer resolutions (Araujo et al., 2005). None-
theless, our projections to future management scenarios might still be
affected by model uncertainties, particularly when considering amphi-
bians and reptiles. Despite the application of a high threshold of pre-
sence records for these taxa to improve model quality, the ectothermic
physiology of these taxa (Huey and Stevenson, 1979) coupled with the
omission of climatic variables might lead to an overestimation of po-
tential distributions. In fact, the relevance of climatic factors on
shaping/limiting the distribution and dispersal of these taxonomic
groups is well known, including them among the most vulnerable taxa
to evaluate climate change impacts on biodiversity (Carvalho et al.,
2011, 2010). As such, the predicted habitat availability for these groups
(generally higher and more variable in comparison to bird species; see
Fig. 5 and Fig. F.2 in Appendix F) must be interpreted with caution.
4.2. Fire regime under fire-smart management scenarios
Despite the abovementioned limitations, our simulations showed
that land-use management policies aimed at promoting agricultural
areas would significantly reduce the potential area burned by large fires
when compared to the business-as-usual scenario (see ‘HNVf’ scenarios
in Fig. 3, and Appendix E). These results are especially relevant given
the high intensity at which large fires burn and their associated en-
vironmental and socioeconomic impacts (Fernandes et al., 2016a)
Fig. 4. Carbon sequestrated per decade between 2011 and 2050 under each
management scenario (see acronyms in Table 2). For all plots, colored lines
indicate mean values while the transparent colored areas indicate the error
limits defined by the median range values.
S. Pais, et al. Ecosystem Services 44 (2020) 101143
8
Fig. 5. Habitat availability (in km
2
) for vertebrate species with and without protection status under different management scenarios within and outside protected
areas (see scenario acronyms in Table 2). For all plots, colored lines indicate mean values while the transparent colored areas indicate the error limits defined by the
median range values. Two protection criteria are considered: the protection under the Birds and Habitats (for amphibians and reptiles) European directives (top) and
the regional IUCN conservation status in Portugal and Spain (middle). For the IUCN criteria, species with status of “Least Concern” and “Near threatened” are
grouped as non-threatened, while species with status of Vulnerable, Endangered and Critically Endangered are grouped as threatened. Endemic and non-endemic
vertebrates from Iberian Peninsula are also represented (bottom).
S. Pais, et al. Ecosystem Services 44 (2020) 101143
9
(including fire fatalities; Molina-Terrén et al., 2019). Our results are in
line with recent studies suggesting that the creation of new agriculture
patches (at least 100 km
2
per year in a region of 32.100 km
2
) would be
required to effectively reduce the total burned area in NE Spain
(Aquilué et al., 2019; Moreira and Pe’er, 2018).
According to our simulations, fire-smart management strategies
characterized by large-scale forest conversions to more fire-resistant
forests (i.e. dominated by native oak species) would not be enough to
effectively reduce potential burned area (Fig. 3). The few fire-sup-
pression opportunities derived by the fire-smart forest conversion can
be explained by the limited extent covered by forest in the Gerês-Xurés
(GX) mountains (less than 15%), which undermines the capacity to
affect fire regime. This might not be the case for other Mediterranean
mountains dominated by large forest areas, wherein fire-smart forest
conversion could be much more effective at reducing fire hazard (see
review in Fernandes 2013). In our study area, the large-scale fire-smart
forest conversion would be only effective in term of fire suppression if
embedded in a landscape mosaic with increased agricultural areas (see
‘HNVf + FireSmart’ scenarios in Fig. 3). Our findings contribute to the
mounting evidence that agricultural policies have a great potential to
reduce future wildfire impacts (Fernandes et al., 2014; Moreira and
Pe’er, 2018; Sil et al., 2019b). In light of these results, other fuel-re-
duction practices such as prescribed burning should be also explored in
future studies given the dominance of shrublands and the traditional
use of fire by rural communities to manage the landscape in northern
Portugal (Fernandes et al., 2013; Úbeda et al., 2018).
4.3. Carbon sequestration under fire-smart management scenarios
Our simulations showed that land-use and fire management policies
have the potential to affect the regional supply of the climate regulation
ecosystem service (CRES) —measured through carbon sequestration
over the next decades (Fig. 4). In particular, a higher supply of the CRES
in the study area would be expected if landscape-scale forest conversion
to native oak woodlands was promoted (‘FireSmart’ scenario in Fig. 4;
see Appendix C; Roces-Díaz et al., 2017) or if the current land aban-
donment processes (i.e. agricultural abandonment, shrubland en-
croachment and native forest expansion) would continue over the
coming decades (e.g. under an ecological rewilding initiative; see ‘BAU’
scenario in Fig. 4;Sil et al., 2017). However, an increase of fuel load
and continuity favored by land abandonment processes could also lead
to larger burned areas in the future (Fig. 3), with the subsequent in-
crease in the release of CO
2
to the atmosphere (Hurteau et al., 2008;
van der Werf et al., 2006).
In the absence of more efficient fire-suppression polices, our results
suggest that the integration of large-scale forest conversion (from fast-
growing tree plantations to native oak woodlands) within agricultural
policies would promote carbon sequestration in the GX mountains with
a reduced fire hazard (see ‘HNVf + Fire-Smart’ scenarios in Figs. 3 and
4). Nevertheless, it would be worth exploring other fuel-reduction
practices (e.g. forest thinning or prescribed burning) to simultaneously
reduce fire hazard and ensure the CRES under the business-as-usual
scenario (Hurteau et al., 2008; Vilén and Fernandes, 2011). Manage-
ment efforts either to reduce the fire hazard or promote the supply of
the CRES through carbon sequestration should also consider the spatial
planning of such practices by identifying priority areas where to in-
tervene (Ascoli et al., 2018), as well as by promoting practices directed
to the maintenance of an heterogeneous landscape —ensuring that fu-
ture sustainability of the provision of ecosystem services and hazard
reduction is achieved (Turner et al., 2013). In addition, an economic
assessment of both ecosystem services (e.g. potential payments for
ecosystem services) and potential economic impacts of fire on eco-
system services (either positive or negative) should be carried out to
complement these results and raise awareness among landowners,
managers and decision-makers (Sil et al., 2019a,b).
4.4. Biodiversity conservation under fire-smart management scenarios
Despite some differences amongst taxonomic groups, particularly
for amphibians that usually select shadier vegetated habitats (e.g.
forested areas) during the day due to their water/humidity de-
pendencies (Loureiro et al., 2008), our results predicted an overall
decline in species’ habitat suitability over the coming decades under the
business-as-usual scenario (Fig. 5), expressing rural abandonment and
subsequent forest expansion (Fig. B.4). Bird species, being more ha-
bitat-specific in relation to herpetiles, respond differently depending on
the scenario. According to our simulations, the bird guild most exposed
to the changes of rural abandonment would be the group of species
breeding in open habitats (especially for farmland and mountain spe-
cies; see Fig. F.3 in Appendix F). Our predictions are in line with po-
pulation declines observed for farmland and mountain bird species for
Europe and North America (Lehikoinen et al., 2018; Rosenberg et al.,
2019; Schipper et al., 2016). Expectedly, reptile species would benefit
from policies inspired by the ‘HNVf’ and ‘HNVf + Firesmart’ scenarios,
since several of the modelled species are found in – and able to adapt to
– humanized habitats with extensive agricultural activities (Loureiro
et al., 2008; Martinez-Freiria et al., 2019; Pleguezuelos et al., 2002). In
fact, these policies might promote the heterogeneity of the landscape
matrix, which might be advantageous for several reptile species by
providing opportunities for thermoregulation, shelter and food avail-
ability. Our results also suggest that land-use policies promoting the
expansion of farmland areas would be extremely important for the
conservation of vertebrate diversity within the protected areas of the
Biosphere Reserve. This management scenario is particularly advanta-
geous for species of conservation concern (for legally protected,
threatened and endemic species), since increasing farmland areas
would potentially prevent a continuous and drastic loss of habitat
availability in the Biosphere Reserve (Fig. 5 and Fig. F.1 in Appendix F).
4.5. Trade-offs between fire mitigation, biodiversity conservation and
carbon sequestration
Our results confirm the urgent need for policies promoting farmland
areas in the GX mountains, both in terms of future fire-suppression
opportunities and biodiversity conservation (Figs. 3 and 5). A large
amount of strategically allocated cropland areas (at least 1200 ha per
year) should be gradually incorporated to the Reserve’s landscape along
the next decades to significantly affect fire regime in the medium term
(Fig. 3). These policies would be also positive for conservation objec-
tives since most of the species would benefit for the recovery of habitats
associated with agricultural activities (Fig. 5). In terms of long-term
supply of the climate regulation ecosystem service (through carbon
sequestration), our models predicted the best outcomes under large-
scale fire-smart forest conversion (Fig. 4). However, the integration of
this fire-smart landscape conversion would be only acceptable for bio-
diversity conservation and fire prevention if embedded in landscape
matrix characterized by increasing agricultural areas over the next
decades (Figs. 3 and 5). Further studies including the socioeconomic
assessment of the different scenarios and other ecosystem services such
as timber production or food provision are needed to better inform
decision makers on the feasibility of each management scenario.
Land abandonment has been proposed by some authors as an op-
portunity for biodiversity conservation in Europe (Cerqueira et al.,
2015; Navarro and Pereira, 2012; Perino et al., 2019). However, when
analyzed at regional and local scales, land abandonment was revealed
by our results as one of the worst scenarios in terms of habitat suit-
ability for species of conservation concern (i.e. in our case mostly open
habitat species). Additionally, this process may also be the driver of
more flammable and fire-prone vegetation, increasing homogeneity and
landscape continuity leading to more severe fire regimes (Azevedo
et al., 2011; Fernandes et al., 2016b; Moreira et al., 2011; Sil et al.,
2019b). Our results are in line with other studies (see e.g. Fernandes
S. Pais, et al. Ecosystem Services 44 (2020) 101143
10
et al., 2014) which describe similar changes in fire regime due to
agricultural abandonment (Azevedo et al., 2011). This general pattern
highlights the relationship between agricultural abandonment and fire
hazard in the Mediterranean region (Moreira et al., 2011). Therefore,
farming activities should continue to play a key role in supporting the
sustainable development of rural territories as well as fire hazard
management (Moreira et al., 2011; Sil et al., 2019b). In the worst-case
scenario, where agricultural policies cannot be implemented in practice
(i.e. ongoing land abandonment), other fuel-reduction strategies (such
as prescribed burning or forest thinning operations) should be tested to
explore how rewilding initiatives could be partially navigated to en-
hance their positive effects on carbon storage/sequestration and reduce
their negative impacts on biodiversity and wildfire hazard.
5. Conclusions
This study highlights the urgency for transnational policy co-
ordination promoting farmland areas in transboundary mountain re-
gions of Southern Europe. Our results show how an effective im-
plementation of agricultural policies would reduce fire hazard while
simultaneously ensuring biodiversity conservation. In addition, our
results suggest that, although fire-smart forest conversion strategies
would be beneficial for a long-term supply of carbon sequestration,
their implementation should be integrated within agricultural policies
to jointly reduce fire hazard and conserve local biodiversity adapted to
these semi-natural systems. Our study evidences the benefits of in-
tegrating fire hazard control, ecosystem service supply and biodiversity
conservation to better inform decision-making in mountain landscapes
of Southern Europe.
Declaration of Competing Interest
The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influ-
ence the work reported in this paper.
Acknowledgements
This research work was funded by national funds through the FCT –
Foundation for Science and Technology, I.P., under the FirESmart
project (PCIF/MOG/0083/2017) and the project INMODES (CGL2017-
89999-C2-2-R) funded by the Spanish Ministry of Science and
Innovation. A.R. was funded by the Xunta de Galicia (postdoctoral
fellowship ED481B2016/084-0) and IACOBUS program (INTERREG V-
A España – Portugal, POCTEP 2014-2020). J.D. and A.R. thanks the
support of Xunta de Galicia ED431B 2018/36. Â. Sil received support
from the Portuguese Foundation for Science and Technology (FCT)
through Ph.D. Grant SFRH/BD/132838/2017, funded by the Ministry
of Science, Technology and Higher Education, and by the European
Social Fund - Operational Program Human Capital within the 2014-
2020 EU Strategic Framework. FM-F has a contract from FCT (ref.
DL57/2016/CP1440/CT0010). We thank to Adrián Lamosa Torres,
Xosé Pardavila and Alberto Gil for their help during fieldwork in Xurés
and Rafael Vázquez for providing additional data for amphibians and
reptiles.
Appendix A. . Fire-vegetation model simulations: processes, state variables, and initialization
Model purpose
The REMAINS model integrates the main landscape-level processes driving fire-vegetation dynamics in mountain landscapes of northern Iberian
Peninsula. It includes the main anthropogenic and natural (abiotic and biotic) drivers of landscape change to study their spatiotemporal interactions
and feedbacks effects. The model allows investigating the interlinked effects of wildfires (including fire suppression), land-use change and man-
agement strategies on fire regime and overall landscape composition. Ultimately, the model aims to generate spatially explicit scenarios of landscape
dynamics according to pre-designed scenario storylines. Future landscape configurations will be used as input for subsequent biodiversity and
ecosystem services model simulations.
State variables
REMAINS is supported by three state variables (updated in each model time step) and seven auxiliary variables.
State variables:
1. LCM (the land-use/cover map) describes the main land use/cover types of the study area: agricultural land, scrublands, rocky areas with sparse
vegetation, coniferous forest, oak woodland, water and urban. All of them are considered dynamics, except water and urban that are static (Fig.
A.1).
2. TSChg indicates the time (in years) since the last land-cover transition.
3. TransType records which type of transitions has more recently taken place in each grid cell: a stand-replacing fire, an anthropogenic driven
transition (i.e., rural abandonment – from cropland to scrubland –, agriculture intensification – from scrubland to cropland –, forest plantation –
from scrubland to coniferous forest–, and fire-smart plantation – from coniferous forest to oak woodland –), or a natural successional transition
(i.e. afforestation – from scrubland to oak woodland – and vegetation encroachment – from rocky areas with sparse vegetation to scrubland –).
Auxiliary variables:
4. AdminRegion: the study area encompasses 2 countries, Portugal and Spain.
5. RoadDens: road density (used in the logistic model to predict fire ignition probability; Fig. A1).
6. UTM: 1 × 1 km UTM grid (used to dynamically compute ignition probability).
7. Elevation: digital elevation model (DEM; in m).
8. Aspect: aspect derived from the DEM (1 - north, 2 - east, 3 - south, and 4 - west).
9. SlopeDegree: slope derived from the DEM (in degrees).
10. ProbIgni: ignition probability (Fig. A.1).
S. Pais, et al. Ecosystem Services 44 (2020) 101143
11
The spatial resolution of all spatial variable is 30 m
2
and the temporal resolution of the model is set at 1 year.
Process overview
The processes that REMAINS simulates can be grouped in three major drivers of change: (1) land-use changes (e.g. rural abandonment, agri-
culture conversion and coniferous plantations), (2) wildfires and fire suppression, and (3) vegetation dynamics such as post-fire regeneration,
vegetation encroachment and woody colonization of scrublands (i.e. afforestation) (Fig. A2).
I. Land use changes
In the current version of the model, and to answer the proposed research questions, four anthropogenic-driven land cover transitions are
modelled: (1) market-oriented forest plantations, the transition of scrublands to coniferous plantations, (2) ‘fire-smart’ plantations, the transition of
coniferous plantations to oak woodlands, (3) rural abandonment, the conversion of crops and grasslands to semi-natural vegetation areas identified
as scrublands, and (4) agriculture intensification, the conversion of scrublands to cultivated land (Fig. 1). In our modelling framework a land-cover
Fig. A1. Fire ignition probability layer derived from a logistic regression that includes topography (altitude and slope; source: DEM), accessibility (road density;
source: OpenStreetMap) and wildland-urban interfaces (source: Landsat-derived maps) as main factors.
Fig. A2. Land-cover types for the Gerês-Xurés Transboundary Natural Park and both anthropogenic and natural driven land-cover transformations. Potential post-fire
state is indicated in red for both the dynamic and the static land-cover types.
S. Pais, et al. Ecosystem Services 44 (2020) 101143
12
transition (e.g. rural abandonment) is defined by a unique target land-cover (e.g. scrublands) and all the land-covers that may undergo change (e.g.
croplands and grasslands). Land-cover transitions are modelled using a demand-allocation approach. The demand or quantity-of-change by time step
has to be provided by the user (based on e.g. historical trends or to emulate landscape scale management policies). Changes will occur in locations
with a higher likelihood to be transformed to the target land-cover. We initialized the likelihood for each land-cover transition as the neighbourhood
factor of the target land-cover introduced by (Verburg et al., 2004). The neighbourhood factor of a land-cover accounts for the relative abundance of
such land-cover within the neighbourhood and the overall landscape. By initializing the likelihood of conversion by the neighbourhood factor we
assume that land-cover transitions tend to homogenize the landscape, by happening where the target land-cover is already more abundant. The
allocation of the changes follows an algorithm that recognizes the emergence and contagion character of such processes (Aquilué et al., 2017).
II. Wildfires and fire suppression
Fires are the main stand-replacing natural disturbance currently included in REMAINS. A given landscape implicitly has a fire regime associated,
even though it may be altered or modified by extreme climatic conditions (Duane et al., 2019). We adopted a top-down approach to model the fire
regime at the landscape level (that currently does not depend on climate). At each time step, each fire event ignites, spreads, and stops to reach a
predefined target annual area that is specific for each administrative region of the study area. We used fire statistical data (obtained from the Spanish
Ministry of Agriculture and Fisheries, Food and Environment) between 1983 and 2010 at the municipal level, and data in shapefile format for
Portugal (obtained from the Institute of Nature Conservation and Forests - ICNF) for the period 1990–2010. From these data, annual burnt areas were
quantified to create a burnt area distribution and thus incorporate historical fire regime into the landscape-fire model. The target fire sizes are a
model input too, but the final fire size emerges from the spatial interaction between the location of the fire ignition, the landscape composition, the
topography and the fire suppression when applied. The probability of fire ignition is a function of human-related and biophysical variables. In the
current version, fire ignition is modelled as a multivariate logistic regression model where explanatory variables are elevation, density of roads, and
neighbourhood configuration that is described at 1 × 1 km cells (matching the UTM grid).
The logistic regression model:
logit P( )
ignition non ignition|
= -0.01613 + 0.002047·RoadDens − 0.001321·Elevation + 1.987·UrbNat + 1.568·CrpNat
being ‘UrbNat’, the interface between urban and natural areas; and ‘CrpNat’, the interface between cropland and natural areas, that are dyna-
mically updated in each time step.
Thus, the spatial distribution of fire ignitions depends on landscape configuration, elevation and accessibility, while fire spread depends on slope,
aspect, and land-cover flammability. The spread rate is formulated as a polynomial expression with three factors (slope, aspect and fire-proneness of
each LCT) adapted from Duane et al. (2016).
Final fire sizes can be explicitly reduced if a fire suppression strategy (or a combination of strategies) is activated. Currently two firefighting actions
are designed: (1) a fuel-based strategy that take advantage of fire spreading situations bellow a specific fire spread threshold to suppress the fire and (2) a
landscape-based strategy that uses open mosaics to stop the advancing fronts. Fuel-based threshold is the maximum spread rate at which fuel-based
suppression can be activated and the landscape-based is the number of contiguous agricultural lands (1 hectare) that have to burn before the landscape-
based suppression strategy can be activated. These are scenario parameters and when are set at 0 suppression is not activated (Table A1).
Fire strategy defined at the administrative level (counties) according to the calibration procedure:
III. Vegetation dynamics
For simplification, we assume that natural vegetation in mountain landscapes always follows the same successional pathway, from rocky areas to
closed scrublands, and then to oak woodlands. The transformation from one state to the following only occurs after a fixed period of time since the
last transition and depends on the presence of potential colonizers in a circular neighbourhood around the target location.
Fire spreads and burns all land covers except urban areas and inland water, these are oak woodland, coniferous plantations, scrublands, rocky or
open scrublands, grasslands and agricultural lands. After fire, agricultural lands, scrublands and open scrublands persist (i.e. there is no land-cover
change because of fire) while forest stands may partially change state to scrublands or to rocky scrublands (Fig. 1). The variable ‘time since last
change’ (proxy of the biomass) always turns to 0. A percentage of forest areas (coniferous and oaks) return to the pre-fire state after some years after
fire while some forest areas remain scrublands or rocky vegetation for a while (Table A2).
1. Afforestation is the colonization of scrublands by oak species, occurs at a certain annual rate (2.6%), only can take place after a time period since
the last land-cover transformation (9 years) and it depends on the percentage of oak woodland found in a circular neighbourhood of radius (13
cells, approx. 400 m) around the target location.
2. Vegetation encroachment is the colonization of open vegetation areas by scrublands, occurs at a certain annual rate (2.2%), only can take place
after a time period since the last land-cover transformation (4 years) and it depends on the percentage of scrublands in a circular neighbourhood
of radius (4 cells, approx. 120 m) around the target location.
3. Post-fire recovery: non-forest areas (agricultural land, scrublands and open vegetation) persist after fire, that is there is no post-fire transition for
these land-covers. For coniferous forests and oak woodlands, the post-fire regeneration is stochastic, even if a percentage of the regeneration
occurs by contagion, that is mimicking the post-fire pathway undertake by a close burnt neighbour. However, when any of these forest species are
set to persist (i.e. remain coniferous or oaks, respectively) these take a certain time (i.e., number of years) to effectively recover the forest canopy
(Table A.3). In order to estimate the average post-fire recovery profiles for each land cover class, the normalized difference vegetation index
Table A1
Firefighting effectiveness levels for each administrative region and fire suppression strategy.
Administrative Region Fuel-based Th (%) Landscape-based Th (ha)
Spain 50 1
Portugal 30 1
S. Pais, et al. Ecosystem Services 44 (2020) 101143
13
(NDVI) was extracted from the MODIS Vegetation Indices product (MOD13Q1, Collection 6, 250 m, 16-day). Burned area masks were extracted
from the MODIS Burned Area Product (MCD64A1, Collection 6, 500 m, monthly) for the years 2001–2016. All MODIS image time-series were re-
projected to WGS84/UTM29N reference system, and resampled to 250 m, using nearest neighbour method. Then, the NDVI time-series was used
to calculate an approximation of the fraction of vegetation cover (FVC), by applying the formula in Gutman & Ignatov, 1998:
=FVC NDVI NDVI NDVI NDVI(( )/( ))
min max min 2
where NDVI
min
and NDVI
max
correspond to values for bare soil (FVC = 0) and dense vegetation (FVC = 1), respectively. The values of this NDVI-
based FVC were then used to extract post-fire regeneration curves for each pixel in the study area identified as burned in each of the years 2004,
2005, or 2006, but not in any of the preceding years (from 2001). Median post-fire regeneration profiles, with values for the 10 years following the
fire year, were then calculated from the aggregation of pixels with a percentage area of each land cover class in 2000 equal or greater than 60%, as
well as for each relevant land cover class transitions between 2000 and 2010 (Fig. A.3).
Table A2
The post-fire probability for each LC class.
pre\post crop shrub coniferous oak rckyveg
crop 100 0 0 0 0
shrub 0 100 0 0 0
coniferous 0 19 42 0 39
oak 0 13 0 54 33
rckyveg 0 0 0 0 100
Table A3
Post-fire recovery rates: % pixels and time to recover to the pre-fire LC class (expressed in years).
Initial state Final state Transition Rate
Coniferous Coniferous 42% 8 years
Rocky 39%
Shrub 19%
Deciduous Deciduous 53% 9 years
Rocky 33%
Shrub 13%
Shrub Shrub 77% 4 years
Deciduous 11% 9 years
Rocky Rocky 78% 4 years
Shrub 13% 4 years
Cropland Cropland 100% 1 year
Fig. A3. Average profiles of post-fire recovery to the pre-fire levels for each land cover class computed from fraction of vegetation cover (FVC).
S. Pais, et al. Ecosystem Services 44 (2020) 101143
14
References
Aquilué, N., De Cáceres, M., Fortin, M.-J., Fall, A., Brotons, L., 2017. A spatial allocation procedure to model land-use/land-cover changes:
Accounting for occurrence and spread processes. Ecol. Modell. 344, 73–86. https://doi.org/10.1016/j.ecolmodel.2016.11.005
Duane, A., Aquilué, N., Canelles, Q., Morán-Ordoñez, A., De Cáceres, M., Brotons, L., 2019. Adapting prescribed burns to future climate change in
Mediterranean landscapes. Sci. Total Environ. 677, 68–83. https://doi.org/10.1016/j.scitotenv.2019.04.348
Duane, A., Aquilué, N., Gil-Tena, A., Brotons, L., 2016. Integrating fire spread patterns in fire modelling at landscape scale. Environ. Model.
Softw. 86, 219–231. https://doi.org/10.1016/j.envsoft.2016.10.001
Verburg, P.H., de Nijs, T.C.M., van Eck, J.R., Visser, H., de Jong, K., 2004. A method to analyse neighbourhood characteristics of land use
patterns. Comput. Environ. Urban Syst. 28, 667–690. https://doi.org/10.1016/j.compenvurbsys.2003.07.001
Appendix B. . Land-use/cover analysis and model simulations.
To analyze land-use/cover changes at landscape level, we cross-tabulated the remote sensing data-derived maps in order to obtain a transition
matrix for quantifying the spatial extent that had been lost or gained by a given land cover type over the entire study period (from 1990 to 2010)
(Fig. B.1).
Fig. B1. Circular plot illustrating the land-cover type transitions between 1990 and 2010, in hectares (ha). The size of the lines is proportional in width to the
contribution of each land-cover type to the change. The colors refer to the land-cover types.
S. Pais, et al. Ecosystem Services 44 (2020) 101143
15
During model calibration exercises, different values of the different model parameters were tested until the reference values were achieved. For
instance, different afforestation and natural succession rate values (from the initial values obtained from the remote sensing analyses) were tested by
running model simulations of the landscape from 1987 to 2010 under the business-as-usual scenario until the total amount of each LC type were
achieved (Fig. B.2).
Model simulations were carried out under each management scenario to predict the amount of each LC class since 2010 to 2050, at annual
timescale. Model outputs are available in raster spatial (Fig. B.3) and table format (Fig. B.4).
Fig. B2. Example of a model calibration exercise. Model simulation from 1987 to 2010 under different natural succession and afforestation rates estimated from
historic LC trends. Red color line represents the target values to be achieved, represented by hectares for each LC type in year 2010.
S. Pais, et al. Ecosystem Services 44 (2020) 101143
16
4600000 4640000 4680000
BAU
0
1
2
3
4
5
6
7
8
9
HNV
0
1
2
3
4
5
6
7
8
9
540000580000
4600000 4640000 4680000
FireSmart
0
1
2
3
4
5
6
7
8
9
540000580000
HNV_FireSmart
0
1
2
3
4
5
6
7
8
9
Fig. B3. Spatial representation of one of the 100 model simulations under different future management scenarios.
S. Pais, et al. Ecosystem Services 44 (2020) 101143
17
Appendix C. . Detailed description of methods and results for carbon sequestration simulations under fire-smart management scenarios
The climate regulation ecosystem service (CRES)
Terrestrial ecosystems play an important role on global climate by controlling the concentrations of gases in the atmosphere, such as carbon
dioxide, due to their ability to remove it from the atmosphere through plant photosynthesis process (i.e. carbon sequestration) and store it as carbon
into plants biomass, litter and soil, to then release it back into the atmosphere through auto- and heterotrophic respiration or due to disturbances
processes, such as fire or land cover change (Ciais et al., 2013). Carbon sequestration by terrestrial ecosystems (between 1960 and 2017) represents
approximately 30% of the total carbon dioxide uptake in the earth system (Le Quéré et al., 2018). This capability of terrestrial ecosystems is an
important regulating function (Petorelli et al., 2017) that enables the provision of the climate regulation ecosystem service (CRES) (Haines-Young
and Potschin, 2018), which can benefit human well-being at global scale by avoiding potential damage costs related to carbon emissions and climate
change impacts (Tol, 2018).
Methods
CRES assessment framework
We conducted a biophysical assessment of the climate regulation ecosystem service (CRES) by applying the InVEST (Integrated Valuation of
Ecosystem Services and Tradeoffs) model (Sharp et al., 2018) to the Transboundary Biosphere Reserve Gerês-Xurés (TBR-GX) in order to evaluate the
impact of land cover (LC) changes under scenarios of fire and land management on this ecosystem service over a period of 63 years (1987 – 2050).
The total carbon sequestered (Tg C), i.e. the carbon sequestered by all carbon pools accumulated over time, and the total carbon sequestration rate
(Tg yr
−1
), i.e. the carbon sequestered per year by all carbon pools, were assumed as proxies of the CRES. Therefore, we compared the total carbon
sequestration and sequestration rate among pathways of landscape evolution given by the simulated management scenarios, in order to assess how
land cover change may affect the potential supply of the CRES in the study area between 1987 and 2050. Uncertainty in future carbon sequestration
estimates was addressed by running the InVEST carbon module for 30 replicates per four decades in each of the four future LC scenarios. Therefore, a
total of n= 482 simulations were run (1 simulation × 2 past dates + 30 replicates × 4 decades × 4 future scenarios).
Model description, data collection and carbon modelling
The carbon sequestration and storage module of the InVEST model (Sharp et al., 2018) was used to perform the simulations. The carbon module
Fig. B4. Amount of each LC class (in Hectares) predicted from 1987 to 2050 under each management scenario. The intra-scenario variability is computed from the
different values obtained from each of the model simulations (total of 100 replicates), and represents the uncertainty associated with fire stochasticicy.
S. Pais, et al. Ecosystem Services 44 (2020) 101143
18
links the carbon stocks in four carbon pools above- and belowground biomass (AGB and BGB, respectively), litter (DOM) and soil organic carbon
(SOC) to each land cover class type available in the study area, returning the carbon stored in the landscape, and compares levels of carbon over time
based on LC spatial data to estimate the carbon sequestered. LC spatial databases available of the TBR-GX for the years 1987, 2000 and 2010 (30 m
spatial resolution, Table C1) and the simulated landscape scenarios (2020 – 2050) classified in five major LC classes (i.e., crops, shrubs, pines, oaks,
and rocky) were used to feed spatial requirements of the carbon storage and sequestration module of the InVEST model. Carbon data on AGB and
BGB, DOM and SOC for each of the major LC classes was collected from data available in published scientific literature at local or regional scale, and
in the official statistics from the Portuguese and Spanish national forestry inventories (Table C2) and used to estimate the carbon stocks in each of
these pools per LC class required by InVEST carbon module (Table C3).
The carbon stocks in AGB and BGB in forest cover classes were computed based on the application of biomass allometric equations (Montero
et al., 2005) to estimate the biomass available for the species occurring within the area, and then converted into carbon through applying a carbon
content factor (Montero et al., 2005) as shown in Table S2. In addition, data on carbon in AGB available from the fifth Portuguese national forest
inventory (IFN5) was directly used after applying a conversion factor (from CO
2
equivalent to C: 12 kg C/44 kg CO
2
= 0.2727). Carbon stocks in
each carbon pool for all the LC classes were maintained constant over time (assuming that that carbon pools are in a steady state), which means that
the carbon sequestration or emission only occurs when a pixel of a given land cover class type change between dates, while if the land cover class
type is kept unchanged between dates, the carbon sequestration/emission rate will be zero for that time period.
Table C1
Total carbon stored (Mg C) and average carbon density (Mg C/ha) per LULC date in the Biosphere Reserve Gerês-Xurés.
LULC date Total carbon stored (Mg C) Average carbon density (Mg C/ha)
1987 25.960.068,95 94,02
2000 26.637.466,45 96,48
2010 28.831.918,75 104,43
Table C2
Total carbon sequestered (Mg C) and carbon sequestration rate (Mg C/ha/yr) per LULC period in the Biosphere Reserve
Gerês-Xurés.
LULC period Total carbon sequestered (Mg
C)
Sequestration rate (Mg C/ha/
yr)
1987–2000 679.494,09 0,19
2000–2010 2.194.451,66 0,79
Table C3
Sources and procedures for the estimation of carbon content in each carbon pool (AGB, BGB, DOM, and SOC) per LULC class.
S. Pais, et al.
Ecosystem Services 44 (2020) 101143
19
Results
The land cover changes (observed and simulated) occurred in the Transboundary Biosphere Reserve Gerês-Xurés (TBR-GX) influenced the supply
of the climate regulation ecosystem service (CRES) over time (Fig. C.1). Between 1987 and 2010 was sequestered a total of 2.87 Tg C, at an average
rate of 0.12 Tg C yr
−1
(Table S2.1 and S2.2). Regarding the future scenarios, the results indicate that the BAU and the Firesmart_4 scenarios present
the highest estimates for the total carbon sequestered (3.63 ± 0.27 and 4.79 ± 0.23 Tg C, respectively) and for the carbon sequestration rate
(0.36 ± 0.03 and 0.48 ± 0.02 Tg C yr
−1
, respectively), while the HNV_FireSmart_2 and the HNV_4 scenarios present the lowest total carbon
sequestered (1.23 ± 0.17 and 0.27 ± 0.13 Tg C, respectively) and the carbon sequestration rate (0.12 ± 0.02 and 0.03 ± 0.01 Tg C yr
−1
,
respectively).Fig. C2
Considering the full period of analysis (1987 – 2050), the results indicate that the GX-BR landscape will keep supplying the climate regulation
ecosystem service (CRES) in the future. However, there were differences among the scenarios (Fig. 4), which may indicate potential trade-offs
between the fire and land management scenarios and the supply of the ACCRES in the area. Therefore, a higher supply of the CRES is expected if in
the future the TBR-GX landscape follows the Firesmart_4 scenario (7.66 ± 0.23 Tg C and 0.12 ± 0.004 Tg C yr
−1
) and the BAU scenario
(6.50 ± 0.27 Tg C and 0.10 ± 0.004 Tg C yr
−1
), while the HNVf_FireSmart_2 (4.10 ± 0.17 Tg C and 0.065 ± 0.003 Tg C yr
−1
) and the HNVf_4
are predicted as the less suitable scenarios for supplying the ACCRES (see Fig. 4 on the main body of the manuscript).
Fig. C1. Spatial representation of total carbon sequestered (Mg C) and carbon sequestration rate (Mg C/ha/yr) per LULC period in the Biosphere Reserve Gerês-Xurés.
S. Pais, et al. Ecosystem Services 44 (2020) 101143
20
References
Autoridade Florestal Nacional. 1998. IFN4 − 4° Inventário Florestal Nacional (Portuguese National Forest Inventory - IFN4). Dados Base do IFN4.
1995–1998. Available: http://www.icnf.pt/portal/florestas/ifn/ifn4
Autoridade Florestal Nacional. 2010. IFN5 − 5° Inventário Florestal Nacional (Portuguese NationalForest Inventory - IFN5). Apresentação do
Relatório Final. Available: http://www.icnf.pt/portal/florestas/ifn/ifn5
P, Sabine C, Bala G, Bopp L, Brovkin V, Canadell J, Chhabra A, DeFries R, Galloway J, Heimann M, et al. 2013. Carbon and other biogeochemical
cycles. In: Stocker, TF, Qin D, Plattner GK, Tignor M, Allen SK, Boschung J, Nauels A, Xia Y, Bex V Midgley PM. (Eds.). Climate change 2013: the
physical science basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. New
York: Cambridge University Press. p; p. 465–570.
Haines-Young R, Potschin MB. 2018. Common International Classification of Ecosystem Services (CICES) V5.1 and Guidance on the Application
of the Revised Structure. Available from www.cices.eu.
Le Quéré C, Andrew RM, Friedlingstein P, Sitch S, Hauck J, Pongratz J, Pickers PA, Korsbakken JI, Peters GP, Canadell JG, Arneth A, Arora VK, et
at. 2018. Global Carbon Budget 2018. Earth Syst. Sci. Data. 10: 2141–2194.
Madeira M, Ricardo RP, Correia A, Garcez A, Monteiro F, Raposo JA, Constantino AT, Duarte J. 2004. Quantidade de carbono orgânico nos solos
de Portugal Continental e particularidades nos solos do Noroeste e dos montados do Sul. Edafologia. 11: 279–293.
Ministerio de Medio Ambiente. 2008. IFN3 - Tercer inventario forestal nacional 1997–2007 (Spanish National Forest Inventory - IFN3).
Organismo Autónomo Parques Nacionales. Available: https://www.miteco.gob.es/es/biodiversidad/servicios/banco-datos-naturaleza/
informacion-disponible/ifn3.aspx
Ministerio de Medio Ambiente. 2000. IFN2 - Segundo inventario forestal nacional 1986–1996 (Spanish National Forest Inventory - IFN2).
Organismo Autónomo Parques Nacionales. Available: https://www.miteco.gob.es/es/biodiversidad/servicios/banco-datos-naturaleza/
informacion-disponible/ifn2.aspx
Montero G, Ruiz-Peinado R, Muñoz M. 2005. Producción de biomassa y fijación de CO2 por los bosques españoles. Madrid: INIA editor.
Penman J, Gytarsky M, Hiraishi T, Krug T, Kruger D, Pipatti R, Buendia L, Miwa K, Ngara T, Tanabe K. et al. editors. 2003. Good practice
guidance for land use, land-use change and forestry. Kanagawa, Japan: Institute for Global Environmental Strategies (IGES)/Intergovernmental
Panel on Climate Change (IPCC).
Pettorelli N, Schulte to Bühne H, Tulloch A, Dubois G, Macinnis-Ng C, Queirós AM, Keith DA, Wegmann M, Schrodt F, Stellmes M, Sonnenschein
R, Geller GN, Roy S, Somers B, Murray N, Bland L, Geijzendorffer I, Kerr JT, Broszeit S, Leitão PJ, Duncan C, El Serafy G, He KS, Blanchard JL,
Lucas R, Mairota P, Webb TJ, Nicholson E. 2017. Satellite remote sensing of ecosystem functions: opportunities, challenges and way forward.
Remote Sensing in Ecology and Conservation. 4: 71–93.
Proença V. 2009. Galicio-Portuguese oak forest of Quercus robur and Quercus pyrenaica: biodiversity patterns and forest response to fire. Lisboa:
Universidade de Lisboa.
Tol RSJ. 2018. The impact of climate change and the social cost of carbon. Working Paper Series 1318. Department of Economics, University of
Fig. C2. Spatial representation of total carbon stored (Mg C) and average carbon density (Mg C/ha) per LULC date in the Biosphere Reserve Gerês-Xurés.
S. Pais, et al. Ecosystem Services 44 (2020) 101143
21
Sussex Business School.
Sharp R, Tallis HT, Ricketts T, Guerry AD, Wood SA, Chaplin-Kramer R, Nelson E, Ennaanay D, Wolny S, Olwero N, Vigerstol K, Pennington D,
Mendoza G, Aukema J, Foster J, Forrest J, Cameron D, Arkema K, Lonsdorf E, Kennedy C, Verutes G, Kim CK, Guannel G, Papenfus M, Toft J,
Marsik M, Bernhardt J, Griffin R, Glowinski K, Chaumont N, Perelman A, Lacayo M Mandle L, Hamel P, Vogl AL, Rogers L, Bierbower W, Denu D,
Douglass J. 2018. InVEST 3.5.0.post337 + n70cca4fa258b User’s Guide. Stanford CA: The Natural Capital Project, Stanford University,
University of Minnesota, The Nature Conservancy, and World Wildlife Fund.
Silva T, Pereira J, Paúl J, Santos M, Vasconcelos M. 2006. Estimativa de emissões atmosféricas originadas por fogos rurais em Portugal. Silva
Lusitana. 14:239–263.
Viana H, Vega-Nieva DJ, Ortiz Torres L, Lousada J, Aranha J. 2013. Principais características da biomassa e propriedades da cinza de combustão
de matos nativos do Norte de Portugal e NW de Espanha. In: Bento J, Lousada J, Patrício MS (Eds.). 7°Congresso Florestal Nacional - Artigos e
Comunicações. Vila Real e Bragança, Portugal: Sociedade Portuguesa de Ciências Florestais.
Appendix D:. Species information and modelling criteria and evaluation metrics.
Data summary
Of 116 modeled species, 21 are under protection of the European Birds and Habitats Directives (13 bird species under the Birds Directive and four
species of both amphibians and reptiles under the Habitats Directive), and 19 are considered as threatened (two amphibians, two reptile and nine
bird species with Vulnerable status; one reptile and four bird species with Endangered status; and one Critically Endangered bird species) according
to regional IUCN conservation criteria for Portugal and Spain. Five modeled species are endemics from Iberian Peninsula (two amphibians and three
reptile species), representing unique regional biodiversity (Table D.1).
Table D1
Species information (taxonomic group, scientific name and acronyms, conservation status), modelling criteria and evaluation metrics. Model replicates refers to the
number of simulations carried out for each modelling algorithm. Quality threshold indicates the value of the area under the curve (AUC) of a receiver operating
characteristics (ROC) used to select models to be included on the ensemble model. AUC and TSS show the accuracy of final ensemble model.
Species information Models
Taxonomic
group
Acronym Scientific name N Birds/Habitats
Directives
Regional IUCN
status
Iberian
endemic
Habitat Model
replicates
Quality
threshold (AUC)
AUC TSS
Amphibia AOB Alytes obstetricans 66 Yes NT No Wetlands 20 0.7 0.968 0.869
Amphibia BSP Bufo spinosus 190 No LC No Water/generalist 20 0.65 0.93 0.736
Amphibia ECA Epidalea calamita 78 Yes LC No Wetlands 20 0.7 0.943 0.751
Amphibia LBO Lissotriton boscai 79 No LC Yes Wetlands 20 0.65 0.884 0.609
Amphibia PPE Pelophylax perezi 107 No LC No Water/generalist 20 0.7 0.977 0.859
Amphibia RIB Rana iberica 171 Yes VU Yes Wetlands 20 0.7 0.949 0.803
Amphibia SSA Salamandra
salamandra
155 No VU No Wetlands/forest 20 0.7 0.906 0.66
Amphibia TMA Triturus marmoratus 64 Yes LC No Wetlands 20 0.7 0.75 0.394
Aves AAPU Apus apus 175 No LC No Anthropogenic 20 0,7 0.922 0.708
Aves AARV Alauda arvensis 369 No LC No Scrubland 10 0,8 0.913 0.656
Aves ACAM Anthus campestris 236 Yes LC No Scrubland 20 0,7 0.893 0.609
Aves ACAU Aegithalos caudatus 241 No LC No Forest 20 0,7 0.890 0.638
Aves ACHR Aquila chrysaetos 45 Yes EN No Mountain 10 0,8 0.987 0.917
Aves ANOC Athene noctua 22 No LC No Scrubland 10 0,8 0.998 0.983
Aves ARUF Alectoris rufa 210 No DD No Agricultural 20 0,7 0.876 0.589
Aves ASPI Anthus spinoletta 17 No LC No Mountain 10 0,8 0.984 0.922
Aves ATRI Anthus trivialis 109 No NT No Forest 10 0,8 0.950 0.763
Aves BBUB Bubo bubo 11 Yes NT No Mountain 10 0,8 0.999 0.996
Aves BBUT Buteo buteo 392 No LC No Forest 20 0,7 0.955 0.795
Aves CBRA Certhia brachydactyla 295 No LC No Forest 20 0,7 0.875 0.610
Aves CCAN Cuculus canorus 492 No LC No Agricultural 20 0,7 0.927 0.700
Aves CCAR Carduelis carduelis 39 No LC No Agricultural 10 0,8 0.984 0.961
Aves CCHL Carduelis chloris 257 No LC No Forest 10 0,8 0.914 0.656
Aves CCIN Cinclus cinclus 54 No LC No Forest 10 0,8 0.969 0.866
Aves CCOR Corvus corone 275 No LC No Agricultural 20 0,7 0.918 0.660
Aves CCOT Coturnix coturnix 180 No DD No Agricultural 10 0,8 0.940 0.697
Aves CCYA Circus cyaneus 93 Yes CR No Mountain 20 0,7 0.940 0.738
Aves COEN Columba oenas 15 No DD No Forest 20 0,7 1 1
Aves CORAX Corvus corax 68 No LC No Mountain 20 0,7 0.963 0.799
Aves CPAL Columba palumbus 329 No LC No Forest 20 0,7 0.880 0.591
(continued on next page)
S. Pais, et al. Ecosystem Services 44 (2020) 101143
22
Table D1 (continued)
Species information Models
Taxonomic
group
Acronym Scientific name N Birds/Habitats
Directives
Regional IUCN
status
Iberian
endemic
Habitat Model
replicates
Quality
threshold (AUC)
AUC TSS
Aves CPYG Circus pygargus 208 Yes EN No Scrubland 20 0,7 0.804 0.455
Aves DMAJ Dendrocopos major 201 No LC No Forest 20 0,7 0.896 0.623
Aves DURB Delichon urbicum 70 No LC No Anthropogenic 10 0,8 0.967 0.890
Aves ECIR Emberiza cirlus 72 No LC No Agricultural 10 0,8 0.981 0.872
Aves ECIT Emberiza citrinella 59 No VU No Scrubland 10 0,8 0.981 0.862
Aves EHOR Emberiza hortulana 25 Yes DD No Mountain 10 0,8 0.995 0.959
Aves ERUB Erithacus rubecula 559 No LC No Agricultural 20 0,7 0.881 0.591
Aves FCOE Fringilla coelebs 543 No LC No Forest 20 0,7 0.840 0.518
Aves FPER Falco peregrinus 45 Yes VU No Mountain 10 0,8 0.969 0.878
Aves FSUB Falco subbuteo 39 No VU No Forest 20 0,7 0.989 0.924
Aves FTIN Falco tinnunculus 138 No LC No Scrubland 20 0,7 0.819 0.496
Aves GGLA Garrulus glandarius 430 No LC No Forest 20 0,7 0.895 0.617
Aves HDAU Cecropis daurica 22 No LC No Anthropogenic 10 0,8 0.995 0.958
Aves HPOL Hippolais polyglotta 123 No LC No Scrubland 20 0,7 0.929 0.702
Aves HRUS Hirundo rustica 188 No LC No Anthropogenic 10 0,8 0.917 0.675
Aves JTOR Jynx torquilla 13 No DD No Forest 10 0,8 0.998 0.996
Aves LARB Lullula arborea 301 Yes LC No Scrubland 20 0,7 0.880 0.606
Aves LCAN Linaria cannabina 498 No LC No Agricultural 20 0,7 0.893 0.632
Aves LCOL Lanius collurio 223 Yes NT No Scrubland 10 0,8 0.927 0.697
Aves LCRI Lophophanes cristatus 443 No LC No Forest 20 0,7 0.942 0.751
Aves LCUR Loxia curvirostra 18 No VU No Forest 10 0,8 0.982 0.845
Aves LEXC Lanius excubitor 32 No LC No Scrubland 20 0,7 0.991 0.927
Aves LMEG Luscinia
megarhynchos
71 No LC No Scrubland 10 0,8 0.978 0.884
Aves MALB Motacilla alba 176 No LC No Agricultural 10 0,8 0.948 0.761
Aves MCAL Emberiza calandra 45 No LC No Agricultural 10 0,8 0.984 0.920
Aves MCIN Motacilla cinerea 115 No LC No Forest 10 0,8 0.943 0.726
Aves MFLA Motacilla flava 23 No LC No Mountain 10 0,8 0.992 0.951
Aves MMIG Milvus migrans 19 Yes NT No Forest 10 0,8 0.998 0.991
Aves MSAX Monticola saxatilis 65 No EN No Mountain 10 0,8 0.947 0.754
Aves MSOL Monticola solitarius 39 No LC No Mountain 10 0,8 0.992 0.940
Aves OHIS Oenanthe hispanica 31 No VU No Mountain 10 0,8 0.950 0.790
Aves OOEN Oenanthe oenanthe 74 No LC No Mountain 10 0,8 0.957 0.784
Aves OORI Oriolus oriolus 170 No LC No Forest 10 0,8 0.930 0.725
Aves OSCO Otus scops 55 No DD No Scrubland 20 0,7 0.970 0.867
Aves PAPI Pernis apivorus 42 Yes VU No Forest 10 0,8 0.803 0.490
Aves PATE Periparus ater 505 No LC No Forest 20 0,7 0.880 0.592
Aves PBON Phylloscopus bonelli 159 No LC No Forest 10 0,8 0.904 0.658
Aves PCAE Cyanistes caeruleus 346 No LC No Forest 10 0,8 0.959 0.775
Aves PCOL Phylloscopus collybita 237 No LC No Forest 20 0,7 0.912 0.663
Aves PDOM Passer domesticus 207 No LC No Anthropogenic 10 0,8 0.958 0.792
Aves PIBE Phylloscopus ibericus 136 No LC No Forest 20 0,7 0.956 0.773
Aves PMAJ Parus major 373 No LC No Forest 10 0,8 0.880 0.604
Aves PMOD Prunella modularis 590 No LC No Scrubland 20 0,7 0.881 0.630
Aves PMON Passer montanus 39 No LC No Agricultural 10 0,8 0.986 0.934
Aves POCH Phoenicurus ochruros 300 No LC No Anthropogenic 10 0,8 0.887 0.570
Aves PPIC Pica pica 20 No LC No Agricultural 10 0,8 0.998 0.979
Aves PPYR Pyrrhula pyrrhula 232 No LC No Forest 10 0,8 0.895 0.637
Aves PRUP Ptyonoprogne
rupestris
45 No LC No Mountain 10 0,8 0.984 0.908
Aves PVIR Picus viridis 317 No LC No Forest 10 0,8 0.893 0.617
Aves PYRR Pyrrhocorax
pyrrhocorax
90 Yes EN No Mountain 10 0,8 0.950 0.788
Aves RIGN Regulus ignicapilla 387 No LC No Forest 20 0,7 0.889 0.605
Aves RRIP Riparia riparia 30 No LC No Scrubland 10 0,8 0.992 0.972
Aves SALU Strix aluco 92 No LC No Forest 20 0,7 0.953 0.778
Aves SATR Sylvia atricapilla 566 No LC No Scrubland 20 0,7 0.888 0.602
Aves SBOR Sylvia borin 74 No VU No Forest 10 0,8 0.944 0.763
Aves SCAN Sylvia cantillans 28 No LC No Scrubland 10 0,8 0.98 0.93
Aves SCOM Sylvia communis 288 No LC No Scrubland 10 0,8 0.928 0.705
Aves SEUR Sitta europaea 138 No LC No Forest 20 0,7 0.940 0.723
Aves SMEL Sylvia melanocephala 48 No LC No Scrubland 10 0,8 0.976 0.869
Aves SRUB Saxicola rubetra 17 No VU No Mountain 10 0,8 0.998 0.984
Aves SSER Serinus serinus 373 No LC No Agricultural 10 0,8 0.867 0.576
Aves STOR Saxicola torquatus 678 No LC No Scrubland 20 0,7 0.880 0.577
Aves STUR Streptopelia turtur 245 No VU No Forest 10 0,8 0.953 0.746
Aves SUND Sylvia undata 569 Yes LC No Scrubland 20 0,7 0.905 0.630
(continued on next page)
S. Pais, et al. Ecosystem Services 44 (2020) 101143
23
Appendix E. . Simulated burned area represented by three fire-size classes under each management scenario between 2011 and 2050.
The results showed that policies promoting farmland and cropland areas (i.e. HNVf scenarios) would lead to a significant reduction of the burned
area in relation to the business-as-usual scenario (see ‘BAU’ and 'HNVf_4′ in Fig. E.1). According to our simulations, the area to be burned by large
fires (> 1,000 hectares) could be potentially reduced from 20,000 hectares under the ‘BAU’ scenario up to 10,000 ha under the ‘HNF_4′ scenarios
(i.e. reduction of 50%; see ‘LF’ between 2031 and 2050 in Fig. E.1). However, the reduction in the area burned by small fires (< 500 hectares) would
only range from approx. 150,000 to 130,000 hectares (i.e. reduction of 13%, see ‘SF’ between 2031 and 2050 in Fig. E.1).
Table D1 (continued)
Species information Models
Taxonomic
group
Acronym Scientific name N Birds/Habitats
Directives
Regional IUCN
status
Iberian
endemic
Habitat Model
replicates
Quality
threshold (AUC)
AUC TSS
Aves SUNI Sturnus unicolor 97 No LC No Anthropogenic 10 0,8 0.944 0.801
Aves TALB Tyto alba 11 No LC No Anthropogenic 10 0,8 1 1
Aves TMER Turdus merula 687 No LC No Agricultural 20 0,7 0.946 0.805
Aves TPHI Turdus philomelos 87 No NT No Forest 10 0,8 0.947 0.776
Aves TTRO Troglodytes
troglodytes
686 No LC No Scrubland 20 0,7 0.902 0.654
Aves TVIS Turdus viscivorus 222 No LC No Forest 20 0,7 0.709 0.346
Aves UEPO Upupa epops 84 No LC No Agricultural 10 0,8 0.822 0.505
Reptilia AFR Anguis fragilis 110 No LC No Grasslands 20 0.65 0.951 0.809
Reptilia CAU Coronella austriaca 76 Yes VU No Open woodlands 20 0.7 0.886 0.614
Reptilia CGI Coronella girondica 77 No LC No Generalist 20 0.7 0.932 0.733
Reptilia CST Chalcides striatus 162 No LC No Grasslands 20 0.7 0.882 0.563
Reptilia LSC Lacerta schreiberi 498 Yes NT Yes Shrublands 20 0.65 0.776 0.388
Reptilia MMO Malpolon
monspessulanus
126 No LC No Generalist 20 0.7 0.937 0.743
Reptilia NAS Natrix astreptophora 184 No LC No Water/generalist 20 0.65 0.883 0.632
Reptilia NMA Natrix maura 163 No LC No Water/generalist 20 0.7 0.71 0.313
Reptilia PAL Psammodromus
algirus
206 No LC No Generalist 10 0.8 0.933 0.725
Reptilia PBO Podarcis bocagei 418 No LC Yes Rocky/generalist 20 0.7 0.969 0.826
Reptilia PGU Podarcis
guadarramae
296 No LC Yes Rocky/generalist 20 0.7 0.964 0.813
Reptilia TLE Timon lepidus 387 Yes LC No Shrublands 20 0.7 0.918 0.679
Reptilia VLA Vipera latastei 102 No VU No Rocky/shrublands 10 0.8 0.961 0.807
Reptilia VSE Vipera seoanei 64 Yes EN No Rocky/shrublands 10 0.8 0.973 0.85
Reptilia ZSC Zamenis scalaris 41 No LC No Open woodlands 10 0.8 0.981 0.912
S. Pais, et al. Ecosystem Services 44 (2020) 101143
24
Appendix F. . Changes in habitat availability for vertebrates under different management scenarios according to the taxonomic group and
conservation status.
This appendix shows the predicted changes in habitat availability for vertebrates under different management scenarios according to different
conservation criteria (expressed in % of change in relation to 2010; Fig. F.1), taxonomic group (expressed in km
2
;Fig. F.2) and habitat preferences
for bird species (expressed in km
2
;Fig. F.3). Our models predicted a wide range of contrasting species responses to management scenarios, with
marked differences among taxonomic groups (Fig. F.2). According to our model projections, habitat availability for birds and reptiles would slightly
Fig. E1. Burned area (expressed in thousands of hectares) grouped by fire-size classes under each management scenario (see acronyms in Table 2) between 2011 and
2050. Results are presented for three fire-size classes: large fires (LF), medium fires (MF), and small fires (SF).
S. Pais, et al. Ecosystem Services 44 (2020) 101143
25
Fig. F1. Habitat availability (% of change in relation to 2010) for vertebrate species with and without protection status under different management scenarios inside
and outside protected areas (see scenario acronyms in Table 2). For all plots, colored lines indicate mean values while the transparent colored areas indicate the error
limits defined by the median range values. Two protection criteria are considered: the protection under the Birds and Habitats (for amphibians and reptiles) European
directives (top) and the regional IUCN conservation status in Portugal and Spain (middle). For the IUCN criteria, species with status of “Least Concern” and “Near
threatened” are grouped as non-threatened, while species with status of Vulnerable, Endangered and Critically Endangered are grouped as threatened. Endemic and
non-endemic vertebrates from Iberian Peninsula are also represented (bottom).
S. Pais, et al. Ecosystem Services 44 (2020) 101143
26
Fig. F3. Habitat availability (km2) for bird species, grouped according to respective habitat preferences, inside and outside the protected areas. For all plots, colored
lines indicate mean values while the transparent colored areas indicate the error limits defined by the median range values.
Fig. F2. Habitat availability (km
2
) for amphibians, birds and reptiles inside and outside the protected areas. For all plots, colored lines indicate mean values while the
transparent colored areas indicate the error limits defined by the median range values.
S. Pais, et al. Ecosystem Services 44 (2020) 101143
27
decrease under the business-as-usual and fire smart scenarios (see ‘BAU’ and ‘Firesmart’ scenarios in Figs. F.2 and F.3). Policies aimed at promoting
agricultural areas would favor both taxonomic groups, progressively recovering the initial values of 2010 (see ‘HNVf’ scenario in Figs. F.2 and F.3).
However, our model projections reveal considerable losses of habitat availability for amphibian species under all management scenarios. None-
theless, contrarily to the other taxa, the business-as-usual and fire smart scenarios could be the best option to ensure long-term habitat availability for
amphibians. Outside protected areas, habitat availability is predicted to decrease considerably for all taxonomic groups.
References
Abatzoglou, J.T., Williams, A.P., Boschetti, L., Zubkova, M., Kolden, C.A., 2018. Global
patterns of interannual climate-fire relationships. Glob. Change Biol. 24, 5164–5175.
https://doi.org/10.1111/gcb.14405.
Agee, J.K., Skinner, C.N., 2005. Basic principles of forest fuel reduction treatments. For.
Ecol. Manage. 211, 83–96. https://doi.org/10.1016/j.foreco.2005.01.034.
Aquilué, N., De Cáceres, M., Fortin, M.-J., Fall, A., Brotons, L., 2017. A spatial allocation
procedure to model land-use/land-cover changes: accounting for occurrence and
spread processes. Ecol. Modell. 344, 73–86. https://doi.org/10.1016/j.ecolmodel.
2016.11.005.
Aquilué, N., Fortin, M.-J., Messier, C., Brotons, L., 2019. The potential of agricultural
conversion to shape forest fire regimes in mediterranean landscapes. Ecosystems.
https://doi.org/10.1007/s10021-019-00385-7.
Araujo, M.B., Thuiller, W., Williams, P.H., Reginster, I., 2005. Downscaling European
species atlas distributions to a finer resolution: implications for conservation plan-
ning. Glob. Ecol. Biogeogr. 14, 17–30. https://doi.org/10.1111/j.1466-822X.2004.
00128.x.
Archibald, S., Lehmann, C.E.R., Gómez-Dans, J.L., Bradstock, R.A., 2013. Defining pyr-
omes and global syndromes of fire regimes. Proc. Natl. Acad. Sci. U. S. A. 110,
6442–6447. https://doi.org/10.1073/pnas.1211466110.
Ascoli, D., Russo, L., Giannino, F., Siettos, C., Moreira, F., 2018. Firebreak and Fuelbreak,
in: Encyclopedia of Wildfires and Wildland-Urban Interface (WUI) Fires. Springer
International Publishing, Cham, pp. 1–9. https://doi.org/10.1007/978-3-319-51727-
8_70-1.
Azevedo, J.C., Moreira, C., Castro, J.P., Loureiro, C., 2011. Agriculture abandonment,
land-use change and fire hazard in Mountain Landscapes in Northeastern, Portugal.
In: Landscape Ecology in Forest Management and Conservation. Springer, Berlin,
Heidelberg, Berlin, Heidelberg, pp. 329–351. https://doi.org/10.1007/978-3-642-
12754-0_14.
Brotons, L., Aquilué, N., de Cáceres, M., Fortin, M.J., Fall, A., 2013. How fire history, fire
suppression practices and climate change affect wildfire regimes in Mediterranean
Landscapes. PLoS One 8, e62392. https://doi.org/10.1371/journal.pone.0062392.
Calviño-Cancela, M., Chas-Amil, M.L., García-Martínez, E.D., Touza, J., 2016. Wildfire
risk associated with different vegetation types within and outside wildland-urban
interfaces. For. Ecol. Manage. 372, 1–9. https://doi.org/10.1016/j.foreco.2016.04.
002.
Calvo, L., Santalla, S., Marcos, E., Valbuena, L., Tárrega, R., Luis, E., 2003. Regeneration
after wildfire in communities dominated by Pinus pinaster, an obligate seeder, and in
others dominated by Quercus pyrenaica, a typical resprouter. For. Ecol. Manage. 184,
209–223. https://doi.org/10.1016/S0378-1127(03)00207-X.
Calvo, L., Tarrega, R., Luis, E., 2002. Regeneration patterns in a Calluna vulgaris
heathland in the Cantabrian mountains (NW Spain): effects of burning, cutting and
ploughing. Acta Oecologica 23, 81–90. https://doi.org/10.1016/S1146-609X(02)
01137-2.
Carvalho-Santos, C., Sousa-Silva, R., Gonçalves, J., Honrado, J.P., 2015. Ecosystem ser-
vices and biodiversity conservation under forestation scenarios: options to improve
management in the Vez watershed, NW Portugal. Reg. Environ. Change. 1–14.
https://doi.org/10.1007/s10113-015-0892-0.
Carvalho, S.B., Brito, J.C., Crespo, E.G., Watts, M.E., Possingham, H.P., 2011.
Conservation planning under climate change: toward accounting for uncertainty in
predicted species distributions to increase confidence in conservation investments in
space and time. Biol. Conserv. 144, 2020–2030. https://doi.org/10.1016/j.biocon.
2011.04.024.
Carvalho, S.B., Brito, J.C., Crespo, E.J., Possingham, H.P., 2010. From climate change
predictions to actions – conserving vulnerable animal groups in hotspots at a regional
scale. Glob. Chang. Biol. 16, 3257–3270. https://doi.org/10.1111/j.1365-2486.2010.
02212.x.
Catry, F.X., Rego, F.C., Bação, F.L., Moreira, F., 2009. Modeling and mapping wildfire
ignition risk in Portugal. Int. J. Wildl. Fire 18, 921–931. https://doi.org/10.1071/
WF07123.
Cerqueira, Y., Araújo, C., Vicente, J., Pereira, H.M., Honrado, J., 2010. Ecological and
Cultural Consequences of Agricultural Abandonment in the Peneda-Gerês National
Park (Portugal), in: Natural Heritage from East to West. Springer Berlin Heidelberg,
Berlin, Heidelberg, pp. 175–183. https://doi.org/10.1007/978-3-642-01577-9_22.
Cerqueira, Y., Navarro, L.M., Maes, J., Marta-pedroso, C., Honrado, J.P., Pereira, H.M.,
2015. Ecosystem services: the opportunities of rewilding in Europe. Rewilding
European Landscapes. 47–64. https://doi.org/10.1007/978-3-319-12039-3.
Chas-Amil, M.L., Prestemon, J.P., McClean, C.J., Touza, J., 2015. Human-ignited wildfire
patterns and responses to policy shifts. Appl. Geogr. 56, 164–176. https://doi.org/10.
1016/j.apgeog.2014.11.025.
Chas-Amil, M.L., Touza, J., Prestemon, J.P., 2010. Spatial distribution of human-caused
forest fires in Galicia (NW Spain), in: Perona, G., Brebbia, C.A. (Eds.), Modelling,
Monitoring and Management of Forest Fires II. pp. 247–258.
Depietri, Y., Orenstein, D.E., 2019. Fire-regulating services and disservices with an ap-
plication to the Haifa-Carmel Region in Israel. Front. Environ. Sci. 7, 107. https://doi.
org/10.3389/fenvs.2019.00107.
Domínguez, J., Arenas, M., Tapia, L., 2005. Guía de las Aves del Parque Natural Baixa
Limia-Serra do Xurés y Lic Baixa Limia. Consellería de Medio Ambiente.
Domínguez, J., Lamosa, A., Pardavila, X., Martínez-Freiría, F., Regos, A., Gil, A., Vidal,
M., 2012. Atlas de los vertebrados terrestres reproductores en el parque natural Baixa
Limia-Serra do Xurés y ZEPVN-LIC Baixa Limia. 9788469565445.
Duane, A., Aquilué, N., Canelles, Q., Morán-Ordoñez, A., De Cáceres, M., Brotons, L.,
2019. Adapting prescribed burns to future climate change in Mediterranean land-
scapes. Sci. Total Environ. 677, 68–83. https://doi.org/10.1016/j.scitotenv.2019.04.
348.
Duane, A., Aquilué, N., Gil-Tena, A., Brotons, L., 2016. Integrating fire spread patterns in
fire modelling at landscape scale. Environ. Model. Softw. 86, 219–231. https://doi.
org/10.1016/j.envsoft.2016.10.001.
Fall, A., Fall, J., 2001. A domain-specific language for models of landscape dynamics.
Ecol. Modell. 141, 1–18. https://doi.org/10.1016/S0304-3800(01)00334-9.
Fernandes, P.M., 2013. Fire-smart management of forest landscapes in the Mediterranean
basin under global change. Landsc. Urban Plan. https://doi.org/10.1016/j.land-
urbplan.2012.10.014.
Fernandes, P.M., Barros, A.M.G., Pinto, A., Santos, J.A., 2016a. Characteristics and con-
trols of extremely large wildfires in the western Mediterranean Basin. J. Geophys.
Res. Biogeosciences 121, 2141–2157. https://doi.org/10.1002/2016JG003389.
Fernandes, P.M., Davies, G.M., Ascoli, D., Fernández, C., Moreira, F., Rigolot, E., Stoof,
C.R., Vega, J.A., Molina, D., 2013. Prescribed burning in southern Europe: developing
fire management in a dynamic landscape. Front. Ecol. Environ. 11. https://doi.org/
10.1890/120298.
Fernandes, P.M., Loureiro, C., Guiomar, N., Pezzatti, G.B., Manso, F.T., Lopes, L., 2014.
The dynamics and drivers of fuel and fire in the Portuguese public forest. J. Environ.
Manage. 146, 373–382. https://doi.org/10.1016/j.jenvman.2014.07.049.
Fernandes, P.M., Monteiro-Henriques, T., Guiomar, N., Loureiro, C., Barros, A.M.G.,
2016b. Bottom-up variables govern large-fire size in Portugal. Ecosystems 19,
1362–1375. https://doi.org/10.1007/s10021-016-0010-2.
Fernandes, P.M., Pacheco, A.P., Almeida, R., Claro, J., 2016c. The role of fire-suppression
force in limiting the spread of extremely large forest fires in Portugal. Eur. J. For. Res.
135, 253–262. https://doi.org/10.1007/s10342-015-0933-8.
Fielding, A.H., Bell, J.F., 1997. A review of methods for the assessment of prediction
errors in conservation presence/absence models. Environ. Conserv. 24, 38–49.
https://doi.org/10.1017/S0376892997000088.
Franks, S.E., Roodbergen, M., Teunissen, W., Carrington Cotton, A., Pearce-Higgins, J.W.,
2018. Evaluating the effectiveness of conservation measures for European grassland-
breeding waders. Ecol. Evol. 8, 10555–10568. https://doi.org/10.1002/ece3.4532.
Gonçalves, A.C., Sousa, A.M.O., 2017. The Fire in the Mediterranean Region: A Case
Study of Forest Fires in Portugal. Mediterr. Identities – Environ. Soc. Cult. https://
doi.org/10.5772/intechopen.69410.
Hirsch, K., Kafka, V., Tymstra, C., McAlpine, R., Hawkes, B., Stegehuis, H., Quintilio, S.,
Gauthier, S., Peck, K., 2001. Fire-smart forest management: a pragmatic approach to
sustainable forest management in fire-dominated ecosystems. For. Chron. 77,
357–363. https://doi.org/10.5558/tfc77357-2.
Huey, R.B., Stevenson, R.D., 1979. Integrating thermal physiology and ecology of ec-
totherms: a discussion of approaches. Integr. Comp. Biol. 19, 357–366. https://doi.
org/10.1093/icb/19.1.357.
Hurteau, M.D., Koch, G.W., Hungate, B.A., 2008. Carbon protection and fire risk reduc-
tion: toward a full accounting of forest carbon offsets. Front. Ecol. Environ. 6,
493–498. https://doi.org/10.1890/070187.
INCF, n.d. Incêndios Rurais — ICNF [WWW Document]. URL http://www2.icnf.pt/
portal/florestas/dfci/inc (accessed 3.4.20).
Keeley, J., Bond, W., Bradstock, R., Pausas, J., Rundel, P., 2012. Fire in Mediterranean
Ecosystems: Ecology, Evolution and Management. Cambridge University Press,
Cambridge, U.K.
Lasanta, T., Arnáez, J., Pascual, N., Ruiz-Flaño, P., Errea, M.P., Lana-Renault, N., 2016.
Space–time process and drivers of land abandonment in Europe. Catena. https://doi.
org/10.1016/j.catena.2016.02.024.
Lehikoinen, A., Brotons, L., Calladine, J., Campedelli, T., Escandell, V., Flousek, J.,
Grueneberg, C., Haas, F., Harris, S., Herrando, S., Husby, M., Jiguet, F., Kålås, J.A.,
Lindström, Å., Lorrillière, R., Molina, B., Pladevall, C., Calvi, G., Sattler, T., Schmid,
H., Sirkiä, P.M., Teufelbauer, N., Trautmann, S., 2018. Declining population trends of
European mountain birds. Glob. Chang. Biol. 25, gcb.14522. https://doi.org/10.
1111/gcb.14522.
Loepfe, L., Martinez-Vilalta, J., Oliveres, J., Piñol, J., Lloret, F., 2010. Feedbacks between
fuel reduction and landscape homogenisation determine fire regimes in three
Mediterranean areas. For. Ecol. Manage. 259, 2366–2374. https://doi.org/10.1016/j.
foreco.2010.03.009.
Lomba, A., Alves, P., Jongman, R.H.G., McCracken, D.I., 2015. Reconciling nature con-
servation and traditional farming practices: a spatially explicit framework to assess
the extent of High Nature Value farmlands in the European countryside. Ecol. Evol. 5,
1031–1044. https://doi.org/10.1002/ece3.1415.
Loureiro, A., Almeida, N.F. de, Carretero, M.A., Octávio, S.P., 2008. Atlas dos Anfíbios e
Répteis de Portugal. Lisboa.
S. Pais, et al. Ecosystem Services 44 (2020) 101143
28
MacDonald, D., Crabtree, J., Wiesinger, G., Dax, T., Stamou, N., Fleury, P., Gutierrez
Lazpita, J., Gibon, A., 2000. Agricultural abandonment in mountain areas of Europe:
Environmental consequences and policy response. J. Environ. Manage. 59, 47–69.
https://doi.org/10.1006/jema.1999.0335.
MAPAMA, 2018. ESTADÍSTICA GENERAL DE INCENDIOS FORESTALES, National Forest
Corporation.
Martínez-Freiría, F., Crochet, P.-A., Fahd, S., Geniez, P., Brito, J.C., Velo-Antón, G., 2017.
Integrative phylogeographical and ecological analysis reveals multiple Pleistocene
refugia for Mediterranean Daboia vipers in north-west Africa. Biol. J. Linn. Soc. 122,
366–384. https://doi.org/10.1093/biolinnean/blx038.
Martinez-Freiria, F., Lorenzo, M., Lizana, M., 2019. Zamenis scalaris prefers abandoned
citrus orchards in Eastern Spain. Ecological insights from a radio-tracking survey.
Amphibia-Reptilia 40, 113–119. https://doi.org/10.1163/15685381-18000003.
Martínez-Freiría, F., Velo-Antón, G., Brito, J.C., 2015. Trapped by climate: interglacial
refuge and recent population expansion in the endemic Iberian adder Vipera seoanei.
Divers. Distrib. 21, 331–344. https://doi.org/10.1111/ddi.12265.
Mclver, J.D., Stephens, S.L., Agee, J.K., Barbour, J., Boerner, R.E.J., Edminster, C.B.,
Erickson, K.L., Farris, K.L., Fettig, C.J., Fiedler, C.E., Haase, S., Hart, S.C., Keeley, J.E.,
Knapp, E.E., Lehmkuhl, J.F., Moghaddas, J.J., Otrosina, W., Outcalt, K.W., Schwilk,
D.W., Skinner, C.N., Waldrop, T.A., Weatherspoon, C.P., Yaussy, D.A., Youngblood,
A., Zack, S., 2013. Ecological effects of alternative fuel-reduction treatments: high-
lights of the National Fire and Fire Surrogate study (FFS). Int. J. Wildl. Fire 22, 63.
https://doi.org/10.1071/WF11130.
Molina-Terrén, D.M., Xanthopoulos, G., Diakakis, M., Ribeiro, L., Caballero, D., Delogu,
G.M., Viegas, D.X., Silva, C.A., Cardil, A., 2019. Analysis of forest fire fatalities in
Southern Europe: Spain, Portugal, Greece and Sardinia (Italy). Int. J. Wildl. Fire 28,
85. https://doi.org/10.1071/WF18004.
Montero, G., Ruiz-peinado, R., Muñoz, M., 2005. Producción de biomasa y fijación de
CO2 por los bosques españoles. Monografías INIA.
Morán-Ordóñez, A., Bugter, R., Suárez-Seoane, S., de Luis, E., Calvo, L., 2013. Temporal
changes in socio-ecological systems and their impact on ecosystem services at dif-
ferent governance scales: a case study of heathlands. Ecosystems 16, 765–782.
https://doi.org/10.1007/s10021-013-9649-0.
Moreira, F., Ascoli, D., Safford, H., Adams, M., Moreno, J.M., Pereira, J.C., Catry, F.,
Armesto, J., Bond, W.J., Gonzalez, M., Curt, T., Koutsias, N., McCaw, L., Price, O.,
Pausas, J., Rigolot, E., Stephens, S., Tavsanoglu, C., Vallejo, R., Van Wilgen, B.,
Xanthopoulos, G., Fernandes, P., 2019. Wildfire management in Mediterranean-type
regions: paradigm change needed. Environ. Res. Lett. https://doi.org/10.1088/1748-
9326/ab541e.
Moreira, F., Pe’er, G., 2018. Agricultural policy can reduce wildfires. Science 359, 1001.
https://doi.org/10.1126/science.aat1359.
Moreira, F., Viedma, O., Arianoutsou, M., Curt, T., Koutsias, N., Rigolot, E., Barbati, A.,
Corona, P., Vaz, P., Xanthopoulos, G., Mouillot, F., Bilgili, E., 2011. Landscape –
wildfire interactions in southern Europe: implications for landscape management. J.
Environ. Manage. 92, 2389–2402. https://doi.org/10.1016/j.jenvman.2011.06.028.
Navarro, L.M., Pereira, H.M., 2012. Rewilding abandoned landscapes in Europe.
Ecosystems 15, 900–912. https://doi.org/10.1007/s10021-012-9558-7.
Omi, P.N., 2015. Theory and practice of wildland fuels management. Curr. For. Reports 1,
100–117. https://doi.org/10.1007/s40725-015-0013-9.
Pausas, J.G., Keeley, J.E., 2019. Wildfires as an ecosystem service. Front. Ecol. Environ.
17, 289–295. https://doi.org/10.1002/fee.2044.
Perino, A., Pereira, H.M., Navarro, L.M., Fernández, N., Bullock, J.M., Ceaușu, S., Cortés-
Avizanda, A., van Klink, R., Kuemmerle, T., Lomba, A., Pe’er, G., Plieninger, T., Rey
Benayas, J.M., Sandom, C.J., Svenning, J.-C., Wheeler, H.C., 2019. Rewilding com-
plex ecosystems. Science (80-.). 364, eaav5570. https://doi.org/10.1126/science.
aav5570.
Pimenta, M., Santarém, M.L., 1996. Atlas das Aves do Parque. Nacional da Peneda-Gerês.
Pleguezuelos, J.M., Márquez, R., Lizana, M., 2002. Atlas y Libro Rojo de los Anfibios y
Reptiles de España. Madrid, Spain.
Regos, A., Domínguez, J., Gil-Tena, A., Brotons, L., Ninyerola, M., Pons, X., 2016. Rural
abandoned landscapes and bird assemblages: winners and losers in the rewilding of a
marginal mountain area (NW Spain). Reg. Environ. Change. 16, 199–211. https://
doi.org/10.1007/s10113-014-0740-7.
Regos, A., Ninyerola, M., Moré, G., Pons, X., 2015. Linking land cover dynamics with
driving forces in mountain landscape of the Northwestern Iberian Peninsula. Int. J.
Appl. Earth Obs. Geoinf. 38, 1–14. https://doi.org/10.1016/j.jag.2014.11.010.
Renard, D., Rhemtulla, J.M., Bennett, E.M., 2015. Historical dynamics in ecosystem
service bundles. Proc. Natl. Acad. Sci. 112, 13411–13416. https://doi.org/10.1073/
pnas.1502565112.
Ribeiro, P.F., Santos, J.L., Bugalho, M.N., Santana, J., Reino, L., Beja, P., Moreira, F.,
2014. Modelling farming system dynamics in High Nature Value Farmland under
policy change. Agric. Ecosyst. Environ. 183, 138–144. https://doi.org/10.1016/j.
agee.2013.11.002.
Roces-Díaz, J.V., Burkhard, B., Kruse, M., Müller, F., Díaz-Varela, E.R., Álvarez-Álvarez,
P., 2017. Use of ecosystem information derived from forest thematic maps for spatial
analysis of ecosystem services in northwestern Spain. Landsc. Ecol. Eng. 13, 45–57.
https://doi.org/10.1007/s11355-016-0298-2.
Rosenberg, K. V., Dokter, A.M., Blancher, P.J., Sauer, J.R., Smith, A.C., Smith, P.A.,
Stanton, J.C., Panjabi, A., Helft, L., Parr, M., Marra, P.P., 2019. Decline of the North
American avifauna. Science (80-.). 366, 120–124. https://doi.org/10.1126/science.
aaw1313.
San-Miguel-Ayanz, J., Moreno, J.M., Camia, A., 2013. Analysis of large fires in European
Mediterranean landscapes: lessons learned and perspectives. For. Ecol. Manage. 294,
11–22. https://doi.org/10.1016/j.foreco.2012.10.050.
Schipper, A.M., Belmaker, J., de Miranda, M.D., Navarro, L.M., Böhning-Gaese, K.,
Costello, M.J., Dornelas, M., Foppen, R., Hortal, J., Huijbregts, M.A.J., Martín-López,
B., Pettorelli, N., Queiroz, C., Rossberg, A.G., Santini, L., Schiffers, K., Steinmann,
Z.J.N., Visconti, P., Rondinini, C., Pereira, H.M., 2016. Contrasting changes in the
abundance and diversity of North American bird assemblages from 1971 to 2010.
Glob. Change Biol. 22, 3948–3959. https://doi.org/10.1111/gcb.13292.
Sharp, R., Chaplin-Kramer, R., Wood, S., Guerry, A., Tallis, H., Ricketts, T., 2018. InVEST
User’s Guide. The natural capital project. Nat. Cap. Proj. Stanford Univ. Univ.
Minnesota. Nat. Conserv. andWorldWildlife Fund 14, 81–87. https://doi.org/10.
1007/s13398-014-0173-7.2.
Sil, Â., Azevedo, J.C., Fernandes, P.M., Regos, A., Vaz, A.S., Honrado, J.P., 2019a. (Wild)
fire is not an ecosystem service. Front. Ecol. Environ. 17, 429–430. https://doi.org/
10.1002/fee.2106.
Sil, Â., Fernandes, P.M., Rodrigues, A.P., Alonso, J.M., Honrado, J.P., Perera, A., Azevedo,
J.C., 2019b. Farmland abandonment decreases the fire regulation capacity and the
fire protection ecosystem service in mountain landscapes. Ecosyst. Serv. 36, 100908.
https://doi.org/10.1016/j.ecoser.2019.100908.
Sil, Â., Fonseca, F., Gonçalves, J., Honrado, J., Marta-Pedroso, C., Alonso, J., Ramos, M.,
Azevedo, J.C., 2017. Analysing carbon sequestration and storage dynamics in a
changing mountain landscape in Portugal: insights for management and planning.
Int. J. Biodivers. Sci. Ecosyst. Serv. Manage. 13, 82–104. https://doi.org/10.1080/
21513732.2017.1297331.
Sil, Â., Rodrigues, A.P., Carvalho-Santos, C., Nunes, J.P., Honrado, J., Alonso, J., Marta-
Pedroso, C., Azevedo, J.C., 2016. Trade-offs and synergies between provisioning and
regulating ecosystem services in a mountain area in portugal affected by landscape
change. Mt. Res. Dev. 36, 452–464. https://doi.org/10.1659/MRD-JOURNAL-D-16-
00035.1.
Soares, C., Álvares, F., Loureiro, A., Sillero, N., Arntzen, J.W., Brito, J.C., 2005. Atlas of
the amphibians and reptiles of Peneda-Gerês National Park, Portugal.
Tedim, F., Leone, V., Amraoui, M., Bouillon, C., Coughlan, M., Delogu, G., Fernandes, P.,
Ferreira, C., McCaffrey, S., McGee, T., Parente, J., Paton, D., Pereira, M., Ribeiro, L.,
Viegas, D., Xanthopoulos, G., 2018. Defining extreme wildfire events: difficulties,
challenges, and impacts. Fire 1, 9. https://doi.org/10.3390/fire1010009.
Tedim, F., Leone, V., Xanthopoulos, G., 2016. A wildfire risk management concept based
on a social-ecological approach in the European Union: Fire Smart Territory. Int. J.
Disaster Risk Reduct. 18, 138–153. https://doi.org/10.1016/j.ijdrr.2016.06.005.
Tedim, F., Remelgado, R., Borges, C., Carvalho, S., Martins, J., 2013. Exploring the oc-
currence of mega-fires in Portugal. For. Ecol. Manage. 294, 86–96. https://doi.org/
10.1016/j.foreco.2012.07.031.
Thuiller, W., Lafourcade, B., Engler, R., Araújo, M.B., 2009. BIOMOD – a platform for
ensemble forecasting of species distributions. Ecography (Cop.) 32, 369–373. https://
doi.org/10.1111/j.1600-0587.2008.05742.x.
Turco, M., Bedia, J., Di Liberto, F., Fiorucci, P., von Hardenberg, J., Koutsias, N., Llasat,
M.-C., Xystrakis, F., Provenzale, A., 2016. Decreasing fires in mediterranean Europe.
PLoS One 11, e0150663. https://doi.org/10.1371/journal.pone.0150663.
Turco, M., Jerez, S., Augusto, S., Tarín-Carrasco, P., Ratola, N., Jiménez-Guerrero, P.,
Trigo, R.M., 2019. Climate drivers of the 2017 devastating fires in Portugal. Sci. Rep.
9, 13886. https://doi.org/10.1038/s41598-019-50281-2.
Turner, M.G., Donato, D.C., Romme, W.H., 2013. Consequences of spatial heterogeneity
for ecosystem services in changing forest landscapes: priorities for future research.
Landsc. Ecol. 28, 1081–1097. https://doi.org/10.1007/s10980-012-9741-4.
Úbeda, X., Pereira, P., Badía, D., 2018. Prescribed fires. Sci. Total Environ. 637–638,
385–388. https://doi.org/10.1016/j.scitotenv.2018.04.272.
van der Werf, G.R., Randerson, J.T., Giglio, L., Collatz, G.J., Kasibhatla, P.S., Arellano,
A.F., 2006. Interannual variability in global biomass burning emissions from 1997 to
2004. Atmos. Chem. Phys. 6, 3423–3441. https://doi.org/10.5194/acp-6-3423-2006.
van der Zanden, E.H., Verburg, P.H., Schulp, C.J.E., Verkerk, P.J., 2017. Trade-offs of
European agricultural abandonment. Land Use Policy 62, 290–301. https://doi.org/
10.1016/j.landusepol.2017.01.003.
Verburg, P.H., de Nijs, T.C.M., van Eck, J.R., Visser, H., de Jong, K., 2004. A method to
analyse neighbourhood characteristics of land use patterns. Comput. Environ. Urban
Syst. 28, 667–690. https://doi.org/10.1016/j.compenvurbsys.2003.07.001.
Vilén, T., Fernandes, P.M., 2011. Forest fires in mediterranean countries: CO2 emissions
and mitigation possibilities through prescribed burning. Environ. Manage. 48,
558–567. https://doi.org/10.1007/s00267-011-9681-9.
Whelan, C.J., Şekercioğlu, Ç.H., Wenny, D.G., 2015. Why birds matter: from economic
ornithology to ecosystem services. J. Ornithol. 156, 227–238. https://doi.org/10.
1007/s10336-015-1229-y.
S. Pais, et al. Ecosystem Services 44 (2020) 101143
29
