1. Introduction
Human-induced habitat loss and fragmentation are the largest global threats to biodiversity (Kaszta, Cushman, & Macdonald, 2020; Mohammadi et al., 2018). Habitat loss and fragmentation can impact ecosystems and species by reducing habitat carrying capacity and increasing mortality risk preventing dispersal of individuals, and thus their genes, across landscapes (Kaszta et al., 2020) This synergistically increases the risk of local extinction (Kaszta et al., 2020; Khosravi, Hemami, & Cushman, 2018). Large carnivores are particularly vulnerable to habitat fragmentation and habitat loss (Broekhuis, Cushman, & Elliot, 2017). They live in low densities and typically have large home ranges (Carroll & Miquelle, 2006; Hilty, Brooks, Heaton, & Merenlender, 2006). Carnivores’ large area requirements demand vast and connected habitat areas where they are protected from human persecution. Increasing land use change and habitat fragmentation have threatened carnivore populations by reducing habitat areas and increasing their isolation, leading to a synergy of increased direct human-caused mortality, reduced local carrying capacity, and reduced ability for populations to be integrated by dispersal (Cushman, Elliot, Macdonald, & Loveridge, 2016).
Large carnivores are also particularly vulnerable to vehicle collisions and barrier effects because of their life-history characteristics (Mohammadi & Kaboli, 2016). (They have low population densities, low fecundity, and relatively large home ranges (Grilo, Bissonette, & Santos-Reis, 2009; Mohammadi et al., 2018; Mohammadi & Kaboli, 2016; Parchizadeh et al., 2018). Vehicle collision poses a substantial threat to wildlife species in central Iran (Shahnaseri et al., 2019). Roads, especially those with high traffic, disrupt both structural and functional connectivity for large carnivores, including grey wolf (Canis lupus ), striped hyena (Hyaena hyaena ) and golden jackal (Canis aureus ), and can lead to reduced gene flow among meta-populations (Shahnaseri et al., 2019). Thus, transportation managers need reliable data to identify when and where particular species are susceptible to high road-kill rates to implement mitigation measures during the road design planning and/or exploration stage (Grilo et al., 2009; Santos et al., 2015).
Large carnivore conservation requires both protection of extensive core areas and the establishment of movement corridors among them (Cushman et al., 2018), particularly when core habitat patches are isolated by road networks (McClure, Ware, Carlisle, & Barber, 2017). Connectivity is critical for long-term species conservation and plays a crucial role in maintaining the genetic and demographic processes that ensure long-term viability (Bennett & Saunders, 2010). Connectivity of populations is of paramount importance to both conserve species locally and to secure their range shifts in response to future hazards such as land use change (Cushman, McRae, et al., 2013), and climate change (Karami, Rezaei, Shadloo, & Naderi, 2020; T. Wasserman, Cushman, Shirk, Landguth, & Littell, 2012; T. N. Wasserman, Cushman, Littell, Shirk, & Landguth, 2013). Enhancing connectivity in conservation networks may reduce the negative impacts of habitat loss and fragmentation (Betts et al., 2014).
Connectivity models provide practical tools for assessing potential fragmentation effects of roads on wildlife and help inform management and conservation planning (Almasieh, Rouhi, & Kaboodvandpour, 2019). A wide variety of methods have been proposed for connectivity analysis, including least-cost path modelling (Adriaensen et al., 2003), current flow (McRae, Dickson, Keitt, & Shah, 2008), factorial least-cost path density (Cushman, McKelvey, & Schwartz, 2009), resistant kernels (Compton, McGarigal, Cushman, & Gamble, 2007) and randomized shortest path algorithm (Panzacchi et al., 2016). The factorial least-cost path and cumulative resistant kernel approaches are strong methods to be used in combination to accurately identify core habitats, fracture zones and corridors across a broad landscape (Cushman et al., 2018; Cushman, Lewis, & Landguth, 2014; Moqanaki & Cushman, 2017).
Understanding the different factors that affect species distribution and habitat selection is important for carnivore conservation (Khosravi et al., 2018; Shahnaseri et al., 2019; Mohammadi et al., 2021). Many other habitat suitability models are available. Among which machine-learning models such as random forests (RF) may perform better than the regression-based algorithms (Cushman, Macdonald, Landguth, Malhi, & Macdonald, 2017; Rodriguez-Galiano, Ghimire, Rogan, Chica-Olmo, & Rigol-Sanchez, 2012). Furthermore, ensemble modeling, in which several species distribution models (SDMs) are combined to quantify a range of predictions across more than one set of uncertainty sources, has been found to increase often the accuracy of model predictions (Araújo & New, 2007; Shahnaseri et al., 2019) and decrease the uncertainty associated with using a single SDM (Shirk et al., 2018).
The central region of Iran accommodates a variety of carnivore species. Grey wolf, golden jackal and striped hyena are the most widely distributed carnivores in central Iran. Most conservation efforts for conserving wildlife diversity in Iran have relied on establishing Protected Areas (hereafter, PAs). However, the existing PA network is not efficient for the long-term conservation of most carnivores (Shahnaseri et al., 2019). Due to the reduction in wild prey species density in Iranian PAs (Behdarvand et al., 2014; Mohammadi, Kaboli, & López-Bao, 2017; Mohammadi, Kaboli, Sazatornil, & López-Bao, 2019), occurrence of carnivore species across inhabited rural areas has increased (Mohammadi et al., 2019). Anthropogenic food resources, notably livestock and garbage, also contribute to these carnivores’ diet and incentivize carnivores moving to high-risk locations in the landscape near human habitations (Babrgir, Farhadinia, & Moqanaki, 2017; Behmanesh, Malekian, Hemami, & Fakheran, 2019; Mohammadi et al., 2019).
In this study, we addressed three main objectives regarding grey wolf and golden jackal status and vulnerability in central Iran. First, we determined the most significant environmental and anthropogenic factors influencing habitat suitability for both species. Second, we defined core areas for each species using resistant kernel modeling, and identified corridor routes among these core areas using factorial least-cost path modeling. Third, we used spatial randomization of vehicle collision locations to test the predictive ability of resistant kernel and factorial least-cost path predictions of movement (Cushman et al., 2014). The results provide clarity on the drivers of habitat quality for multiple carnivore species, and the patterns of habitat extent and connectivity for these species across Central Iran which is critical for conservation management planning of carnivores in Iran.