ArticlePDF Available

Abstract and Figures

The wild boar is an important crop raider as well as a reservoir and agent of spread of swine diseases. Due to increasing densities and expanding ranges worldwide, the related economic losses in livestock and agricultural sectors are significant and on the rise. Its management and control would strongly benefit from accurate and detailed spatial information on species distribution and abundance, which are often available only for small areas. Data are commonly available at aggregated administrative units with little or no information about the distribution of the species within the unit. In this paper, a four-step geostatistical downscaling approach is presented and used to disaggregate wild boar population density statistics from administrative units of different shape and size (polygons) to 5 km resolution raster maps by incorporating auxiliary fine scale environmental variables. 1) First a stratification method was used to define homogeneous bioclimatic regions for the analysis; 2) Under a geostatistical framework, the wild boar densities at administrative units, i.e. subnational areas, were decomposed into trend and residual components for each bioclimatic region. Quantitative relationships between wild boar data and environmental variables were estimated through multiple regression and used to derive trend components at 5 km spatial resolution. Next, the residual components (i.e., the differences between the trend components and the original wild boar data at administrative units) were downscaled at 5 km resolution using area-to-point kriging. The trend and residual components obtained at 5 km resolution were finally added to generate fine scale wild boar estimates for each bioclimatic region. 3) These maps were then mosaicked to produce a final output map of predicted wild boar densities across most of Eurasia. 4) Model accuracy was assessed at each different step using input as well as independent data. We discuss advantages and limits of the method and its potential application in animal health.
This content is subject to copyright.
RESEARCH ARTICLE
Wild boar mapping using population-density
statistics: From polygons to high resolution
raster maps
Claudia Pittiglio, Sergei Khomenko, Daniel Beltran-Alcrudo*
Animal Production and Health Division, Food and Agriculture Organization of the United Nations, Viale delle
Terme di Caracalla, Rome, Italy
*Daniel.BeltranAlcrudo@fao.org
Abstract
The wild boar is an important crop raider as well as a reservoir and agent of spread of swine
diseases. Due to increasing densities and expanding ranges worldwide, the related eco-
nomic losses in livestock and agricultural sectors are significant and on the rise. Its manage-
ment and control would strongly benefit from accurate and detailed spatial information on
species distribution and abundance, which are often available only for small areas. Data are
commonly available at aggregated administrative units with little or no information about the
distribution of the species within the unit. In this paper, a four-step geostatistical downscaling
approach is presented and used to disaggregate wild boar population density statistics from
administrative units of different shape and size (polygons) to 5 km resolution raster maps by
incorporating auxiliary fine scale environmental variables. 1) First a stratification method
was used to define homogeneous bioclimatic regions for the analysis; 2) Under a geostatisti-
cal framework, the wild boar densities at administrative units, i.e. subnational areas, were
decomposed into trend and residual components for each bioclimatic region. Quantitative
relationships between wild boar data and environmental variables were estimated through
multiple regression and used to derive trend components at 5 km spatial resolution. Next,
the residual components (i.e., the differences between the trend components and the origi-
nal wild boar data at administrative units) were downscaled at 5 km resolution using area-to-
point kriging. The trend and residual components obtained at 5 km resolution were finally
added to generate fine scale wild boar estimates for each bioclimatic region. 3) These maps
were then mosaicked to produce a final output map of predicted wild boar densities across
most of Eurasia. 4) Model accuracy was assessed at each different step using input as well
as independent data. We discuss advantages and limits of the method and its potential
application in animal health.
Introduction
The wild boar (Sus scrofa, L. 1758) is a generalist and opportunistic species found from western
Europe and the Mediterranean Basin to eastern Russian Federation and Japan, throughout
PLOS ONE | https://doi.org/10.1371/journal.pone.0193295 May 16, 2018 1 / 19
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Pittiglio C, Khomenko S, Beltran-Alcrudo
D (2018) Wild boar mapping using population-
density statistics: From polygons to high resolution
raster maps. PLoS ONE 13(5): e0193295. https://
doi.org/10.1371/journal.pone.0193295
Editor: Stephanie S. Romanach, U.S. Geological
Survey, UNITED STATES
Received: March 17, 2017
Accepted: February 8, 2018
Published: May 16, 2018
Copyright: ©2018 Food and Agriculture
Organization of the United Nations (FAO). This is an
open access article distributed under the terms of
the Creative Commons Attribution IGO License
(http://creativecommons.org/licenses/by/3.0/igo/
legalcode), which permits unrestricted use,
distribution, and reproduction in any medium,
provided the original work is properly cited.
Data Availability Statement: All relevant data are
within the paper and its Supporting Information
files.
Funding: Funded by European Union’s Seventh
Framework Programme (FP7/2007-2013) under
grant agreement no. 311931 (ASFORCE Targeted
research effort on African swine fever).
Competing interests: The authors have declared
that no competing interests exist.
southeast Asia. Wild boar occupies one of the largest geographic range among all terrestrial
mammals [1,2], including a variety of habitats, vegetation types and climate. Due to a combi-
nation of biological (i.e., species ecological plasticity, high reproduction rate), environmental
(i.e., climate change, mild winters), and anthropogenic factors (i.e., depopulation of rural
areas, reintroduction, lack of large natural predators, change in agricultural practices, reduced
hunting pressure, supplementary feeding and other husbandry practices), the abundance of
the wild boar has continuously increased over the last decades and its distribution expanded
across the whole geographic range [3,4,5,6]. This expansion poses a threat to the agriculture,
conservation and livestock health sectors, as the wild boar is an invasive and pest species caus-
ing substantial economic loss [7,8]. In recent years, there has been an increase in the likelihood
of disease spread, car accidents, damages to crops and to natural vegetation [4]. In particular,
the species may represent a reservoir or play another type of role in the transmission of many
livestock, wildlife and human diseases such African and classical swine fever, brucellosis,
tuberculosis, salmonellosis, Aujeszky’s disease and foot and mouth disease [9]. Domestic pigs
and wild boar share most diseases. A review of viral diseases of the European wild boar with a
direct effect on wild boar and an economic impact on domestic pig production systems identi-
fied 17 viral agents [9]. The involvement of the species in the recent expansion and persistence
of African swine fever in eastern Europe and the Caucasus, despite control efforts, has
attracted considerable international attention, stressing the difficulties of managing animal
diseases in wild populations [10].
The management and control of wild boar populations require accurate and detailed spatial
information on species distribution and abundance. Wild boar density varies from 0.01 to 43
animals / km
2
, following an east-west biogeographical gradient and exhibiting high inter-
annual fluctuation due to their high reproductive potential [2,11]. However, accurate and
detailed wild boar population statistics, such as census and hunting data, are not available for
large-scale studies [11,12,13,14]. Two recent studies have extrapolated and predicted wild boar
distribution and expansion at global level using wild boar data available across parts of the geo-
graphical range of the species [5,6]. This is probably due to the ecological traits of the species
(e.g. complex social structure, nocturnal activity pattern, preference for dense vegetation, and
high inter- and intra- annual variability in reproduction rate), that make direct observations
over large geographic areas more difficult and expensive than the collection of indirect signs,
such as faecal droppings, tracks, etc. [15]. Species distribution models based on indirect signs
such as presence/absence or presence-only data have been recently applied to predict the wild
boar occurrence from environmental and climatic covariates [6,12,13]. Although habitat suit-
ability models are useful management and conservation tools, their capacity to estimate species
abundance is highly controversial as the occupancy may not reflect species density and stability
[16,17,18,19,20]. In addition, abundance estimates are needed to study the epidemiology of
wildlife diseases, rather than suitability and distribution maps [15]. Wild boar population sta-
tistics are mainly available at aggregated spatial level, i.e. by census, hunting or administrative
units [7,8,13,15], with little or no information about the distribution of the species within these
units. The units can be very heterogeneous in shape, size as well as in terms of land use and
environmental characteristics. Aggregated population data does not account for this variation
among and within the units [21], and they are difficult to model [22]. Specific disaggregation
methods are required to predict population statistics from coarse to fine scale. Recently, multi-
ple linear regression has been used to predict wild boar density at global level [5]. Geostatistics
has been successfully used for mapping wild species distribution, including elusive marine
mammals, from observations acquired at point or transect level [23], as well as for spatial
downscaling and data disaggregation from coarse to fine resolution in land cover mapping
[24], human population mapping estimation [21], image sharpening [25], disease mapping
Wild boar mapping
PLOS ONE | https://doi.org/10.1371/journal.pone.0193295 May 16, 2018 2 / 19
[26], and precipitation estimate [27]. Recent studies demonstrated the utility of incorporating
the residuals of the regression models to increase interpolation accuracy. In particular, multi-
ple regression and area-to-point kriging of the regression residuals were applied to estimate
human population density from coarse census blocks to fine resolution land use maps [21], as
well as to downscale precipitation data from coarse to fine resolution rasters [27]. However, to
the best of our knowledge, the method has not yet been applied to disaggregate wild boar pop-
ulation data from polygons/units of different shape and size to high resolution raster maps.
We present a geostatistical method based on regression modelling and area-to-point resid-
ual kriging to disaggregate wild boar statistics, map and predict its abundance from spatially
heterogeneous administrative units (polygons) to high resolution raster maps at 5 km using
climatic and environmental covariates. We discuss advantages and limits of the method and its
potential application in animal health.
Materials and methods
Study area
The wild boar population density was estimated from western Europe to central-northern Asia
(Fig 1). This area encompasses Mediterranean, Temperate, and Boreal bioclimatic zones [28]
with a wide range of vegetation types, climate and elevation [29], which are known to influence
the distribution of the species regionally and locally [11].
Fig 1. Wild boar density by administrative units (original input data) and wild boar range.
https://doi.org/10.1371/journal.pone.0193295.g001
Wild boar mapping
PLOS ONE | https://doi.org/10.1371/journal.pone.0193295 May 16, 2018 3 / 19
Wild boar population data and geographic extent
The wild boar population density was estimated using a dataset available at FAO, including
hunting and census data available from different sources and at different level of administra-
tive units [30], for a total of 46 countries and 509 sub-national units (Fig 1 and S1 Table). Cen-
suses and population estimates were acquired for 74% of the countries and were available
particularly for eastern Europe, while hunting data were available for 83% of the countries,
and mainly for western Europe. Both types of data were acquired for 57% of the countries.
This subset was used to calculate the proportion of hunted animals and estimate a correction
factor to convert hunting data to densities for those countries presenting hunting data only
(n = 8)(S1 Table) [30]. The data were spatially and temporally heterogeneous. National totals
were acquired for 27 countries, while sub-national data (level 1 or higher, n = 480) were avail-
able for 19 countries. Except for Turkey, Kazakhstan, Turkmenistan, Uzbekistan, the countries
with national totals were similar in size to the sub-national units of surrounding countries,
thus making the dataset regionally homogeneous and regular in terms of the size and shape
of the spatial units. The time frame of the dataset ranged between 1993 and 2011, with 84% of
the data acquired after the year 2007. Data earlier than 2000 were obtained from Azerbaijan
and Spain (4%). Population, hunting data, correction factors and source data are reported in
S1 Table.
The International Union for Conservation of Nature (IUCN) wild boar occurrence map
[31] was updated by digitizing detailed maps for Spain, Italy, Iran, Greece and ex-USSR coun-
tries found through a literature review [32,33,34,35,36,37] and used to exclude unsuitable wild
boar areas inside each administrative unit using a simple GIS overlay operation in ArcGIS (Fig
1). The population density data, expressed as numbers per km
2
of suitable habitat, were nor-
malized to approximate to a normal distribution using the fourth root power (x1
4) transforma-
tion prior to perform the statistical analysis [38].
Predictors
Based on a literature review of environmental and climatic determinants of the wild boar dis-
tribution, 18 bioclimatic variables, 3 continuous vegetation cover (i.e. percentage of tree cover,
herbaceous vegetation and bare ground) and 2 topographic variables (elevation and slope)
were selected as the most significant covariates for predicting wild boar density across the
study area [2,5,7,11,12,39,40]. The bioclimatic variables were obtained from the WorldClim
online database at 30 arc-seconds (~1 km) spatial resolution for the period 1950–2000 [41].
The vegetation cover grids at 500 m resolution were downloaded from the Global Land Cover
Facility [42]. The elevation and slope were derived from the Shuttle Radar Topography Mis-
sion (SRTM), 30 arc-second pixel [43]. Average values of the predictors by suitable habitat by
wild boar administrative unit were normalized and standardized and used in the statistical
analysis. The predictors and properties are listed in Table 1.
All GIS layers were re-projected to the Polar Lambert Azimuthal Equal Area and re-sam-
pled (with the bilinear interpolation method) to 1 ×1 km
2
before performing spatial analysis.
Spatial analysis was performed in ArcGIS 10.0 (ESRI) and statistical analysis in R 3.1.0 [44].
Wild boar modelling approach
A flowchart illustrates the four-step modelling approach developed in this study (Fig 2). First,
we calculated the average value of each predictor by wild boar administrative unit. Given the
large geographical and high bioclimatic heterogeneity of the study area, a given predictor vari-
able may be associated quite differently with the species densities in different bioclimatic
Wild boar mapping
PLOS ONE | https://doi.org/10.1371/journal.pone.0193295 May 16, 2018 4 / 19
regions [45]. We therefore applied a stratification method [46] to define regions with similar
environmental conditions for the wild boar occurrence and group the wild boar units accord-
ingly. Second, for each bioclimatic region, we applied a two-step downscaling geostatistical
approach based on multiple regression and area-to-point residual kriging [27] to disaggregate
wild boar densities from coarse (polygons) to fine resolution raster maps (5 km). Within the
geostatistical framework, the wild boar density was decomposed into trend and residual com-
ponents. Quantitative relationships between wild boar data and environmental variables by
administrative unit were estimated through multiple regression analysis and the coefficients
were used to derive the trend components at 1 and 5 km spatial resolution. Then, the residual
components i.e., the differences between the trend components and the original wild boar data
by administrative units were downscaled at 5 km resolution using area-to-point kriging. The
trend and residual components at 5 km resolution were finally added to generate fine scale
wild boar estimates for each bioclimatic region. The trend components for each bioclimatic
region were also extrapolated to the whole study area. Third, we produced three different out-
put maps of predicted wild boar density for the whole study area: (a) a “geostatistical mosaicked
model” by mosaicking (merging) the predicted wild boar densities obtained for each biocli-
matic region; (b) an “averaged trend model” by averaging the trend components extrapolated
to the whole study area and (c) an averaged geostatistical model” by adding the map to the
kriged residual components. Fourth, we validated the three final output models using the
Table 1. Predictor variables included in the wild boar model-building process.
Variable Name Description Resolution Unit
Bioclimatic
BIO1 Annual Mean Temperature 1 km Degrees Celsius
BIO2 Mean Diurnal Range (Mean of monthly (max temp—min temp)) 1 km Degrees Celsius
BIO3 Isothermality (BIO2/BIO7) (100) 1 km Percent
BIO4 Temperature Seasonality (standard deviation 100) 1 km Percent
BIO6 Min Temperature of Coldest Month 1 km Degrees Celsius
BIO7 Temperature Annual Range (BIO5-BIO6) 1 km Degrees Celsius
BIO8 Mean Temperature of Wettest Quarter 1 km Degrees Celsius
BIO9 Mean Temperature of Driest Quarter 1 km Degrees Celsius
BIO10 Mean Temperature of Warmest Quarter 1 km Degrees Celsius
BIO11 Mean Temperature of Warmest Quarter 1 km Degrees Celsius
BIO12 Annual Precipitation 1 km Millimeter
BIO13 Precipitation of Wettest Month 1 km Millimeter
BIO14 Precipitation of Driest Month 1 km Millimeter
BIO15 Precipitation Seasonality (Coefficient of Variation) 1 km Percent
BIO16 Precipitation of Wettest Quarter 1 km Millimeter
BIO17 Precipitation of Driest Quarter 1 km Millimeter
BIO18 Precipitation of Warmest Quarter 1 km Millimeter
BIO19 Precipitation of Coldest Quarter 1 km Millimeter
Vegetation cover
Percentage Tree cover Modis Vegetation continuous fields 500 m Percent
Percentage herbaceous Modis Vegetation continuous fields 500 m Percent
Percentage bare ground Modis Vegetation continuous fields 500 m Percent
Topography
Elevation Elevation 1 km Meter
Slope Slope 1 km Degree
https://doi.org/10.1371/journal.pone.0193295.t001
Wild boar mapping
PLOS ONE | https://doi.org/10.1371/journal.pone.0193295 May 16, 2018 5 / 19
original wild boar input data as well as an independent dataset. Each step of the process is
detailed below. The spatial resolution of 5 km was chosen to match the spatial resolution of
livestock distribution maps produced by FAO, including the domestic pig density map [38] to
facilitate the analysis of diseases’ transmission among wild and domestic species.
Study area stratification: Definition of the wild boar bioclimatic regions. The normal-
ized and standardized average values of the predictors by administrative units were input in a
principal component analysis (PCA) to reduce the dimensionality of the environmental data-
set into a set of linearly uncorrelated and independent components. The first four principal
components (eigenvector >1) were input in the k-means cluster analysis to classify the study
area in homogeneous bioclimatic regions. The elbow method was used to define the number
of bioclimatic regions [47](Fig 2, box 1).
Two-step downscaling approach: Disaggregating and predicting wild boar densities
from polygons to 5 km resolution maps. For each bioclimatic region, we applied a two-step
downscaling geostatistical approach [27] based on multiple regression and area-to-point resid-
ual kriging to disaggregate wild boar population data from coarse scale administrative units
Fig 2. Flowchart illustrating the 4 main steps of the wild boar (WB) density modelling approach.
https://doi.org/10.1371/journal.pone.0193295.g002
Wild boar mapping
PLOS ONE | https://doi.org/10.1371/journal.pone.0193295 May 16, 2018 6 / 19
(polygons) to fine resolution raster maps (5 km) using auxiliary fine resolution bioclimatic and
environmental information. The method decomposed wild boar densities D(u) in: (a) a deter-
ministic, trend component m(u) (at coarse scale, i.e., polygon level), which indicates the wild
boar density (trend) influenced by the bioclimatic and geographic variables, and (b) a stochas-
tic residual component R(u) that accounts for the spatial correlation information of the input
wild boar residuals (see Eq 1).
DðuÞ ¼ mðuÞ þ RðuÞ ð1Þ
The trend component was estimated from the statistical relation between wild boar density
and auxiliary environmental variables by administrative units through multiple regression.
Under the assumption that attribute values at a coarse scale are linear averages of their constit-
uent fine scale point values, these relationships were applied to the environmental variables in
order to estimate the trend component at a fine scale. The stochastic residual component was
estimated at finer scale by interpolating the regression residuals using the area-to-point kriging
[48]. Then the trend component was added to the residual component to produce a fine resolu-
tion wild boar density map (hereafter ATP-based WB predicted density) for each region using
the Eq 1. This approach presents the mass preservation property, i.e. it can reproduce the origi-
nal wild boar density values when the downscaling results at a fine scale are re-aggregated to
the coarse scale [27] and it accounts for irregular geographic units (shape, size)[26] (Fig 2,
box 2).
Theory on regression-based interpolation and area-to-point residual kriging. Consider
a study area of kwild boar density data by irregular and coarse administrative units {z(v
k
),
k= 1,. . .,K}, where v
k
=v(u
k
) is the kth data with its centroid u
k
, and Mauxiliary fine scale
environmental variables fyk
iðunÞ;i¼1;...;M;n¼1;. . . ;Ngwithin each kth unit. Nis the
number of discretizing points within each unit and its determination depends on the prede-
fined finer scale value. If the environmental variables show linear relationships with the wild
boar density values, the latter at both coarse and fine scales can be expressed in terms of the
environmental variables via multiple linear regression as:
zðvkÞ ¼ aþX
M
i¼1
biyiðvkÞ þ RðvkÞ;ð2Þ
zkðunÞ ¼ aþX
M
i¼1
biyk
iðunÞ þ RkðunÞ;
where aand b
i
are regression coefficients for the intercept and slope of the ith variable, respec-
tively. z
k
(u
n
) is the downscaled wild boar value at a target finer scale within the kth unit. R(v
k
)
and R(u
n
) are the residual components at coarse and fine scales, respectively, which cannot be
accounted for by environmental variables. If the original wild boar data, environmental vari-
ables, and the residual components at a coarse scale can be expressed by the average values of
Nfine scale data within each unit, respectively, Eq (2) can be reformulated as (3)
zðvkÞ ¼ aþX
M
i¼1
bi
1
NX
N
n¼1
yk
iðukÞ
" #þ1
NX
N
n¼1
Rkun
ð Þ ð3Þ
zðvkÞ ¼ 1
NX
N
n¼1
aþX
M
i¼1
biyk
iðunÞ þ RkðunÞ
" #
Wild boar mapping
PLOS ONE | https://doi.org/10.1371/journal.pone.0193295 May 16, 2018 7 / 19
zðvkÞ ¼ 1
NX
N
n¼1
zkðunÞ
The trend component at a finer scale was estimated by applying the regression relationships
obtained at course scale to the environmental variables. The residual component at finer scale
was predicted using area-to-point simple kriging [27] of the residual components available at a
coarse scale. Area-to-point simple kriging predicts the residual component values at a fine
scale by a linear combination of neighboring attribute values at a coarse scale [26,27]. Briefly,
variogram deconvolution was applied to the regression residuals for each region to estimate
the unknown point-support variogram of the residuals.
A detailed explanation of the area-to-point kriging can be found elsewhere [21,26,27].
Area-to point residual kriging was implemented using SpaceStat 2.0 (BioMedware).
The normalized and standardized predictors were tested for multicollinearity prior to per-
forming the multiple regression because multicollinearity may violate statistical assumptions
and may alter model predictions [49]. Different methods exist to evaluate multicollinearity in
regression analysis, including correlation matrix, PCA, Tolerance and Variance Inflation fac-
tor [50]. In this study we chose the VIF. Predictors with VIF larger than 2.5 were excluded
from the multiple regression [51,52]. The variables were tested one by one. A (forwards) step-
wise variable selection procedure was applied to select a parsimonious set of predictors for the
multiple regression modelling. The method is explained in detail elsewhere [50,51,52].
Mapping the wild boar predicted densities for the whole study area (5 km). The first
output (“geostatistical mosaicked model”, hereafter mosaicked model) was obtained by
mosaicking the ATP-based wild boar density maps predicted for each bioclimatic region in a
single, seamless map for the study area (see Fig 2, box 3). In order to minimize the abrupt
changes along the boundaries of adjacent regions, a blending image processing technique was
applied to the input maps. Specifically, a buffer of 100 km was built around the boundary of
each region and used as a mask to average inside the wild boar predicted densities obtained
from models of adjacent regions.
Two additional modelling outputs were generated and compared with the mosaicked
model: 1) the averaged trend model (output 2, Fig 2, box 3), which was derived by averaging
the trend components of each bioclimatic region extrapolated at 1 km resolution to the whole
study area and resampled at 5 km resolution; 2) the averaged geostatistical model (output 3,
Fig 2, box 3), obtained by adding the kriged residual component to the averaged trend model.
The residual component was generated from an ordinary kriging of all regression residuals.
Model evaluation and accuracy assessment. The predictive performance of the models
was assessed using the wild boar densities of the original input data as well as an independent
dataset (Fig 2 box 4). Specifically, the overall accuracy of the averaged trend model, averaged
geostatistical model and the mosaicked model were measured by calculating the averaged pre-
dicted densities by administrative unit and then relating these averaged predicted density to
the averaged observed density of the input data. Given the mass preservation of the ATP-based
method [26], a 1:1 relationship was expected. Model performance was also evaluated for each
region. In addition, the accuracy across the whole wild boar study area was independently
assessed using the density data reported by Melis et al. [11] for 54 locations in Eurasia. These
locations were georeferenced and overlaid with the three output models. The predicted wild
boar densities extracted at those locations were therefore related to the densities observed by
Melis et al. [11].
Wild boar mapping
PLOS ONE | https://doi.org/10.1371/journal.pone.0193295 May 16, 2018 8 / 19
The Pearson correlation coefficient (r
p
), coefficient of determination (R
2
) and root mean
square error (RMSE) calculated between predicted and observed densities were used to assess
model accuracy [53].
Results
The k-means cluster analysis grouped the wild boar units into four main homogeneous biocli-
matic regions for the wild boar occurrence: 1) Asian, 2) eastern; 3) western, and 4) southern
(Fig 3). The units within each region were similar in size and shape, except for southern,
which included the largest units (Iran, Turkey, Kazakhstan, Uzbekistan and Turkmenistan).
Elevation and slope were highly correlated with most of the bioclimatic variables (Pearson
correlation coefficient, p<0.05) and were therefore excluded from the geostatistical modelling.
The results of multiple regression models performed for each bioclimatic region are reported
in Table 2. The variance explained by the trend components ranged between 49 and 53%, indi-
cating that regression alone is not sufficient to predict the wild boar density in each region.
The predictors were reduced to 2–6 significant variables and differed among regions. In the
Fig 3. Bioclimatic regions for the wild boar as defined by PCA and cluster analysis. The cluster plot of the first and second components is shown in the inset. The
symbols represent the administrative units grouped in the 4 clusters/regions: Asian (red), eastern (pink), western (green) and southern (blue).
https://doi.org/10.1371/journal.pone.0193295.g003
Wild boar mapping
PLOS ONE | https://doi.org/10.1371/journal.pone.0193295 May 16, 2018 9 / 19
Asian region (pink, Fig 3) wild boar density was positively related to annual mean temperature
(BIO1) and negatively related to both precipitation of coldest quarter (BIO19) and continuous
herbaceous cover. Instead, in the eastern region (blue) wild boar density was negatively related
to both temperature annual range (BIO7) and precipitation of wettest month (BIO13), but
positively associated with mean temperature of wettest quarter (BIO8) and continuous tree
cover. In the southern region (yellow), a significant positive association was found with mean
diurnal range (BIO2), minimum temperature of coldest month (BIO6), and continuous tree
cover. In the western region (orange) wild boar density was positively related to annual mean
temperature (BIO1) and continuous tree cover, but negatively related to mean diurnal range
(BIO2), mean temperature of wettest quarter (BIO8), precipitation seasonality (BIO15) and
precipitation of coldest quarter (BIO19). Given the lower variance explained by the regression
model for the Asian region, as well as the large residuals associated with the largest units in the
southern region (i.e. Iran, Kazakhstan, Uzbekistan and Turkmenistan), the Asian model was
excluded from the geostatistical modelling and the boundary of the southern region was
delimited along by the Ural mountains and the Caspian sea.
The trend components generated by applying the three regression relationships to the envi-
ronmental variables at 1 km resolution and extrapolated for the whole study area at 5 km reso-
lution are displayed in Fig 4A–4C. The results of the variogram deconvolution for each region
are shown in S2 Table, while an example of the area-to-point residual kriging is given for the
eastern region in S1 Fig. The predicted wild boar population-densities by region obtained by
adding trend and residual components are shown in S2 Fig. When the downscaled results at 5
km were re-aggregated at the original administrative unit level, the preservation of the coher-
ence (mass) property was highest for the eastern region (Pearson correlation coefficient r
p
=
0.7), western (r
p
= 0.6) and lowest for the southern region (r
p
= 0.4). The wild boar densities
predicted in the Iberian Peninsula (southern region) were less accurate than those obtained
using the regression coefficient of the western region. Consequently, the western region was
modified to include the Iberian Peninsula and the mosaic mask for generating the final
Table 2. Results of the multiple regression models by bioclimatic region: Standardized coefficients and standard errors (in brackets), adjusted R
2
, sample size (N),
F value and degrees of freedom (dfs), residual standard error (RSE) and p-values.
Predictors Asian Eastern Southern Western
(Intercept) 1.21 7.19 -0.87 2.13
Annual Mean Temperature (BIO1) 0.04 0.08
Mean Diurnal Range (BIO2) 1.45 -1.11
Min Temperature of Coldest Month (BIO6) 0.03
Temperature Annual Range (BIO7) -3.60
Mean Temperature of Wettest Quarter (BIO8) 0.02 -0.01
Precipitation of Wettest Month (BIO13) -0.92
Precipitation Seasonality (BIO15) -0.61
Precipitation of Coldest Quarter (BIO19) -0.05 -0.07
Continuous herbaceous cover -0.04
Continuous tree cover 0.25 0.28 0.80
R
2
adjusted 0.51 0.53 0.50 0.49
N26 176 94 171
F9.67 50.2 31.5 28.5
dfs (3)(22) (4)(171) (3)(90) (6)(164)
RSE 0.07 0.17 0.23 0.22
p<0.05 <0.05 <0.05 <0.05
https://doi.org/10.1371/journal.pone.0193295.t002
Wild boar mapping
PLOS ONE | https://doi.org/10.1371/journal.pone.0193295 May 16, 2018 10 / 19
prediction map was adjusted accordingly (Fig 4C). The final wild boar population-density
models, i.e. averaged trend, average geostatistical and mosaicked models, are shown in Fig
5A–5C. Since population density cannot be less than 0, a small proportion of negative esti-
mates (1%), mainly located in mountainous areas (i.e., Pyrenees, Alps, Caucasus) was adjusted
to 0.
The model performance assessed using the wild boar input data increased substantially
from the average trend model (t = 17.38, df = 456, p<0.05, Pearson correlation coefficient r
p
=
0.63), to averaged geostatistical model (t = 24.5, df = 456, p<0.05, r
p
= 0.75), and it was highest
for the mosaicked model (t = 32.1, df = 456, p<0.05, r
p
= 0.83), though a 1:1 relation was not
found. We found a significant positive correlation between the mosaicked model wild boar
predicted densities and the densities reported by Melis et al. [11] (t = 9.16, df = 43, p<0.001;
r
p
= 0.81).
Table 3 shows the r, MAE and RMSE of the validation analysis for the three models with
input and independent data.
Fig 4. Trend components derived from the regression relationship obtained for (A) the eastern, (B) western and (C) southern regions and extrapolated to the whole
study area respectively. (D) Redefined bioclimatic regions and blending zones. The Asian region is not shown as it was excluded from the geostatistical analysis.
https://doi.org/10.1371/journal.pone.0193295.g004
Wild boar mapping
PLOS ONE | https://doi.org/10.1371/journal.pone.0193295 May 16, 2018 11 / 19
Discussion
Wild boar population statistics are mainly available at aggregated level, or as scattered observa-
tions, particularly for large geographical areas [5,7]. However, the management and control of
wild boar populations require accurate and detailed spatial information on species distribution
Fig 5. Model outputs: average trend (A), average geostatistical model (B) and mosaicked model (C).
https://doi.org/10.1371/journal.pone.0193295.g005
Wild boar mapping
PLOS ONE | https://doi.org/10.1371/journal.pone.0193295 May 16, 2018 12 / 19
and abundance. Two recent studies have extrapolated and predicted wild boar distribution
and expansion at global level using multiple linear regression [5] and Bayesian approaches [6]
based on wild boar data available across parts of geographical range of the species. In our
study, a multi-step geostatistical approach was proposed, tested and validated to disaggregate
wild boar population density statistics from polygons of irregular size and shape to fine spatial
resolution maps. The approach represents a valuable downscaling tool that could be applied to
any population statistics, including other wildlife species, livestock, human, as well as disease
data. Strengths and limitations of the approach are discussed below.
The first strength of the approach is the stratification step. Differently from other studies
[5,6,7,8], our novel method was able to account for the large geographical and high bioclimatic
heterogeneity of the study area and to define four different biogeographic/bioclimatic regions
—Asian, southern, eastern, and western—reflecting different spatiotemporal patterns of food
resources, shelter/cover, available to the species. By analyzing each region separately, a parsi-
monious set of specific predictors and limiting factors for wild boar distribution and density
was identified for each region: winter harshness in the Asian and eastern regions (as indicated
by the negative relation with Precipitation of Wettest Month (BIO13) and Precipitation of
Coldest Quarter (BIO19) respectively); forest productivity and shelter in western, eastern and
southern regions (as indicated by the positive relation with tree vegetation cover); extreme
temperature and precipitation variability (e.g., droughts, floods) in western region (as indi-
cated by the negative relation with Precipitation seasonality (BIO15) and Annual Mean Diur-
nal Range (BIO2)); low habitat heterogeneity and lack of shelter in the Asian region (as
indicated by the negative relation with the herbaceous vegetation cover). These results con-
firmed that winter severity, temperature and precipitation anomalies, as well as vegetation
structure, are main macroecological determinants of the wild boar distribution and abundance
in northern and temperate latitudes, respectively [5,7,11,12,14], as they affect population
dynamics, particularly the survival of newborn piglets. Differently from previous studies
[12,39], slope and elevation were not significant predictors, suggesting that terrain is an impor-
tant species determinant at local scale but not at regional scale.
The main advantage of the approach is the geostatistical framework (step 2), which decom-
posed the wild boar density in trend and residual components. Under this framework, the
quantitative relationship found between wild boar densities and environmental variables at
coarse scale (administrative units) through the regression analysis was used to derive the trend
component at finer scale (1km and 5 km). The low accuracy of the trend models was in agree-
ment with the results obtained by other studies [21,27,54], indicating that regression alone
cannot account for the within-units environmental heterogeneity and its influence on species
distribution and abundance.
The accuracy of the regression-based trend models varied among the regions: the western
and the eastern regions showed the lowest (R
2
= 0.49) and highest (R
2
= 0.53) accuracy respec-
tively. The wild boar density in the western region was mainly estimated from hunting data,
while the southern region was mainly characterized by larger and irregular units, suggesting
that low quality of the input data (hunting vs. census, [4]), differences in years covered by the
data series [11], as well as units variability in size and shape can impact model performance.
Table 3. Validation results.
Model name r
p
adjusted R
2
RMSE p
Output 1: Mosaicked model 0.83 0.69 0.17 <0.001
Output 2: Averaged Trend model 0.63 0.40 0.13 <0.001
Output 3: Averaged Geostatistical model 0.75 0.57 0.21 <0.001
https://doi.org/10.1371/journal.pone.0193295.t003
Wild boar mapping
PLOS ONE | https://doi.org/10.1371/journal.pone.0193295 May 16, 2018 13 / 19
As expected, the predictive performance of the models increased with the incorporation of
the regression residuals through the area-to-point kriging. This result was in line with the find-
ing of other studies [21,27] due to the ability of the Area-to-point kriging to account for the
stochastic component and the spatial autocorrelation of the aggregated input data. The accu-
racy was highest for the mosaicked model (r= 0.83), and lowest for the averaged trend model
(r= 0.63), meaning that the statistical relations between wild boar abundance and bioclimatic
variables cannot be extrapolated to areas outside the training regions [55]. This is clearly
shown by the trend maps (Fig 4A–4C), which display different distribution patterns of wild
boar among the bioclimatic regions and low predictability outside the training regions. This
finding suggests that model transferability remains an issue in species distribution models
[56]. This result further corroborates the importance of defining bioclimatic regions based on
statistical stratification methods. However, high environmental variability in certain areas, as
well as anthropogenic factors, may also impact the accuracy of the models [5]. In our study,
the model performance for Spain based on the regression coefficients of the southern region
was low, but improved when we applied the coefficient of the adjacent western region. The
two model outputs were discussed with wild boar experts in Spain, who provided a visual vali-
dation of the two maps, confirming the higher accuracy of the western model. This result was
explained by the high environmental and topographic heterogeneity of the country [12] as well
as the wide use of supplementary feeding in some areas [45], which could not be captured by
the predictors of the southern region. High environmental heterogeneity was a positive predic-
tor for the wild boar distribution and range expansion in recent studies [5,6,8].
Although we found a strong relation between predicted and observed wild boar densities
(r= 0.83), the expected 1:1 relation was not found. As highlighted by Liu et al. [21], the mass-
preserving property of the area-to-point kriging is lost when negative estimates are reset to 0.
The model accuracy was very high even when assessed against independent wild boar data
[11]. The biogeographical West-East gradient in wild boar density observed by Melis et al. [11]
was also found in our model. This longitudinal decline in wild boar densities is mainly related
to milder climatic conditions as well as higher vegetation productivity and biodiversity (i.e.,
tree species with edible seeds for wild boar) in western and southern Europe as compared to
the eastern Eurasian range [11].
In synthesis, the approach optimized the use of available coarse resolution abundance data
for producing high resolution density maps. Although the geostatistical method accounted for
units of different size and shape, the quality of the input data and high variability in size and
shape impacted model performance. In this study the approach was tested using main climatic
and environmental predictors for the wild boar. Additional predictors such as metrics of land-
scape fragmentation [32] could increase model performance [8]. The application of Poisson
regression could be tested to improve the regression-based trend interpolation [57]. In addi-
tion to the mean value, other measures of the variability of the predictors by administrative
unit (e.g., range, coefficient of variation) could be explored to improve model prediction, par-
ticularly for regions with high environmental heterogeneity. Given the availability of new
areas, more recent or more precise wild boar and geographic data, the model can be easily
updated, as well as applied to other geographic areas where the species occurrence is expand-
ing (e.g., China, United States of America).
Wild boar distribution and density maps can become useful tools to assess the growing
threat that these populations pose to agriculture (i.e. crop damage), conservation, road traffic,
and health (livestock, wildlife, and even human). Such maps will allow to first assess the situa-
tion, and then implement management actions accordingly, in an attempt to solve, or at least
minimize, the negative effects. The next section expands on how such assessment and manage-
ment applies specifically to animal health, which is a topic of particular urgency and concern
Wild boar mapping
PLOS ONE | https://doi.org/10.1371/journal.pone.0193295 May 16, 2018 14 / 19
given the current progressive spread of ASF in Europe, which often involves wild boar. An
important objective of the FAO is to disseminate the data. The predicted wild boar densities
based on the mosaicked model are freely available and can be downloaded from S1 Geodataset
of the Supporting Information, provided the original authors and source are credited.
Applications to animal health management
Wildlife and livestock connect through different paths, which allows for disease transmission
in both directions. This has clear implications in health management, as the objective is to
keep livestock and wildlife healthy, by preventing the introduction of diseases from one popu-
lation to another. Sometimes, there is also a public health component/concern. Wild boar dis-
tribution and density maps can be an extremely useful tool for veterinary services, wildlife
managers and epidemiologists to prevent and control animal diseases. When incorporated
into risk analysis or disease risk modelling, such maps allow the identification and assessment
of the specific pathways and risks posed by wild boar in the introduction, spread and mainte-
nance of animal diseases in a certain region, and their potential subsequent spread to livestock
(usually domestic pigs) and vice versa. In this line, spatial models have been recently developed
to simulate the introduction and spread of classic swine fever and foot-and-mouth disease in
wild pigs in Australia, allowing for the testing of the effectiveness of different control measures
and surveillance strategies [58,59,60]. Similar models based on accurate maps of wild boar
densities will allow identifying the potential corridors of introduction, the areas of highest den-
sities, or where wild boar-domestic pig interactions are more likely (e.g. where backyard or
free-ranging pigs, and other low biosecurity production systems exist; unsecured dumping
sites which may contain infected pig products; or through hunters). As a result, early detection
strategies in high-risk areas can be planned and implemented, e.g. strengthening passive sur-
veillance, testing hunted animals or road kills, or through targeted surveillance (non-invasive
sampling or capture and release). Prevention measures can also be applied in high risk areas.
Those applied to the domestic pig sector will be most effective in preventing infection in both
directions, e.g. double-fencing, permanent enclosure of animals, proper disposal of kitchen
and slaughtering waste, and other biosecurity improvements. Although controversial and still
subjected to extensive debate, there are intervention measures to prevent, or at least minimize,
the entry of infected wild boar into certain areas (repellents, fences, hunting pressure, etc.).
There are also management options if a disease becomes established in wild boar: vaccination,
carcass removal, ban of supplementary feeding, or hunting strategies.
The precise set of measures will depend to a certain extent on the mechanism of transmis-
sion of the disease, but in all the cases, knowing the numbers and distribution of wild boar will
greatly help to plan the strategies and estimate the efforts/resources needed.
Supporting information
S1 Fig. ATP residual kriging for the eastern region.
(TIF)
S2 Fig. Predicted wild boar density by region based on ATP-regression models: (A) east-
ern, (B) southern, (C) western, and (D) original input density data by administrative unit. Leg-
end: Low (green)–High (red) density values.
(TIF)
S1 Table. Wild boar source data.
(DOCX)
Wild boar mapping
PLOS ONE | https://doi.org/10.1371/journal.pone.0193295 May 16, 2018 15 / 19
S2 Table. Results of the variogram deconvolution for the three bioclimatic regions and rel-
ative charts.
(DOCX)
S1 Geodataset. Predicted wild boar densities based on the mosaicked model.
(ZIP)
Acknowledgments
We acknowledge the European Union’s Seventh Framework Programme (FP7/2007-2013)
under grant agreement no. 311931 (ASFORCE–Targeted research effort on African swine
fever) for funding. We would also like to acknowledge Vittorio Guberti (ISPRA, Italy), Sophie
Rossi (ONCFS, France) and Tsviatko Alexandrov (BFSA, Bulgaria), who facilitated acquisition
of the most up-to-date sub-national statistical data on some wild boar populations. Timothy
Robinson (FAO) conducted the first experimental modeling trails with the earlier versions of
wild boar population dataset using disaggregation methodology applied to livestock, which
helped to identify ways for developing original downscaling approach presented in this paper.
The views expressed in this information product are those of the author(s) and do not neces-
sarily reflect the views or policies of FAO.
Author Contributions
Conceptualization: Claudia Pittiglio, Sergei Khomenko, Daniel Beltran-Alcrudo.
Data curation: Claudia Pittiglio, Sergei Khomenko.
Formal analysis: Claudia Pittiglio, Sergei Khomenko.
Funding acquisition: Daniel Beltran-Alcrudo.
Investigation: Claudia Pittiglio.
Methodology: Claudia Pittiglio, Sergei Khomenko.
Project administration: Daniel Beltran-Alcrudo.
Resources: Sergei Khomenko.
Software: Claudia Pittiglio.
Supervision: Daniel Beltran-Alcrudo.
Validation: Claudia Pittiglio, Sergei Khomenko, Daniel Beltran-Alcrudo.
Writing original draft: Claudia Pittiglio, Daniel Beltran-Alcrudo.
Writing review & editing: Claudia Pittiglio, Daniel Beltran-Alcrudo.
References
1. Ballari SA, Barrios-Garcı
´a MN (2014) A review of wild boar Sus scrofa diet and factors affecting food
selection in native and introduced ranges. Mammal Review 44: 124–134.
2. Massei G, Genov PV (2004) The environmental impact of wild boar. Galemys 16.
3. Apollonio M, Anderson R, Putman R, editors (2010) European Ungulates and their Management in the
21st Century. Cambridge: Cambridge University Press.
4. Massei G, Kindberg J, Licoppe A, GačićD, S
ˇprem N, Kamler J, et al. (2015) Wild boar populations up,
numbers of hunters down? A review of trends and implications for Europe. Pest Management Science
71: 492–500. https://doi.org/10.1002/ps.3965 PMID: 25512181
Wild boar mapping
PLOS ONE | https://doi.org/10.1371/journal.pone.0193295 May 16, 2018 16 / 19
5. Lewis JS, Farnsworth ML, Burdett CL, Theobald DM, Gray M, Miller RS (2016) Biotic and abiotic factors
predicting the global distribution and population density of an invasive large mammal. Sci Rep. 2017;
7:44152. https://doi.org/10.1038/srep44152 PMID: 28276519
6. Sales LP, Ribeiro BR, Hayward MW, Paglia A, Passamani M, Loyola R (2017). Niche conservatism and
the invasive potential of the wild boar. Journal of Animal Ecology In Press:DOI: https://doi.org/10.1111/
1365-2656.12721 PMID: 28656732
7. McClure ML, Burdett CL, Farnsworth ML, Lutman MW, Theobald DM, et al. (2015). Modeling and map-
ping the probability of occurrence of invasive wild pigs across the contiguous United States. PLoS ONE
10(8):e0133771. https://doi.org/10.1371/journal.pone.0133771 PMID: 26267266
8. Snow NP, Jarzyna MA, VerCauteren KC (2017) Interpreting and Predicting the Spread of Invasive Wild
Pigs. Journal of Applied Ecology https://doi.org/10.1111/1365-2664.12866
9. Ruiz-Fons F, Segale
´s J, Gorta
´zar C (2008) A review of viral diseases of the European wild boar: Effects
of population dynamics and reservoir role. The Veterinary Journal 176: 158–169. https://doi.org/10.
1016/j.tvjl.2007.02.017 PMID: 17420149
10. EFSA (European Food Safety Authority), Depner K, Gortazar C, Guberti V, Masiulis M, More S et al.
(2017). Scientific report on the epidemiological analyses of African swine fever in the Baltic States and
Poland. EFSA Journal 2017; 15(11):5068, 59 pp. https://doi.org/10.2903/j.efsa.2017.5068
11. Melis C, Szafrańska PA, Jędrzejewska B, BartońK (2006) Biogeographical variation in the population
density of wild boar (Sus scrofa) in western Eurasia. Journal of Biogeography 33: 803–811.
12. Bosch J, Mardones F, Pe
´rez A, de la Torre A, Muñoz MJ (2014) A maximum entropy model for predict-
ing wild boar distribution in Spain. Spanish Journal of Agricultural Research 12: 16.
13. Bosch J, Peris S, Fonseca C, Martinez M, de la Torre A, Iglesias I, et al. (2012) Distribution, abundance
and density of the wild boar on the Iberian Peninsula, based on the CORINE program and hunting statis-
tics. Folia Zooligica 61: 138–151.
14. Acevedo P, Quiro
´s-Ferna
´ndez F, Casal J, Vicente J (2014) Spatial distribution of wild boar population
abundance: Basic information for spatial epidemiology and wildlife management. Ecological Indicators
36: 594–600.
15. Acevedo P, Vicente J, Ho
¨fle U, Cassinello J, Ruiz-Fons F, Gortazar C (2007) Estimation of European
wild boar relative abundance and aggregation: a novel method in epidemiological risk assessment. Epi-
demiology and Infection 135: 519–527. https://doi.org/10.1017/S0950268806007059 PMID: 16893488
16. He F, Gaston KJ (2000) Estimating species abundance from occurrence. The American Naturalist 156
553–559. https://doi.org/10.1086/303403 PMID: 29587511
17. Jime
´nez-Valverde A, Diniz F, De Azevedo EB, Borges PAV (2009) Species distribution models do not
account for abundance: The Case of Arthropods on Terceira Island. Annales Zoologici Fennici 46:
451–464.
18. Boyce MS, Johnson CJ, Merrill EH, Nielsen SE, Solberg EJ, van Moorter B (2015) Can habitat selection
predict abundance? Journal of Animal Ecology: n/a-n/a.
19. Yin D, He F (2014) A simple method for estimating species abundance from occurrence maps. Methods
in Ecology and Evolution 5: 336–343.
20. Estrada A, Arroyo B (2012) Occurrence vs abundance models: Differences between species with vary-
ing aggregation patterns. Biological Conservation 152 37–45.
21. Liu XH, Kyriakidis PC, Goodchild MF (2008) Population-density estimation using regression and area-
to-point residual kriging. International Journal of Geographical Information Science 22: 431–447.
22. Tranmer M, Steel D (1998) Using census data to investigate the causes of the ecological fallacy. Envi-
ronment & Planning A 30: 817–831.
23. Alessi J, Fiori C (2014) From science to policy–a geostatistical approach to identifying potential areas
for cetacean conservation: a case study of bottlenose dolphins in the Pelagos sanctuary (Mediterranean
Sea). Journal of Coastal Conservation 18, 1–10
24. Boucher A, Kyriakidis PC (2006) Super-resolution land cover mapping with indicator geostatistics.
Remote Sensing of Environment 104: 264–282.
25. Pardo-Igu
´zquiza E, Chica-Olmo M, Atkinson PM (2006) Downscaling cokriging for image sharpening.
Remote Sensing of Environment 102: 86–98.
26. Goovaerts P (2008) Kriging and Semivariogram Deconvolution in the Presence of Irregular Geographi-
cal Units. Math Geosci 40: 101–128.
27. Park N-W (2013) Spatial Downscaling of TRMM Precipitation Using Geostatistics and Fine Scale Envi-
ronmental Variables. Advances in Meteorology 2013: 1–9.
28. Peel MC, Finlayson BL, McMahon TA (2007) Updated world map of the Ko
¨ppen-Geiger climate classifi-
cation. Hydrol Earth Syst Sci 11: 1633–1644.
Wild boar mapping
PLOS ONE | https://doi.org/10.1371/journal.pone.0193295 May 16, 2018 17 / 19
29. Rivas-Martı
´nez, A. Penas & T.E.
´az (2004) Bioclimatic and biogeographic maps of Europe. www.
globalbioclimatics.org/form/maps.htm (accessed November 2014). Chartographic Service, Univeristy
of Leon, Spain.
30. FAO (2013) African swine fever in the Russian Federation: risk factors for Europe and beyond.
EMPRES WATCH, Vol. 28, May 2013. Rome
31. IUCN (2008) Sus scrofa. The IUCN Red List of Threatened Species. Version 2014.3 http://maps.
iucnredlist.org/map.html?id=41775.
32. Carnevali L, Pedrotti L, Riga F, Toso S (2009) Banca Dati Ungulati:Status, distribuzione, consistenza,
gestione e prelievo venatorio delle popolazioni di Ungulatiin Italia. Rapporto 2001–2005. Biologia e Con-
servazione della Fauna 117: 1–168.
33. Rossell C, Herrero J (2007) Sus scrofa Linnaeus,1758. In: Palomo LJ, Gisbert J, Blanco JC, editors.
Atlas y libro rojo de los mamı´feros de España Madrid: Direccio
´n General para la Biodiversidad-
SECEM-SECEMU. pp. 348–351.
34. Tsachalidis EP, Hadjisterkotis E (2009) Current distribution and population status of wild boar (Sus
scrofa L.) in Greece. Acta Silvatica and Lignaria Hungarica 5.
35. Opsteegh M, Swart A, Fonville M, Dekkers L, van der Giessen J (2011) Age-Related Toxoplasma gondii
Seroprevalence in Dutch Wild Boar Inconsistent with Lifelong Persistence of Antibodies. PLoS ONE 6:
e16240. https://doi.org/10.1371/journal.pone.0016240 PMID: 21283764
36. Karami M, Mohammadi H, Cheraghi S (2010) Present distribution, abundance and problems of wild pig
(Sus scrofa) in Iran. 2010 International Wild Pig Symposium.
37. Saulich MIAoa (2007) Area of distribution and damage of Sus scrofa Linnaeus. Interactive Agricultural
Ecological Atlas of Russia and Neighboring Countries http://wwwagroatlasru/en/content/pests/Sus_
scrofa/map/
38. Quinn GP, Keough MJ (2002) Experimental Design and Data Analysis for Biologists. Cambridge: Cam-
bridge University Press.
39. Acevedo P, Escudero M, Muńoz R, Gorta
´zar C (2006) Factors affecting wild boar abundance across an
environmental gradient in Spain. Acta Theriologica 51: 327–336.
40. Merli E, Meriggi A (2006) Using harvest data to predict habitat-population relationship of the wild boar
Sus scrofa in Northern Italy. Acta Theriologica 51: 383–394.
41. Hijmans RJ, Cameron SE, Parra JL, Jones PG, Jarvis A (2005) Very high resolution interpolated climate
surfaces for global land areas. International Journal of Climatology 25: 1965–1978.
42. DiMiceli CM, Carroll ML, Sohlberg RA, Huang C, Hansen MC, Townshend JRG (2011) Annual Global
Automated MODIS Vegetation Continuous Fields (MOD44B) at 250 m Spatial Resolution for Data
Years Beginning Day 65, 2000–2010, Collection 5 Percent Tree Cover University of Maryland, College
Park, MD, USA.
43. Rodrı
´guez E, Morris CS, Belz JE (2006) A global assessment of the SRTM performance. Photogram-
metric Engineering & Remote Sensing 72: 249–260.
44. R Development Core Team (2014) R: A language and environment for statistical computing. Vienna,
Austria: R Foundation for Statistical Computing.
45. Robinson TP, Wint GRW, Conchedda G, Van Boeckel TP, Ercoli V, Palamara E, et al. (2014) Mapping
the Global Distribution of Livestock. PLoS ONE 9: e96084. https://doi.org/10.1371/journal.pone.
0096084 PMID: 24875496
46. Metzger MJ, Bunce RGH, Jongman RHG, Sayre R, Trabucco A, Zomer R (2013) A high-resolution bio-
climate map of the world: a unifying framework for global biodiversity research and monitoring. Global
Ecology and Biogeography 22: 630–638.
47. Bholowalia P, Kumar A (2014) EBK-Means: A Clustering Technique based on Elbow Method and K-
Means in WSN. International Journal of Computer Applications 105: 17–24.
48. Kyriakidis PC (2004) A Geostatistical Framework for Area-to-Point Spatial Interpolation. Geographical
Analysis 36: 259–289.
49. Fourcade Y, Engler JO, Ro¨dder D, Secondi J (2014). Mapping Species Distributions with MAXENT
Using a Geographically Biased Sample of Presence Data: A Performance Assessment of Methods for
Correcting Sampling Bias. PLoS ONE 9, e97122. https://doi.org/10.1371/journal.pone.0097122 PMID:
24818607
50. Dormann CF, Elith J, Bacher S, Buchmann C, Carl G, Carre
´G, et al. (2013), Collinearity: a review of
methods to deal with it and a simulation study evaluating their performance. Ecography, 36: 27–46.
https://doi.org/10.1111/j.1600-0587.2012.07348
51. Pittiglio C, Skidmore AK, van Gils HAMJ, Prins HHT (2012) Identifying transit corridors for elephant
using a long time-series. International Journal of Applied Earth Observation and Geoinformation 14:
61–72.
Wild boar mapping
PLOS ONE | https://doi.org/10.1371/journal.pone.0193295 May 16, 2018 18 / 19
52. Allison PD (1999) Logistic Regression Using the SAS System: Theory and Applications. NC: SAS Insti-
tute Inc., Cary.
53. Liu X, Liu Y (2008) The accuracy assessment in areal interpolation: An empirical investigation. Science
in China Series E: Technological Sciences 51: 62–71.
54. Tranmer M, Holt D, Steel D (1997) Logistic regression analysis with aggregate data: tackling the eco-
logical fallacy.; 1997.
55. Barbosa AM, Real R, Vargas JM (2009) Transferability of environmental favourability models in geo-
graphic space: the case of the Iberian desman (Galemys pyrenaicus) in Portugal and Spain. Ecological
Modelling 220: 747–754.
56. Wenger SJ, Olden JD (2012) Assessing transferability of ecological models: an underappreciated
aspect of statistical validation. Methods in Ecology and Evolution 3: 260–267.
57. Potts JM, Elith J (2006) Comparing species abundance models. Ecological Modelling 199: 153–163.
58. Leslie E, Cowled B, Graeme Garner M, Toribio JALML, Ward MP (2014) Effective Surveillance Strate-
gies Following A Potential Classical Swine Fever Incursion in a Remote Wild Pig Population in North-
Western Australia. Transboundary and emerging diseases 61: 432–442. https://doi.org/10.1111/tbed.
12044 PMID: 23294519
59. Cowled BD, Garner MG, Negus K, Ward MP (2012) Controlling disease outbreaks in wildlife using lim-
ited culling: modelling classical swine fever incursions in wild pigs in Australia. Veterinary research 43:
1–16.
60. Ward MP, Garner MG, Cowled BD (2015) Modelling foot-and-mouth disease transmission in a wild pig–
domestic cattle ecosystem. Australian Veterinary Journal 93: 4–12. https://doi.org/10.1111/avj.12278
PMID: 25622702
Wild boar mapping
PLOS ONE | https://doi.org/10.1371/journal.pone.0193295 May 16, 2018 19 / 19
... Understanding how bTB moves through wildlife communities is essential for managing OBs [9]. In countries such as Italy, diagnosing and monitoring bTB in wild boars is especially crucial in establishing comprehensive eradication schemes [53,58]. However, accurate and detailed population statistics, such as census and hunting data, are often not available, with a persistent lack of comprehensive information of interactions at the wildlife-livestock interface [2,58]. ...
... In countries such as Italy, diagnosing and monitoring bTB in wild boars is especially crucial in establishing comprehensive eradication schemes [53,58]. However, accurate and detailed population statistics, such as census and hunting data, are often not available, with a persistent lack of comprehensive information of interactions at the wildlife-livestock interface [2,58]. In addition, the diagnosis of bTB in free-ranging wildlife is undoubtedly more difficult than in domestic animals [53]. ...
Article
Full-text available
Bovine tuberculosis (bTB) is a zoonotic disease with consequences for public health as well as the economy. In the EU, compulsory eradication programmes have been applied, and most territories in Italy have been reported as disease-free (FTs). However, outbreaks (OBs), i.e., an officially confirmed occurrence of bTB in one or more animals in an establishment, have continued to be reported. In this study we provide an overview of bTB in terms of OB numbers in cattle from Italian FTs. Legislative sources were collected to find the FTs, the relevant declaration of free status year (FSY), and regional control and surveillance plans. Then, descriptive and statistical analyses were applied to the collected OBs. A total of 12 regions and 19 provinces were declared FTs in the 20 years from 2003 to 2023. Differences in regional plans were observed with respect to the percentages of herds that were annually controlled (control frequency). Overall, 370 OBs were recorded. A non-statistically significant decrease in the OB incidence rate after the FSY was declared. However, a notable increase in OBs detected at slaughterhouses after the FSY suggests that control systems (serological tests) at the herd level are not completely effective. Differences in the herds’ control frequencies among FTs seem to not have had a significant influence on the observed OB number. The Tuscany region was the most affected FT based on the OB numbers after the FSY (especially in the last year). Epidemiologically relevant primary determinants seem to be the farming system (semi-extensive and adjacent herds) and the cattle movements from positive incidence areas (trade and animal fairs). The role of wild boars in the disease maintenance cannot be excluded. The results of this study stress the need to revise bTB eradication and surveillance plans based on risk analysis.
... This issue makes it difficult to differentiate between territories where the species is present or even abundant but it is not hunted, from those places where the species is actually absent and therefore no hunting yield is reported (Imperio et al., 2010). Second, hunting yield is usually reported at different spatial resolutions depending on local or regional administrations, from hunting estates to whole provinces (Lindström & Bergqvist, 2020;Pittiglio et al., 2018). This makes the comparison among regions more complicated, hindering the datapooling from different regions or countries. ...
Article
Full-text available
Harvest data have the potential to be used as an abundance index due to its widespread availability and long‐term collection across large geographical areas. However, challenges such as the lack of hunting effort information, varying data resolutions and reporting biases hinder its direct use as an abundance proxy. Here, we present the game target‐group, a statistical approach based on a thinned inhomogeneous Poisson point process, to estimate animal abundance at fine‐scale resolution from hunting data. We employ a Bayesian hierarchical framework to borrow information from harvest data on related species to overcome issues due to the lack of hunting management information. We conducted a simulation study to explore model performance and parameter identifiability under different scenarios (sample size, species catchability/abundance and unmodelled heterogeneity) and assessed the method on a real case study with four species in central Spain. The simulation study confirmed that with a large enough sample size (n > 5000), high catchability and lack of unmodelled heterogeneity in the abundance process, the model was able to obtain unbiased estimations for total abundance parameters. In the case study, our model successfully captured species‐habitat relationships and produced reliable estimates of total abundance at regional scale. Internal validation with independent test data and external validation with fieldwork data confirmed the model's ability to predict hunting yields and estimate species total abundance accurately. Our approach provides a flexible and valuable tool for large‐scale monitoring programs relying on harvest data with potential applications in wildlife management and conservation. However, the method should be applied with caution when there is unmodelled heterogeneity, low catchability or the sample size is small (<5000).
... Wild boar populations have increased recently, particularly in Europe, encroaching on crops, pastures, and peri-urban areas. In these regions, the animals may be susceptible to infection with the parasite, which could result in the transmission of the parasite to other animals and humans [14,15]. ...
Article
Full-text available
Background Macracanthorhynchus hirudinaceus (Pallas, 1781) is a zoonotic acanthocephalan that parasitizes the small intestine of wild boars. It is a pathogenic that causes economic losses, and poses a public health threat due to increased emergence. Purpose The aims of this study is describes histopathologically the damage caused by M. hirudinaceus in the small intestine of wild boar Sus scrofa Linnaeus, 1758, and molecularly characterize this parasite (sequences, haplotypes, phylogeny) for the first time in Elazig city, Türkiye. Results A high prevalence of infection was obtained. Upon separating the worms, it was discovered that there were ulcers resembling craters in the center, of the small intestine mucosa, surrounded by edema. The intestine wall where the parasite attached was damaged, with the villi epithelium and lamina propria in the mucosa being destroyed. The genomic DNA was isolated from all M. hirudinaceus samples, and PCR amplified the 489 bp gene fragments were sequenced and confirmed that all 21 sequences were M. hirudinaceus. The haplotype analysis of the sequences revealed the presence of a central star-shaped haplotype, in addition to four other haplotypes. Conclusion After conducting sequence analysis, the genetic differences between the M. hirudinaceus sequences obtained in this study and those reported from Europe and Japan suggest that this parasite is endemic to Türkiye's local wild boar population. Also, four haplotypes were identified, distinguishing it from other haplotypes by 1-5 mutation steps. It is essential to consider the worm's sequences and the formation of haplotypes, since these intrinsic characteristics may impact in the epidemiology and pathology of the worm in the future.
... Area-to-point kriging was used to disaggregate the residuals of the linear regression. Pittiglio et al. [80] disaggregated wild boar population density from the administrative level to 5 km 2 grids, following a similar approach, utilizing used temperature, precipitation, vegetation cover, and topography variables such a slope and elevation as proxies. ...
Article
Full-text available
National-level climate action plans are often formulated broadly. Spatially disaggregating these plans to individual municipalities can offer substantial benefits, such as enabling regional climate action strategies and for assessing the feasibility of national objectives. Numerous spatial disaggregation approaches can be found in the literature. This study reviews and categorizes these. The review is followed by a discussion of the relevant methods for the disaggregation of climate action plans. It is seen that methods employing proxy data, machine learning models, and geostatistical ones are the most relevant methods for the spatial disaggregation of national energy and climate plans. The analysis offers guidance for selecting appropriate methods based on factors such as data availability at the municipal level and the presence of spatial autocorrelation in the data. As the urgency of addressing climate change escalates, understanding the spatial aspects of national energy and climate strategies becomes increasingly important. This review will serve as a valuable guide for researchers and practitioners applying spatial disaggregation in this crucial field.
... To investigate the potential correlation between genetic diversity within a province and the corresponding density of wild boar populations, the Pearson correlation coefficient was calculated. The wild boar density in each province was obtained with the following approach in QGIS 3.22 Białowieża LTR: in the study area, a buffer of 10.38 km [average dispersal range of adult wild boars according to Keuling et al. (2010) from the borders of each municipality was calculated; the average density of wild boars within each obtained buffer was extracted from the wild boar density raster estimated by Pittiglio et al. (2018)]; the average density was used as an aggregated measure for each province. ...
Article
Full-text available
Introduction Porcine circovirus 2 (PCV-2) is a key pathogen for the swine industry at a global level. Nine genotypes, differing in epidemiology and potentially virulence, emerged over time, with PCV-2a, -2b, and -2d being the most widespread and clinically relevant. Conversely, the distribution of minor genotypes appears geographically and temporally restricted, suggesting lower virulence and different epidemiological drivers. In 2022, PCV-2e, the most genetically and phenotypically divergent genotype, was identified in multiple rural farms in North-eastern Italy. Since rural pigs often have access to outdoor environment, the introduction from wild boars was investigated. Methods Through a molecular and spatial approach, this study investigated the epidemiology and genetic diversity of PCV-2 in 122 wild boars across different provinces of North-eastern Italy. Results Molecular analysis revealed a high PCV-2 frequency (81.1%, 99/122), and classified the majority of strains as PCV-2d (96.3%, 78/81), with sporadic occurrences of PCV-2a (1.2%, 1/81) and PCV-2b (2.5%, 2/81) genotypes. A viral flow directed primarily from domestic pigs to wild boars was estimated by phylogenetic and phylodynamic analyses. Discussion These findings attested that the genotype replacement so far described only in the Italian domestic swine sector occurred also in wild boars. and suggested that the current heterogeneity of PCV-2d strains in Italian wild boars likely depends more on different introduction events from the domestic population rather than the presence of independent evolutionary pressures. While this might suggest PCV-2 circulation in wild boars having a marginal impact in the industrial sector, the sharing of PCV-2d strains across distinct wild populations, in absence of a consistent geographical pattern, suggests a complex interplay between domestic and wild pig populations, emphasizing the importance of improved biosecurity measures to mitigate the risk of pathogen transmission.
... The study was conducted on wild boars from different regions of Greece. Most samples were collected from northern areas of the country due to the high wild boar population density and the proximity to the European inland and other Balkan countries with high wild boar populations (Pittiglio et al. 2018;ENETwild Consortium et al. 2018). During the 2019-2020 hunting season (i.e., October 2019 to January 2020), in the frame of a previously published study (Dimzas et al. 2021), diaphragm samples were collected from 128 wild boars. ...
Article
Full-text available
The genus Sarcocystis includes protozoan parasites with an indirect life cycle. Sarcocystis spp. can infect various animal species and humans, causing sarcocystosis, a parasitosis of economic importance and zoonotic concern. Wild boars can act as intermediate hosts for Sarcocystis miescheriana and the zoonotic Sarcocystis suihominis that infects humans by consumption of raw or undercooked infected swine meat. In the present study, the diaphragmatic muscle tissue of 123 wild boars hunted in Greece was examined to determine the frequency of Sarcocystis spp. The samples were examined by tissue compression and molecular techniques. Under light microscopy, 34 out of 123 (27.6%) wild boars tested positive for Sarcocystis spp., while a higher infection prevalence (75%) was revealed by multiplex PCR performed in 100 of the samples. The partial mtDNA cox1 gene (~ 1100 bp) of 20 samples tested positive for S. miescheriana by multiplex PCR was amplified and sequenced. Sarcocystis miescheriana was identified as the only species involved in these infections. This is the first study on the prevalence of Sarcocystis spp. in wild animals in Greece. Further, large-scale surveys are needed to assess the prevalence and species of this parasite in Greece and to design efficient control and preventive measures in a One Health perspective.
Article
Tuscany region (Italy) hosts a large wild boar (Sus scrofa) population. The spread of these large mammals in agricultural areas poses significant challenges for farmers, local communities, and ecosystems. Moreover, it represents a health risk by favoring the spread of bacterial and viral pathogens. In this study, we investigated the genetic variability of wild boar from Tuscany using short tandem repeats (STRs) markers. Sixty wild boars hunted across four provinces (Pisa, Siena, Livorno, and Grosseto) during the 2018/2019 hunting season were processed. DNA was extracted from kidney and 15 microsatellites commonly used for genetic analysis (S0005, S0090, S0101, S0155, SW24, SW240, SW857, SW72, SW936, SW911, S0227, S0228, S0386, S0355, and SW951) were employed. For each marker, we calculated the number of alleles, allelic frequencies, effective number of alleles, observed heterozygosity, and polymorphism information content. We also computed molecular co-ancestry coefficients, kinship distance, inbreeding coefficients, and genetic similarities. Population structure was inferred using STRUCTURE. Pairwise Wright’s Fixation Index (FST) and the gene flow estimation (Nm) comparisons were performed between sample from all provinces, providing insight into the degree of genetic exchange and connectivity. Analysis of molecular variance (AMOVA) was used to examine the partitioning of genetic diversity among populations from the four provinces. In total, 87 alleles were found; two markers (S0355 and SW951) were monomorphic, while the other 13 were polymorphic, with allele counts ranging from three (S0228) to 20 (S0005), and a mean of 6.54 (± 4.351). Results indicated an unstructured and homogeneous population due to the lack of significant geographical barriers and the short distances between sampling sites.
Article
Full-text available
African swine fever (ASF) is a highly contagious disease affecting wild and domestic pigs, characterised by severe haemorrhagic symptoms and high mortality rates. Originally confined to Sub-Saharan Africa, ASF virus genotype II has spread to Europe since 2014, mainly affecting Eastern Europe, and progressing through wild boar migrations and human action. In January 2022, the first case of ASF, due to genotype II, was reported in North-western Italy, in a wild boar carcass. Thereafter, numerous positive wild boars were identified, indicating an expanding wild epidemic, severely threatening Italian pig farming and trade. This study focused on the mapping of the suitable habitats for wild boars and their potential dispersal corridors in Northern Italy, using species distribution models and landscape connectivity analysis. The resulting maps identified areas with higher likelihood of wild boar presence, highlighting their preferential pathways crossing Northern Italy. The distribution of ASF positive wild boars along the major corridors predicted by the model suggests the obtained maps as valuable support to decision-makers to improve ASF surveillance and carcass early detection, aiming for eradication. The applied framework can be easily replicated in other regions and countries.
Preprint
Full-text available
Colour polymorphism, the presence of multiple colour variants within a population, is a common example of intraspecific phenotypic variation and has served as a model for studying drivers of diversity. Climatic factors can influence the distribution and abundance of colour variants, yet research often focuses on lineages where sexual selection covaries with the climate-colouration associations. Research has also focussed disproportionately on vertebrates and a few insects, neglecting other taxa where polymorphism is widespread, like arachnids. Here, we investigated climatic factors influencing colour variation in a widely distributed colour polymorphic spider using a combination of controlled experiments, field measurements and tests for macroecological associations. We showed in the lab and field that dark colouration is unlikely to provide biologically relevant thermal benefits; yet different colour morphs occupy distinct climatic niches across Australia. Precipitation, rather than temperature, appears to be the strongest driver of colour morph frequencies: there were higher frequencies of dark individuals in regions with higher precipitation while orange individuals were more common in drier areas. We discuss potential mechanisms related to background matching or pathogen protection in humid environments. Our findings highlight the role of climatic factors in shaping colour variation, and demonstrate the value of combining macroecological, field, and laboratory approaches.
Article
Full-text available
EFSA assisted four countries in the analysis of epidemiological data on African swine fever (ASF), collected until September 2017. The temporal analysis demonstrated that the average proportions of PCR and antibody-ELISA positive samples from the hunted wild boar remained below 3.9 and 6.6, respectively. A peak in the ASF incidence was observed 6 months after the first observed case, followed by a significant reduction of the number of cases and low levels of African swine fever virus (ASFV) circulation at the end of 38 months follow-up period at different spatial resolutions. The spatial analysis concluded that human-mediated spread of ASFV continues to play a critical role in the ASF epidemiology, despite all measures currently taken. 'Wild boar density', 'total road length' (as proxy for human activity) and 'average suitable wild boar habitat availability' were identified as predictors for the occurrence of ASF in Estonia by a Bayesian hierarchical model, whereas 'wild boar density' and 'density of pig farms' were predictors according to a generalised additive model. To evaluate the preventive strategies proposed in EFSA's Scientific Opinion (2015) to stop the spread of ASFV in the wild boar population, a simulation model, building on expert knowledge and literature was used. It was concluded that reduction of wild boar population and carcass removal to stop the spread of ASFV in the wild boar population are more effective when applied preventively in the infected area. Drastic depopulation, targeted hunting of female wild boar and carcass removal solely implemented as measures to control ASF in the wild boar population need to be implemented in a highly effective manner (at or beyond the limit of reported effectivity in wild boar management) to sustainably halt the spread of ASF.
Article
Full-text available
Niche conservatism, i.e. the retention of a species’ fundamental niche through evolutionary time, is cornerstone for biological invasion assessments. The fact that species tend to maintain their original climate niche allows predictive maps of invasion risk to anticipate potential invadable areas. Unravelling the mechanisms driving niche shifts can shed light on the management of invasive species. Here, we assessed niche shifts in one of the world's worst invasive species: the wild boar Sus scrofa . We also predicted potential invadable areas based on an ensemble of three ecological niche modelling methods, and evaluated the performance of models calibrated with native vs. pooled (native plus invaded) species records. By disentangling the drivers of change on the exotic wild boar population's niches, we found strong evidence for niche conservatism during biological invasion. Ecological niche models calibrated with both native and pooled range records predicted convergent areas. Also, observed niche shifts are mostly explained by niche unfilling, i.e. there are unoccupied areas in the exotic range where climate is analogous to the native range. Niche unfilling is expected as result of recent colonization and ongoing dispersal, and was potentially stronger for the Neotropics, where a recent wave of introductions for pig‐farming and game‐hunting has led to high wild boar population growth rates. The invasive potential of wild boar in the Neotropics is probably higher than in other regions, which has profound management implications if we are to prevent their invasion into species‐rich areas, such as Amazonia, coupled with expansion of African swine fever and possibly great economic losses. Although the originally Eurasian‐wide distribution suggests a pre‐adaptation to a wide array of climates, the wild boar world‐wide invasion does not exhibit evidence of niche evolution. The invasive potential of the wild boar therefore probably lies on the reproductive, dietary and morphological characteristics of this species, coupled with behavioural thermoregulation.
Article
Full-text available
Biotic and abiotic factors are increasingly acknowledged to synergistically shape broad-scale species distributions. However, the relative importance of biotic and abiotic factors in predicting species distributions is unclear. In particular, biotic factors, such as predation and vegetation, including those resulting from anthropogenic land-use change, are underrepresented in species distribution modeling, but could improve model predictions. Using generalized linear models and model selection techniques, we used 129 estimates of population density of wild pigs (Sus scrofa) from 5 continents to evaluate the relative importance, magnitude, and direction of biotic and abiotic factors in predicting population density of an invasive large mammal with a global distribution. Incorporating diverse biotic factors, including agriculture, vegetation cover, and large carnivore richness, into species distribution modeling substantially improved model fit and predictions. Abiotic factors, including precipitation and potential evapotranspiration, were also important predictors. The predictive map of population density revealed wide-ranging potential for an invasive large mammal to expand its distribution globally. This information can be used to proactively create conservation/management plans to control future invasions. Our study demonstrates that the ongoing paradigm shift, which recognizes that both biotic and abiotic factors shape species distributions across broad scales, can be advanced by incorporating diverse biotic factors.
Article
Full-text available
This is the first attempt to estimate wild boar (Sus scrofa L.) population density and distribution in Greece. The study took place in 2004, in all 38 Greek counties of the mainland, as well as in the larger forested Greek islands: Crete, Rodos, Samos, Lesvos, Limnos, Thasos, Corfu, Kefalonia and Zakinthos. It was observed that the species was present in every county in continental Greece, with the exception of the two counties Attica and Evia. Attica, which includes the capital city of Athens, is densely inhabited not allowing much space for wildlife. Evian is an island connected to the mainland with a bridge. Wild boar can be found in an area around 19,495 km2, which is about 14% of the country. The mean distribution altitude reaches 686 ± 266 m. The average population number was estimated at 19,033 (0.98 ± 0, 21 ind/km2) individuals with maximum population level 23,030 animals, and a minimum of 16,536. The highest mean density was observed in the prefecture of Sterea Hellas (1.26 ind/km2) and the lowest in the prefecture of Thrace (0.89 ind/km2). The maximum density was found in Sterea Hellas (3.14 ind/km2) and the minimum density in Thrace (0.13 ind/km2). The above results are expected to provide valuable information for the management of wild boar in Greece.
Article
1. The eruption of invasive wild pigs (IWPs) Sus scrofa throughout the world exemplifies the need to understand the influences of exotic and nonnative species expansions. In particular, the continental USA is precariously threatened by a rapid expansion of IWPs, and a better understanding of the rate and process of spread can inform strategies that will limit the expansion. 2. We developed a spatially and temporally dynamic model to examine three decades (1982–2012) of IWP expansion, and predict the spread of IWPs throughout the continental USA, relative to where IWPs previously inhabited. We used the model to predict where IWPs are likely to invade next. 3. The average rate of northward expansion increased from 6�5 to 12�6 km per year, suggesting most counties in the continental USA could be inhabited within the next 3–5 decades. The spread of IWPs was primarily associated with expansion into areas with similar environmental characteristics as their previous range, with the exception of spreading into colder regions. We identified that climate change may assist spread into northern regions by generating milder winters with less snow. Otherwise, the spread of IWPs was not dependent on agriculture, precipitation or biodiversity at the county level. The model correctly predicted 86% of counties that were invaded during 2012, and those predictions indicate that large portions of the USA are in immediate danger of invasion. 4. Synthesis and applications. Anti-invasion efforts should focus along the boundaries of current occupied range to stop natural expansion, and anti-invasion policies should focus on stopping anthropogenic transport and release of invasive wild pigs. Our results demonstrate the utility of a spatio-temporal examination to inform strategies for limiting the spread of invasive wild pigs.
Article
The spatial prediction of point values from areal data of the same attribute is addressed within the general geostatistical framework of change of support; the term support refers to the domain informed by each datum or unknown value. It is demonstrated that the proposed geostatistical framework can explicitly and consistently account for the support differences between the available areal data and the sought-after point predictions. In particular it is proved that appropriate modeling of all area-to-area and area-to-point covariances required by the geostatistical framework yields coherent (mass-p reserving or pycnophylactic) predictions. In other words, the areal average (or areal total) of point predictions within any arbitrary area informed by an areal-average (or areal-total) datum is equal to that particular datum. In addition, the proposed geostatistical framework offers the unique advantage of providing a measure of there liability (standard error) of each point prediction. It is also demonstrated that several existing approaches for area-to-point interpolation can be viewed within this geostatistical framework. More precisely, it is shown that (i) the choropleth map case corresponds to the geostatistical solution under the assumption of spatial independence at the point support level; (ii) several forms of kernel smoothing can be regarded as alternative (albeit sometimes incoherent) implementations of the geostatistical approach; and (iii) Tobler's smooth pycnophylactic interpolation, on a quasi- infinite domain without non-negativity constraints, corresponds to the geostatistical solution when the semivariogram model adopted at the point support level is identified to the free-space Green's functions (linear in 1-D or logarithmic in 2-D) of Poisson's partial differential equation. In lien of a formal case study, several I-D examples are given to illustrate pertinent concepts.