One of the most significant ecological impacts of roads and railways are their barrier effects for terrestrial wildlife (Forman and Alexander 1998; O’Brien 2006; Beckman and Hilty 2010; Barrientos and Borda-de-Água 2017). By obstructing movements and, thereby, restricting the access to resources and the opportunities for migration and dispersal, linear infrastructures may inhibit the individual fitness and genetic diversity of wildlife and negatively impact population demography and conservation status. After the emergence and growth of the applied scientific field of road ecology in the last decades (for example. Forman et al. 2003; van der Ree et al. 2015), the barrier effects for large wildlife, such as ungulates and large carnivores, are now well recognised in countries worldwide (Clevenger and Huijser 2011; Wingard et al. 2014; Georgiadis et al. 2015, 2018; Collinson and Patterson-Abrolat 2016; van der Grift et al. 2018; Hlaváč et al. 2019). Accordingly, transport agencies increasingly construct adapted culverts, tunnels and vegetated bridges to provide wildlife with safe opportunities to cross major roads and railways (Iuell et al. 2003; Clevenger and Ford 2010; Rijkswaterstaat 2011; Smith et al. 2015).

Monitoring of over- and underpasses for large wildlife has provided frequent proof that they are used by a variety of species (van der Ree et al. 2007; Smith et al. 2015). In general terms, larger (wider, higher) constructions are used by larger species, by a broader array of taxa and by a larger proportion of target populations (Rodriguez et al. 1996; Clevenger and Waltho 2000; Bhardwaj et al. 2020), although other aspects of their design may affect the frequency of use, such as human disturbances, occurrence of vegetation and cover and siting in relation to preferred habitats (Clevenger and Waltho 2000, 2005; Ascensão and Mira 2007; Glista et al. 2009; van der Ree and van der Grift 2015; Andis et al. 2017).

Despite having recognised both the problem with barrier effects and its potential solution, in infrastructure planning practice, many transport agencies still seem to consider crossing structures for wildlife to entail external or unexpected costs. Accordingly, such constructions have to be argued for on a case-by-case basis and often end up being rather few. In response, environmental planners tend to advocate for as large wildlife crossing structures as possible and put much effort into finding the ideal locations for those crucial constructions. This situation is reflected not least in current European guidelines for mitigation of barrier effects at transport infrastructures; many of these have their focus on methods to identify major wildlife corridors and state ideal rather than optimal dimensions of crossing structures (Iuell et al. 2003; Alterra 2008; Jędrzejewski et al. 2009; Nowak et al. 2010; Vejdirektoratet 2011; Statens Vegvesen 2014; Ciabò et al. 2015; Reck et al. 2018; Hlaváč et al. 2019).

While crossing structures may be necessary measures to safeguard the connectivity for wildlife across large linear infrastructures, they inevitably create bottlenecks for animal movements, irrespective of location and size. Funnelling animals from larger areas into movement bottlenecks may have a number of ecological disadvantages, for example, increased predation (Little et al. 2002; Mata et al. 2015) or exaggerated social interactions between animals. Moreover, the concept of natural movement corridors has been criticised for lacking solid theoretical and empirical foundations (Simberloff et al. 1992) and that its frequent application in land-use planning satisfies political and economic interests rather than ecological requirements (Van Der Windt and Swart 2008; Shilling 2020). For large terrestrial wildlife, well-defined, predictable migratory paths do occur in some populations (Andersen 1991; Berger et al. 2006; Kauffman et al. 2018), but seem to be the exception rather than the rule to how animals move between areas.

The size is one of the most cost-driving factors for crossing structures and, in the infrastructure planning reality, the cost-effectiveness of measures has to be considered. Wildlife crossing structures, from culverts to viaducts and green bridges, may range in investment cost by orders of magnitude (Sijtsma et al. 2020; see also Fig. 1) and considerable savings can be made if the optimal trade-off is found between number and size of crossing structures with the aim of reaching the maximum infrastructure permeability for wildlife. While some guidelines for wildlife measures at transport infrastructures do acknowledge that a large number of narrow wildlife crossings may be more effective than a single, wide one (Iuell et al. 2003; Jakobi and Adelsköld 2011; Reck et al. 2018), the required cost-benefit analyses are rarely conducted (Sijtsma et al. 2020).

Figure 1.

Examples of differently sized crossing structures in Sweden used by large wildlife, with rough estimates of investment costs. The precise costs depend on a number of site-specific factors, and values given are intended to serve as indications. Images by courtesy of Trafikverket and PEAB.

The question of size vs. number of wildlife crossing structures is analogous with that of the so-called SLOSS dilemma in conservation, i.e. the question whether a Single Large Or Several Small protected areas would be more effective for species conservation (Diamond 1975; Simberloff and Abele 1976). That question remains a dilemma as it has no universal answer; the best strategy depends on, inter alia, to what extent the smaller areas share species, on the environmental variability in and amongst areas and on the distance between areas (Simberloff and Abele 1976; Akcakaya and Ginzburg 1991). The SLOSS dilemma of road ecology – the trade-off between single large or several small crossing structures (Karlson et al. 2017) – is likely to share many characteristics with that of protected area designation.

The issue of SLOSS wildlife crossing structures has previously been addressed by Karlson et al. (2017), using a theoretical approach to compare the outcome in model landscapes with different levels of habitat contrast and aggregation. They concluded that in homogenous (low-contrast, low-aggregation) landscapes, a number of smaller crossing structures are better than one large, given that each still meets minimum ecological design criteria. This conclusion derived simply from geometry; with passage opportunities evenly distributed along an infrastructure, the distance to a crossing structure from an average point in the landscape will be shorter. In heterogeneous landscapes, on the other hand, the outcome will depend on the habitat quality in and around the crossing structures; fewer animals would cross through a structure located in low quality habitat. Accordingly, in heterogeneous landscapes, more care must be taken to the location of crossing structures in relation to the habitat requirements of target species.

In this paper, I develop the SLOSS dilemma of road ecology using the case of crossing structures for large wildlife in Sweden. Based on ecological and pragmatical arguments, I list a number of situations where it may be more beneficial to distribute the conservation efforts in the landscape by constructing several small crossing structures rather than one or a few large. I argue that the situations described for Sweden are not unique, but may apply to other taxa and geographical regions. I conclude by suggesting how the SLOSS discussion could provide information for planning of wildlife mitigation at linear infrastructures and by proposing some directions for future research in the field.

Populations of many large mammals are currently relatively strong in Sweden and species such as moose (Alces alces), deer (red deer Cervus elaphus, fallow deer Dama dama, roe deer Capreolus capreolus), wild boar (Sus scrofa) and large carnivores (wolf Canis lupus, bear Ursus arctos, lynx Lynx lynx) range over large parts of the country (Bergström and Danell 2008; Liberg et al. 2010; Chapron et al. 2014). Natural or semi-natural habitats, such as managed forest, wetland or mountain make up some 80% of the Swedish land area (Gerell et al. 1996). While most large mammals do show some preferences for forested areas, they also use agricultural land and built-up areas, particularly in night-time when the human disturbance is low (Winsa 2008; Godvik et al. 2009; Milleret et al. 2018; Fattebert et al. 2019; Richter et al. 2020) or during seasons with available crop (Thurfjell et al. 2009; Olsson et al. 2011). Accordingly, these species tend to occur in most habitats and most landscapes and their movements are less likely to be strongly funnelled to specific habitat corridors. One exception may be seasonally migratory ungulates in the north (primarily moose and semi-domestic reindeer Rangifer tarandus), which follow routes along river valleys and other topographic landscape elements that may be maintained between generations or even decades (Sweanor and Sandegren 1988; Andersen 1991; Singh et al. 2012; Lindberg 2013; St John et al. 2016).

Within the managed boreal forest, ungulates may prefer certain stand types, for example, clear-cuts, young or dense forest stands and linear landscape elements, such as riparian areas and edge zones (Winsa 2008; Thurfjell et al. 2009; Bjørneraas et al. 2011). However, the spatial distribution of forest stands is likely to change over decades, i.e. within the expected lifespan of a bridge or culvert, due to forest growth or management activities. Additionally, in less intensively managed landscapes, habitats are expected to undergo changes due to natural disturbances, succession or climate change, with potential change in animal movement patterns over time as a result.

Animal movements may also change due to sudden human influences in the surrounding landscape, such as new housing, mining or industry and increased outdoor recreation adjacent to crossing structures (Singh et al. 2012). While such developments should be addressed in landscape level physical plans and environmental impact assessment (Clevenger and Ford 2010; Ryegård and Åkerskog 2020), not all can be foreseen during the planning stage of fauna mitigation schemes. Moreover, transport agencies have limited authority over the land use outside the road or railway right-of-way, so the long-term functionality of a wildlife crossing structure depends on the compliance of surrounding landowners and land users.

Extensive site-specific empirical data on wildlife movements are in short supply, in Sweden as in other countries (Clevenger and Ford 2010; Helldin and Souropetsis 2017). Identification of movement corridors – which is often required in the planning practice – has to rely on the distribution of natural or wildlife habitat, wildlife accident data or expert opinion (van der Grift and Pouwels 2006; Clevenger and Ford 2010; Olsson et al. 2019). However, such indirect approaches have their flaws (Clevenger and Ford 2010; Helldin and Souropetsis 2017; Sjölund et al. 2020) and the true spatial distribution of wildlife movements remain obscure, with few and localised exceptions.

Some Scandinavian mammals are territorial, amongst these being roe deer and large carnivores (Linnell and Andersen 1998; Mattisson et al. 2011) and may, therefore, expel other individuals of the same species and gender from a crossing structure. Similarly, interspecific competition occurs frequently amongst ungulates (Latham et al. 1997; Feretti 2011; Pfeffer 2021; La Morgia et al. in review) and amongst carnivores (Mattisson et al. 2011), which may lead to a dominant species effectively expelling subdominants. Although such “ecological plugs” are probably only partial, they could inhibit the movement of subdominant individuals or species through a crossing structure.

In addition, game and prey species, such as ungulates, may adapt their spatial distribution, habitat choice and activity patterns to the risk of being hunted or predated (Cromsigt et al. 2013; Lone et al. 2014, 2015; Zbyryt et al. 2018). Similarly, hunting and poaching are main causes of mortality for large carnivores in Scandinavia (Andrén et al. 2006; Liberg et al. 2012) and, consequently, these species avoid human interaction (Ordíz et al. 2011; Carricondo-Sanchez et al. 2020). Hunting in the direct vicinity of over- or underpasses occurs in Sweden (own observations), but how frequent this happens is not known. Incidents of natural predation on ungulates near wildlife crossing structures have been reported, but appear to be rare (Little et al. 2002; Plaschke et al. 2021). Yet, only the presence of ambushing predators or hunters in the area may temporarily inhibit the structure´s effectiveness for target species (Mata et al. 2015).

This Swedish case of large wildlife ecology describes a number of situations that – each individually and all taken together – suggest that distributing conservation efforts on several small crossing structures may perform better than a single large crossing, namely:

• In relatively intact or homogenous landscapes, where animal movements are dispersed.

• Where animal movement routes are expected to gradually change over time due to landscape changes.

• Where future human development cannot be controlled and natural habitats surrounding crossing structures may suddenly deteriorate.

• Where animal movement habits simply are not known.

• When wildlife mitigation needs to target multiple species with different habitat choices and no ideal site can be appointed.

• When target species are territorial or competitors and there is a risk that some individuals or species monopolise the area in and around the crossing structure.

• When target species are sensitive to hunting, poaching or predation and enemies (human or natural predators) may ambush at sites where movements of prey are pinched.

The Swedish Transport Administration (STA), the responsible manager for the public road and railway network in Sweden, currently works along a strategy for landscape connectivity for large wildlife that partly take a SLOSS approach. According to the national ecological standards (Trafikverket 2019), safe passageways for large mammals (ungulates and large carnivores) should be provided at a maximum distance of 6 km along all major roads and railways; a requirement based on the assumption that large mammal movements are ubiquitous and dispersed or at least ought to be so. Via supporting documents (Seiler et al. 2015 and references therein), the standards point out moose and roe deer as focal species (sensu Lambeck 1997); moose, in particular, because it is supposedly one of the most demanding large mammal species in Sweden when it comes to crossing structure design and one of the most problematic when it comes to wildlife-vehicle accidents and barrier effects.

The standards describe a range of larger to smaller crossing structures as suitable for moose and roe deer (Seiler et al. 2015; Trafikverket 2021a, 2021b) and it also takes into account the predicted wildlife connectivity provided by bridges constructed for other purposes, for example, watercourses, trails and low-traffic roads (Seiler et al. 2015). Accordingly, the standards provide a framework allowing, but not requiring, that trade-offs are made between functionality and number of crossing structures on the level of a longer road section or an infrastructure network.

Due to the lack of an explicit SLOSS approach in the planning for large wildlife mitigation, opportunities for better ecological function and more cost-effective mitigation measures may still be missed. For example, regional differences in data availability, plasticity in animal movements or target species for mitigation would imply different output depending on the region. In northern Sweden, investing in few large crossing structures at major migration routes may be warranted. Thorough ecological data should be collected and compiled to identify the ideal sites for these crossing structures and considerable efforts should be made to secure their long-term effectiveness through adapted management of the surrounding landscape. In more southern parts of the country, however, sufficient overall permeability of infrastructures may be achieved by several smaller crossing structures, including non-wildlife bridges which tend to be plentiful along most major roads and railways.

Though based on the specific case of Swedish large wildlife, I believe that many of the situations described above are what infrastructure planning often faces. Site-specific knowledge of animal movement patterns tends to be sparse (Clevenger and Ford 2010) and, in many biomes, it is likely that movement routes will change over time due to natural landscape dynamics or anthropogenic impacts. With mitigation schemes targeted to multiple wildlife species, it will be difficult to find the perfect site for a crossing structure and target species are likely to interact at the site. In these cases, the connectivity delivered by each individual crossing structure cannot be guaranteed and distributing investments over several structures would convey a risk diversification. Moreover, this is not only an economical or practical consideration; transport agencies should aim at allowing dispersed or flexible animal movements wherever they occur and avoid the ecological predicaments that pinched animal movements may entail. In principle, these aspects could apply similarly to other animal taxa that are frequent targets for crossing structures at roads and railways, such as medium-sized mammals and amphibians (Iuell et al. 2003; Langton 2015).

Following this line of argument and with support from the results from the modelling approach adopted by Karlson et al. (2017), I suggest a default strategy for transport agencies to construct several small crossing structures rather than concentrating the passage opportunities along major transport infrastructures to a single large structure. What to choose should, however, depend on the context: for example, the degree of habitat heterogeneity (aggregation and contrast), habitat predictability, the dimension requirements of target species and the spatial overlap between species (Mata et al. 2005; Karlson et al. 2017). Single large structures may be selected at sites where it is likely that the crossing structure can serve a large proportion of target animals (species and individuals), for example, where animal movements follow distinct routes and where target species have a large overlap in habitat requirements and little social or trophic interference. However, going for a single large structure should require that the long-term functionality of the crossing structure could be guaranteed, for example, in areas that are legally protected or when solid agreements can be made with adjacent land-users to protect the crossing structure and its surroundings from significant impacts. There may be situations where an intermediate or mixed (single large combined with several small) approach may be the best choice.

A planning strategy aiming at several smaller crossing structures rather than a single large structure could facilitate the planning of wildlife mitigation in a few ways. It may not be necessary to put as much effort into finding the best siting or design of each crossing structure, which may save both time and costs at early planning stages. Instead crossing structures may have a standard design and be spaced out on pre-defined intervals along the infrastructure or where the ground conditions (topography and soil) are ideal from a technical perspective. Non-wildlife bridges or culverts used by wildlife may also be included in the wildlife mitigation plan. While the goal of wildlife mitigation plans should not be to save money, but to minimise wildlife-traffic conflicts, the SLOSS issue will open the question of how to get the most out of available investments or how to reach conservation goals with a minimum of cost and it may, therefore, help the matter by redirecting the focus in planning from costs to savings.

Trade-offs between size and number of crossing structures in wildlife mitigation schemes may require that road ecology research take a somewhat different angle than that currently prevailing. Research and monitoring of over- and underpasses during the last decades have provided a basic understanding of how well different type of structures correspond to the demands of different species or taxa (Jędrzejewski et al. 2009; Clevenger and Ford 2010; Smith et al. 2015), but comprehensive comparisons of structures of different size and design are still few (but see Clevenger and Waltho 2005; Mata et al. 2005; Taylor and Goldingay 2010; Cramer 2012; Bhardwaj et al. 2020; Sijtsma et al. 2020). Moreover, the costs for the constructions, including any costs for planning, traffic diversion during construction, long-term maintenance etc., are rarely integrated into the analyses (Sijtsma et al. 2020). Seiler et al. (2016) and Sijtsma et al. (2020) point out some directions for how cost-effectiveness analyses of wildlife crossing structures can be set up, but the field needs to be further explored. Monitoring of wildlife-use of crossing structures should be conducted following a standardised protocol to be able to make a just comparison of the performance of a range of crossing structures and to be able to add new monitoring results over time to a global analysis (Helldin and Olsson 2015).

A strategy to construct several small crossing structures should entail an increased demand for research on how to make also narrower crossing structures more functional for wildlife, for example, by adapting vegetation and limiting human disturbance. However, squeezing down the size of crossing structures would also mean approaching a lower limit for functionality and, in the light of this, a much better understanding of the ecology of narrow crossing structures is needed.

I suggest a stronger emphasis in monitoring of crossing structures, not only on how different species use them differentially (such as described by, for example, Cramer 2012; Mata et al. 2015), but also differences between animal categories within species, for example, between sexes and ages, individuals of different status or condition and individuals in groups of different size and composition. It is likely that different animal individuals or categories show differences in vigilance and sensitivity to disturbance (Liley and Creel 2008) and crossing structures that deter certain categories of animals are less likely to provide functional connectivity for the population, irrespective of the absolute number of individuals using the structure.

To this, we need better knowledge of what happens between animals at crossing structures, for example, predation risk (real and perceived), interference competition, territoriality, dominance and other antagonistic types of behaviour that can expel some target animals from the sites. The well-developed methods, using camera traps and video surveillance of crossing structures, provide opportunities for studying both animal categories and types of behaviour to a larger extent than is currently done.

Finally, I call for more efforts in wildlife ecology research to develop the knowledge of animal movements, to specifically address the questions posed by road and railway planning, of movement routes (importance of certain routes, their stability over time and reliable methods to map them) and potential movement distances along fences to find safe passages (Bissonette and Adair 2008). While this has been studied for some large and charismatic species (e.g. moose in Sweden), these aspects are largely unknown for most species, including important target species for wildlife crossing structures.