You are on page 1of 9

Ecological Indicators 71 (2016) 627–635

Contents lists available at ScienceDirect

Ecological Indicators
journal homepage: www.elsevier.com/locate/ecolind

The reliability of a composite biodiversity indicator in predicting bird


species richness at different spatial scales
Francesco Valerio a,b , Marco Basile b,c,d,∗ , Rosario Balestrieri b,c , Mario Posillico c,e ,
Sergio Di Donato c , Tiziana Altea d , Giorgio Matteucci f
a
CIBIO/InBIO-UE- Research Center in Biodiversity and Genetic Resources, Pole of Évora Applied population and community ecology laboratory, University
of Évora UBC - Conservation Biology Lab, Department of Biology, Mitra, 7002-554, Évora, Portugal
b
Coordinamento MITO2000, via Trento 49, I-43100 Parma, Italy
c
Consiglio Nazionale delle Ricerche, Istituto di Biologia Agroambientale e Forestale, Montelibretti, RM, Italy
d
Chair of Wildlife Ecology and Management, University of Freiburg, Tennenbacher Str. 4, D-79106 Freiburg, Germany
e
Corpo Forestale dello Stato, Ufficio Territoriale Biodiversità di Castel di Sangro, Centro Ricerche Ambienti Montani, Via Sangro 45, I-67031 Castel di
Sangro, AQ, Italy
f
Consiglio Nazionale delle Ricerche, Istituto per i Sistemi Agricoli e Forestali del Mediterraneo, Rende, CS, Italy

a r t i c l e i n f o a b s t r a c t

Article history: Several biodiversity features can be linked to landscape heterogeneity, that, in turn, can be informative
Received 15 March 2016 for management and conservation purposes. Usually, the more the landscape is complex the more the
Received in revised form 9 July 2016 biodiversity increases. Biodiversity indicators can be a useful tool to assess biodiversity status, in func-
Accepted 25 July 2016
tion of landscape heterogeneity. In this study, we developed a biodiversity indicator, based on Shannon
Available online 4 August 2016
diversity index and built from distribution maps of protected species. With such an approach, we seek
to evaluate the feasibility of using a combination of target species as a surrogate for assessing the status
Keywords:
of the whole bird community. Our approach was spread over multiple spatial scales, to determine which
Landscape metrics
Species distribution model
was the most informative. We selected four species protected by European regulation and generated
Heterogeneity a presence-absence map from species distribution modelling. We, therefore, used the FRAGSTATS bio-
Bird community diversity metric to calculate Shannon index for the overlapped presence-absence maps, at two spatial
Umbrella species scales (500 m and 1000 m). Then, the relationships with the whole community was assessed through gen-
eralised least square models, at the spatial scale of 4 ha, 9 ha and 25 ha. Results showed that the higher
rate of variability of community was explained by the biodiversity indicator at 1000 m scale. Indeed, the
more informative spatial scale for the whole bird community was 9 ha. In addition, a pattern emerged
about the relationships between biodiversity indicator and community richness, that is worth of further
research. Our study demonstrates that the usefulness of surrogate species for biodiversity and community
assessment can become clear only at a certain spatial scales. Indeed, they can be highly predictive of the
whole community, and highly informative for conservation planning. Moreover, their use can optimize
biodiversity monitoring and conservation, focusing on a small number of noteworthy species.
© 2016 Elsevier Ltd. All rights reserved.

1. Introduction occurring species (Simberloff, 1998). The promising advantages


of this concept became interesting especially for conservation-
The concept of umbrella species has appealed researchers and ists and practitioners (Barua, 2011). Hence, an umbrella species
conservationist since its early appearance in scientific publications. shows key ecological relationships with the environment, on which
An umbrella species is defined as a species with such ecological other species depend, or which range covers those of other species.
demands that preserving its habitat will benefit a large set of co- The use of umbrella species in biological conservation has grown
faster in the last decades, till its acceptance among ecologists
(Fleishman et al., 2001; Roberge and Angelstam, 2004). Applied
researches has covered, for instance, the role of salamanders in
∗ Corresponding author at: Chair of Wildlife Ecology and Management, University
the food web (Davic and Welsh, 2004), the importance of wood-
of Freiburg, Tennenbacher Str. 4, D-79106 Freiburg, Germany.
E-mail address: marcob.nat@gmail.com (M. Basile).
peckers to supply micro-habitats to other species (Brambilla et al.,

http://dx.doi.org/10.1016/j.ecolind.2016.07.043
1470-160X/© 2016 Elsevier Ltd. All rights reserved.
628 F. Valerio et al. / Ecological Indicators 71 (2016) 627–635

2013; Bütler et al., 2004), the role of flying squirrel in predict- be even more difficult, due to the difficulty of gaining reli-
ing the presence of old-growth forest associated species (Hurme able data over large areas (Simberloff, 1998). Main complications
et al., 2008). A consequence of the relationship with the envi- arise in remote regions, where occurrence data are not available.
ronment is that often umbrella species are considered indicator To circumvent this problem, surrogate species should be tested
species too (Welsh and Droege, 2001). Their presence/absence can over data-rich areas, through high predictive models, including
indicate the occurrence of particular ecological conditions, and ecologically relevant environmental predictors (Ferrier, 2002). Fre-
supply information on species richness or ecosystem functioning quently, the use of landscape- scale umbrella species imply a
(Fleishman et al., 2005; Mac Nally and Fleishman, 2004; Welsh multi-species approach. Such an approach can ensure the exporta-
and Hodgson, 2013). The shift from umbrella to indicator defini- bility of the methods, regardless of landscape heterogeneity and
tion imply the use of the species as a measure of environmental the prediction of species richness in data-poor areas (Banks-
conditions (Welsh and Droege, 2001). Commonly, researchers and Leite et al., 2011; Fleishman et al., 2005). Another approach
practitioners are interested in biodiversity indicators, i.e. those regards the use of landscape metrics or environmental predic-
species which presence is related to high species richness. How- tors as a biodiversity indicator (Lindenmayer et al., 2002; Noss,
ever, it can be difficult to identify a standard set of conditions to rely 1999). However, the response of a species to landscape depends
on for choosing such species (Branton and Richardson, 2011; Lawler on several conditions dictated by its own life history (Ducci
and White, 2008), but experimental methods showed that bio- et al., 2015; Schindler et al., 2013). Therefore, a multi-species
diversity indicators can be regionally or taxonomically restricted approach is needed to address the response of landscape-scale
(Lambert, 2011; Stoch et al., 2009). In most cases, however, bio- indicators, taking into account the interaction of multiple spatial
diversity indicators are considered endemic or otherwise typical scales (Bellamy et al., 2013; Fearer et al., 2007; McAlpine et al.,
of the region of interest and/or the most common in their taxo- 2008).
nomic group (Lambert, 2011; Sinsch et al., 2011), limiting de facto In this study, we tested a multi-species biodiversity index as a
the use of the umbrella species or biodiversity indicators. There- predictor of the whole species richness (Fig. 1). In doing so, we com-
fore, a generalisation of the concepts is not yet feasible (Favreau muted the environmental suitability (ES) of surrogate species into
et al., 2006). the index at different spatial scales. Our aim is to provide support
The usefulness of a surrogate species (i.e. an umbrella or a bio- to the idea that a set of surrogate species, representative of differ-
diversity indicator) depends also on the spatial extent at which ent habitats, can predict the species richness across the landscape.
it is used to infer about other species, since species and envi- We seek to deduce useful guidelines for landscape management.
ronments can interact over many spatial scales (Wiens, 1989). In doing this, we selected four bird species, protected by Euro-
Landscape-scale conservation is a major issue for practition- pean regulations (Birds Directive), inhabiting all the habitat types
ers, especially in the face of increasing land use (Lindenmayer present in the study area:
et al., 2008). Finding umbrella species at landscape level can

Fig. 1. Workflow describing the modelling process. Species occurrences and environmental layers are gathered together to produce presence/absence maps. These, in turn,
are processed through FRAGSTATS to obtain a biodiversity indexes, and then, the indexes are related to species richness. In our case, (a) we obtained the species occurrence
of the four candidate species (fig. x) and the environmental layers (see Methods section and Supplementary materials). (b) Species occurrence undergone a bias-correction
procedure. (c) Presence/absence maps were extracted using a 10th percentile threshold. (d) A multispecies presence/absence map was developed, discretising as separate
patches the overlaps between the four PA maps. (e) The biodiversity index was calculated over the multispecies PA map, implementing two different spatial scales (moving
windows). (f) In parallel, bird community count data were aggregated over three spatial scales (three grids). (g) Other three grids were implemented for ecotonal areas only.
(h) GLS were built to test the relationship between the biodiversity index and the bird community richness.
F. Valerio et al. / Ecological Indicators 71 (2016) 627–635 629

Fig. 2. Study area. The site is located in central Italy, centroid coordinates 41◦ 50 N; 13◦ 57 E.

1) The rock partridge Alectoris graeca, typical of high mountain influenced by landscape structure and, not the least, by the occur-
slopes and rocky prairies (Brichetti and Fracasso, 2004); rence of other species (Kissling et al., 2012; Wisz et al., 2013). As
2) The Eurasian nightjar Caprimulgus europaeus, typical of forest a consequence, landscape metrics can be used as a proxy of biodi-
edges (Brichetti and Fracasso, 2006); versity indicators (Schindler et al., 2013). Indeed, we computed a
3) The collared flycatcher Ficedula albicollis, typical of beech forests biodiversity index as a spatial metric, where each species represent
(Brichetti and Fracasso, 2013); a specific habitat. We hypothesize that the combination of species
4) The red-backed shrike Lanius collurio, typical of shrubland with different habitat requirements can predict species richness
(Brichetti and Fracasso, 2011). over the whole landscape. To test this, we approached to such an
issue at different spatial scales, comparing the results.

We evaluated the ES using species distribution models (SDMs).


SDMs are statistical techniques used to predict the distribution of a 2. Metodhs
species, based upon a set of occurrences and environmental predic-
tors (Austin, 2007; Elith and Leathwick, 2009). Their usefulness in 2.1. Study area
the broad context of conservation biology is widely demonstrated
(Guisan and Thuiller, 2005; Visconti et al., 2015). We used the pres- The study site (centroid coordinates 41◦ 50 N; 13◦ 57 E) is the
ence probability distribution of the four selected birds to obtain a easternmost part of the Site of Community Importance IT7110205
single biodiversity landscape metric. Many ecological processes are “Parco Nazionale d’Abruzzo” (Fig. 2). It is a mountainous area, ca.
630 F. Valerio et al. / Ecological Indicators 71 (2016) 627–635

Table 1
variables selected for species distribution modelling (i.e. variables used by MaxEnt to derive species’ distribution) and relative importance. NA = not applicable, variable not
included in model.

Variable Rock partridge Eurasian nightjar Collared flycatcher Red-backed shrike

CORINE Land Cover 24.9 5.4 47.7a 51.5b


Forest type 9 10.4 NA NA
Forest structure 0.1 62.3 20.4 NA
Elevation 49.7 0 22.4 38.6
Slope 8.8 5.2 1.8 6
Aspect 7.5 16.6 NA NA
Patch richness NA NA NA 3.9
Core area NA NA 7.6 NA
a
Merged with forest type map and reclassified layer, comprising 10 classes.
b
Reclassified into 8 classes.

83 km2 wide, with a mainly limestone bedrock, located in the cen- tor without the specific mesh and comparing the results with the
tral Apennines, L’Aquila province, Abruzzo Region, Italy. Elevation estimated richness for the whole grid. Comparison showed that
ranges from 1061 to 2285 m a.s.l., with mountain ridges oriented any mesh contributed significantly more to species richness. How-
NNW to SSE. Total yearly precipitations averaged across 20 years ever, species richness may increase in marginal areas more than
from four meteorological stations located immediately outside the in core areas, due to the proximity of different habitats (McGarigal
study site, ranged from ca. 850 to >1100 mm with a monthly min- and McComb, 1995) and a positive effects of habitat fragmentation
imum from June to August and maximum precipitations recorded (Fahrig, 2003). For this reasons, a subset of meshes covering eco-
in Autumn and Winter, usually as snow from late-November until tones were extracted from every grid. Ecotones were those patches
March. Annual average temperature ranges from 6.8 to 10.8 ◦ C, that were not recognised as core area by FRAGSTATS, i.e. those area
peaking in August (from 20.4 to 15.8 ◦ C), and usually falling below that represented the interior area of patches after a 200 m edge
0 ◦ C in winter months (from −1.3 to 3.0 ◦ C on January). Morphology buffer was eliminated.
is variable, from flat to gently-rolling mid-to-high elevation pas-
tures, to steep and usually forested mountain slopes, from small
and medium-sized gorges to karstic plateaus. Only one, usually 2.3. Statistical analyses
permanent, stream crosses the area at its western limit, though
many springs are scattered and used by livestock breeders. A high The environmental suitability (ES) of the selected species A.
elevation (1881 m a.s.l.) small (2 ha) glacial lake is present. Tran- graeca, C. europaeus, F. albicollis and L. collurio was calculated using a
shumance characterize the area, which hosts a few thousands of machine learning technique based on the maximum entropy (Max-
sheep from mid-June to early-October, as well as a few cattle and Ent) algorithm (Phillips et al., 2006) (Fig. 1a–d). The building of
horses (mostly belonging to resident breeders). Beech woods (Fagus the SDMs took advantage of the presence data in the databases of
sylvatica) (30%) a wide array of both primary and secondary pas- the Forest Service (Corpo Forestale dello Stato − Ufficio Territori-
tures and grasslands (50%), bare rocks, screes and rocky grassland ale Biodiversità di Castel di Sangro), the National Research Council
(ca. 10%) with a few conifer reforestations and cultivations, as well (Consiglio Nazionale delle Ricerche) and the MITO2000 project
as limited shrubland characterize the vegetation and land use of (Fornasari et al., 2010) (Fig. 3). The whole dataset was processed to
the area; a fairly widespread network of unimproved pasture and reduce spatial autocorrelation and multicollinearity through SDM-
forest roads (but for its southernmost and highest elevation por- toolbox (Brown, 2014) (Fig. 1 b). The processed datasetconsisted
tion) crosses the area, which hosts a very small human settlement of 108 presence records, divided as such: 14 for A. graeca; 16 for C.
near its northern edge comprising some small family raised farms. europaeus; 42 for F. albicollis; and 36 for L. collurio. The environmen-
Among wildlife, the brown bear (Ursus arctos) and the wolf (Canis tal predictors used to derive ES maps consisted of 6 layers (Table 1).
lupus) are present, along with the red- and the roe-deer (Cervus Firstly, the land cover (CORINE 2006 III level) (EEA, 2012; European
elaphus, Capreolus capreolus, respectively) and the wild boar (Sus Union, 1994), a forest typology map (including 26 categories) and a
scrofa). Hunting is mainly limited at the very southern and eastern forest structure map (including 4 categories) (Marchetti et al., 2009)
border of the area. were used as categorical habitat maps. A digital elevation model
(provided by the national Institute for Environmental Protection
and Research, available at http://www.sinanet.isprambiente.it/it)
2.2. Richness estimation was used to derive the elevation, slope and aspect, trough GIS. In
addition, landscape metrics of patch richness and core area were
Presence data were supplied by ad hoc surveys and by the implemented through FRAGSTAT v. 4.2.1.603 (McGarigal et al.,
database of the Forest Service (Corpo Forestale dello Stato − Uffi- 2012), with a moving window of 500 m and 200 m, respectively
cio Territoriale Biodiversità di Castel di Sangro) (Anon, 2008). (details are provided in Appendix S1). According to their ecological
Such database include data collected usually with two differ- requirements, not all layers were used for every species (Table 1).
ent techniques: passive hearing or playback stimulation. Layout Analyses were carried out at the finest possible scale, i.e. 20 × 20 m
of sampling points followed stratified-systematic or stratified- grid, due to the absence of relevant information in the literature,
random approach. Presence data were aggregated over multiple except for L. collurio, for which was used a 100 × 100 m grid that was
grids. Specifically, three grids were used: one with mesh area of 4 ha proved consistent by Brambilla et al. (2009). Every dataset included
(GR4), one with mesh area of 9 ha (GR9) and one with mesh area of also 10.000 background points and was split into a training dataset,
25 ha (GR25) (Fig. 1f–g). Species richness for every grid was evalu- including a randomly selected 70% of the data, and a testing dataset,
ated using the incidence-based coverage estimators (ICE) (Lee and including the remaining 30%. For each species, model ran 50 times
Chao, 1994). Since no difference emerged between empirical and and each run was iterated at most 1.000 times, through bootstrap
estimated richness, simple counts were considered exhaustive of estimation, and results averaged. The area under the curves (AUC)
the local species richness (Table 2). The contribution of every mesh formed by the receiver operating characteristics (ROC) was used to
to overall species richness was evaluated running the ICE estima- assess model fit (Hanley and McNeil, 1982).
F. Valerio et al. / Ecological Indicators 71 (2016) 627–635 631

Fig. 3. Species occurrences employed in the distribution modelling and potential distribution within the study area. Shaded area = potential distribution; dot = presence
point; a = rock partridge; b = Eurasian nightjar; c = collared flycatcher; d = red-backed shrike. The inlets show occurrence points outside the study area. The arrow indicate a
point relocated on the map for better visualization.

Table 2
grid characteristics. Richness is estimated through ICE estimator. Standard deviation of observed richness per mesh is in brackets.

Acronym Mesh area N. of meshes Observed species richness Estimated richness Mean observed richness per mesh

GR4 4 ha 164 57 56.20 9.10 (6.12)


GR4eco 4 ha 153 58 58.43 8.42 (6.00)
GR9 9 ha 148 59 59.20 8.92 (5.85)
GR9eco 9 ha 137 59 59.71 7.65 (5.53)
GR25 25 ha 118 59 59.39 9.58 (6.01)
GR25eco 25 ha 121 59 59.74 8.79 (5.97)

The final ES map of every species was used to generate a with FRAGSTATS, the Shannon’s diversity metric (L127 Shannon’s
multi-species biodiversity indicator (Fig. 1e). First, the outputs Diversity Index SHDI in FRAGSTATS) were computed, generating
of MaxEnt were translated in presence-absence map, using the an output where small values indicate landscape homogeneity for
threshold of the 10th percentile of the training presence. Then, a the presence of few umbrella species, and high values indicate land-
new categorical map was generated from the overlap of discretized scape heterogeneity for the presence of many umbrella species. This
presence-absence maps, with as many categories as species and a procedure was implemented at two spatial scales: a 1000 m mov-
category representing the suitability of two or more species. Again ing window (H1000) and a 500 m moving window (H500). Since
632 F. Valerio et al. / Ecological Indicators 71 (2016) 627–635

Table 3 Table 4
Spatial autocorrelation and heteroscedasticity within each dataset. GLS parameter estimation for every dataset, with Likelihood Ratio Test (LRT) per-
formed over the null model and pseudo-R2 to measure explained variability. Dataset
Dataset Moran I p-value Levene F p-value acronyms are explained in Table 2.
GR4 0.47 <0.001 2.97 0.054
Dataset Parameter Estimates LRT pseudo-R2
GR4eco 0.49 <0.001 18.01 <0.001
GR9 0.52 <0.001 9.05 <0.001 GR4 i 7.75*** 0.53 −0.014
GR9eco 0.57 <0.001 21.62 <0.001 H500 2.83
GR25 0.49 <0.001 2.27 0.11 GR9 i 4.81*** 14.3*** 0.345
GR25eco 0.53 <0.001 12.61 <0.001 H500 3.57***
GR25 i 7.2*** 1.48 0.116
H500 1.07
every target species was representative of a specific habitat, in this GR4eco i 5.52*** 14.15*** 0.083
way we assessed also the presence of different habitats using two H500 2.94***
GR9eco i 4.51*** 15.69*** 0.27
different spatial scales. High values, representing portions of terri- H500 3.54***
tory where combination of the species co-occur, indicate also area GR25eco i 7.23*** 2.81 0.093
of high conservation value. H500 1.1
Finally, six datasets were obtained and replicated for the two GR4 i 5.27*** 6.10* −0.007
H1000 2.83**
spatial scales of the biodiversity indicators, H1000 and H500
GR9 i −4.16*** 30.72*** 0.876
(Table 2). Taking into account the strong spatial structure of the H1000 10.75***
richness data, the relationship between landscape diversity (as GR25 i −0.73 17.88*** 0.732
expressed by the biodiversity index) and species richness (within H1000 7.15***
the whole community) was investigated through generalised least GR4eco i 1.76 29.7 0.068
H1000 5.77 *** ***
square models (GLS) (Beale et al., 2010). Before applying GLS, data GR9eco i −3.27*** 0.83
64.63***
were tested for spatial autocorrelation through Moran’s I statistic H1000 9.90***
(Cliff and Ord, 1981; Moran, 1950). Data were also tested for het- GR25eco i 3.96*** 11.4 0.336
eroscedasticity with a Levene F-test, dividing the dataset into three H1000 3.87*** ***

groups of high, medium and low richness, separated by first and i = intercept.
*
third quartiles (Fox, 2008). In this way, we assessed the heterogene- p < 0.05.
**
ity among central and extreme values of richness. Tests and GLS p < 0.01.
***
p < 0.001.
were applied for every dataset. GLS was built maximising the log-
likelihood, allowing to evaluate the predictability against the null
model through a likelihood ratio test. In addition, pseudo-R2 was
developed for every model as a measure of explained variability
(Griffis and Stedinger, 2007). Statistical analyses were conducted
with R software and packages ‘nlme’ and ‘spdep’ (Bivand and Piras,
2015; Pinheiro et al., 2015; R Core Team, 2015).

3. Results

The goodness of the distribution modelling were not similar


among the 4 target species. The model for the rock partridge
showed a high predictive power (AUC = 0.963) and a high affinity
for high altitude prairie and grassland, even if the elevation was the
most important variable (Table 1). The nightjar model showed an
AUC = 0.818, suggesting a lower predictive power than the former
model. However, the nightjar resulted more influenced by forest
structure and type (Table 1), being linked to semi-open habitat,
with broadleaved trees and open pastures. The collared flycatcher
distribution model was quite predictive (AUC = 0.879), with a major
importance of land cover, especially beech forests (Table 1). Finally,
the model for the red-backed shrike resulted in an AUC = 0.87, with
land cover (mainly ecotones) and elevation being the most impor-
tant variables in predicting its presence (Table 1). The resulting
distribution maps showed an almost complementary distribution Fig. 4. Linear model of species richness as a function of the biodiversity indicator
commuted at a 1000 m spatial scale (i.e. H1000) over the 9 ha grids (i.e. total area
(Fig. 3)
GR9 and ecotones only GR9eco).
The computing of H1000 resulted in a mean diversity of 0.97
(±0.28 SD), while the H500 had a diversity of 0.77 (±0.35 SD).
Every dataset showed a Moran’s I > 0.45 (p < 0.001), suggesting (p < 0.01), while H500 did not result significant for the models GR4
a strong spatial structure of data. The Levene F-test indicated the (p = 0.45), GR25 (p = 0.18), GR25eco (p = 0.085) (Table 4). Likelihood
absence of homoscedasticity in every dataset (p ≤ 0.05) except GR4 ratio test confirmed the predictability of every model, except for the
(F = 2.97; p = 0.054) and GR25 (F = 2.27; p = 0.11) (Table 3). However, three non-significant ones (Table 4). However, explained variance
p values were ignored, due to small values, and heteroscedastic- was high only for the models of GR9 and GR9eco built with H1000
ity considered present in every datasets. As a consequence of this (Table 4). Even if these latter models are highly significant, the lin-
two statistics, both spatial autocorrelation and heteroscedastic- ear model showed a large variability for higher values (approx. > 10)
ity were included in the GLS model building. The influence of the of species richness (Fig. 4). A characteristic trend emerged from
two biodiversity indicators (H1000 and H500) varied among grids. the analyses, where the value of GR9 model are higher than the
H1000 influence on species richness was significant for every model other ones, the ecotones explain less richness than the total area
F. Valerio et al. / Ecological Indicators 71 (2016) 627–635 633

nity, exploring three spatial scales. We tested the goodness of the


explanatory variable (i.e. the biodiversity indicator) in predicting
the response variable (i.e. community species richness) through
linear models that take in account the spatial structure of the data.
Results showed that the greatest portion of richness variability was
explained at the medium spatial scale.
Surrogate species have a widespread application in biological
conservation, even though their usefulness was and still is much
debated (Barua, 2011). The use of single or a small set of umbrella
species allows to study fewer species while getting inference that
can assure protection of many species (Simberloff, 1998). How-
ever, to be a valuable umbrella species many assumptions must
be met, from the probability of persistence in the region or the sta-
tus of adjacent lands (Berger, 1997), to the co-occurrence with a
large number of other species (Andelman and Fagan, 2000). In the
case of birds these assumptions may be easily met and serve as a
solid base for testing. Indeed, we met these criteria in conducting
our research in a protected area surrounded by protected lands,
while the co-occurrence of many species was tested a priori. More-
over, one species cannot always be a perfect surrogate for an entire
community (Breckheimer et al., 2014), or it can show responses
that are difficult to interpret (Hoffmann et al., 2016). In fact, single
Fig. 5. Comparison of the pseudo-R2 values. Grey = ecotones-only models; umbrella species can be less predictive of species richness, even
Dots/solid lines = H1000 models (biodiversity indicator commuted at a 1000 m spa-
if for birds a clear distinction exists among small- and large-sized
tial scale); Triangles/dashed lines = H500 models (biodiversity indicator commuted
at a 500 m spatial scale).
species (Branton and Richardson, 2011). Indeed, our approach is
devoted to the predictability of the results and the exportability
of the methodologies to both other species and regions. The pre-
dictability was obtained through the modelling of species richness
over different spatial scales. Such a tool is robust as well as it stays
within the taxonomic limits of the co-occurring species (Fleishman
et al., 2005). The exportability, indeed, is assured by the fact that the
developed methodology is easy to implement. Although, it have to
be remarked that such a methodology cannot be exported without
modification, due to the impossibility of generating universal bio-
diversity indicator, valid over a wide range of regions and species
(Favreau et al., 2006).
Taking into account landscape heterogeneity is of fundamental
importance to assess the conservation value of surrogate species
(Ferrier, 2002). In this paper, we addressed this issue combin-
ing multiple species that exploit different habitats. Indeed, their
distribution showed slight overlaps, in which can fall important
conservation area of multiple habitat. Also, the quasi- comple-
mentarity in the distributions highlight the good choice of species
employed in the multi-species biodiversity indicator. At landscape
scale this procedure translated in a better definition of ␤ diversity,
and its relationships with local ␣ diversity. However, our results
depends on the selected spatial scales, but a regular pattern in
the outputs was apparent, that may subside to an empirical rela-
Fig. 6. Observed versus fitted values of richness, according to the model over the
tionship, mirroring similar study in adjacent region (Morelli et al.,
9 ha grid (i.e. GR9) as a function of the biodiversity indicator commuted at a 1000 m 2013). Species richness at a small (GR4) or large (GR25) spatial scale
spatial scale (i.e. H1000). Line shows the least square regression. is poorly predicted by our biodiversity indicator. Therefore, it can
be argued that species richness respond to landscape heterogene-
ity at a specific spatial scale, which also has a relationship with
and the H1000 is more closely related to the total richness (Fig. 5). habitat-specific umbrella species (Bellamy et al., 2013). The rela-
The comparison among raw data, diversity maps and model pre- tionships between landscape structure and species occurrence can
diction showed that the correspondence is quite high, but for some have clear patterns, due to the fragmentation, the degree of het-
area in which high values of species richness are reported for low erogeneity or the land-use policy (Atauri and De Lucio, 2001; Di
values of diversity. This results is highlighted in Fig. 6 where the Febbraro et al., 2015; Fahrig, 2003; Schindler et al., 2013). How-
fitted versus observed plot shows an increase in point dispersion ever, our use of umbrella species diversity went a step further
for higher values. toward the usual development of biodiversity indicators, since it
included both the landscape structure and the species’ response
4. Discussion to the landscape. Contrastingly with our expectations, our biodi-
versity indicator worked better over the whole area, than over
We built a biodiversity index commuting the environmental ecotones only. Usually, ecotones are sites of higher richness, due
suitability of four habitat-specific surrogate species and tested to the adjacency of several habitats (Fahrig, 2003; Ries et al., 2004).
its ability to explain the species richness of the whole commu- Our results can demonstrate that the high richness of ecotones can
634 F. Valerio et al. / Ecological Indicators 71 (2016) 627–635

be an artefact of the spatial scales used. The use of habitat specific Berger, J.O., 1997. Population constraints associated with the use of black rhinos as
indicator can greatly improve inference on biodiversity, since the an umbrella species for desert herbivores. Conserv. Biol. 11, 69–78.
Bivand, R., Piras, G., 2015. Comparing implementations of estimation methods for
selection of the indicator species is crucial to obtain informative spatial econometrics. J. Stat. Softw. 63.
output (Larsen et al., 2011). Brambilla, M., Casale, F., Bergero, V., Crovetto, G.M., Falco, R., Negri, I., Siccardi, P.,
Bogliani, G., 2009. GIS-models work well, but are not enough: habitat
preferences of Lanius collurio at multiple levels and conservation implications.
4.1. Implication for conservation Biol. Conserv. 142, 2033–2042, http://dx.doi.org/10.1016/j.biocon.2009.03.033.
Brambilla, M., Bassi, E., Bergero, V., Casale, F., Chemollo, M., Falco, R., Longoni, V.,
Our results can have significant consequences in conservation Saporetti, F., Viganò, E., Vitulano, S., 2013. Modelling distribution and potential
overlap between boreal owl Aegolius funereus and black woodpecker
planning. The use of surrogate species can greatly optimize data col- Dryocopus martius: implications for management and monitoring plans. Bird
lection, but the choice of the species is fundamental (Berger, 1997). Conserv. Int. 23, 502–511, http://dx.doi.org/10.1017/S0959270913000117.
The use of species distribution models for conservation planning Branton, M., Richardson, J.S., 2011. Assessing the value of the umbrella-species
concept for conservation planning with meta-analysis. Conserv. Biol. 25, 9–20,
is now widespread (Ferrier et al., 2007; Maiorano et al., 2015) and http://dx.doi.org/10.1111/j.1523-1739.2010.01606.x.
our results promotes their use for this purpose, by linking the dis- Breckheimer, I., Haddad, N.M., Morris, W.F., Trainor, A.M., Fields, W.R., Jobe, R.T.,
tribution of few species with the presence of many. The feasibility Hudgens, B.R., Moody, A., Walters, J.R., 2014. Defining and evaluating the
umbrella species concept for conserving and restoring landscape connectivity.
to assess the relationship between fragmentation and low species Conserv. Biol. 28, 1584–1593, http://dx.doi.org/10.1111/cobi.12362.
richness can increase, with obvious advantages for land managers Brichetti, P., Fracasso, G., 2004. Ornitologia Italiana Vol. 2 Tetraonidae −
and practitioners. Scolopacidae. Oasi Alberto Perdisa Editore, Bologna, Italy.
Brichetti, P., Fracasso, G., 2006. Ornitologia Italiana Vol. 3 Stercoraridae −
Caprimulgidae. Oasi Alberto Perdisa Editore, Bologna, Italy.
Acknowledgements Brichetti, P., Fracasso, G., 2011. Ornitologia Italiana Vol. 7 − Paridae-Corvidae. Oasi
Alberto Perdisa Editore, Bologna, Italy.
Brichetti, P., Fracasso, G., 2013. Ornitologia Italiana Vol. 8 Sturnidae − Fringillidae.
We are grateful to Corpo Forestale dello Stato who provided Oasi Alberto Perdisa Editore, Bologna, Italy.
bird community data for our study area (mainly recorded by A. Brown, J.L., 2014. SDMtoolbox: a python-based GIS toolkit for landscape genetic,
Mancinelli). This work was co-funded by the European Agricultural biogeographic and species distribution model analyses. Methods Ecol. Evol. 5,
694–700, http://dx.doi.org/10.1111/2041-210X.12200.
Fund for Rural Development (FEASR: fondo europeo agricolo per Cliff, A.D., Ord, J.K., 1981. Spatial Process: Models and Applications. Taylor &
lo sviluppo rurale) to Regione Abruzzo by measure 3.2.3, action 1 Francis.
of the Common Agricultural Policy 2007-2013 (Piano di Sviluppo Davic, R.D., Welsh, H.H.J., 2004. On the ecological roles of salamanders. Annu. Rev.
Ecol. Evol. Syst. 35, 405–434, http://dx.doi.org/10.1146/annurev.ecolsys.35.
Rurale), and through contract N.125-29/01/2013 bewteen Regione 112202.130116.
Abruzzo and National Research Council (CNR-IBAF). R. Bucci, F. Di Febbraro, M., Roscioni, F., Frate, L., Carranza, M.L., De Lisio, L., De Rosa, D.,
La Civita, A. Mancinelli assisted during field-work and sampling. Marchetti, M., Loy, A., 2015. Long-term effects of traditional and
conservation-oriented forest management on the distribution of vertebrates in
R. Toffoli and P. Culasso kindly contributed with some European
Mediterranean forests: a hierarchical hybrid modelling approach. Divers.
nightjar occurrences recorded in 2013. F. Buoninconti kindly con- Distrib. 21, 1141–1154, http://dx.doi.org/10.1111/ddi.12362.
tributed with manuscript preparation. Ducci, L., Agnelli, P., Di Febbraro, M., Frate, L., Russo, D., Loy, A., Carranza, M.L.,
Santini, G., Roscioni, F., 2015. Different bat guilds perceive their habitat in
different ways: a multiscale landscape approach for variable selection in
Appendix A. Supplementary data species distribution modelling. Landsc. Ecol., http://dx.doi.org/10.1007/
s10980-015-0237-x.
EEA, 2012. Corine Land Cover 2006 raster data.
Supplementary data associated with this article can be found, Elith, J., Leathwick, J.R., 2009. Species distribution models: ecological explanation
in the online version, at http://dx.doi.org/10.1016/j.ecolind.2016. and prediction across space and time. Annu. Rev. Ecol. Evol. Syst. 40, 677–697,
07.043. http://dx.doi.org/10.1146/annurev.ecolsys.110308.120159.
European Union, 1994. CORINE Land Cover Technical Guide.
EUR-OP/OOPEC/OPOCE.
References Fahrig, L., 2003. Effects of habitat fragmentation on biodiversity. Annu. Rev. Ecol.
Evol. Syst. 34, 487–515, http://dx.doi.org/10.1146/annurev.ecolsys.34.011802.
Andelman, S.J., Fagan, W.F., 2000. Umbrellas and flagships: efficient conservation 132419.
surrogates or expensive mistakes? Proc. Natl. Acad. Sci. U. S. A. 97, 5954–5959, Favreau, J.M., Drew, C.A., Hess, G.R., Rubino, M.J., Koch, F.H., Eschelbach, K. a., 2006.
http://dx.doi.org/10.1073/pnas.100126797. Recommendations for assessing the effectiveness of surrogate species
Anon, 2008. Relazione sullo status delle specie animali e vegetali di interesse approaches. Biodivers. Conserv. 15, 3949–3969, http://dx.doi.org/10.1007/
comunitario e degli habitat prioritari nell’alta Val di Sangro. In: Prodotto s10531-005-2631-1.
Identificabile Dell’Azione D10 Del Progetto Life Tutela Dei Siti Natura 2000 Fearer, T.M., Prisley, S.P., Stauffer, D.F., Keyser, P.D., 2007. A method for integrating
Gestiti Dal Corpo Forestale Dello Stato (LIFE04NAT/IT/000190). Corpo Forestale the Breeding Bird Survey and Forest Inventory and Analysis databases to
dello Stato- Commissione Europea, pp. Pp. 285. evaluate forest bird-habitat relationships at multiple spatial scales. For. Ecol.
Atauri, J. a., De Lucio, J.V., 2001. The role of landscape structure in species richness Manag. 243, 128–143, http://dx.doi.org/10.1016/j.foreco.2007.02.016.
distribution of birds, amphibians, reptiles and lepidopterans in Mediterranean Ferrier, S., Manion, G., Elith, J., Richardson, K., 2007. Using generalized dissimilarity
landscapes. Landsc. Ecol. 16, 147–159, http://dx.doi.org/10.1023/ modelling to analyse and predict patterns of beta diversity in regional
A:1011115921050. biodiversity assessment. Divers. Distrib. 13, 252–264, http://dx.doi.org/10.
Austin, M., 2007. Species distribution models and ecological theory: a critical 1111/j.1472-4642.2007.00341.x.
assessment and some possible new approaches. Ecol. Modell. 200, 1–19, Ferrier, S., 2002. Mapping spatial pattern in biodiversity for regional conservation
http://dx.doi.org/10.1016/j.ecolmodel.2006.07.005. planning: where to from here? Syst. Biol. 51, 331–363.
Bütler, R., Angelstam, P., Ekelund, P., Schlaepfer, R., 2004. Dead wood threshold Fleishman, E., Blair, R.B., Murphy, D.D., 2001. Empirical validation of a method for
values for the three-toed woodpecker presence in boreal and sub- Alpine umbrella species selection. Ecol. Appl. 11, 1489–1501, http://dx.doi.org/10.
forest. Biol. Conserv. 119, 305–318, http://dx.doi.org/10.1016/j.biocon.2003.11. 2307/3060934.
014. Fleishman, E., Thomson, J.R., Mac Nally, R., Murphy, D.D., Fay, J.P., 2005. Using
Banks-Leite, C., Ewers, R.M., Kapos, V., Martensen, A.C., Metzger, J.P., 2011. indicator species to predict species richness of multiple taxonomic groups.
Comparing species and measures of landscape structure as indicators of Conserv. Biol. 19, 1125–1137, http://dx.doi.org/10.1111/j.1523-1739.2005.
conservation importance. J. Appl. Ecol. 48, 706–714, http://dx.doi.org/10.1111/ 00168.x.
j.1365-2664.2011.01966.x. Fornasari, L., Londi, G., Buvoli, L., Tellini Florenzano, G., La Gioia, G., Pedrini, P.,
Barua, M., 2011. Mobilizing metaphors: the popular use of keystone, flagship and Brichetti, P., de Carli, E., 2010. Distribuzione geografica e ambientale degli
umbrella species concepts. Biodivers. Conserv. 20, 1427–1440, http://dx.doi. uccelli comuni nidificanti in Italia, 2000–2004 (dati del progetto MITO2000).
org/10.1007/s10531-011-0035-y. Avocetta 34, 5–224.
Beale, C.M., Lennon, J.J., Yearsley, J.M., Brewer, M.J., Elston, D.A., 2010. Regression Fox, J., 2008. Applied Regression Analyses and Gemeralized Linear Models, second.
analysis of spatial data. Ecol. Lett. 13, 246–264, http://dx.doi.org/10.1111/j. ed. Sage.
1461-0248.2009.01422.x. Griffis, V.W., Stedinger, J.R., 2007. The use of GLS regression in regional hydrologic
Bellamy, C., Scott, C., Altringham, J., 2013. Multiscale, presence-only habitat analyses. J. Hydrol. 344, 82–95, http://dx.doi.org/10.1016/j.jhydrol.2007.06.
suitability models: fine-resolution maps for eight bat species. J. Appl. Ecol. 50, 023.
892–901, http://dx.doi.org/10.1111/1365-2664.12117.
F. Valerio et al. / Ecological Indicators 71 (2016) 627–635 635

Guisan, A., Thuiller, W., 2005. Predicting species distribution: offering more than McGarigal, K., Cushman, S.A., Ene, E., 2012. Fragstats, http://dx.doi.org/10.1093/
simple habitat models. Ecol. Lett. 8, 993–1009, http://dx.doi.org/10.1111/j. ntr/nts298.
1461-0248.2005.00792.x. Moran, P.A.P., 1950. Notes on continuous stochastic phenomena. Biometrika 37,
Hanley, J., McNeil, B., 1982. The meaning and use of the area under a receiver 17–23, http://dx.doi.org/10.2307/2332142.
operating characteristic (ROC) curve. Radiology 143, 29–36. Morelli, F., Pruscini, F., Santolini, R., Perna, P., Benedetti, Y., Sisti, D., 2013.
Hoffmann, J., Wittchen, U., Stachow, U., Berger, G., 2016. Moving Window Landscape heterogeneity metrics as indicators of bird diversity: determining
Abundance −A method to characterise the abundance dynamics of farmland the optimal spatial scales in different landscapes. Ecol. Indic. 34, 372–379,
birds: the example of Skylark (Alauda arvensis). Ecol. Indic. 60, 317–328, http://dx.doi.org/10.1016/j.ecolind.2013.05.021.
http://dx.doi.org/10.1016/j.ecolind.2015.06.037. Noss, R.F., 1999. Assessing and monitoring forest biodiversity: a suggested
Hurme, E., Mönkkönen, M., Sippola, A.L., Ylinen, H., Pentinsaari, M., 2008. Role of framework and indicators. For. Ecol. Manag. 115, 135–146, http://dx.doi.org/
the Siberian flying squirrel as an umbrella species for biodiversity in northern 10.1016/S0378-1127(98)00394-6.
boreal forests. Ecol. Indic. 8, 246–255, http://dx.doi.org/10.1016/j.ecolind.2007. Phillips, S.J., Anderson, R.P., Schapire, R.E., 2006. Maximum entropy modeling of
02.001. species geographic distributions. Ecol. Modell. 190, 231–259, http://dx.doi.org/
Kissling, W.D., Dormann, C.F., Groeneveld, J., Hickler, T., Kühn, I., McInerny, G.J., 10.1016/j.ecolmodel.2005.03.026.
Montoya, J.M., Römermann, C., Schiffers, K., Schurr, F.M., Singer, A., Svenning, Pinheiro, J., Bates, D., DebRoy, S., Sarkar, D., EISPACK, R-core, 2015. Package nlme.
J.-C., Zimmermann, N.E., O’Hara, R.B., 2012. Towards novel approaches to R Core Team, 2015. R: A Language and Environment for Statistical Computing.
modelling biotic interactions in multispecies assemblages at large spatial Ries, L., Fletcher, R.J., Battin, J., Sisk, T.D., 2004. Ecological responses to habitat
extents. J. Biogeogr. 39, 2163–2178, http://dx.doi.org/10.1111/j.1365-2699. edges: mechanisms, models, and variability explained. Annu. Rev. Ecol. Evol.
2011.02663.x. Syst. 35, 491–522, http://dx.doi.org/10.1146/annurev.ecolsys.35.112202.
Lambert, J.E., 2011. Primate seed dispersers as umbrella species: a case study from 130148.
Kibale National Park Uganda, with implications for Afrotropical forest Roberge, J., Angelstam, P.E.R., 2004. Usefulness of the umbrella species concept.
conservation. Am. J. Primatol. 73, 9–24, http://dx.doi.org/10.1002/ajp.20879. Conserv. Biol. 18, 76–85, http://dx.doi.org/10.1111/j.1523-1739.2004.00450.x.
Larsen, J.L., Heldbjerg, H., Eskildsen, A., 2011. Improving national habitat specific Schindler, S., von Wehrden, H., Poirazidis, K., Wrbka, T., Kati, V., 2013. Multiscale
biodiversity indicators using relative habitat use for common birds. Ecol. Indic. performance of landscape metrics as indicators of species richness of plants,
11, 1459–1466, http://dx.doi.org/10.1016/j.ecolind.2011.03.023. insects and vertebrates. Ecol. Indic. 31, 41–48, http://dx.doi.org/10.1016/j.
Lawler, J.J., White, D., 2008. Assessing the mechanisms behind successful ecolind.2012.04.012.
surrogates for biodiversity in conservation planning. Anim. Conserv. 11, Simberloff, D., 1998. Flagships, umbrellas, and keystones: is single-species
270–280, http://dx.doi.org/10.1111/j.1469-1795.2008.00176.x. management passé in the landscape era? Biol. Conserv. 83, 247–257, http://dx.
Lee, S., Chao, A., 1994. Estimating population size via sample coverage for closed doi.org/10.1016/S0006-3207(97)00081- 5.
capture-recapture models. Biometrics 50, 833–836. Sinsch, U., Greenbaum, E., Kusamba, C., Lehr, E., 2011. Rapid assessment of
Lindenmayer, D.B., Cunningham, R.B., Donnelly, C.F., Lesslie, R., 2002. On the use of montane anuran communities in the Albertine Rift: hyperolius castaneus Ahl,
landscape surrogates as ecological indicators in fragmented forests. For. Ecol. 1931 as an umbrella species for conservation. Afr. Zool. 46, 320–333, http://dx.
Manag. 159, 203–216, http://dx.doi.org/10.1016/S0378-1127(01)00433- 9. doi.org/10.3377/004.046.0211.
Lindenmayer, D., Hobbs, R.J., Montague-Drake, R., Alexandra, J., Bennett, A., Stoch, F., Artheau, M., Brancelj, A., Galassi, D.M.P., Malard, F., 2009. Biodiversity
Burgman, M., Cale, P., Calhoun, A., Cramer, V., Cullen, P., Driscoll, D., Fahrig, L., indicators in European ground waters: towards a predictive model of
Fischer, J., Franklin, J., Haila, Y., Hunter, M., Gibbons, P., Lake, S., Luck, G., stygobiotic species richness. Freshw. Biol. 54, 745–755, http://dx.doi.org/10.
MacGregor, C., McIntyre, S., Mac Nally, R., Manning, A., Miller, J., Mooney, H., 1111/j.1365-2427.2008.02143.x.
Noss, R., Possingham, H., Saunders, D., Schmiegelow, F., Scott, M., Simberloff, Visconti, P., Bakkenes, M., Baisero, D., Brooks, T., Butchart, S.H.M., Joppa, L.,
D., Sisk, T., Tabor, G., Walker, B., Wiens, J., Woinarski, J., Zavaleta, E., 2008. A Alkemade, R., Di Marco, M., Santini, L., Hoffmann, M., Maiorano, L., Pressey, R.L.,
checklist for ecological management of landscapes for conservation. Ecol. Lett. Arponen, A., Boitani, L., Reside, A.E., van Vuuren, D.P., Rondinini, C., 2015.
11, 78–91, http://dx.doi.org/10.1111/j.1461-0248.2007.01114.x. Projecting global biodiversity indicators under future development scenarios.
Mac Nally, R., Fleishman, E., 2004. A successful predictive model of species richness Conserv. Lett. 00, http://dx.doi.org/10.1111/conl.12159, n/a–n/a.
based on indicator species. Conserv. Biol. 18, 646–654, http://dx.doi.org/10. Welsh, H.H., Droege, S., 2001. A case for using plethodontid salamanders for
1111/j.1523-1739.2004.00328 18 3.x. monitoring biodiversity and ecosystem integrity of north american forests.
Maiorano, L., Amori, G., Montemaggiori, a., Rondinini, C., Santini, L., Saura, S., Conserv. Biol. 15, 558–569, http://dx.doi.org/10.1046/j.1523- 1739.2001.
Boitani, L., 2015. On how much biodiversity is covered in Europe by national 015003558.x.
protected areas and by the Natura 2000 network: insights from terrestrial Welsh, H.H., Hodgson, G.R., 2013. Woodland salamanders as metrics of forest
vertebrates. Conserv. Biol. 29, 956–995, http://dx.doi.org/10.1111/cobi.12535. ecosystem recovery: a case study from California’s redwoods. Ecosphere 4,
Marchetti, M., Chiavetta, U., Santuopoli, G., 2009. La cartografia forestale su base 1–25, http://dx.doi.org/10.1890/Es12-00400.1.
tipologica della Regione Abruzzo: dai prodromi alla carta forestale dell’Italia Wiens, J.A., 1989. Spatial scaling in ecology. Funct. Ecol. 3, 385–397, http://dx.doi.
centrale. In: Collalti, D., D’alessandro, L., Marchetti, M., Sebastiani, A. (Eds.), La org/10.2307/2389612.
Carta Tipologico-Forestale Della Regione Abruzzo. Regione Abruzzo, L’Aquila, Wisz, M.S., Pottier, J., Kissling, W.D., Pellissier, L., Lenoir, J., Damgaard, C.F.,
Italy. Dormann, C.F., Forchhammer, M.C., Grytnes, J.A., Guisan, A., Heikkinen, R.K.,
McAlpine, C. a., Rhodes, J.R., Bowen, M.E., Lunney, D., Callaghan, J.G., Mitchell, D.L., Høye, T.T., Kühn, I., Luoto, M., Maiorano, L., Nilsson, M.C., Normand, S.,
Possingham, H.P., 2008. Can multiscale models of species’ distribution be Öckinger, E., Schmidt, N.M., Termansen, M., Timmermann, A., Wardle, D. a.,
generalized from region to region? A case study of the koala. J. Appl. Ecol. 45, Aastrup, P., Svenning, J.C., 2013. The role of biotic interactions in shaping
558–567, http://dx.doi.org/10.1111/j.1365-2664.2007.01431.x. distributions and realised assemblages of species: implications for species
McGarigal, K., McComb, W.C., 1995. Relationships between landscape structure distribution modelling. Biol. Rev. 88, 15–30, http://dx.doi.org/10.1111/j.1469-
and breeding birds in the oregon coast range. Ecol. Monogr. 65, 235–260. 185X.2012.00235.x.

You might also like