Connectivity modelling is a powerful tool for strengthening the link between landscape and species conservation. This approach often relies on expert knowledge of connectivity indicators or limited, small numbers and small-scale monitoring data, for example during animal foraging activities. However, integrating larger-scale movement data, including dispersal or seasonal movements, is crucial to making conservation relevant by covering the entire life cycle of species. Using Resource Selection Function, the movement patterns of greater horseshoe bats (GHB) studied on a local scale were transposed to a regional scale to model the connectivity in western France. GHB is highly sensitive to loss of connectivity and makes seasonal regional migrations. How the local landscape heterogeneity influenced the conductance parameters estimation for modelling was examined using gap-crossing method at four different sites with variable landscape composition. The inferred parameters were used to create a regional connectivity map based on circuit theory. To validate this map, acoustic monitoring was conducted during the autumn migration to assess its effectiveness in identifying connectivity gradients. Finally, the resulting connectivity map was superimposed on the Natura 2000 network. Firstly, it can be assumed that connectivity parameters are identical whatever the landscape context. Secondly, the regional connectivity model identified the main potential corridors connecting all the major sites in the region. Finally, based on acoustic sampling, the number of GHB in transit was significantly higher in areas of higher connectivity. In terms of overlap with conservation, the functional connectivity of the GHB have been variously addressed in the current Natura 2000 network, with an overall lack of representativeness. Studying pathways during high mobility periods is one of the main missing elements for effective conservation, particularly for small species such as bats. An acoustic, stratified sampling at both local and large scale provided sufficient spatial and temporal accuracy to model connectivity throughout the life cycle of bats. This framework can easily be applied to other bat species to improve our understanding of connectivity, in order to explicitly integrate this crucial aspect for highly mobile species into the protection network.

Empirical estimation of local-scale RSF in different landscapes

Study sites

To calculate conductance, RSF was estimated using the gap-crossing approach, i.e., the probability of crossing a gap in a given hedgerow as a function of the width of the gap. To determine whether variations in landscape context lead to significant variations in RSF estimation, gap-crossing probabilities were estimated at four sites differing in hedgerow density (from less than 20 to 200 m.ha^-1) (see Fig.S2). Data were recorded when most females were lactating, in July 2016 for the Annepont site and late June and July 2018 for three other sites: Les Verchers-sur-Layon, Réaumur, and Azay-sur-Thouet.

Acoustic sampling

The gap-crossing approach was used to estimate the probability of crossing a discontinuity (gap) in a connecting feature (hedgerow) as a function of the width of this gap, following Pinaud et al. (2018). At the four sites centred on a GHB maternity, recorders were placed overnight near the colony (< 2 km) in hedgerows with varying gap widths ranging from 5 to 150 m). BatClassify software (Scott, 2012) was then used to identify GHB in the recordings, which were then manually confirmed with Kaleidoscope software (version 5.3.0, Wildlife Acoustics, USA).

Modelling the connectivity map

Definition of the conductance raster

Based on the gap-crossing model, the RSF spatially predicted the probability of the presence of GHB as a function of the distance to a connecting element such as hedgerows or forests that facilitate movement, producing the conductance values in a raster for the entire study area. Connecting features (woods, vineyards, orchards, hedgerows, peri-urban areas such as villages and farms) were first mapped on a 10x10m resolution raster extracted from BD TOPO® provided by the French National Geographical Institute (IGN) in 2018. The probability of the presence of GHB as a function of the distance to the nearest connecting feature was then predicted from the gap-crossing model, giving the conductance values. From this model, since the probability of crossing at d=0 (intercept) was not equal to 1, the probabilities were rescaled from 0 to 1. As important barriers for bats, major roads and large urban areas were set to a conductance value of 0. Then, in order to facilitate movement within the connecting features (conductance initially set at 1, identical at locations a short distance away from them), the conductance values of the connecting features were set at 10, by analogy to classical studies (see Sawyer et al., 2011).

Calculation of the connectivity map

From this conductance raster, we used Circuitscape v5 software to create cumulative current maps (McRae et al., 2008) that identify functional connectivity of GHBs at a regional scale. For the layer of nodes to connect, we used the main roosts (maternity and wintering) with >10 individuals known from long-term monitoring of the GHB population in this region. We calculate the connectivity between all types of roost because the early results of an ongoing CMR study with automatic readers and >7000 GHBs fitted with PIT tags since 2016 showed that large-scale GHB movements were not unidirectional and occurred in both directions between all types of sites during short transit periods, both in autumn and spring. When calculating connectivity, for computational reasons due to the large number of sites (n = 418 sites: 56 for maternity and 361 for hibernation), the conductance raster was aggregated with 100 × 100 m cells by calculating the median conductance per aggregated 10 x 10 m cells. The use of the median for cell aggregation was validated using independent acoustic and radiotracking datasets from the Annepont study (see Pinaud et al., 2018), showing no difference in the ability of the model to predict the presence of GHB based on the flow of cumulative currents (i.e., connectivity).

Validation of the regional connectivity map using independent field data

Sampling design

To evaluate the biological relevance of the regional connectivity map, we aimed to relate the number of GHB detected (nbGHB) to connectivity values (cumulative currents) calculated from Circuitscape. As nbGHB also depends on the probability of GHB presence at a location, we expected:

   nbGHB = P(presence) x Connectivity (eq. 1)

We used a stratified sampling design where strata were chosen to maximise the probability of finding a GHB performing regional movements that occur in autumn between maternity and hibernation sites. This approach was adopted over pure random sampling as it is known to optimize efforts in collecting ecological spatial data, particularly with high density heterogeneity (see, for example, Smith et al. 2017). In our case, according to studies using passive bat recorder networks in this region (i.e., “VigieChiro”, Kerbiriou et al. 2018), the probability of contacting a GHB during a whole night far from roosting sites throughout the study area is low (between 0.1 and 0.5), leading to an expected large number of zeros (absence of GHB nearby, regardless of connectivity values) and consequently, according to a power analysis, to a higher sampling effort. With this design, if at least one GHB is detected in a stratum (in a corridor site), it means that this stratum is being used (i. e.   in eq. 1), so we will be able to directly relate the number of GHB passes to the connectivity predicted by the model. Passive ultrasonic recorders were placed at high-intensity corridor sites as indicated on the connectivity map. To maximize the probability of contacting a GHB in transit rather than foraging close to a roost, we placed the recorders at least 10 km away from known maternity roosts and hibernation sites, to match the maximum distance bats travel when foraging around roosts (Finch et al., 2020; Pinaud et al., 2018). Of the 46 corridor sites that met this criterion, 32 were randomly selected as strata for sampling. At each corridor site, recorders were placed (143 locations in total) in hedgerows at locations chosen to cover the entire gradient of connectivity values at that site.

Acoustic survey

To evaluate GHB activity around selected corridors, passive recorders fitted with ultrasonic microphones were placed overnight in the field from 6 pm to 6 am in late August 2020, during peak transit periods, with favorable weather conditions (no rain or strong wind). The recordings were classified using BatClassify software (Scott, 2012) and manually verified. The number of GHB detected at the sampling locations was measured as the number of GHB in a 5 seconds pass per night.