Introduction

After being widespread in Europe, the Eurasian beaver (Castor fiber) was reduced to 1200 individuals at the beginning of the 20th century, mainly due to overhunting (Nolet and Rosell 1998; Halley and Rosell 2002). These individuals were divided into eight relict populations, five of them occurring in Europe (Halley et al. 2012). Following legal protection, hunting restrictions, and several translocation programs, the species started a recovery, gradually recolonizing much of its former habitat (Nolet and Rosell 1998), and the more recent estimate now indicates a population of approximately 1,200,000 individuals in Europe (Wróbel 2020).

Legal protection started in the 18th and 19th centuries in most European countries, while the first reintroduction occurred in Sweden in the 1920s (Nolet and Rosell 1998). Subsequent reintroductions concerned Norway, Finland, Russia, Estonia, Latvia, Lithuania, Poland, Netherlands, Germany, Switzerland, France, Austria, and the Czech Republic, initially motivated by the aim of reintroducing the species for hunting purposes (Nolet and Rosell 1998). Even if some reintroductions failed, most of them resulted in the establishment of stable populations (Nolet and Rosell 1998). Among the factors that favor its recovery, are the habitat availability and the ecological plasticity of the species, capable of using various food sources and selecting even sub-optimal habitats (Haarberg and Rosell 2006; John et al. 2010; Krojerová-Prokešová et al. 2010; Swinnen et al. 2017).

Similar to the situation in other parts of Europe, the Eurasian beaver in Italy and Spain was hunted extensively for meat, fur, and castoreum, a secretion produced by peri-anal glands used by humans for medical purposes (Nolet and Rosell 1998). In Italy, the species appeared in the Early Pleistocene, with fossils found in both Northern and Central Italy (Barisone et al. 2006). Late Pleistocene fossils have been found predominantly in Northern Italy (Salari et al. 2020). Similarly, most of the Holocene records of the species came from Northern Italy, Lombardy, and North-Eastern Italy, with fewer numbers from Central Italy (Salari et al. 2020). The extinction of the Eurasian beaver in Italy is dated between the 15th and 16th centuries, with a specific extinction date in 1541 (Halley et al. 2012). Regarding the Iberian Peninsula, the beaver has been present since the Pliocene, with its first appearance in the Early Pleistocene, followed by population fluctuations during subsequent periods (Cuenca-Bescós et al. 2021). The extinction in Spain is dated to the 17th century (Halley and Rosell 2002), while no historical information is available for Portugal.

In 2003, eighteen beavers were illegally reintroduced in Spain within the Ebro basin (Ceña et al. 2004). Attempts to remove the individuals failed, and the species subsequently expanded its range, establishing a permanent presence in the Ebro basin (Halley et al. 2021). Currently, there are more than 1000 individuals, spanning approximately 300 km along the Ebro River and its primary tributaries (Halley et al. 2021). Following the derogation from the Habitats Directive which allowed the Spanish authorities to remove the species, in 2018 a letter from the European Commission Environment Directorate reversed the position, affirming beavers as a historically native species, and once again extending protection under the Habitats Directive. More recently, the species was reported in the western Iberian Peninsula, bordering Portugal, in the Douro basin, as a result of a probable unauthorized release (Calderón et al. 2022).

Only recently, the Eurasian beaver reappeared in north-eastern Italy (Friuli Venezia Giulia region), with a natural recolonization right from one of the reintroduced populations in Austria (Pontarini et al. 2018). Another individual was observed in Alto Adige in 2020 (Pucci et al. 2021). Since then, the species has been noted in different disjointed areas in Central Italy, probably because of multiple illegal reintroductions (Pucci et al. 2021). Currently, the Eurasian beaver in Italy occurs in Friuli-Venezia-Giulia (Tarvisio forest), Alto-Adige, Tuscany (Tevere, Arno, Merse and Ombrone river basins), Emilia Romagna (Reno river basin), Umbria (Tevere river basin), Latium (Tevere river basin), Abruzzi (Aterno river basin), between Molise and Campania (Volturno river basin), and in Piedmont (Po and Toce river basin) (Mori et al. 2021; Pucci et al. 2021; Capobianco et al. 2023).

While the return of the Eurasian beaver in Northern Italy may be attributed to natural recolonization from the populations from Austria, the occurrences from Central Italy are likely the result of multiple illegal reintroductions, like the ones occurred in the western Iberian Peninsula (Pucci et al. 2021; Calderón et al. 2022; Capobianco et al. 2023). In this context of rapid beaver expansion, it is of paramount importance to assess where the species could further expand next. Furthermore, considering the abundance and environmental impacts, of the dams settled in the European rivers (Belletti et al. 2020), it is relevant to identify the specific barriers that should be targeted for mitigation or management actions to reduce river’s fragmentation for the Eurasian beaver.

Previous ecological modeling studies have indicated that both the Italian and Iberian Peninsulas retain several suitable patches for the beaver, particularly in Central Italy, Northern Spain, and Central Portugal (Serva et al. 2023a). Thus, further waves of natural expansions from the reproductive populations in Central Italy can be expected, as well as for the populations in the Ebro basin. In fact, this process is still ongoing as demonstrated by the recent new occurrences in both countries. To assess how the Eurasian beaver could expand, structural and functional connectivity is critical. In recent years, several techniques have been developed to evaluate functional connectivity. Circuit theory is among the most popular techniques, and Circuitscape is the reference software (McRae et al. 2008; Dickson et al. 2019). Circuitscape is based on the random-walk assumption (i.e., individuals moving across a landscape have no knowledge of surrounding habitat patches (McRae et al. 2008; Dickson et al. 2019). The Omniscape algorithm, applies Circuitscape in ‘advanced’ mode, iteratively, by using a moving window with a user-specified search radius, to evaluate omnidirectional connectivity in the landscape without considering source/destination areas (McRae et al. 2016; Landau et al. 2021). Both were recently used for semi-aquatic species (Dertien and Baldwin 2023; Falaschi et al. 2023; Leoncini et al. 2023). To explore structural and functional connectivity only considering each river where it has reported the presence of the Eurasian beaver, we also used the recently developed R package ‘riverconn’ (Baldan et al. 2022). The package is based on graph theory and allows the user to calculate different metrics of connectivity and fragmentation within each river (at reach-level), prioritizing the barriers present (Baldan et al. 2022).

In this study, we first aim to model functional connectivity in Italy and in the Iberian Peninsula, using circuit-theory techniques. Moreover, we apply the ‘riverconn’ R package to explore the functional and structural connectivity inside each river where the beaver presence is reported, and to highlight the potential barrier to the beaver’s expansion.

Materials and methods

Target species and study areas

The target species is the Eurasian beaver, among the larger rodents on Earth, listed as ‘least concern’ on the IUCN Red List. The species is protected under Annex II and IV of the European Habitats Directive, as under the Federal Act for the Protection of Nature. We focus on two different study areas, the Italian and the Iberian Peninsulas, two regions where the beavers are now expanding, after having been illegally reintroduced recently.

Environmental predictors

A resistance surface was generated as a layer, to describe the relationship between landscape structure and the Eurasian beaver movement. In this resistance surface, each pixel is assigned a value reflecting the resistance to the movement of the target species (Adriaensen et al. 2003). Considering the ecology of the Eurasian beaver, we integrated six environmental variables: Corine Land Cover, the hydrographic network, the distance from the hydrographic network, the Profile Curvature (which acts as a proxy of acceleration or deceleration of water flow, following Bernabò et al. (2022), the Tree Cover, and the position of dams.

The CORINE Land Cover layer was obtained from the Copernicus repository (https://land.copernicus.eu/pan-european/corine-land-cover/clc2018), the hydrographic network data from the HydroSHED dataset (https://www.hydrosheds.org/hydroatlas), the dams shapefile from OpenStreetMap, and then compared to other existing dams’ datasets (https://www.globaldamwatch.org/), while the Tree Cover was accessed from the Copernicus repository (https://land.copernicus.eu/pan-european/high-resolution-layers/forests/tree-cover-density). We used the ‘Euclidean Distance’ tool available in ArcGIS Pro (Inc 2023) to calculate the distance from the river network, and the ‘Surface Parameters’ function to obtain the Profile Curvature layer, starting from a Digital Elevation Model (downloaded from the Copernicus repository, European Digital Elevation Model, EU-DEM, version 1.1 https://land.copernicus.eu/imagery-in-situ/eu-dem). All the raster layers were processed at a 100 m resolution. Each variable was reclassified to a common scale, ranging from 1 (indicating the lowest suitability, and maximum resistance) to 10 (indicating the highest suitability and minimum resistance), with the ‘Reclassify’ tool in ArcGIS Pro. We opted for an expert-based approach to assign a resistance value to each pixel, and the such-obtained results were finally compared to published literature. This method was chosen because using inferential data applied to peripheral or new areas of a species’ expanding range might not effectively distinguish between suitable and unsuitable patches, as such areas may not yet be utilized as pathways (Clevenger et al. 2002). Moreover, expert-based opinion is among the most used approach to estimate resistance to movement (Zeller et al. 2012).

Given the semi-aquatic nature of beavers, the hydrographic network was assigned the lowest resistance value. Different resistance values were applied based on the Strahler order of rivers (higher resistance for rivers with Strahler order equal to 1, indicating torrential flow and unsuitability for beaver dams with a higher probability of running dry). We reclassified the distance from the hydrographic network following the reported terrestrial distances preferences (Haarberg and Rosell 2006; Graf et al. 2016; Swinnen et al. 2017), with different intervals: 0–50 m (lower resistance), 50–100 m, 100–200 m, 200–400 m, 400–20,000 m (higher resistance).

Considering the habitat selection of the beaver (Pinto et al. 2009; John et al. 2010; Smeraldo et al. 2017; Swinnen et al. 2017; Alakoski et al. 2020), we reclassified the Corine Land Cover (see Table S1 in Online Resource 1 for full details) assigning higher resistance values to artificial sites, marine waters, arable land, permanent crops and open space with little or no vegetation. Medium resistance values to pastures, heterogeneous agricultural areas, and shrub or herbaceous vegetation associations. Lower resistance values to forest areas, inland waters, and inland wetlands. The Tree Cover layer was reclassified linearly, i.e., from 1 to 100 to 1 to 10, according to beaver’s habitat preferences (Pinto et al. 2009; Smeraldo et al. 2017; Swinnen et al. 2017). We chose to use Profile Curvature, rather than Slope alone, as a predictor: this layer reports the acceleration and deceleration of the water flow and is useful to identify areas where slow- (or fast-) flowing water-adapted species may occur. We reclassified the Profile Curvature layer, following the beaver’s preferences (Pinto et al. 2009; Steyaert et al. 2015), giving low resistance values for areas with low acceleration/deceleration, thus indicating flat areas, and high resistance values for areas with high acceleration/deceleration values.

Finally, the dams’ layer was assigned a unique value of 1 (highest resistance), giving their potential role as a barrier for beaver dispersal (Nolet and Rosell 1998).

Resistance surface

We combined the predictors through the ‘Weighted Overlay’ tool in ArcGIS Pro. This tool lets the user combine different layers that share the same scale of values, assigning to each a specific percentage weight. We deliberately excluded the dam layer from this process to avoid modifying the effect (i.e., the resistance value) of these artificial barriers. Instead, we added the dams’ information after merging the other layers through the ‘Mosaic to New Raster’ function in ArcGIS Pro. Then, we transformed the final habitat suitability layer into a resistance layer, using a negative exponential transformation function (Keeley et al. 2016):

$$\text{R}= \text{1,000}-999\times \frac{1-\text{e}\text{x}\text{p}(-\text{c}\times \text{h})}{1-\text{exp}\left(-\text{c}\right)}$$

In this function, R represents the final resistance value of a pixel, h is the habitat suitability value for the same pixel, and c represents a constant factor determining the curvature of the negative exponential function. Previous studies have demonstrated that moderate values of the constant factor, c, provide the best performance (Keeley et al. 2016). The resistance layer expresses the degree to which a grid cell facilitates or impedes movement of the target species (Spear et al., 2010), a technique already employed in previous connectivity studies (Iannella et al. 2021a).

This transformation assigns high resistance values to the lowest habitat suitability values, and it has been shown to be a more accurate representation of landscape resistance, considering that habitat suitability may not be correlated with movement probability and landscape permeability (Keeley et al. 2016; Zeller et al. 2018).

We finally compared the resistance layer so obtained with the suitability maps found in Serva et al. (2023a) and Falaschi et al. (2023).

Omniscape analysis

We used Omniscape v.0.5.8 in Julia to compute landscape connectivity (Landau et al. 2021). Omniscape calculates omnidirectional connectivity using circuit theory, by using as input a resistance surface (McRae et al. 2016; Landau et al. 2021). This software operates iteratively within a moving window framework, with a user-specified search radius, where the center pixel is used as the destination and is set to ‘ground’, while pixels within the moving window are the sources where the current is injected, and the user could decide whether to use all the pixels as sources or only those with certain resistance values (Landau et al. 2021). Once defined the target pixel and the sources, Omniscape applies Circuitscape in “advanced” mode (Landau et al. 2021). These steps are repeated for every potential target pixel. As with other circuit theory software, the computed current is comparable to the probability of movement of the target species, in this case, the Eurasian beaver. Omniscape generates three different outputs: cumulative current flow, potential current flow, and normalized current flow. We set the search radius of the moving window to 25 km, as it is reported to be the maximum distance dispersal for the beaver (Mayer et al. 2017).

Connectivity toward current suitable patches

Considering that further beaver expansion in the Italian and Iberian Peninsulas is plausible (Halley et al. 2021; Serva et al. 2023a), we took advantage of the suitable patches in those areas identified by Serva et al. (2023a), to perform Circuitscape (v5.13.1, in Julia) runs in ‘advanced’ mode (Hall et al. 2021). Circuitscape requires as input a resistance layer and some source/target locations (i.e., the nodes). In the ‘advanced’ mode, users can select multiple sources (i.e., where current is injected) and destinations (i.e., where current exits the system) across a landscape, and they are activated simultaneously. We designated the basins with the highest suitability values and treated them as destination nodes. The areas where the beaver’s presence is confirmed (Halley et al. 2021; Mori et al. 2021; Pucci et al. 2021; Calderón et al. 2022; Capobianco et al. 2023) were used as sources. The resulted connectivity map represents the expected net probability of an organism moving from the nodes to the destinations.

Circuitscape analysis

Even though, within the two study areas, the beaver occurs only in a few territories, we decided to model a directional connectivity assessment, by using Circuitscape with the ‘pairwise’ mode. The main difference compared to Omniscape is that in this case, the software produces a connectivity assessment between source/destination nodes, i.e., a directional connectivity. In contrast to other connectivity analyses which produce only one path, Circuitscape produces multiple paths between nodes. The resistance map is read as an electric circuit, where the current flows between nodes, and the pixel value represents the passage cost of a ‘random walker’ (McRae et al. 2008). In the ‘pairwise’ mode, each node is both the source and the destination, and a current map is calculated iteratively for every pair of nodes, with a cumulative final map for all nodes representing the expected net probability of an organism to move between nodes.

Structural connectivity assessment

To assess structural connectivity within the river systems where the Eurasian beaver has been reported, we employed the recently developed ‘riverconn’ R package (Baldan et al. 2022). This package allows user to calculate different indices related to connectivity and fragmentation, at catchment or reach scale, distinguishing between structural (resulting from the physical structure of the river network) and functional connectivity (Baldan et al. 2022). The river network is treated as a graph (Erős et al., 2012), with an assemblage of edges and vertices, with edges representing either barriers or confluences between reaches, and vertices representing reaches. Reaches are river sections placed either between two confluences or between two barriers (Baldan et al. 2022). Furthermore, it is possible to prioritize the main barriers (such as the dams in this case) and include in the indices’ calculation additional data relevant to the target species (Baldan et al. 2022). We opted to include in those the tree cover density, as tree abundance is a critical factor for the expansion process of the Eurasian beaver and in defining whether a habitat will be occupied (Pinto et al. 2009; John et al. 2010; Swinnen et al. 2017). Specifically, we decided to calculate the following indices: DCI (Dendritic Connectivity Indices) where n ode weights are reaches lengths, IIC (Integral Index of Connectivity) also with uniform weights, and PCI (Population Connectivity Indices) both normal and with weighted usable length, associating the Tree Cover data. All the analyses were conducted in R software (R Core Team 2023), and the required shapefiles were managed in ArcGIS Pro (Inc 2023).

Results

Omniscape cumulative current map

The resistance layers (Fig. 1a) generally show the importance of the hydrographic network as corridors for the Eurasian beaver expansion, especially in flat areas and with medium-high tree density.

Fig. 1
figure 1

a Resistance surfaces for the study areas and b current landscape connectivity obtained from Omniscape cumulative current map

Full size image

In Italy there is a generally robust connectivity network, particularly in the Northern Apennines and the regions surrounding Garda, Como, and Maggiore lakes (Fig. 1b). In the proximity of the latter (Toce river) the presence of the beaver has been confirmed, and from this area, the Ticino River basin shows favorable connectivity values extending towards the Po basin (Fig. 1b). High connectivity values are also evident in the western Po basin, where the presence of beavers is verified. In Central Italy, an area in which the Eurasian beavers are found, a general high connectivity is found, although exceptions should be made. For instance, the occurrences found in the Tevere basin in the Umbria Region and the ones in the Aterno basin (Fig. 1b).

In the Iberian Peninsula, the highest connectivity values occur in the northwestern regions (Fig. 1b). In the Ebro basin, where beavers have been present since 2003, connectivity is generally lower, with higher values in the northern tributaries of the Ebro and the southern tributaries leading towards the Sistema Iberico (Fig. 1b). The recent occurrence from the Douro basin is more connected southwards, possibly through the Huebra river (Fig. 1b).

At the basin level, the Omniscape current mean map for the Iberian Peninsula reveals greater connectivity in the northern areas and the northern side of the Ebro basin, where beavers occur (Supplementary Figure S1 in Online Resource 1). A similar pattern is observed in the Omniscape current mean map for the Italian Peninsula at the basin level (Supplementary Figure S1 in Online Resource 1), highlighting high connectivity values in the Northern Apennines and in the proximity of major northern lakes. For Central Italy, this analysis indicates that most of the occurrences are situated in basins with medium to high levels of connectivity (Supplementary Figure S1 in Online Resource 1).

Connectivity toward current suitable patches

When considering the connectivity derived from Circuitscape in ‘advanced’ mode towards the currently suitable patches, it becomes evident that in Italy, the northern occurrences in Piedmont and Alto-Adige have the potential to make the Eurasian beaver expand southwards through the Piave and Tagliamento rivers, even considering that the presence in the Po River is in close proximity to various suitable patches (Fig. 2). In Central Italy, there are several corridors, with a notably stronger connection originating from the Tevere and Volturno basins (Fig. 2). Furthermore, from the Ombrone and the Aterno basins, there are good connectivity values linked to the same patches well-connected to the Tevere and Volturno basins (Fig. 2). From these projections, it is very likely that these patches could become areas of stable presence in the near future. This pattern of connectivity is highlighted also from the connectivity map obtained from Circuitscape in ‘pairwise’ mode, where high connectivity is found in Central Italy, between the Tevere and the Ombrone basins, and between the Aterno and the Volturno basins (Supplementary Figure S2 in Online Resource 1).

Regarding the Iberian Peninsula, numerous corridors establish connections between the Ebro occurrences and nearby suitable patches in both northern and southern directions (Fig. 2). In addition, the Douro occurrence is well-connected to the suitable eastern patches across the Tormes river, but with the important barrier of the Almendra dam (Fig. 2). From the connectivity map obtained from Circuitscape in ‘pairwise’ mode great connectivity values occur in the Ebro basin, with important corridors toward the west and south, especially in the Esla river, which could potentially link the occurrence in the Douro basin (Supplementary Figure S2 in Online Resource 1).

Fig. 2
figure 2

Circuitscape cumulative current map using advanced mode for both Peninsulas, with beaver occurrences as sources and suitable patches as destination nodes. In the Iberian Peninsula, the recent occurrence from the Douro basin is well-connected eastward, while the occurrences in the Ebro basin are in proximity with several suitable northern patches. In Italy, higher connectivity values occur in the Tevere basin, and from Volturno and Toce rivers

Full size image

Connectivity assessment within PAs

Among PAs, high connectivity values occur in the Po basin, as for the Alto Garda Bresciano Natural Park (In Fig. 3 indicated as AGB), or in the Val Grande National Park (VG) and neighboring PAs, and in the Northern Apennines, as in the Val Boreca Monte Lesina Special Protection Areas (VB-ML), in the Antola Regional Natural Park (An) and in the Aveto Regional Natural Park (Av) (Fig. 3a). Connectivity remains high across the Apennine range up to the Site of Community Importance (SCI) Forests of the upper Arno basin (UAB). In the Eastern Alps Basin, high connectivity values are found in two SCIs, Rio Bianco di Taipana and Gran Monte (RBT-GM), and in the Middle Tagliamento Valley (MTV). In Central Italy, high connectivity values are found in fewer PAs, such as the Alto Merse and Val di Farma Nature Reserves (AM-VF), or in the lower Valnerina and in the Boschi di Terne Special Protection Areas (LV-BT) (Fig. 3a).

For what concern the Iberian Peninsula, the highest connectivity values are found in some protected areas in the northern area of the peninsula, such as the Urbasa y Andía (UA) or Robledales de Ultzama y Basaburua Natural Park (RUB), the Gorbeia Natural Park (Go) and towards the west through the Redes Natural Park (R) up to Ancares - Courel (A-C) site (Fig. 3a). Furthermore, in the Catalonian basin there are great connectivity values from the Garrotxa Natural Park (Ga) to the Montseny Natural Park (M), and between the Ebro and Douro basins, in the Sierra de la Demanda (SD) natural park there are good connectivity values (Fig. 3a).

At the basin scale, higher connectivity values occur in the Arno in Italy, and in the Cantabrian-Miño and Mondego basins in the Iberian Peninsula (Fig. 3b). The lowest connectivity values in the Andalusian basin are mainly due to climatic factors: this basin hosts the hottest and driest conditions in the Iberian Peninsula. Overall, PAs have slightly higher levels of connectivity when compared to the basin itself, especially in the Northern Apennines and the Volturno basin in Italy, while in the Iberian Peninsula PAs have higher values only in the Catalonian and Cantabrian-Miño basins (Fig. 3b). Therefore, as regards the state of connectivity, the PAs of the Iberian Peninsula in general have lower values when compared to their catchment area (Fig. 3a).

Fig. 3
figure 3

Omniscape connectivity values a within PAs for both peninsulas, and b mean connectivity values for Italian and Iberian Peninsulas at basin level, and only for the PAs occurring in each basin

Full size image

Riverconn connectivity assessment

In Fig. 4, we consider four hydrographic basins within the Italian Peninsula, along with the two basins where beaver presence has been reported in the Iberian Peninsula, introducing the results of the IIC and PCI indices only for the sake of brevity. Indeed, all the calculated indices for each individual basin can be found in Online Resource 1 (Figs. S4-S11), accompanied by prioritization rankings for reaches and dams.

In the Iberian Peninsula, the Ebro basin shows a high concentration of dams, contributing significantly to fragmentation in both the northern and southern tributaries, particularly when considering the PCI index (Fig. 4a). Conversely, the primary segment of the Ebro demonstrates higher fragmentation according to both indices, even though the prioritization of reaches indicates good connectivity (Fig. 4a). It is important to note that most of the northern dams with high fragmentation due to PCI index in the Ebro basin are near some PAs (Fig. 4a). In the Douro basin, higher connectivity values at the reach level are observed mainly on the eastern side, whereas the other rivers exhibit more significant fragmentation (Fig. 4a). Despite a lower abundance of dams, they still contribute significantly to the fragmentation of the Douro, especially in the southern portion of the basin (Fig. 4a).

Turning to the Italian Peninsula, the prioritization of dams reveals that all four considered basins (ordered from north to south: Arno, Ombrone, Tevere, and Aterno) have at least one dam contributing to fragmentation when both indices are considered, with the Tevere basin being the most impacted one (Fig. 4b). The dam’s PCI index shows higher fragmentation, while the IIC indices reports the fragmentation only in the Tevere and Ombrone basins (Fig. 4b). When prioritizing reaches in Italy, the central sections of the Arno and Ombrone rivers, as well as the southern section of the Aterno river, exhibit high connectivity when both indices are considered, while the Tevere basin displays greater fragmentation, especially in its tributaries (Fig. 4b). Additionally, in the southern Tevere basin, connectivity is hindered in the Aniene river and the Salto river, the latter being due to the presence of the Salto dam (Fig. 4b).

Fig. 4
figure 4

Habitat and barriers prioritization using ‘riverconn’ package: a for the basin in the Iberian Peninsula where the Eurasian beaver occurs, and b for the basins in Central Italy where the species is recently expanding. Classification was made using PCI (for functional connectivity) and IIC (for structural connectivity) indices. The Iberian Peninsulas appear as the more impacted, considering the greater number of barriers, especially in the Ebro basin. In Italy, important barriers for the beaver occur mainly in the Tevere basin

Full size image

Discussion

In this study, we explored the functional and structural connectivity of the Eurasian beaver in the Italian and the Iberian Peninsula, two areas where the species is expanding, both through natural recolonization and through illegal reintroductions (Mori et al. 2021; Pucci et al. 2021; Calderón et al. 2022; Capobianco et al. 2023). After being close to extinction in the last centuries, the species is recovering and recolonizing much of its formal range (Halley et al. 2021). Moreover, it is probable that this expansion will continue in the future. A prior study utilizing species distribution models at the European scale has shown that several suitable areas in Southern Europe remain unoccupied (Serva et al. 2023a). Our connectivity models, together with predicted suitable areas (Serva et al. 2023a), highlight areas where the beaver may undergo further expansion in the near future, allowing management actions to be focused on these. The EU biodiversity strategy aims to reconnect 25,000 km of European rivers by 2030. However, the alterations to river flow resulting from historical human activities, primarily due to the presence of barriers, especially dams, pose a significant challenge, with an estimated 0.74 barriers per kilometer (Belletti et al. 2020). Therefore, it is important to know what barrier management actions focus on to restore river connectivity. Our results on barrier prioritization allowed us to identify the ones with the most severe effects on the Eurasian beaver.

For both study areas, higher connectivity is detected for areas with high tree cover, broadleaved forests, and flat terrains, including valleys within the Apennine range, the Cantabrian Mountains, and the Basque range (Fig. 1b).

In Italy, the Omniscape cumulative current map reveals elevated current values along the Apennines, particularly in the northern segment, and in the vicinity of the larger lakes located in the northernmost part of the study area (Fig. 1b). Some well-connected patches emerge in the Po basin, primarily in its western reaches (Fig. 1b). Despite this good connectivity, a natural expansion from the northern occurrences (Alto Adige and Friuli-Venezia-Giulia Regions) to the central ones, seems not viable, even considering the dispersing capacity of the beaver and the long-distance traveled per year, up to 70–80 km (Fustec et al. 2001; Smeraldo et al. 2017), capable to swim for 20 km at night (Nolet and Rosell 1994). Furthermore, the considerable distance between the central Italian occurrences could also be linked to the Eurasian beaver’s spatially discontinuous dispersal pattern, which tends to favor optimal sites further from less suitable ones located closer by (Nolet and Rosell 1994; John and Kostkan 2009).

Considering the occurrences in Central Italy, both Omniscape cumulative current and Circuitscape pairwise current show that these occurrences are well-connected, and the barriers present (i.e., man-made dams) do not seem to significantly impede the corridor’s accessibility (Figs. 1b and 2, Supplementary Figure S2 in Online Resource 1). It is possible that for the Eurasian beaver, some artificial dams could be a tangible threat to movements (Nolet and Rosell 1998). Although, in some cases, dams occur near the springs, where the steep slopes made the habitat unsuitable for the beaver, thus not influencing the connectivity, as also found from Leoncini et al. (2023). For dams located in valleys, there may exist multiple aquatic and terrestrial routes that compensate for their presence, a phenomenon similar to what has been reported for the otters (Lutra lutra) (Leoncini et al., 2023). In the context of possible expansion toward the suitable patches, most of them are in close proximity to the current beaver occurrences (Fig. 2). Consequently, it is highly probable that the species could reach these areas in the near future. Thus, these regions merit special attention.

In the Iberian Peninsula, the Eurasian beaver was reintroduced in 2003 in the Ebro basin (Ceña et al. 2004). In this basin, an expansion of approximately 15 km per year has been calculated, thus considering the new occurrence in the Douro basin of anthropic origin (Calderón et al. 2022). This is supported by the fact that no signs attributable to beaver activity were found between the two sites of presence. Our connectivity models support this scenario, even if the connectivity map obtained from Circuitscape in the ‘pairwise’ mode finds a corridor in the Esla river northward, and in the Douro river eastward (Supplementary Figure S2 in Online Resource 1). Our analysis of habitat connectivity demonstrates that specific parts of the Iberian Peninsula are both connected as well as suitable (Serva et al. 2023a). The higher connectivity values detected from Omniscape in the Iberian Peninsula, such as those in the Cantabrian and Mino basins are mainly due to the higher values in tree cover. From the well-established Ebro basin population, there is a potential southward expansion path toward the Júcar basin (Fig. 1b). Instead, from the recent occurrence in the Douro basin, the beaver could expand southwards to the near Mondego basin, in Portugal, or to the Tajo basin, between Spain and Portugal (Fig. 1b). The greater connectivity towards the northern Iberian Peninsula is accompanied by the presence of suitable patches in proximity to the occurrences from the Ebro basin (Fig. 2). Still, some smaller suitable patches occur also southwards and westwards, being well-connected despite the long distance from the Ebro river. Additionally, some suitable patches are located between the Ebro and Douro occurrences, aligning with the rivers highlighted by Circuitscape (Fig. 2, Supplementary Figure S2 in Online Resource 1).

PAs are of crucial importance for biodiversity conservation (Pacifici et al. 2020; Chen et al. 2022), and it has been estimated that without those, the rate of extinction for vertebrates could have been 20% higher (Hoffmann et al. 2010). Despite in some basins PAs maintain higher connectivity values compared to unprotected areas, in other basins, such as Po and Aterno in Italy, and Douro and Tajo in the Iberian Peninsula, PAs host low connectivity values. Additionally, some PAs are affected by the presence of dams that could hinder structural ecological connectivity. Consequently, management measures should prioritize addressing the impact of these barriers within PAs, given their significance in facilitating biodiversity dispersal movements (Santini et al. 2016).

The analysis of barriers and habitat prioritization conducted through the ‘riverconn’ package highlights the reaches most fragmented and the primary barriers contributing to this fragmentation (Baldan et al. 2022). These metrics can serve as valuable tools in the restoration of river connectivity, pinpointing the specific sites where interventions are necessary. In the Tevere basin, the major barriers are situated in its central part, both to the east and west (Fig. 3b). Interestingly, these barriers, primarily dams, do not appear to significantly impede beaver connectivity (Fig. 1b). However, to enhance river connectivity in the Italian Peninsula, efforts should be focused between the Arno and Ombrone basins, and between the Arno and the upper Tevere basin. Conversely, the Ebro basin exhibits strong structural connectivity values in its main channel, with tree cover playing a more significant role in the northern parts of the basin (Fig. 3a). However, the high structural connectivity at reach level detected in the Ebro, despite the high presence of dams, is linked to the ‘passability’ (a parameter of the ‘riverconn’ package) of the barriers, considering the semi-aquatic ecology of the beaver (Baldan, 2023 pers. comm.). The barriers identified to the north could account for the limited beaver expansion into these suitable areas (Figs. 1b and 3a). Thus, to restore connectivity in the Ebro, specific measures should focus on the northern dams, and to the southern tributaries, while considering the Douro basin, efforts to improve connectivity should primarily focus on the Meseta Central area, where several barriers contribute to its fragmentation.

Recently, in Europe several species are making a comeback (Nolet and Rosell 1998; Chapron et al. 2014; Ledger et al. 2022; Serva et al. 2023b). The return of the Eurasian beaver to Italy and its expansion in the Iberian Peninsula could potentially lead to conflicts with human activities (Auster et al. 2021). These conflicts are primarily due to plantation damage, alteration of artificial channels, and inundations (Taylor and Singleton 2014; Swinnen et al. 2017). A prior study in Italy found that, while local communities, in general, do not perceive the beaver as a threat, some express concerns about its presence due to certain impacts, such as flooding and damage to crops and local species (Viviano et al. 2023). However, we must not forget that the Eurasian beaver is a protected species in most European countries, and its presence is generally associated with positive effects on the ecosystem and biodiversity (Puttock et al. 2017; Brazier et al. 2021), especially in the context of climate change which is negatively impacting on ecosystem services (e.g., De Simone et al. 2020; Iannella et al. 2021b). In fact, beavers can increase local species richness by increasing riparian habitat, and improving the hydrogeological safety of rivers, helping in preventing floods (Rosell et al. 2005; Puttock et al. 2017). Furthermore, beaver activity contributes to creating various habitats used by other species, including taxa of conservation concern (Dalbeck et al. 2007; Nummi et al. 2019). These different habitats contribute to increased environmental heterogeneity. Despite the major positive effects across multiple levels of biodiversity, some negative effects could be observed for some taxa (Rosell et al. 2005; Stringer and Gaywood 2016). Among invertebrates, Trichoptera and Plecoptera could be negatively influenced by the transition to lentic habitat (Stringer and Gaywood 2016), while for amphibians, some salamanders preferring fast-flowing, highly-oxigenated waters could not utilize beaver ponds (Dalbeck et al. 2007). For what concerns mammals, a previous meta-analysis comparing the effects of beaver activity, reported mainly positive effects; though, some negative impacts occur, such as for bats (Stringer and Gaywood 2016). Semiaquatic mammals, like otters (Lutra lutra), benefit from beaver ponds, which provide stable water levels and prey abundance (Rosell et al. 2005). Furthermore, also ungulates generally profit from the presence of beavers due to the greater availability of vegetation (Rosell et al. 2005). Beavers cut down trees and some trees die from flooding events, becoming attractive for woodpeckers (Grover and Baldassarre 1995), and other ecosystem engineers, due to the use of their holes by a range of secondary cavity-nesting species (Robles and Martin 2014). Negative effects have been observed only for some bird species, such as horned grebe (Podiceps auritus, which, however, does not occur in the study areas) and common whitethroat (Curruca communis) (Elmeros et al. 2003; Kuczynski et al. 2012). Especially for the Iberian Peninsula, the Eurasian beaver presence could help in preventing fires, as beaver dams (in this case Castor canandensis) enhance the fire resistance of riparian vegetation (Fairfax and Whittle 2020). In particular, Portugal is among the countries of the European Union most affected by fires, which burned an annual average of 147,000 hectares of land between 2000 and 2018 (San-Miguel-Ayanz et al. 2018).

Therefore, the best strategy is to prevent and limit conflicts with human activities. For instance, given that beaver activity is concentrated near the water’s edge, restoring a small width zone of riparian habitat could be sufficient to limit some damages to plantations (Nolet and Rosell 1998), even if could not be realistic in more urbanized landscapes. Informing local stakeholders about the Eurasian beaver’s presence is another tool that can help maintain local support (Swinnen et al. 2017). Predicting the areas that the Eurasian beaver could colonize is an important tool for monitoring signs of presence and damage, as well as for implementing any necessary management measures.

Though the robustness of both the algorithms used for our analyses and the corresponding high performances, it is important to make some considerations about our results and the datasets used. Considering the scale of the study areas and the accessibility of published data, we relied on the use of the Corine Land Cover map from Copernicus at 100 m precision, buffering some possible forest cover-limitations also including the Tree Cover layer in the dataset preparation. The HydroRIVERS dataset does not contain rivers with less than 10 km2 catchment area and/or river flow less than 0.1 m3/s. However, we did not consider this an important limit considering the target species studied. Similarly, the HydroBASINS dataset is available at a resolution of ~ 500 m. All this considered, studies on smaller study areas, as well as research on peculiar territories, should include finer-scale predictors to obtain more specific outcomes. Also, higher-resolution datasets are urgently needed to improve ecological modeling studies.

We did not incorporate a road layer into the resistance surface, as roadkill is not a significant risk for the Eurasian beaver, with a very low number of individuals being road-killed (Balčiauskas et al. 2022). Furthermore, in some cases, beavers might be attracted to roads, likely due to the presence of herbaceous roadside vegetation (Forman and Alexander 1998; Steyaert et al. 2015; Ciach et al. 2023). The low effect of anthropic environments is also corroborated by other studies that have shown the Eurasian beaver’s tendency not to avoid urban areas, demonstrating its capacity to survive even in urbanized environments (Dewas et al. 2012; Korbelová et al. 2016).

In Italy the colonization process is still in its initial phase, considering that it takes about 30–50 years (Hartman 1995; Halley and Rosell 2002), while in the Iberian Peninsula, where beavers have been present for 20 years, the colonization process has led the species to spread almost only in the Ebro basin, and further expansion is therefore probable.

Conclusion

The ability to move without barriers is a necessary condition for animal species to disperse, reproduce, and survive (Turner et al. 2001; Nathan et al. 2008). Increasingly, ecological connectivity is limited by the presence of human structures that contribute to habitat fragmentation (Haddad et al. 2015). In our study, we analyzed habitat connectivity on multiple levels, for the Eurasian beaver in two areas where the species has been expanding recently. Despite the presence of some barriers, particularly dams, which can negatively affect the beaver, we found good connectivity values, especially in areas associated with good forest cover and flat relief. Furthermore, despite the great distance, the occurrences in Central Italy are still well connected through the hydrographic network. In the near future, it is plausible that there will be a further expansion of the species in these countries, considering the suitable patch available (Serva et al. 2023a). Identifying key barriers in watersheds where beaver presence has been reported can help mitigate the impact of these barriers on beaver expansion. Our results may help predict further expansion of the beaver into the Italian and Iberian peninsulas, helping to understand this recolonization process, regardless of the initial causes of the return of this keystone species.