also avoided potential discontinuities induced by the conversion of linear elements
in a raster format (Adriaensen et al. 2003). Altogether, thirteen landscape categories
were obtained (Table 13.1).
Landscape Graph Construction
The nodes of the graphs corresponding to the habitat patches were defined as the
aquatic habitat units (breeding ponds) adjacent to an area of potential terrestrial
habitat. The quality of the habitat patches, i.e. capacity, was defined as the amount of
terrestrial habitat and suitable elements around ponds. The links between nodes were
defined in cost distance because the ability of the tree frog to disperse depends
greatly on the surrounding matrix. The ecological literature and expert opinions were
thus used to classify each landscape category according to its resistance to movement (Table 13.1). Radio tracking experiments (Pellet et al. 2004; Vos and Stumpel
1996) have concluded that wooded grasslands and linear elements like hedgerows or
forest edges facilitate movement and often provided the species’ terrestrial habitat.
Rivers and wetlands are less favorable because their permeability depends on the
density of vegetation. Conversely, grasslands, roads and railways tend to constrain
movement. Finally, the cores of forest patches, bare ground, buildings and motorways are considered highly impassable and are mostly avoided by the tree frog.
Clauzel et al. (2013) performed several tests by varying the cost values and the
number of classes to find the model that best explained the occurrence of the tree
frog in the Franche-Comté region. The results showed that highly contrasting
values between favorable and unfavorable landscape categories were the most
relevant. Consequently, in the present study (Table 13.1), aquatic and terrestrial
habitats were assigned a cost of 1. Suitable elements such as wetlands or wooded
grasslands were assigned a low cost (10), unfavorable landscape elements a cost of
100, and highly unfavorable elements a high cost (1000). In the second landscape
map including the HSR line, a cost of 10,000 was assigned to the infrastructure so
as to remove any links crossing it, i.e. considering it as an absolute barrier. From the
two landscape maps, without and with the HSR line, two graphs were built and
thresholded at a distance of 2500 m corresponding approximately to the dispersal
distance for the tree frog. This distance was selected in line with the results by
Clauzel et al. (2013), where several maximum distances were tested. The model
using the distance of 2500 m proved the most relevant. Consequently, only links
shorter than this distance were kept.
Landscape Graph Analysis
For each graph, connectivity metrics were calculated, and then compared over time
by computing their rates of variation. The magnitude of the difference between
metric values provided an assessment of the impact of the HSR line. Two levels of
220
C. Clauzel
in a raster format (Adriaensen et al. 2003). Altogether, thirteen landscape categories
were obtained (Table 13.1).
Landscape Graph Construction
The nodes of the graphs corresponding to the habitat patches were defined as the
aquatic habitat units (breeding ponds) adjacent to an area of potential terrestrial
habitat. The quality of the habitat patches, i.e. capacity, was defined as the amount of
terrestrial habitat and suitable elements around ponds. The links between nodes were
defined in cost distance because the ability of the tree frog to disperse depends
greatly on the surrounding matrix. The ecological literature and expert opinions were
thus used to classify each landscape category according to its resistance to movement (Table 13.1). Radio tracking experiments (Pellet et al. 2004; Vos and Stumpel
1996) have concluded that wooded grasslands and linear elements like hedgerows or
forest edges facilitate movement and often provided the species’ terrestrial habitat.
Rivers and wetlands are less favorable because their permeability depends on the
density of vegetation. Conversely, grasslands, roads and railways tend to constrain
movement. Finally, the cores of forest patches, bare ground, buildings and motorways are considered highly impassable and are mostly avoided by the tree frog.
Clauzel et al. (2013) performed several tests by varying the cost values and the
number of classes to find the model that best explained the occurrence of the tree
frog in the Franche-Comté region. The results showed that highly contrasting
values between favorable and unfavorable landscape categories were the most
relevant. Consequently, in the present study (Table 13.1), aquatic and terrestrial
habitats were assigned a cost of 1. Suitable elements such as wetlands or wooded
grasslands were assigned a low cost (10), unfavorable landscape elements a cost of
100, and highly unfavorable elements a high cost (1000). In the second landscape
map including the HSR line, a cost of 10,000 was assigned to the infrastructure so
as to remove any links crossing it, i.e. considering it as an absolute barrier. From the
two landscape maps, without and with the HSR line, two graphs were built and
thresholded at a distance of 2500 m corresponding approximately to the dispersal
distance for the tree frog. This distance was selected in line with the results by
Clauzel et al. (2013), where several maximum distances were tested. The model
using the distance of 2500 m proved the most relevant. Consequently, only links
shorter than this distance were kept.
Landscape Graph Analysis
For each graph, connectivity metrics were calculated, and then compared over time
by computing their rates of variation. The magnitude of the difference between
metric values provided an assessment of the impact of the HSR line. Two levels of
220
C. Clauzel