No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Spatiotemporal variation in resource selection: Insights from the American marten (Martes americana)
Ecological Applications
Abstract
Behavioral and genetic adaptations to spatiotemporal variation in habitat conditions allow species to maximize their biogeographic range and persist over time in dynamic environments. An understanding of these local adaptations can be used to guide management and conservation of populations over broad extents encompassing diverse habitats. This understanding is often achieved by identifying covariates related to species' occurrence in multiple independent studies conducted in relevant habitats and seasons. However, synthesis across studies is made difficult by differences in the model covariates evaluated and analytical frameworks employed. Furthermore, inferences may be confounded by spatiotemporal variation in which habitat attributes are limiting to the species' ecological requirements. In this study, we sought to quantify spatiotemporal variation in resource selection by the American marten (Martes americana) in forest ecosystems of the Pacific Northwest, USA. We developed resource selection functions for both summer and winter based on occurrence data collected in mesic and xeric forest habitats. Use of a consistent analytical framework facilitated comparisons. Habitat attributes predicting marten occurrence differed strongly between the two study areas, but not between seasons. Moreover, the spatial scale over which covariates were calculated greatly influenced their predictive power. In the mesic environment, marten resource selection was strongly tied to riparian habitats, whereas in the xeric environment, marten responded primarily to canopy cover and forest fragmentation. These differences in covariates associated with marten occurrence reflect differences in which factors were limiting to marten ecology in each study area, as well as local adaptations to habitat variability. Our results highlight the benefit of controlled metareplication studies in which analyses of multiple study areas and seasons at varying spatial scales are integrated into a single framework.
... Hegel et al. 2010). This is unlikely to be the case given that different limiting factors can affect the occurrence of species in different locations of its range, different times, and in different contexts of the surrounding landscape (e.g., Cushman and Lewis 2010;Cushman et al. 2013;Short Bull et al. 2011;Shirk et al. 2014). More recently, research has shown that factors such as spatial scale (Cushman and McGarigal, 2003;Grand et al. 2004;Graf et al., 2005, De Knegt et al., 2010, McGarigal et al., 2016, Wan et al. 2019, spatial accuracy (Hanspach et al., 2011;A.B. ...
... Many studies in the literature ignore or improperly account for nonstationarity and invdividual heterogeneity, yet the consequences can be severe (Muff et al. 2019). With the growing recognition of the importance of nonstationary to factors as diverse as gene flow (Short Bull et al. 2011), movement (Cushman et al. 2013;Kaszta et al. 2019, Kaszta et al. 2021, habitat use (Wan et al. 2019;Shirk et al. 2014;Jones et al. 2023) and responses to climate changes (e.g. A.B. Smith et al., 2019) there have been a few attempts to evaluate nonstationarity in wildlife studies. ...
... Importantly, the classical Gaussian hypervolume assumes no difference in a species responses in different contexts of limiting factors. However, numerous studies have found that species environmental relationships change in nonstationary and nonlinear ways when conditions change or when particular resources become limiting (e.g., Cushman et al. 2011Cushman et al. , 2013Short Bull et al. 2011;Shirk et al. 2014;Vergara et al. 2017). Thus, the high heterogeneity we observed in our analysis among the response curves of different individual wildcat hybrids may reflect spatial nonstationarity in environmental conditions leading to geographically varying limiting factors that interact nonlinearly. ...
The ecological niche and the species-environment relationship are both cornerstones of contemporary ecological science. The realized habitat niche defines the conditions in which a species occurs, is adapted and can thrive, and quantification of the species-environment relationship is a means to describe the realized habitat niche. A frequent, if unspoken, assumption in analyses of species-environment relationships and the ecological niche is that there is a common, stationary and stable relationship between a species and its environment. This implies, additionally, that this relationship applies to the species as a whole, or rather to all individuals of the species. However, another cornerstone of ecological science is that populations, and even individuals, differ in their genetic characteristics and the environmental influences that shape their behavior. Therefore, the species-environment relationship and the realized habitat niche are likely to vary intra-specifically. Uniformity in behaviour under different ecological circumstances or genetic homogeneity in response to spatially varying limiting factors are assumptions that should be investigated and tested. In this paper, using European wildcat (Felis silvestris) x domestic cat (F.catus) hybrids in Scotland as a policy-relevant exemplar, we explore what ecological modellers call nonstationary habitat use, and what field ecologists call individual variation, or, with an evolutionary perspective, intra-specific variation. We analyze the occurrence patterns and ecological response curves of 14 individual wildcat hybrids distributed across Scotland to assess how much individual variation there is in expressed patterns of habitat association across multiple environmental variables. We propose three conceptual models corresponding to three divergent patterns of habitat association for the sampled population: stationary generalist, stationary specialist, and nonstationary specialist. Each of these alternative hypotheses of habitat selection for wildcat hybrids had unique expectations for the shape and overlap of response curves along environmental variables, and for the degree of overlap between used and available habitat among individuals. We were able to show a high degree of individual heterogeneity and specialization across our small but geographically widespread sample. Our results support the hypothesis that wildcat hybrids in Scotland are nonstationary habitat specialists. That is, the habitat associations of wildcat hybrids are highly heterogeneous at an individual level, and that pooled analyses across individuals fails to completely represent the range or variation of individual responses, and also fails to represent the actual habitat selection response curves of any individual. This provides a compelling example of the highly variable and heterogenous nature of habitat association within a single species. Our results support other recent studies where species demonstrate local adaptations to available conditions.
... We also searched for spatially dynamic parameters with meta-replicated study areas, studies in which landscape analyses were conducted across different landscape replicates (McGarigal et al., 2016). Meta-replication can be important to address spatial heterogeneity, context-dependence, and nonstationary limiting factors (Shirk et al., 2023(Shirk et al., , 2014Vergara et al., 2016;Short Bull et al., 2011;Unnithan Kumar et al., 2022). ...
... The existence of a response to a fine or broad scale for a particular variable did not eliminate the possibility of importance of another variable at a different scale for the same species (Gallo et al., 2018;Chambers et al., 2016). Indeed, one of the main messages to emerge from recent multi-scale habitat modeling research is that a given species often selects different resources each at unique scales (Shirk et al., 2023(Shirk et al., , 2014Wasserman et al., 2010;Wan et al., 2017). Despite the apparent mobility of volant species, fine spatial scales (<100 m) were dominant in a number of studies (e.g., Pauwels et al., 2019;O'Keefe et al., 2013;Bunkley et al., 2015). ...
... Methods used for meta-replicated study areas (sensu Shirk et al., 2023;Shirk et al., 2014) were diverse but help illustrate the need to examine species and their interactions to landscape configurations such as varying areas of forest or distance from hedgerows (Pardini et al., 2009;Lacoeuilhe et al., 2016). A meta-replicated study in Panama and Brazil found that species richness in forested land patches experienced lower loss of species when compared to actual islands in a lake with similar forest. ...
Landscape-scale analysis is an evolving approach to quantify the effects of landscape structure (composition and configuration) on focal species. Bats—a remarkably rich and diverse group—use habitat from fine (< 0.5 km) to broad (> 4 km) scales, and thus identifying their responses to changing landscapes can highlight a variety of management implications. We conducted a literature review of >170 peer-reviewed studies from five continents of landscape-scale studies in bats. We used cluster analysis to highlight study trends and identify biases and knowledge gaps in landscape-scale studies of bats. Species in the families Vespertilionidae and Phyllostomidae, which represent 51 % of extant bat diversity, were the focus of two thirds of studies; other families were underrepresented. Tropical and subtropical Africa and Asia, notable for their high bat species richness, were underrepresented
in studies. Although considered by few studies, context-dependent demographic data, including temporal and behavioral parameters (e.g., age, season) were important for explaining bat-landscape interactions. No one-size-fits-all set of variables or scales exists for bats, and even closely related species vary in their responses
to variable-scale combinations. However, variables that quantify habitat size and presence of patch edges were often strong predictors of bat use. Based on this review, researchers should consider a range of scales including broad scales (>5 km), landscape and bioclimatic variables, and archiving data for future studies across temporal scales. We provide a list of recommendations that can help researchers and conservationists improve inferences in determining the landscape associations of bats species and other taxa.
... Implicitly, these approaches assume a stationary realized niche that is common to all individuals of the same species However, intra-specific variation in behavioural ecology is common in nature, especially amongst Carnivora (e.g. Macdonald, 1983), and recent research has illustrated that species-environment relationships and the realized habitat niches that they attempt to measure are often highly variable through space and time (e.g., Cushman 2010; Vergara et al. 2017;Shirk et al. 2014;Kaszta et al. 2019). This intra-specific variability in habitat selection often reflects differing patterns of resource availability and mortality (e.g. ...
... We analyzed 15 variables identified as being important (Cushman et al., submitted c) at seven scales (50 m, 100 m, 200 m, 400 m, 800 m, 1600 and 3200 m) and used optimization to select the best scale for each variable in individual wildcat hybrid models, as suggested by McGarigal et al. (2016). Many studies have found that habitat selection is a highly scale-dependent phenomenon, with different habitat features being selected at different scales (e.g., Grand et al. 2004, Shirk et al. 2014, Wasserman et al. 2012, McGarigal et al. 2016, Hearn et al. 2018, 2020. Thus, we selected the most effective scale for each variable in predicting used vs. available locations for each individual wildcat hybrid model, based on highest MIR. ...
... Vergara et al. (2017) conducted a similar meta-replicated study of stone marten landscape genetics in the Iberian Peninsula, and explored how differences in predictions were associated with different ecological contexts and limiting factors. Shirk et al. (2014) conducted a meta-replicated study of American marten habitat selection in two disjunct study areas and found dramatic differences in the apparent association of marten occurrences with environmental variables and spatial scales in the two areas that were possibly related to spatially varying limiting factors. Our study is consistent with these and takes the additional step of a larger sample of meta-replicated models across 14 individuals while also comparing two different widely used species distribution modeling approaches. ...
Little is known about the factors that drive nonstationarity and inter-individual differences in realized habitat niches and species-environment relationships. We explored this topic by developing individual habitat selection models for 14 wildcat hybrids distributed across Scotland, and assessed how differences in their predicted probabilities of occurrence were related to factors including (1) geographic distance, (2) multivariate ecological
distance, (3) difference in degree of hybridization and (4) difference in sex (male vs female). We found that the individual models were exceptionally effective in predicting the habitat use and occurrence of the particular individuals on whose data they were trained, but were generally highly divergent and not transferable among individuals. We conducted a reciprocal validation approach where we calculated the AUC for each individual model, predicting the occurrence patterns of the 13 other individuals. We then fit regression and nonparametric
splines to evaluate the impacts of geographical distance, ecological distance, hybridization distance and difference
in the sex of individuals in the ability of individual wildcat hybrid habitat models to predict the occurrences of other individuals. We found that, of the four factors assessed, ecological distance was supported as being inversely related to ability of a model from one individual to predict occurrence of another individual. The other
three factors were not strongly related to differences in reciprocal model predictive ability. This suggests that ecological differences where individual wildcat hybrids reside drive differences in their habitat selection, but that geographical distance, degree of genetic hybridization and difference in the sex of individuals are not consistently associated with differences in model prediction or reciprocal validation performance. These results highlight the effect of ecological limiting factors, and the importance of nonstationary limiting factors in determining the habitat they select, their expressed species-environment relationship and the description of their realized habitat niches.
... We also predicted that genet latrines would be predominantly found in riverine corridors (Prediction 3) (Palomares 1993;Espírito-Santo et al. 2007). They have been described as natural highways for many species (Shirk et al. 2014;Carvalho et al. 2016a), but also boast a high amount of food resources and micro-habitats favourable for resting (Buesching and Jordan 2022). Finally, we also anticipated genet latrines to respond to important scale-independent elements in the landscape, such as the proximity of (i) other latrines to allow marking even if some are destroyed (Prediction 4a) (Espírito-Santo et al. 2007;Buesching and Jordan 2022); (ii) dirt roads where marking behaviour along their edges has been recorded (Carvalho et al. 2011), which may increase the chances of scent-mark detection due to frequent animal passages (Prediction 4b) (Buesching and Jordan 2022); and (iii) water sources as a crucial element in semi-arid landscapes (Prediction 4c) (Carvalho et al. 2016a). ...
... These corridors boast a high availability of potential latrine structures, mostly burrows, hollow trunks, and fallen tree branches (Fig. 3). They also provide water and a high abundance of preys (e.g., small mammals and insects) for small carnivores (Palomares 1993;Madikiza et al. 2010;Sikade 2017), and buffer extreme weather conditions (Shirk et al. 2014). Riverine forests are genets' olfactory marking hotspots, where several latrines are clustered along them (Palomares 1993;Espírito-Santo et al. 2007). ...
... The spread of latrines along ecotones also suggests that latrines may play a role in territory defence (Piñeiro and Barja 2015). In fact, riverine habitats are intensively used as movement corridors by animals, increasing the chances of encountering conspecifics, prey, and predators (Shirk et al. 2014). However, the concealment offered by dense vegetation cover, coupled with a high number of evasion routes (trails), provides safety while marking (Bist et al. 2021). ...
Context
Understanding how species select resources often requires assessing the environment at different spatial scales. Although the function of latrines in animal communication and social interactions has been studied in several carnivores, latrine site selection remains under-researched.
Objectives
We aimed to describe the characteristics of latrines and determine the environmental factors, operating at different scales, that drive latrine site selection by two sympatric genet species (Genetta genetta and G. tigrina) in an Albany Thicket landscape (South Africa).
Methods
We used a multi-scale modelling framework to investigate latrine site selection by comparing environmental characteristics at used latrines with that of two random points at four different scales. We then adapted a match-control design to derive the probability of latrine site selection.
Results
At the fine scale, genets selected latrine sites located in steeper slopes and boasting a higher availability of favourable micro-structures (e.g., burrows, termite mounds, hollow branches). At the landscape scale, latrines were positively associated with riverine forest corridors where they occurred in clusters. Genets avoided open areas and high terrain ruggedness to place their latrines. The best scale-independent model revealed the higher importance of edge habitats like riverine corridors and, to a lesser extent, dirt roads.
Conclusions
This study outlines the importance of including fine scale variables in multi-scale selection studies, as they may reveal features that are crucial for animal marking behaviour. Overall, our findings highlight the scales at which various factors influence latrine site selection the most. Based on our results, we suggest management practices that maintain animal communication by preserving riverine habitats across landscapes.
... Even when the correct variables are included in the model, if the scale at which scaledependant characteristics operate is incorrect, this can lead to a dramatically different interpretation of which factors are actually influencing species occurrence (Shirk et al., 2014;Bellamy et al., 2013;Girvetz and Greco, 2009). As there is no a priori way of inferring the grain and extent at which each environmental predictor is most strongly related to species presence (Shirk et al., 2014), habitat suitability modelling is moving towards increasingly complex multiscale models to reveal the true grain at which species respond to the landscape. ...
... Even when the correct variables are included in the model, if the scale at which scaledependant characteristics operate is incorrect, this can lead to a dramatically different interpretation of which factors are actually influencing species occurrence (Shirk et al., 2014;Bellamy et al., 2013;Girvetz and Greco, 2009). As there is no a priori way of inferring the grain and extent at which each environmental predictor is most strongly related to species presence (Shirk et al., 2014), habitat suitability modelling is moving towards increasingly complex multiscale models to reveal the true grain at which species respond to the landscape. Several recent studies conducted on mammals Bellamy et al., 2013;Shirk et al., 2014;Mateo Sánchez et al., 2014) have demonstrated the effectiveness of multi-scale approaches (Vergara et al., 2016). ...
... As there is no a priori way of inferring the grain and extent at which each environmental predictor is most strongly related to species presence (Shirk et al., 2014), habitat suitability modelling is moving towards increasingly complex multiscale models to reveal the true grain at which species respond to the landscape. Several recent studies conducted on mammals Bellamy et al., 2013;Shirk et al., 2014;Mateo Sánchez et al., 2014) have demonstrated the effectiveness of multi-scale approaches (Vergara et al., 2016). These allow more accurate and fine-scale predictions of species occurrence and habitat suitability by systematically varying the scale of analysis to find the dominant scale at which each variable operates to build the models (Shirk et al., 2012). ...
... The human disturbance model measured at the 100-m scale was the most informative for explaining patterns of detections of Pacific marten at trap stations in this study. In particular, marten were less likely to be detected in areas with a relatively greater proportion of young forest associated with clear-cut logging -a negative relationship that has been observed for other coastal populations of the species (Baker 1992;Slauson et al. 2007;Shirk et al. 2014). Marten require downed wood and near-ground structures that are usually absent from regenerating clearcuts (Sherburne & Bissonette 1994;Cushman et al. 2011;Shirk et al. 2014). ...
... In particular, marten were less likely to be detected in areas with a relatively greater proportion of young forest associated with clear-cut logging -a negative relationship that has been observed for other coastal populations of the species (Baker 1992;Slauson et al. 2007;Shirk et al. 2014). Marten require downed wood and near-ground structures that are usually absent from regenerating clearcuts (Sherburne & Bissonette 1994;Cushman et al. 2011;Shirk et al. 2014). A number of prey species for marten on Haida Gwaii, including Hairy Woodpecker (Picoides villosus picoides) and Red-breasted Sapsucker (Sphyrapicus ruber), are much less abundant in clearcuts compared to unharvested forests (Savard et al. 2000;Nagorsen 2006). ...
... Haida Gwaii marten have unique cranial morphology that may be an evolutionary adaptation for crushing the calcareous shells of marine invertebrates (Foster 1963;Giannico & Nagorsen 1989). Mink (Neovison vison), a typically aquatic mustelid, are absent on Haida Gwaii (Foster 1963;Golumbia 2000); thus marten may experience relatively low competition for marine invertebrates (Hodder et al. 2017 Buskirk et al. 1989;Shirk et al. 2014). Furthermore, salmonid fish scavenged from other predators are an important source of food for marten during fall and winter in coastal areas (Nagorsen et al. 1991;Ben-David et al. 1997). ...
... One of the frontiers in habitat modeling is understanding the temporal nonstationarity of habitat relationships and model predictions (e.g., Kaszta et al. 2021). There has been considerable attention paid to spatial nonstationarity, in which metareplicated studies are conducted to understand differences in habitat relationships and limiting factors in different study areas (e.g., Short Bull et al. 2011;Shirk et al. 2014;Wan et al. 2019). In contrast, there has been less attention paid to temporal nonstationarity, except in terms of comparing seasonal habitat models (e.g., Shirk et al. 2014) or post-disturbance changes in limiting factors (e.g., Cushman et al. 2011). ...
... There has been considerable attention paid to spatial nonstationarity, in which metareplicated studies are conducted to understand differences in habitat relationships and limiting factors in different study areas (e.g., Short Bull et al. 2011;Shirk et al. 2014;Wan et al. 2019). In contrast, there has been less attention paid to temporal nonstationarity, except in terms of comparing seasonal habitat models (e.g., Shirk et al. 2014) or post-disturbance changes in limiting factors (e.g., Cushman et al. 2011). Our results identified temporal nonstationarity in jaguar habitat selection, but not in puma habitat selection (Fig. 5). ...
... In tropical regions, food and water availability are subject to spatial and temporal variations, and habitat selection needs to be interpreted in that context. Thus, in systems with high seasonality of habitat use (e.g., Shirk et al. 2014), models developed from occurrence data across the full year may fail to accurately reflect habitat selection in any season within the year. These results suggest that for large felids in Mesoamerica, considering temporal nonstationarity is important for some species but might not be relevant for others. ...
Context
Few habitat modeling studies consider multiple spatial or temporal scales; less identify the operative scale of an organism's response to predictor variables. Optimizing habitat suitability models yields robust, reliable inferences about species-habitat relationships that can inform conservation efforts for species, such as jaguars (Panthera onca) and pumas (Puma concolor).
Objectives
We provide one of the first examples of evaluating temporal nonstationarity between seasons while simultaneously evaluating the effects of spatial and temporal scales on habitat selection. We sought insight into the predictor variables and associated scales determining seasonal distribution.
Methods
We selected predictor variables known to affect felid occurrence, then identified the optimal scale for each variable. We calculated the focal mean at spatial scales ranging from 500 m to 15,000 m. We then developed habitat suitability models and evaluated the effects of temporal scale on species co-occurrence.
Results
Patterns of jaguar and puma habitat selection varied. For jaguars, primary forest and its resources at fine scales were dominant predictors. For pumas, primary forest, secondary forest, and agropecuary lands at broad scales drove habitat selection. We observed divergent seasonal habitat selection, particularly for jaguars. Models confirmed that these sympatric predators might engage in spatial coordination to facilitate coexistence, as increased spatial overlap at a given scale in each season was associated with a diversification of landcover types.
Conclusions
Our results highlight the importance of considering spatial and temporal scales and temporal nonstationarity in habitat modeling. We suggest habitat modeling studies evaluate and optimize spatial and temporal scale relationships.
... Multiple factors drive the species distribution, with each being most influential at a specific spatial scale; thus, the apparent habitat-species relationships may change across spatial scales (Wiens 1989). The inclusion of scales is vital for understanding the species-habitat relationships (Schaefer and Messier 1995;Shirk 2012;Wasserman et al. 2012;Sánchez et al. 2014). The concept of scale in ecology is believed to be much older (e.g., see Schneider 2001) and is now recognized as a central theme in spatial ecology (Schneider 1994;Schneider et al. 1997;Schneider 1998;Cushman and McGarigal 2004). ...
... Our results are consistent with similar studies arguing that habitat selection measured at one specific scale may be insufficient to predict that selection at another scale (Mayor et al. 2009). Similar studies for brown bears (Martin et al. 2012;Sánchez et al. 2014); Dar et al. 2021) and other species (Shirk 2012;Shirk et al. 2014;Wan et al. 2017;Klaassen and Broekhuis 2018) also support the scale-dependent habitat selection. Consistent with these studies, our results indicate that habitat selection occurs across the range of scales for sloth bears, thus supporting our hypothesis of scale-dependent habitat selection in sloth bears. ...
... Our results are consistent with similar studies arguing that habitat selection measured at one specific scale may be insufficient to predict that selection at another scale (Mayor et al. 2009). Similar studies for brown bears (Martin et al. 2012;Sánchez et al. 2014); Dar et al. 2021) and other species (Shirk 2012;Shirk et al. 2014;Wan et al. 2017;Klaassen and Broekhuis 2018) also support the scale-dependent habitat selection. Consistent with these studies, our results indicate that habitat selection occurs across the range of scales for sloth bears, thus supporting our hypothesis of scale-dependent habitat selection in sloth bears. ...
Background
Habitat resources occur across the range of spatial scales in the environment. The environmental resources are characterized by upper and lower limits, which define organisms’ distribution in their communities. Animals respond to these resources at the optimal spatial scale. Therefore, multi-scale assessments are critical to identifying the correct spatial scale at which habitat resources are most influential in determining the species-habitat relationships. This study used a machine learning algorithm random forest (RF), to evaluate the scale-dependent habitat selection of sloth bears ( Melursus ursinus ) in and around Bandhavgarh Tiger Reserve, Madhya Pradesh, India.
Results
We used 155 spatially rarified occurrences out of 248 occurrence records of sloth bears obtained from camera trap captures ( n = 36) and scats located ( n = 212) in the field. We calculated focal statistics for 13 habitat variables across ten spatial scales surrounding each presence-absence record of sloth bears. Large (> 5000 m) and small (1000–2000 m) spatial scales were the most dominant scales at which sloth bears perceived the habitat features. Among the habitat covariates, farmlands and degraded forests were the essential patches associated with sloth bear occurrences, followed by sal and dry deciduous forests. The final habitat suitability model was highly accurate and had a very low out-of-bag (OOB) error rate. The high accuracy rate was also obtained using alternate validation matrices.
Conclusions
Human-dominated landscapes are characterized by expanding human populations, changing land-use patterns, and increasing habitat fragmentation. Farmland and degraded habitats constitute ~ 40% of the landform in the buffer zone of the reserve. One of the management implications may be identifying the highly suitable bear habitats in human-modified landscapes and integrating them with the existing conservation landscapes.
... North American martens (Pacific: M. caurina, and American: M. americana) typically occur at northern latitudes, or high elevations, within areas of contiguous forest cover. Within these areas, the structure and configuration of forested stands, including age, patch size, and fragmentation (Hargis et al. 1999;Fuller and Harrison 2005;Moriarty et al. 2016), and topographic features, such as elevation or complexity (Kirk and Zielinski 2009;Shirk et al. 2014;Sirén et al. 2016), are often cited as characteristic elements of marten habitat. North American martens have generally declined in distribution and abundance due to habitat loss, fragmentation, and the demographic effects of over-harvest during the 19th and early 20th centuries (e.g., Laliberte and Ripple 2004). ...
... Though many studies have quantified habitat selection of martens during a single season (e.g., Drew 1995;Hearn 2007;Wiebe et al. 2014;Tweedy et al. 2019) few studies have concurrently explored resource selection by martens during multiple seasons (Chapin et al. 1997;Shirk et al. 2014;Sirén et al. 2016). Although these studies varied in location, they generally indicated that martens select for complex and contiguous forest structure regardless of season, and select for topographic features associated with deep snow, such as riparian drainages (Shirk et al. 2014) or rugged terrain (Sirén et al. 2016) during snow-covered seasons. ...
... Though many studies have quantified habitat selection of martens during a single season (e.g., Drew 1995;Hearn 2007;Wiebe et al. 2014;Tweedy et al. 2019) few studies have concurrently explored resource selection by martens during multiple seasons (Chapin et al. 1997;Shirk et al. 2014;Sirén et al. 2016). Although these studies varied in location, they generally indicated that martens select for complex and contiguous forest structure regardless of season, and select for topographic features associated with deep snow, such as riparian drainages (Shirk et al. 2014) or rugged terrain (Sirén et al. 2016) during snow-covered seasons. While patterns of resource selection have emerged from these few cross-seasonal efforts, they were limited by locations being collected via VHF telemetry and ground triangulation, which compromises the precision at which animal locations were recorded, as well as the scale of landscape data integrated into RSFs (Chapin et al. 1997;Shirk et al. 2014;Sirén et al. 2016). ...
Context
Characterizing animal space-use and resource selection is central to effective conservation. In seasonally variable systems, animals may alter space-use to minimize risk, mediate physiological costs, and maintain access to resources. However, it is often unclear which environmental features influence space-use across seasons, and whether resource selection of non-migratory animals varies in seasonally snow-covered environments.
Objectives
We quantified space-use and scale-dependent resource selection of Pacific martens (Martes caurina) in northern California to evaluate the relative influence of abiotic (e.g., topography, weather) and biotic (e.g., forest structure) covariates on spatial ecology of martens in ecologically distinct seasons (i.e., snow-covered, snow-free).
Methods
We obtained fine-scale location data from GPS-collared martens (n = 26) in the Cascade and Sierra Nevada mountain ranges in California, USA. We incorporated spatially explicit weather, topographic, and forest structure data in a scale-optimized, seasonal resource selection function framework to determine the relative importance of abiotic and biotic conditions during snow-covered and snow-free periods.
Results
During snow-free periods, martens selected for features associated with complex forest structure, including increasing stem basal area. Conversely, space-use was associated with dense forest structure and topographic features in snow-covered periods. Though the relative influence of abiotic and biotic covariates on resource selection varied by season, the scale at which these variables best explained space-use did not.
Conclusions
Our results highlight seasonality and scale-dependence of resource selection by martens and emphasize the importance of understanding spatio-temporal responses of free-ranging animals to landscape heterogeneity. We suggest behavioral or ecological requirements that differ by season and scale may influence space-use and resource selection patterns, and, consequently, can inform conservation actions.
... For example, at broad extents in Idaho (i.e., multiple home ranges), marten occurrence was negatively affected by habitat fragmentation and roads, and at finer extents (i.e., within home ranges) martens were positively associated with patches of old-growth forests (Shirk et al. 2012;Wasserman et al. 2012). In Washington, marten avoided young, regenerating forests at broad extents but selected riparian forests with complex forest structure at finer extents (Shirk et al. 2014). Marten home ranges tend to occur in older (≥ 80 years) forests (Buskirk and Powell 1994;Powell et al. 2003;Payer and Harrison 2003). ...
... Young-and mid-successional forests that support marten likely contain complex structure (Bowman and Robitaille 1997;Payer and Harrison 2003;Poole et al. 2004;Porter et al. 2005;Hearn et al. 2010). In young-and mid-successional forests, some of these attributes (e.g., snags, downed wood) often have high spatial heterogeneity (Payer and Harrison 2003;Shirk et al. 2014), rendering information from standard forest inventory (often summarized at forest stand-levels and used to inform habitat management) uninformative for understanding marten space use (Silet 2017). ...
... We found hydrographic effects on marten space use, with marten core use areas associated with riparian cover type from LANDFIRE yet occurring farther away from linear hydrographic features like rivers and lakeshores. Shirk et al. (2014) found marten strongly selected for riparian areas in Washington and Oregon, which they attributed to mature forests occurring in riparian zones. Even though our study area is predominately forested (i.e.,~60% in trees on average around marten locations), riparian is the other dominant cover type (3 5% on average; Table S1). ...
Martens (Martes spp.) occupy areas with complex forest structure that can exhibit patchy distribution, particularly in managed forest landscapes. These structures (e.g., downed wood) are often difficult to reliably sample so more easily acquired surrogates may better describe marten habitat. Recent advances in global positioning system (GPS) collars combined with integrated (i.e., imagery with geospatial modeling), resolute, remotely sensed maps offer a potentially efficient means of understanding marten space use. We placed GPS collars on 13 American marten (M. americana), attempted to acquire a locational fix every 15 min, and calculated adaptive kernel home ranges from successful locations. We modeled probability of marten use as a function of covariates derived from remotely sensed data that included proportion of vegetation cover types, and distances to maintained roads and hydrographic features at 2.5, 4.5, and 7.1 ha extents. Average cover type values varied minimally across spatial extents we evaluated, indicating fine-scale homogenization. Amount of tall (> 10 m) deciduous forest, tall and short conifer forests, and riparian forests had positive effects on marten use, whereas distance to maintained roads had a weak negative effect. Broad riparian areas (e.g., scrub–shrub swamps that occur in broad topographic depressions), not necessarily associated with mapped hydrographic lines, were heavily used by marten. We found that core use areas for marten could reliably be predicted from 30 m remotely sensed maps summarized at relatively small extents (2.4–7.1 ha).
... However, there is no methodology to define, a priori, the scales at which a given predictor exerts the strongest influence on species Shirk, Wasserman, Cushman, & Raphael, 2012). In this context, it is important to apply scale opti- Shirk, Raphael, & Cushman, 2014;Shirk et al., 2012;Timm et al., 2016;Vergara et al., 2015;Wan et al., 2017;Wasserman, Cushman, Wallin, et al., 2012). ...
... Fourth, we found a relatively high degree of agreement between the two study areas in terms of variables and scales, with temperature and large extents of grass and sparsely vegetated conditions in upland and ridge topographic settings important for snow leopards in both landscapes. Fifth, the differences we did observe between study areas seemed to be related to differences in the limiting factors in those particular landscapes (e.g., Cushman et al., 2011;Shirk et al., 2014;Short bull et al., 2011). ...
... Environmental factors that are not highly variable, or that vary at broad scales with low local variation within landscapes, tended to be selected at coarser scales. Although snow leopards have a consistent response to landscape topography and composition, the extent to which habitat components vary, in relation to local attributes, lead primarily to a differential scale of effect of such predictors, and secondly to the inclusion of different limiting factors Shirk et al., 2014;Short Bull et al., 2011) as strongest descriptors of habitat in different areas (Tables 3-5, Tables S3 and S4). ...
Replicated multiple scale species distribution models (SDMs) have become increasingly important to identify the correct variables determining species distribution and their influences on ecological responses. This study explores multi‐scale habitat relationships of the snow leopard (Panthera uncia ) in two study areas on the Qinghai–Tibetan Plateau of western China. Our primary objectives were to evaluate the degree to which snow leopard habitat relationships, expressed by predictors, scales of response, and magnitude of effects, were consistent across study areas or locally landcape‐specific. We coupled univariate scale optimization and the maximum entropy algorithm to produce multivariate SDMs, inferring the relative suitability for the species by ensembling top performing models. We optimized the SDMs based on average omission rate across the top models and ensembles’ overlap with a simulated reference model. Comparison of SDMs in the two study areas highlighted landscape‐specific responses to limiting factors. These were dependent on the effects of the hydrological network, anthropogenic features, topographic complexity, and the heterogeneity of the landcover patch mosaic. Overall, even accounting for specific local differences, we found general landscape attributes associated with snow leopard ecological requirements, consisting of a positive association with uplands and ridges, aggregated low‐contrast landscapes, and large extents of grassy and herbaceous vegetation. As a means to evaluate the performance of two bias correction methods, we explored their effects on three datasets showing a range of bias intensities. The performance of corrections depends on the bias intensity; however, density kernels offered a reliable correction strategy under all circumstances. This study reveals the multi‐scale response of snow leopards to environmental attributes and confirms the role of meta‐replicated study designs for the identification of spatially varying limiting factors. Furthermore, this study makes important contributions to the ongoing discussion about the best approaches for sampling bias correction.
... In North America, martens are typically associated with northerly regions exhibiting some degree of contiguous forest cover. Indeed, the structure and configuration of forested stands, including age, patch size, and fragmentation (e.g., Hargis, Bissonette, & Turner, 1999;Moriarty, Epps, & Zielinski, 2016), and topographic features, such as elevation or complexity (Kirk & Zielinski, 2009;Shirk, Raphael, & Cushman, 2014), and should therefore be directly tested to evaluate whether they are helpful for achieving management or conservation goals. ...
... Following Shirk (et al. 2014), we ran univariable logistic regression models using the lme4 package in Program R to estimate the effects of abiotic and biotic covariates on marten space use. For each covariate, we calculated mean values within a moving window using the "focal" function in the raster package in Program R. We re-scaled covariates in 150 m increments at nine scales (30 m to 1230 m) and used AICc scores (Burnham and Anderson 2002) to determine the scale at which the model was optimized (i.e., the scale at which the covariate exhibits greatest influence on space use). ...
... The observed selection for structural features, such as basal area, can inform effective management strategies, particularly when a species or system exhibits consistent responses to dynamic landscape conditions (Heinemeyer et al., 2019;Tweedy et al., 2019). Our observations during snow-covered periods support previous research on seasonal trends of resource selection (Shirk et al., 2014). Indeed, we found that the ruggedness and slope of terrain, which have been used as proxies for snow-depth (Shirk et al., 2014), had relatively greater influence on space use of martens than forest characteristics. ...
Understanding how animals move through landscapes can allow us to elucidate the often complex relationships between behavior, physiology, and landscape conditions. While it can be difficult to clarify these relationships in free-ranging populations of animals, increasingly precise approaches to quantify space use and estimate energetics afford researchers the opportunity to address these complex questions in situ. Using a combination of fine-scale GPS collar data, precise estimates of metabolism, and high-resolution habitat information, we (1) estimated movement characteristics and field metabolic rates (FMR) of Pacific martens (Martes caurina) in a heterogeneous landscape, Lassen National Forest, California, to explore the role of movement and landscape characteristics on energetics and (2) explored the variation in seasonal resource selection of martens, as well as the effect of scale on patterns of resource selection.
In the first chapter, we examined the relationships between Pacific marten movement, expenditure, and landscape features in Lassen. We concurrently used DLW and GPS collars, a relatively new approach, and one that has not been applied in small-bodied, terrestrial animals, to investigate the relationship between movement characteristics and FMR. Movement velocity explained the greatest amount of variation in mass-specific FMR, and we used this relationship to predict expenditures of previously collared martens. We found that predicted mass-specific FMR was highest among males and increased in open patches primarily as a result of increased velocity and more erratic movements. Additionally, martens moving through deep snow also exhibited increased FMR. Our work shows movement metrics can effectively explain variation in FMR and identify landscape features, like forest structure and snow depth, that influence movement with cascading effects on energetics of free-ranging mammals in rapidly changing systems.
In the second chapter, we examined how resource selection of Pacific martens varied by season and scale. To quantify seasonal patterns of space use and resource selection, we obtained fine-scale location data from GPS-collared martens at the interface of the Cascade and Sierra Nevada mountain ranges in California, USA. We incorporated spatially explicit weather, topographic, and forest structure covariates in a scale-optimized resource selection function framework to determine the relative importance of abiotic and biotic conditions in snow-covered and snow-free periods. We found martens strongly select for forest structure features, including stem basal area, during snow-free periods. Conversely, marten space use was associated with topographic features in snow-covered periods with martens selecting for greater slopes and against rugged topography. Our results highlight the seasonality of resource selection by a cryptic carnivore in a dynamic system, and suggest behavioral or ecological requirements that differ by season may drive patterns of space use and resource selection. These findings highlight the importance of understanding fine-scale spatio-temporal responses to landscape heterogeneity and shifting conditions to clarify resource requirements and better inform conservation strategies.
... To evaluate rigorously the relative and interacting effects of core area vs. corridor protection a simulation framework is needed (Cushman 2014(Cushman , 2015Kaszta et al. 2019;Landguth et al. 2017;Shirk et al. 2014). Empirical data are an essential foundation for ecological science, but their analysis involves inductive reasoning where the patterns in the data are associated with alternative hypotheses of processes that generate them, usually through correlation or regression (e.g. ...
... Landguth and Cushman 2010) that drives the relationship, which then enables researchers to conduct modelling experiments (e.g. Balkenhol et al. 2015;Kaszta et al. 2019;Shirk et al. 2014) which can systematically vary parameters or scenario elements thereby allowing researchers to control the patterns and the processes together and therefore rigorously predict the relative effects of different scenarios. ...
CONTEXT: The efficient and effective design of protected areas is a fundamental challenge in landscape ecology, focusing on how spatial patterns of habitat influence conservation outcomes. This has sparked debate about the relative importance of habitat area versus connectivity in maintaining populations across fragmented landscapes. OBJECTIVES: We evaluate the relative importance of habitat area and connectivity by comparing counterfactual scenarios for landscape configuration on Borneo. We examine how habitat area and connectivity influence Sunda clouded leopard population size and genetic diversity across scenarios and dispersal abilities. METHODS: We compared 28 landscape scenarios on Borneo, incorporating all plausible combinations of core areas and movement corridors. Using spatially explicit genetic simulations, we modelled clouded leopard population size and genetic diversity metrics across five dispersal thresholds to compare how area and connectivity influence conservation outcomes. RESULTS: Our analysis reveals a strong, disproportionate relationship between landscape area and population size and genetic diversity. Even when accounting for landscape area, larger areas consistently provide superior conservation outcomes. Corridors showed minimal impact, becoming effective only at the highest dispersal thresholds. Habitat area emerged as the primary driver of conservation success, challenging assumptions about the importance of connectivity and highlighting the complex interactions between landscape configuration and species mobility. CONCLUSIONS: Our findings challenge paradigms in landscape ecology by demonstrating habitat area is more critical for biodiversity conservation than connectivity, especially where corridor lengths exceed species' dispersal abilities. Conservation strategies should therefore prioritise expanding core habitat areas, with corridor investments strategically targeted to highly mobile species.
... Study is a fixed effect of HJA, SIU, and UMP. Each trapping grid, Site, was treated as a normally distributed random effect (Gillies et al., 2006;Shirk et al., 2014) and we allowed unmodeled annual temporal variation to vary by study area using a normally distributed random effect, YearStudy. We determined the scale with the most predictive power by the lowest AIC score (Shirk et al., 2014;Tweedy et al., 2019;Wasserman et al., 2010). ...
... Each trapping grid, Site, was treated as a normally distributed random effect (Gillies et al., 2006;Shirk et al., 2014) and we allowed unmodeled annual temporal variation to vary by study area using a normally distributed random effect, YearStudy. We determined the scale with the most predictive power by the lowest AIC score (Shirk et al., 2014;Tweedy et al., 2019;Wasserman et al., 2010). The spatial scale of PPT, TMin, and TMax was fixed at 4 km 2 from PRISM, and as such we did not conduct scale optimization for these variables. ...
Remote sensing can provide continuous spatiotemporal information about vegetation to inform wildlife habitat estimates, but these methods are often limited in availability or lack adequate resolution to capture the three‐dimensional vegetative details critical for understanding habitat. The Global Ecosystem Dynamics Investigation (GEDI) is a spaceborne light detection and ranging system (LiDAR) that has revolutionized the availability of high‐quality three‐dimensional vegetation measurements of the Earth's temperate and tropical forests. To date, wildlife‐related applications of GEDI data or GEDI‐fusion products have been limited to estimate species habitat use, distribution, and diversity. Here, our goal was to expand the use of GEDI‐based applications to wildlife demography by evaluating if GEDI data fusions could aid in characterizing demographic parameters of wildlife. We leveraged a recently published dataset of GEDI‐fusion forest structures and capture–mark–recapture data to estimate the density and survival of two small mammal species, Humboldt's flying squirrel (Glaucomys oregonensis) and Townsend's chipmunk (Neotamias townsendii), from three studies in western Oregon spanning 2014–2021. We used capture histories in Huggins robust design models to estimate apparent annual survival and density as a derived parameter. We found strong support that both flying squirrel and chipmunk density were associated with GEDI‐fusion forest structures of foliage height diversity and plant area volume density in the 5–10 m strata for flying squirrels and proportionately higher plant area volume density in the 0–20 m strata for chipmunks, as well as other spatiotemporal factors such as elevation. We found weak support that apparent annual survival was associated with GEDI‐fusion forest structures for flying squirrels but not for chipmunks. We demonstrate further utility of these methods by creating spatially explicit density maps of both species that could aid management and conservation policies. Our work represents a novel application of GEDI data to evaluate wildlife demography and produce continuous spatially explicit density predictions for these species. We conclude that aspects of small mammal demography can be explained by forest structure as characterized via GEDI data fusions.
... Indeed, few occupancy studies have evaluated and optimised the multiscale relationships between species presence and environmental predictors (Jiménez-Franco et al. 2019). Modelling species occupancy without optimisation of spatial scales can risk misleading conclusions (Mayor et al. 2009;Shirk, Raphael, and Cushman 2014;Wevers et al. 2021). In recent years, several studies have demonstrated that scale optimisation produces models that not only have superior explanatory power but may also offer different, more realistic, interpretations than those based on models not optimised for scale Mateo Sánchez, Cushman, and Saura 2014;Penjor et al. 2021;Timm et al. 2016). ...
... Our selected environmental and anthropogenic variables provide insight into the resources that may limit species' habitat niches (Blonder 2018) and their relationships to species ecology and community composition (Chiaverini et al. 2022;Macdonald et al. 2020). In addition, heterogeneity in ecological conditions across broad extents might lead to differences in local limiting factors and expressed species-habitat relationships (Cushman and Landguth 2012;Shirk, Raphael, and Cushman 2014;Short Bull et al. 2011). We found that many species responded differently to the same variables in different landscapes in our study, suggesting spatially varying limiting factors. ...
Aim
Myanmar, an Indo‐Burmese biodiversity hotspot, lacks baseline data on species occurrence and distribution. This hinders biodiversity monitoring and optimisation of conservation and development plans. We aim to document baseline mammal occupancy, interactions with environmental factors and scale‐dependent responses.
Location
Hkakaborazi National Park, Htamanthi Wildlife Sanctuary, Alaungdaw Kathapa National Park, Rakhine Yoma Elephant Range, Say Taung and Myinmoletkhat Key Biodiversity Areas distributed across Myanmar.
Methods
Camera trap data throughout Myanmar were used to analyse species occupancy. We conducted a multiscale hierarchical spatial modelling process, using local and pooled data across Myanmar. We also optimised spatial scale across five scales and six predictors, using univariate occupancy models. We then selected scale‐optimised variables for multivariate modelling, repeating this process for each species across local, regional and national datasets.
Results
The study identified 47 terrestrial species and observed strong scale‐dependent nonstationarity in occupancy estimates. Relationships with environmental variables differed among species and were highly scale dependent. Importantly, occupancy estimates produced by pooling data across sites were greatly different from any of the estimates for the individual sites, suggesting that high heterogeneity in occurrence and abundance across sites among species requires local or nested occupancy estimates to account for spatial heterogeneity and variation.
Main Conclusions
Future conservation efforts should focus on Northern Myanmar if range‐restricted and rare species are to be protected, while focus should still be given to common species which serve as potential indicators of overall community structure. The nonstationarity of occupancy results from different datasets underscores the potential for misleading interpretations from aggregated data in nonstationary ecological systems. Metareplicated analyses of local, geographically and ecologically proximal regional datasets provide an important view of spatial variation in occupancy patterns guiding conservation design and improving understanding of the drivers of biodiversity patterns and change across large regions, such as Southeast Asia.
... Habitat selection and the structure of the realized habitat niche are scale-dependent phenomena (Levin, 1992). Therefore, correctly specifying the scale at which organisms simultaneously are influenced by and select for multiple resources and conditions has been shown to be critical to accurate predictions of species-environmental relationships in general (e.g., McGarigal, 2002, 2003) and species distribution models in particular (e. g., Grand et al., 2004;Wasserman et al., 2012;Mateo-Sanchez et al., 2014;Shirk et al., 2014;Hearn et al., 2018;Macdonald et al., 2018Macdonald et al., , 2019. ...
... Another main insight that emerged from the study is that univariate scaling, which is the oldest and most widely used method of scale optimization in multiscale modeling (e.g. Grand et al., 2004;Wasserman et al., 2012;Shirk et al., 2014;Vergara et al., 2017;Macdonald et al., 2019Macdonald et al., , 2021, was the most robust method of variable and scale selection, along with MRMR which performed equivalently. MRMR is widely used in bioinformatics research where highdimensional problems are the rule and variable selection is critical for tractability and interpretability (Saeys et al., 2007). ...
... This phenomenon is known as non-stationarity (Turner et al., 2001;Osborne et al., 2007;Kaszta et al., 2021;Rollinson et al., 2021). This can pose a problem for habitat suitability models, whereby covariates that have been determined to predict a response at one scale or landscape might incorrectly predict the response when projected to a new environment or geographical area (Turner et al., 2001;Cushman et al., 2011;Shirk et al., 2014). This can happen when the relationship between the predictor and covariates changes both spatially and temporally indicating non-stationarity in species-habitat preferences (Dobrowski et al., 2011;Vergara et al., 2017;Kaszta et al., 2021). ...
... This can happen when the relationship between the predictor and covariates changes both spatially and temporally indicating non-stationarity in species-habitat preferences (Dobrowski et al., 2011;Vergara et al., 2017;Kaszta et al., 2021). For example, the scale of effect (i.e., the scale at which a covariate is most important) could change between two similar habitats in different places (Shirk et al., 2014;McGarigal et al., 2016;Atzeni et al., 2020). ...
Habitat fragmentation and loss are major threats to species conservation worldwide. Studying species-habitat relationships is a crucial first step toward understanding species habitat requirements, which is necessary for conservation and management planning. However, some species inhabit a range of habitat types, potentially making the use of range-wide habitat models inappropriate due to non-stationarity in species-habitat preferences. The Mexican spotted owl (Strix occidentalis lucida) (MSO) is a species that inhabits both forests and rocky canyonlands, two habitats with large differences in environmental conditions. It is unclear whether the species uses habitat differently in these two habitat types or if previously-built habitat models for forest-dwelling owls can be used to understand MSO habitat use in rocky canyonlands. To explore this, we developed the first scale-optimized habitat suitability model for this subspecies of spotted owl in rocky canyonlands using an ensemble framework. We then compared our results with a previously-built habitat model for MSO in forested areas. In the rocky canyonland model, slope (800 m scale), cumulative degree days (1200 m scale), insolation (1000 m scale), and monsoon precipitation (100 m scale) were the most important environmental covariates. In contrast, in the forest model, percent canopy cover (100 m scale), percent mixed-conifer (5000 m scale), and slope (500 m scale) were the most important environmental covariates. The rocky canyonland model performed well, while the forest model performed poorly when projected to rocky canyonlands and predicted low suitability across the entire study area, including areas with known nesting locations. These results support the non-stationarity in habitat use for MSOs between rocky canyonland and forest habitats. Hence, when transferring habitat suitability models from one region to another, it is necessary to evaluate the transferability of the model by accounting for non-stationarity in species-habitat preferences.
... By contrast, strategic planning that targets actions in areas that are expected to provide the greatest benefits for species may help maximize efficiency (Arkle et al. 2014). However, this requires an understanding of how species use resources across landscapes (Boyce 2006;McGarigal et al. 2016;Marini et al. 2019;Northrup et al. 2021) and consideration of local variation in habitats and habitat-use relationships among populations (Shirk et al. 2014;Saher et al. 2022). ...
... Models that are developed for specific geographic locations (e.g., populations or sub-populations), rather than an entire species range, can consider unique habitat-use relationships at those sites. Such RSFs may be of great value for strategic restoration planning because they allow differentiation between seemingly similar sites and bring clarity to expected returns on habitat management investments, an important consideration when populations differ in selection response to variable habitat characteristics (e.g., Shirk et al. 2014;Saher et al. 2022). While the resulting suitability layers can be used to identify candidate sites for habitat restoration action, they stop short of indicating where limited restoration resources might be best allocated across the landscape. ...
Efforts to restore habitats and conserve wildlife species face many challenges that are exacerbated by limited funding and resources. Habitat restoration actions are often conducted across a range of habitat conditions, with limited information available to predict potential outcomes among local sites and identify those that may lead to the greatest returns on investment. Using the Gunnison sage-grouse (Centrocercus minimus) as a case study, we leveraged existing resource selection function models to identify areas of high restoration potential across landscapes with variable habitat conditions and habitat-use responses. We also tested how this information could be used to improve restoration planning. We simulated change in model covariates across crucial habitats for a suite of restoration actions to generate heatmaps of relative habitat suitability improvement potential, then assessed the degree to which use of these heatmaps to guide placement of restoration actions could improve suitability outcomes. We also simulated new or worsening plant invasions and projected the resulting loss or degradation of habitats across space. We found substantial spatial variation in projected changes to habitat suitability and new habitat created, both across and among crucial habitats. Use of our heatmaps to target placement of restoration actions improved habitat suitability nearly fourfold and increased new habitat created more than 15-fold, compared to placements unguided by heatmaps. Our decision-support products identified areas of high restoration potential across landscapes with variable habitat conditions and habitat-use responses. We demonstrate their utility for strategic targeting of habitat restoration actions, facilitating optimal allocation of limited management resources to benefit species of conservation concern.
... In two classic papers, Wiens (1989) and Levin (1992) introduced the modern concepts of scale dependence in ecological relationships, and in 2004, Thompson and McGarigal (2002) introduced the concept of functional heterogeneity in habitat relationships, in which each variable could have a different scale (spatial or temporal) of functional influence on the behavior, occurrence or distribution of a species. Since then, there has been copious work to explore the scale-dependence of species-habitat relationships (e.g., McGarigal 2003, Grand et al., 2004;Wasserman et al., 2012;Shirk et al., 2014;Wan et al., 2017;McGarigal et al. 2016). However, to date no one has assessed the nonstationarity (where habitat selection depends on the individual, geographical location and ecological context of that location) or heterogeneity of scale dependence among individuals of a species inhabiting different ecological contexts, and few have compared scale dependence and variable importance in a spatially replicated framework. ...
... Some other research has focused on variation in variable selection and optimal scales for single species in a meta-replicated context. Shirk et al. (2014) built scale optimized predictive models for American marten (Martes americana) in two disjunct and highly different ecological systems in Washington State, USA. They found that there were large differences in the selection of variables, their relative importance, and the spatial scales between the two study areas, and hypothesized the influence of spatially varying limiting factors as an explanation. ...
... Vertical forest structural complexity should be collectively created by many of our predicted plot-level features and may function primarily to alleviate marten mortality risk from larger-bodied terrestrial and avian carnivores (Godbout and Ouellet 2010, Cheveau et al. 2013, Smith et al. 2022b (Hodgman et al. 1997). Consistent with our findings, martens have been commonly described as inhabiting forests with relatively high amounts of canopy cover , Shirk et al. 2014, Tweedy et al. 2019, which offers overhead protection from raptors (e.g., (Thompson 1994, Bull and Heater 2001, McCann et al. 2010. Nonetheless, canopy heterogeneity is likely to be as important as canopy cover for martens, given that some raptors (e.g., goshawks) are adept at hunting in complex environments (Beier and Drennan 1997). ...
... Our study focused on marten associations with forest conditions at a fine spatial resolution and we acknowledge that inference beyond this scale is limited. Importance of scale to understanding habitat use has been widely indicated (Wiens 1989, Mayor et al. 2009) and specifically noted for martens (Shirk et al. 2014, Tweedy et al. 2019 Despite potential advances, field-based data such as ours remains invaluable to characterize fine-scale forest conditions (Zellweger et al. 2014). Indeed, we were specifically interested in providing information at a scale that is relevant to contemporary forest management, particularly in light of management options being considered to reduce the escalating threat of large, severe wildfires to coniferous forests of western North America (Abatzoglou andWilliams 2016, Hagmann et al. 2021). ...
When wildlife species exhibit unexpected associations with vegetation, replication of studies in different locales can illuminate whether patterns of use are consistent or divergent. Our objective was to describe fine‐scale forest conditions used by Pacific martens ( Martes caurina ) at 2 study sites in northern California that differed in forest composition and past timber harvest. We identified denning and resting locations of radio‐marked martens and sampled structure‐ and plot‐level vegetation using standardized forest inventory methods between 2009–2021. Woody structures used by martens were significantly larger than randomly available structures across types (e.g., live tree, snag, log) and at both study sites. Den and rest structures occurred in areas characterized by higher numbers of logs and snags, lower numbers of live trees and stumps, larger diameter live trees and logs, and greater variation in live tree and log diameter. Features of denning and resting locations were largely consistent across study sites and were generally representative of fine‐scale forest heterogeneity and increased structural complexity, conditions that martens have been widely associated with at broader spatial scales (i.e., home range or landscape). The spatial occurrence of denning and resting locations may indicate that fine‐scale structural complexity facilitates marten foraging while reducing predation risk. Our work offers timely and directed information that can guide forest management in the context of increased landscape change.
... Consistent with previous studies on this (Martin et al., 2012;Mateo-Sanchez et al., 2014a;Zarzo-Arias et al., ) and other species (Thompson & McGarigal, 2002;Grand et al., 2004;Shirk, Raphael & Cushman, 2014;Vergara et al., 2015;Chambers et al., 2016;Wan et al., 2017), habitat selection of the brown bear populations in Western Himalaya was scale-dependent. The majority of the habitat variables were selected at the broadest scale tested, indicating that brown bears perceive landscape features mostly at larger scales. ...
... This discrepancy suggests further research to quantify the relative limitations of brown bear distribution driven by climate and human factors independently. This will likely require metareplicated studies (sensu Shirk et al., 2014, ande.g. Short Bull et al., 2011) in which research is repeated in several ecosystems to determine in what circumstances particular habitat variables become limiting to brown bear distribution. ...
Climate change and land use change jointly are the largest drivers of population declines, range contraction and extinction for many species across the globe. Wide‐ranging and large‐bodied species are especially vulnerable to habitat loss and fragmentation due to their typically low population densities, reflecting their need for extensive and connected habitats. We used the multi‐scale Random Forests machine learning algorithm to identify factors driving the habitat selection and future changes in habitat of Himalayan brown bear, an iconic wide‐ranging and large‐bodied species of high conservation interest, across a range of spatial scales. Habitat selection of brown bears was scale‐dependent, with most variables selected at broad scales. Climatic variables such as maximum temperature of coldest month, minimum temperature of warmest month and the potential evapotranspiration of wettest quarter strongly influenced habitat selection of brown bears. Future projections indicate a strong difference between the high and low emission scenarios. Alarmingly, our model suggests that high emission scenarios, with or without land use change, may result in a decline of brown bear habitat of >90% by the end of the century. In contrast, low emission scenarios are projected to reduce brown bear habitat by <23%, with much of the species range shifting to higher elevations. This study provides an integrative understanding of scale‐dependent variables in brown bear habitat selection, providing critical information for prioritizing areas for habitat management and conservation. Most importantly, our future projections imply that traditional conservation efforts, such as in situ conservation, will not be sufficient to protect the species without climate change mitigation. The incorporation of climate change mitigation and adaptation in conservation strategies will be one of the most pressing priorities in biodiversity conservation in this region.
... Taking into account that population dynamics of amphibians is not related to a particular scale of environmental conditions, the use of variables in a multiscale approach is an option to obtaining robust and reliable estimations of the relation between amphibian species and their habitats, as has been shown for other animal groups (Jaskula & Brodman, 2000;McGarigal, Wan, Zeller, Timm, & Cushman, 2016;Moreira et al., 2016;Ribeiro Jr et al., 2018;Shirk, Raphael, & Cushman, 2014;Zeller et al., 2014). Nevertheless, if these estimations do not consider species detectability and the variation associated with environmental seasonality, it is more likely to obtain biased estimations that compromise the success of conservation efforts (Guillera-Arroita, 2017). ...
... Including variables at different spatial and temporal scales might be an efficient approach to achieve reliable estimates of species distribution (Shirk et al., 2014;Zeller et al., 2014), as well as to avoid overestimation (Mazerolle & Villard, 1999), especially for amphibian species that have a two-phase life cycle (Becker et al., 2007). Studies to estimates occupancy probability of species with terrestrial and aquatic life cycles, such as the Montevideo Treefrog (Hypsiboas pulchellus) and the Striped Snouted Tree-frog (Scinax squalirostris) have recognized the importance of the surrounding vegetation type and the conditions of water bodies (Moreira et al., 2016). ...
Numerous amphibian species are at risk of extinction worldwide. Therefore, reliable estimations of the distribution and abundance of these species are necessary for their conservation. Generally, amphibians are difficult to detect in the wild, which compromises the accuracy of long-term population monitoring and management. Occupancy models are useful tools to assess how environmental variables, at a local and at a landscape scale, affect the distribution and abundance of organisms taking into account species imperfect detectability. In this study, we evaluated with an environmental multiscale approach the seasonal variation of the occupation area of the threatened salamander, Ambystoma ordinarium along its distribution range. We obtained readings in 60 streams of physicochemical variables associated with habitat quality and landscape features. We found that detection and occupation probability of A. ordinarium are seasonally associated with different environmental variables. During the dry season, detectability was positively associated with temperature and stream depth, whereas occupancy was positively associated with the proportion of crops in the landscape and stream elevation. In the rainy season , the detection probability was not explained by any variable considered, and occupancy was negatively associated with stream's electrical conductivity and dissolved oxygen. Based on the estimation of occupied sites, we showed that A. ordinarium presents a more restricted distribution range than previously projected. Therefore, our results reveal the importance of evaluating the accuracy of distribution estimates for the conservation of threatened species as A. ordinarium.
... Metareplication study design requires replicate study areas that at least partly differ in land use characteristics (e.g., contrasting habitat configurations), so as to enable conclusions about how the focal species responds to landscape variables generally (Johnson 2002). Importantly, when researchers compare the best-fit optimizations of model parameters (e.g., geographic scale and function) that are independently derived from multiple study areas, cases exist in which even relatively minor changes in local environmental conditions can alter the influence that a given landscape variable has on gene flow owing to ''threshold or connectivity effects'' (see Cushman et al. 2011;Shirk et al. 2014). Through replicated analyses, limiting factors such as presence of human disturbance (Reddy et al. 2019), patch size of disturbed areas (Shirk et al. 2014), habitat connectivity (Castillo et al. 2016), and land cover heterogeneity (Vergara et al. 2017) have been found to affect how landscape variables contribute to resistance or facilitation of gene flow. ...
... Importantly, when researchers compare the best-fit optimizations of model parameters (e.g., geographic scale and function) that are independently derived from multiple study areas, cases exist in which even relatively minor changes in local environmental conditions can alter the influence that a given landscape variable has on gene flow owing to ''threshold or connectivity effects'' (see Cushman et al. 2011;Shirk et al. 2014). Through replicated analyses, limiting factors such as presence of human disturbance (Reddy et al. 2019), patch size of disturbed areas (Shirk et al. 2014), habitat connectivity (Castillo et al. 2016), and land cover heterogeneity (Vergara et al. 2017) have been found to affect how landscape variables contribute to resistance or facilitation of gene flow. For this reason, metareplication in landscape genetics is particularly powerful for extending the scope of insights that can be gained. ...
Context
Landscape genetics can identify habitat features that facilitate or resist gene flow, providing a framework for anticipating the impacts of land use changes on dispersal of individuals. To inform management, a better understanding of how inferences derived from one study region are applicable to other regions is needed.
Objectives
We investigated the manner in which five landscape variables correlated with gene flow among Plethodon mississippi populations in two study regions. We compared order of importance, direction (facilitation vs. resistance of gene flow) and scale of effect, and functional relationships of variables within each study area.
Methods
In forests in Mississippi and Alabama, USA, we tested individual-based genetic distances derived from microsatellite genotypes against effective distances caused by agriculture, hardwoods, pine, manmade structures, and wetlands that were optimized for both scale and transformation using maximum likelihood population effects modeling.
Results
Of the landscape variables, agriculture and wetlands ranked at the top of both study areas’ models. In both forest regions, agriculture was consistently associated with resistance, whereas pine was inferred to facilitate gene flow. However, we found region-specific differences in effects of wetlands, hardwoods, and manmade structures. Configuration of the latter landscape variables differed between forest regions, which may explain the contrasting outcomes.
Conclusions
Our results underscore the value of metareplication in revealing which components of landscape genetics models may be consistent across different portions of a species’ range, and those that have context-dependent impacts on gene flow. We also highlight the need to consider habitat configuration when interpreting the results of landscape genetics analyses.
... Martens can exhibit habitat selection at multiple spatial scales (Slauson et al. 2007, Thompson et al. 2012. We used bivariate spatial scale optimization to identify the optimal spatial scale for each variable, which is a technique used to capture scale-dependent effects of habitat selection for martens (Shirk et al. 2014, Tweedy et al. 2019, Martin et al. 2021. We created six spatial scales represented by buffers around the 2 km grid point (station A) for each survey unit with radii of 50, 270, 500, 750, 1,170, and 3,000 m. ...
... Most applications of species distribution modeling assume context independent, stationary relationships between patterns of species occurrence and environmental variables, such that global predictive models developed across all sampled locations and over time can produce single and stable predictions that can be applied to the full spatial range of the studied population and across time (Hegel et al., 2010). However, extensive recent research has shown that there is often a high degree of nonstationarity in space (e.g., Macdonald 1983, Shirk et al., 2014Vergara et al., 2016) and time (Cushman et al., 2013;Kaszta et al., 2019) in species-environmental relationships, and other ecological pattern-process relationships, such as the factors that predict and control gene flow (e.g., Short Bull et al. 2011;Cushman et al., 2013;Vergara et al., 2017;Reddy et al., 2019) and movement behavior (Elliot et al., 2014a,b;Unnithan Kumar et al. 2022). ...
Species distribution modeling is widely used to quantify and predict species-environment relationships. Most past
applications and methods in species distribution modeling assume context independent and stationary relationships between patterns of species occurrence and environmental variables. There has been relatively little research investigating context dependence and nonstationarity in species distribution modeling. In this paper we explore spatially varying limiting factors in species-environment relationships using high resolution telemetry
data from 14 individual wildcat hybrids distributed across geographical and environmental gradients in Scotland. (1) We proposed that nonstationary limiting factors would be indicated by significant association between statistical measures of variability of predictors and the predictive importance of those variables. (2) We further proposed that most of the limiting factor relationships observed would be related to spatial variation and a lesser amount to mean value of environmental variables within individual study sites. (3) Additionally, we anticipated that when there was a relationship between variation of an environmental factor and its importance as a predictor this relationship would be positive, such that higher variation would be associated with higher importance of the variable as a predictor (following the theory of limiting factors). (4) Conversely, we proposed that when
there was a relationship between the mean value of an environmental variable and its importance as a predictor this relationship would be roughly evenly split between positive and negative relationships, given that environmental variables could become limiting either when they are highly abundant or high value, or when they are rare or low value in a particular landscape, depending on the nature of the species-environment relationship for that ecological variable. (5) Finally, we hypothesized that the frequency of supported limiting factor relationships would differ among variable groups, with variables that were directly related to key environmental resources more likely to be limiting than those that would have more indirect impacts on wildcat hybrid habitat
selection or foraging. Our results show that assumptions of global, stationary habitat associations are likely not met in many habitat models, requiring explicit consideration of scale and context dependence in a nonstationary modeling paradigm. We found that both the mean value and the standard deviation are strong predictors of whether that variable will be limiting and differentially important as a predictor of occurrence. We confirmed that limiting factors become more limiting when it has higher variability across the sampled data, or when it is rare or not abundant.
... avoiding predators, accessing the subnivium) appear disproportionate to the cost. Previous research has identified snow as an important predictor of habitat use [31,63] and behaviour [25,60], but the fitness response we have as shown here links snow to the persistence of populations. ...
Ecological heterogeneity promotes species persistence and diversity. Environmental change has, however, eroded patterns of heterogeneity globally, stifling species recovery. To test the effects of seasonal heterogeneity on a reintroduced carnivore, American martens ( Martes americana ), we compared metrics of local and season-specific heterogeneity to traditional forest metrics on the survival of 242 individuals across 8 years and predicted a survival landscape for 13 reintroduction sites. We found that heterogeneity—created by forest structure in the growing season and snow in the winter—improved survival and outperformed traditional forest metrics. Spatial variation in heterogeneity created a distinct survival landscape, but seasonal change in heterogeneity generated temporal discordance. All translocation sites possessed high forest heterogeneity but there were greater differences in winter heterogeneity; recovery sites with the poorest snow conditions had the lowest viability. Our work links heterogeneity across seasons to fitness and suggests that management strategies that increase seasonal aspects of heterogeneity may help to recover other sensitive species to continuing environmental change.
... Spatial variation in species' niches may occur if there are genetic differences among populations (Pearman et al. 2008) or if a species is facing different degrees and types of competition, predation and diseases across their range (Araújo and Luoto 2007, Daskin and Alford 2012, Chamberlain et al. 2014, Vergnon et al. 2017). There are some indications that the assumption of stationarity may not always hold true (Whittingham et al. 2007, Fink et al. 2010, Schmidt et al. 2014, Shirk et al. 2014, Howard et al. 2015, Laube et al. 2015, Gómez et al. 2016, Zuckerberg et al. 2016, Wan et al. 2017, and while the inclusion of biotic factors in SDMs would likely reduce the problem of non-stationarity, it is often impractical to include them. ...
Species distribution models (SDMs) provide insights into species' ecology and distributions and are frequently used to guide conservation priorities. However, many uses of SDMs require model transferability, which refers to the degree to which a model built in one place or time can successfully predict distributions in a different place or time. If a species' model has high spatial transferability, the relationship between abundance and predictor variables should be consistent across a geographical distribution. We used Breeding Bird Surveys, climate and remote sensing data, and a novel method for quantifying model transferability to test whether SDMs can be transferred across the geographic ranges of 129 species of North American birds. We also assessed whether species' traits are correlated with model transferability. We expected that prediction accuracy between modeled regions should decrease with 1) geographical distance, 2) degree of extrapolation and 3) the distance from the core of a species' range. Our results suggest that very few species have a high model transferability index (MTI). Species with large distributions, with distributions located in areas with low topographic relief, and with short lifespans are more likely to exhibit low transferability. Transferability between modeled regions also decreased with geographical distance and degree of extrapolation. We expect that low transferability in SDMs potentially resulted from both ecological non‐stationarity (i.e. biological differences within a species across its range) and over‐extrapolation. Accounting for non‐stationarity and extrapolation should substantially increase the prediction success of species distribution models, therefore enhancing the success of conservation efforts.
... Survey stations were placed at randomly generated points within high priority townships as described above, maintaining a minimum distance of 6 km between stations. This spacing meets the assumption of independence between our detections for martens (home ranges 6-18 km 2 ; Gosse et al., 2005;Shirk et al., 2014;Simons, 2009) and fishers (home ranges up to 10 km 2 for females and up to 38 km 2 for males; Clark, 1986;Furnas et al., 2017;Powell et al., 2003). Snow depth data were obtained from the National Snow and Ice Data Center Snow Data Assimilation System data products (NOHRSC, 2004). ...
Human land use is a driving force of habitat loss and modification globally, with consequences for wildlife species. The American marten (Martes americana) and fisher (Pekania pennanti) are forest‐dependent carnivores native to North America. Both species suffered population declines due to loss of forested habitat and overharvest for furs, and continued habitat modification is an ongoing threat. Furthermore, the smaller marten may be susceptible to intraguild exclusion where the larger fisher are abundant, and both habitat modification and climate change may reduce spatial refugia available to marten. A detailed understanding of co‐occurrence patterns of marten and fisher in landscapes subjected to intense forest disturbance represents a key knowledge gap for wildlife ecology and management. Maine, in the northeastern United States, supports populations of both these species. It is an extensively forested state, and the vast majority is managed as commercial timberland. We designed a large‐scale field study to understand the relative importance of three sets of predictions for marten and fisher occupancy patterns where commercial silviculture is widespread: (1) The intensity of forest disturbance primarily determined both marten and fisher occupancy rates, (2) fisher occupancy was limited to areas of shallower snow and marten limited by fisher presence, or (3) both species responded to the composition of tree species within forested habitat. We collected data to test these nonmutually exclusive hypotheses via camera‐trap surveys, using an experimental design balanced across a gradient of forest disturbance intensity. We deployed 197 camera stations in both summer and winter over 3 years (2017–2020). We tagged over 800,000 images and found marten at 124 (63%) and fisher at 168 (85%) of the stations. By fitting multiseason occupancy models to the data, we found that the degree of habitat disturbance negatively influenced detection, occupancy, and temporal turnover for both species. Contrary to our expectations, however, we found no evidence of interspecific competition and instead support for positive associations with detection probabilities both spatially and temporally. Both species were positively associated with forest stands containing deciduous trees. Our findings further illustrate the impact that land use has on the occupancy dynamics for these forest‐dependent carnivores.
... Survey stations were placed at randomly generated points within high priority townships as described above, maintaining a minimum distance of 6 k between stations. This spacing meets the assumption of independence between our detections for martens (home ranges 6 k2 to 18 k2 [Gosse et al. 2005, Simons 2009, Shirk et al. 2014] and fishers (home ranges up to 10 k2 for females and up to 38 k2 for males [Clark 1986, Powell et al. 2003, Furnas et al. 2017). ...
Understanding trends in the abundance and distribution of carnivores is important at global, regional and local scales due to their ecological role, their aesthetic and economic value, and the numerous threats to their populations. Carnivores in Maine range from the American black bear (Ursus americanus), to numerous native mesocarnivore species, such as American marten (Martes americana), fisher (Pekania pennanti), coyote (Canis latrans), red fox (Vulpes vulpes), bobcat (Lynx rufus), Canada lynx (Lynx canadensis) and to two small weasel species (Mustela erminea and Neogale frenata). Though smaller than their apex carnivore cousins, Mesocarnivores are essential components of ecosystems and have complex impacts on prey species and intraguild dynamics. However, these species can vary in how they respond to human disturbances, from direct declines due to unregulated harvest and habitat loss, and their ability to adapt to land-use change. Maine is a working landscape which provides habitat for diverse wildlife species coincident with extensive forest harvest industries, as well as tourism and recreation. The intensity, timing, and configuration of harvest activities all interact to modify the landscape, with cascading impacts on the distribution of many animals. Forest management practices have changed through time (Maine Forest Service 2003) with potentially unpredictable outcomes (e.g. Simons 2009). However, the extent to which carnivore species adapt to land use change is a key knowledge gap that needs to be addressed to ensure proper management and conservation going forward. I investigated these patterns by designing a natural experiment across the forested landscape of Maine, and by collecting detection data on multiple species at camera trapping survey stations deployed along a gradient of forest disturbance. My dissertation aims to collect broad-scale, relevant information for carnivore management and conservation, and assess the efficacy of motion-triggered trail cameras for long-term monitoring. My work is divided into four sections, reflected by the four chapters included in the dissertation. My first goal was to determine the optimal number and configuration of camera-trap transects, to balance between reasonable effort expended and high-quality data collection. I used multi-method occupancy analyses to compare between one, two or three camera units spaced either 100 m or 150 m apart. We found that a design with three cameras spaced 100 m apart increased detection probabilities up to five-fold over a single camera trap, and thus used this configuration for the duration of the following research. Once the survey unit was selected, I established a large-scale, multi-year camera trapping regimen across the northern two-thirds of Maine. Survey sites were selected in compliance with a natural experimental design, replicating across all combinations of a) forest disturbance intensity, b) latitude, and c) fur trapping harvest reports for key furbearing species. In the second chapter I present this study design in more detail, and use the resulting data to investigate the interspecies dynamics of marten and fisher, two species of interest to the state of Maine that co-exist in several geographic areas and partition habitat in distinct ways. Both species are sensitive to habitat change resulting from timber harvest, which was a more important factor in occupancy patterns than intraguild dynamics. In chapter three, I took advantage of the large data set I collected to provide a landscape scale understanding of long-tailed and short-tailed weasel distribution patterns in the face of habitat change. Both of these species are poorly studied, and may be in decline in North American. My results indicate that short-tailed weasel are widespread in Maine and do not appear limited by forest harvest practices, while long-tailed weasel are rarer and more apt to be present in southern Maine. Finally in chapter four I ran models incorporating multiple states for species occupancy, beyond mere present or absent, to understand the dynamics of black bears and of black bear reproduction across managed forests in Maine. I found that generally disturbance at a small scale was positively associated with both occupancy and probability of reproduction, while the availability of hardwood trees (an important food source for bears) was also positively linked to the probability of female bears being with cubs. In addition to meeting our stake holder needs for informed management guidelines, I hope that many of my findings will be directly relevant to the broader research community—as camera trapping equipment becomes more affordable, it will become feasible to both monitor and rigorously study wildlife populations in remote locations and under many scenarios of human land-use.
... Our study similarly highlights some of the challenges in identifying habitat. Not only can animals respond differently to the same covariates in different areas but also they respond differently to the same covariates across multiple spatial scales (Shirk et al. 2014, McGarigal et al. 2016, Tweedy et al. 2019 if limiting factors differ with scale (Rettie and Messier 2000, Meyer and Thuiller 2006, Mayor et al. 2009). Thus, meeting the needs of species management and recovery planning, such as contemporary distribution, habitat requirements, and habitat availability, requires a range of analyses with cautious interpretation of spatioenvironmental relationships from studies designed to address specific questions or data gaps. ...
Abstract Conservation and management of animal populations requires knowledge of their occurrence and drivers that influence their distribution. Noninvasive survey methods and occupancy models to account for imperfect detection have become the standard tools for this purpose. Simultaneously addressing both occurrence and occurrence–environment relationships, however, presents multiple challenges, particularly for species with reduced ranges or those recovering from historical declines. Here, we present a comprehensive framework to satisfy the assumption of organism–environment equilibrium, map the range of a species, incorporate camera traps and detection dogs as complementary data sources, and make inference about wildlife space use at multiple scales. To meet these goals, we developed a Bayesian spatial occupancy model for Pacific fishers (Pekania pennanti) in Oregon using data from a large‐scale (64,280 km2) empirical effort combining 1240 camera traps (74,219 trap nights) and 196 detector dog surveys (3 × 3 km units, survey average = 17.3 km/unit). We deployed this model with and without a geoadditive term to improve predicted range map generation and covariate inference, respectively. We used reaction–diffusion models to project recovery trajectories to determine both plausible spatial extents for inclusion in our occupancy model and whether the current distribution can be explained by time‐limited population expansion from historical refugia. To assess nonstationary effects where species–habitat relationships vary spatially, we fit separate models within distinct ecological regions. We confirmed the presence of the native and introduced fisher populations, but populations occupy less area than previously believed. The spatial extent of the introduced population was less than expected except under our lowest growth model, suggesting limiting factors were preventing population expansion. The native population extent matched expectations under several growth scenarios, suggesting that the contemporary distribution is plausibly due to time‐limited expansion. The relationship of fisher occupancy to environmental covariates varied with scale, spatial extent, and ecological region, but fishers consistently selected for old forests at fine spatial scales in the detection model across spatial extents and detection modalities. Collectively, we provide an integration of camera traps and detection dogs into spatial occupancy models and demonstrate how to generate plausible spatial extents to improve inferences for species recovering from range contractions.
... Another challenge in the development of broad-scale HSMs is the presence of geographic variation in speciesenvironment relationships due to regional differences in climate, topography, or vegetation communities (Shirk et al., 2014;Doherty et al., 2016;Wan et al., 2017). Such differences represent an extension of the broader issue of transferability in HSMs, which is a widely recognized problem when predicting to novel spatiotemporal conditions (Guisan & Thuiller, 2005;Yates et al., 2018). ...
Range‐wide species conservation efforts are facilitated by spatially explicit estimates of habitat suitability. However, species‐environment relationships often vary geographically and models assuming geographically constant relationships may result in misleading inferences. We present the first range‐wide habitat suitability model (HSM) for the federally threatened eastern indigo snake (Drymarchon couperi) as a case study illustrating an approach to account for known latitudinal variation in habitat associations. Specifically, we modeled habitat suitability using interactive relationships between minimum winter temperature and several a priori environmental covariates and compared our results to those from models assuming geographically constant relationships. We found that multi‐scale models including interactive effects with winter temperature outperformed single‐scale models and models not including interactive effects with winter temperature. Our top‐ranked model had suitable range‐wide predictive performance and identified numerous large (i.e., ≥1000 ha) potential habitat patches throughout the indigo snake range. Predictive performance was greatest in southern Georgia and northern Florida likely reflecting more restrictive indigo snake habitat associations in these regions. This study illustrates how modeling interactive effects between temperature and environmental covariates can improve the performance of HSMs across geographically varying environmental gradients.
... Within a species' distribution, the location of a study site (in this case, referring to the location of data collection within the context of a species' range) can determine the amount, quality, and configuration of land cover types available to the organism. Studies explicitly examining the influence of site on resource selection results have largely found evidence of site-dependent selection trends (Mcnew et al., 2013;Shirk et al., 2014;Wan et al., 2017). However, less is known about the influence of close and relatively similar sites on the resource selection patterns of generalist species. ...
Quantifying resource selection (an organism's disproportionate use of available resources) is essential to infer habitat requirements of a species, develop management recommendations, predict species responses to changing conditions, and improve our understanding of the processes that underlie ecological patterns. Because study sites, even within the same region, can differ in both the amount and the arrangement of cover types, our objective was to determine whether proximal sites can yield markedly different resource selection results for a generalist bird, northern bobwhite (Colinus virginianus). We used 5 years of telemetry locations and newly developed land cover data at two, geographically distinct but relatively close sites in the south-central semi-arid prairies of North America. We fit a series of generalized linear mixed models and used an information-theoretic model comparison approach to identify and compare resource selection patterns at each site. We determined that the importance of different cover types to northern bobwhite is site-dependent on relatively similar and nearby sites. Specifically, whether bobwhite selected for shrub cover and whether they strongly avoided trees, depended on the study site in focus. Additionally, the spatial scale of selection was nearly an order of magnitude different between the cover types. Our study demonstrates that—even for one of the most intensively studied species in the world—we may oversimplify resource selection by using a single study site approach. Managing the trade-offs between practical, generalized conclusions and precise but complex conclusions is one of the central challenges in applied ecology. However, we caution against setting recommendations for broad extents based on information gathered at small extents, even for a generalist species at adjacent sites. Before extrapolating information to areas beyond the data collected, managers should account for local differences in the availability, arrangement, and scaling of resources.
... fig-4 high levels of canopy cover. Marten populations are typically associated only with relatively dense and increasing canopy cover (Shirk, Raphael & Cushman, 2014) and we posit that a quadratic response to canopy cover by Humboldt martens may be a function of shrub cover. Although additional information is needed to describe fine-scale vegetation associations, forest conditions with a dense understory layer of shrub and mast-producing species represent achievable targets that can guide management or restoration. ...
Background:
Many mammalian species have experienced range contractions. Following a reduction in distribution that has resulted in apparently small and disjunct populations, the Humboldt marten (Martes caurina humboldtensis) was recently designated as federally Threatened and state Endangered. This subspecies of Pacific marten occurring in coastal Oregon and northern California, also known as coastal martens, appear unlike martens that occur in snow-associated regions in that vegetation associations appear to differ widely between Humboldt marten populations. We expected current distributions represent realized niches, but estimating factors associated with long-term occurrence was challenging for this rare and little-known species. Here, we assessed the predicted contemporary distribution of Humboldt martens and interpret our findings as hypotheses correlated with the subspecies' niche to inform strategic conservation actions.
Methods:
We modeled Humboldt marten distribution using a maximum entropy (Maxent) approach. We spatially-thinned 10,229 marten locations collected from 1996-2020 by applying a minimum distance of 500-m between locations, resulting in 384 locations used to assess correlations of marten occurrence with biotic and abiotic variables. We independently optimized the spatial scale of each variable and focused development of model variables on biotic associations (e.g., hypothesized relationships with forest conditions), given that abiotic factors such as precipitation are largely static and not alterable within a management context.
Results:
Humboldt marten locations were positively associated with increased shrub cover (salal (Gautheria shallon)), mast producing trees (e.g., tanoak, Notholithocarpus densiflorus), increased pine (Pinus sp.) proportion of total basal area, annual precipitation at home-range spatial scales, low and high amounts of canopy cover and slope, and cooler August temperatures. Unlike other recent literature, we found little evidence that Humboldt martens were associated with old-growth structural indices. This case study provides an example of how limited information on rare or lesser-known species can lead to differing interpretations, emphasizing the need for study-level replication in ecology. Humboldt marten conservation would benefit from continued survey effort to clarify range extent, population sizes, and fine-scale habitat use.
... Graph-theoretic methods such as population graphs and corresponding conditional genetic distance (cGD) metric (Dyer and Nason 2004) describe how genetic variation is distributed among taxa in space by using conditional covariances to account for contributions from all strata in the study. In addition, established graph-theoretic metrics and approaches can be used alongside population graphs, such as node removal simulations to understand how networks are impacted by the loss of individual nodes, highlighting regions and specific populations that conservation managers can focus on to alleviate concerns of possible extirpation or loss of genetic connectivity (Albert et al. 2013, Zero shown that factors such as species distribution and gene flow can vary between study areas for a species of interest (Short Bull et al. 2011, Shirk et al. 2014. This is of importance for conservation, where range-wide extrapolation of genetic data obtained from a small portion of a species range can lead to ineffective management decisions (Short Bull et al. 2011, Trumbo et al. 2013. ...
Characterizing both genetic and functional connectivity is of paramount concern in the face of ongoing environmental change. Approaches combining both landscape genetic and graph theoretic methodologies have shown promise in allowing for simultaneous identification of strong and vulnerable populations, and the landscape factors that may inhibit or facilitate genetic connectivity. Here we combine landscape genetic and graph theoretic methods to: 1) Assess the genetic structure and genetic connectivity of Parnassius clodius butterflies in three protected regions in the United States: North Cascades National Park (WA), Grand Teton National Park (WY), and Yosemite National Park (CA)), and determine whether these metrics vary with differences in sampling scale between regions. 2) Test the resilience of population connectivity to extirpation, and 3) Test the relative importance of isolation by distance (IBD) vs. isolation by resistance (IBR) by testing the following IBR models: P. clodius habitat suitability model (HSM), Dicentra host plant HSMs, and topographical complexity in the form of terrain roughness and forest cover in limiting genetic connectivity. Both traditional genetic clustering analyses and network analyses revealed fine-scale genetic structure across all three regions. Our network analyses revealed similarity in topology across regions despite significant landscape variation, and network sensitivity analyses revealed that P. clodius subpopulations within the Grand Teton and Yosemite NP regions are more vulnerable to perturbations. Our results suggest environmental variables were more important than geographical distance in mediating genetic connectivity, with P. clodius and Dicentra HSMs playing a larger role than terrain complexity.
... Thus, investigations into habitat selection and movement should be conducted at multiple, ecologically relevant scales (Wiens, 1989;Goodwin and Fahrig, 1998;McGarigal et al., 2016b). The failure to do so can undermine the performance of habitat selection and movement models and their interpretation (Wasserman et al., 2012;Mateo Sanchez et al., 2014;Shirk et al., 2014), potentially leading to errors of inference and application (McGarigal and Cushman, 2002). Reliable knowledge about the multi-variate and multi-scale effects of environmental heterogeneity on organism distribution, abundance and movement can be acquired through robust multi-scale analytical methods supported by empirical data (Cushman et al., 2013;Zeller et al., 2018a). ...
Effective conservation and land management require robust understanding of how landscape features spatially and temporally affect population distribution, abundance and connectivity. This is especially important for keystone species known to shape ecosystems, such as the African elephant (Loxodonta africana). This work investigates monthly patterns of elephant movement and connectivity in Kruger National Park (KNP; South Africa), and their temporal relationship with landscape features over a 12-month period associated with the occurrence of a severe drought. Based on elephant locations from GPS collars with a short acquisition interval, we explored the monthly patterns of spatial-autocorrelation of elephant movement using Mantel correlograms, and we developed scale-optimized monthly path-selection movement and resistant kernel connectivity models. Our results showed high variability in patterns of autocorrelation in elephant movements across individuals and months, with a preponderance of directional movement, which we believe is related to drought induced range shifts. We also found high non-stationarity of monthly movement and connectivity models; most models exhibited qualitative similarity in the general nature of the predicted ecological relationships, but large quantitative differences in predicted landscape resistance and connectivity across the year. This suggests high variation in space-utilization and temporal shifts of core habitat areas for elephants in KNP. Even during extreme drought, rainfall itself was not a strong driver of elephant movement; elephant movements, instead, were strongly driven by selection for green vegetation and areas near waterholes and small rivers. Our findings highlight a potentially serious problem in using movement models from a particular temporal snapshot to infer general landscape effects on movement. Conservation and management strategies focusing only on certain areas identified by temporarily idiosyncratic models might not be appropriate or efficient as a guide for allocating scarce resources for management or for understanding general ecological relationships.
... Species distribution models are a powerful tool for ecological research and biodiversity conservation, but they may be uninformative or misleading if they fail to identify the relevant factors driving species' habitat selection (Pliscoff et al., 2014;Williams et al., 2012). Moreover, there is a longstanding recognition that species-environment relationships occur across a range of spatial scales (Levin, 1992;Wiens, 1989), and assessing environmental factors at a single scale often produces biased estimates or weaker predictive capacity of models (Mateo-Sánchez et al., 2014;Shirk et al., 2014). However, relatively few habitat suitability or species distribution modelling studies have rigorously addressed spatial scale issues and even fewer have applied multi-scale optimization to reliably describe scale dependencies (McGarigal et al., 2016). ...
Sampling bias and autocorrelation can lead to erroneous estimates of habitat selection, model overfitting and elevated omission rates. We developed a multi-scale habitat suitability model of the flammulated owl (Psiloscops flammeolus) in the Northern Rocky Mountains based on extensive but spatially clustered survey data, and then used simulations to evaluate the effects of spatially non-representative and spatially representative sampling strategies on model performance and predictions. Our hypothesis was that models trained with spatially non-representative simulated datasets would suffer from bias in parameter estimates, and would show lower pre-dictive performance. The models trained with the spatially representative simulated datasets greatly out-performed the models trained with the spatially non-representative simulated datasets judged on standard metrics of model performance. However, the spatially non-representative models produced superior predictions based on their ability to identify the correct spatial scales, covariates, signs and magnitudes of the species-environment relationships, when compared to the spatially representative models. Thus, it is likely that representative spatial sampling across a broad range of environmental gradients also resulted in over-dispersion of sampling data, with a higher proportion of samples falling in areas of low probability of presence, leading to lower ability to resolve the relationships between species presence-absence and environmental covariates. In contrast, the spatially non-representative sampling, by concentrating sampling along environmental gradients that are characterized by higher probability of presence of the modelled species, produced predictions that, while seeming to be weaker based on standard measures of model performance (e.g., AUC, Kappa, PCC), greatly outperformed the spatially representative models based on measures of true model prediction (e.g., correctly describing the actual spatial scales, direction and strength of species-environment relationships). Further work using simulation approaches is warranted to more fully evaluate the ability of species distribution modelling techniques to correctly identify scales, driving covariates, signs and magnitudes of relationships between species presence-absence patterns, and environmental covariates.
... The human disturbance model measured at the 100-m scale was the most informative for explaining patterns of detections of Pacific marten at trap stations. In particular, marten were less likely to be detected in areas with a relatively greater proportion of young forest associated with clear-cut logginga negative relationship that has been observed for other coastal populations of the species as well as American marten (Baker, 1992;Slauson et al., 2007;Shirk et al., 2014;Bridger et al., 2016). Density of roads, including active and abandoned roads, and forest edges were both positively correlated with marten detections. ...
Apex mesocarnivores can have a large influence on the functioning of plant and animal communities. Such effects can be more pronounced in relatively depauperate island ecosystems, especially in the context of a legacy of landscape change and introduced species. One such ecosystem is Haida Gwaii, British Columbia, Canada, where a long history of introduced species and forest harvesting has greatly influenced the composition of the plant and animal communities. Most notably, Sitka black-tailed deer (Odocoileus hemionus sitkensis), an introduced and unchecked invasive species, has interacted with forest harvesting to greatly change the understory structure and composition of those temperate rainforests. Little is known of the habitat ecology of Pacific marten (Martes caurina), the small-bodied apex predator on those islands. We used camera traps to monitor the distribution and habitat ecology of marten across 2 winters. We used mixed-effects logistic regression to test 14 a priori hypotheses represented by statistical models formulated at two spatial scales (100 and 1000 m), with four categories of explanatory factors: sampling bias, forest structure, topography, and human disturbance. Relationships between detections and covariates from the top models revealed that at the 100-m scale, marten were more likely to use habitat that was closer to streams and marine shorelines and that contained a small area of forest harvesting. Marten detections were associated with habitat containing some component of forest edge and road, suggesting foraging behavior adapted to invasive deer, an important source of carrion for marten during winter. The broad use of habitats by Pacific marten suggests that intraguild competition and predation might occur with several species at risk including the endemic Haida ermine (Mustela ermine haidarum). Those two species would likely overlap in diet and distribution when co-occurring across low-elevation coastal habitats. Conservation planning on Haida Gwaii should consider measures that decrease the potential for competition or predation between marten and species at risk.
... The importance of incorporating spatial scale into modelling of species-habitat relationships is wellestablished (McGarigal and Cushman 2002;McGarigal et al. 2016). Failing to do so can undermine the performance of habitat selection models and their interpretation (e.g., Wasserman et al. 2012;Mateo-Sanchez et al. 2014;Shirk et al. 2014), potentially leading to errors of inference and application (McGarigal and Cushman 2002). Optimized multiple-scale approaches (see definitions by McGarigal et al. (2016)] have been used to model habitat selection, limiting factors, and threats for a number of species (e.g. ...
The article Optimization of spatial scale, but not functional shape, affects the performance of habitat suitability models: a case study of tigers (Panthera tigris) in Thailand, written by Eric Ash, David W. Macdonald, Samuel A. Cushman, Adisorn Noochdumrong, Tim Redford and Zaneta Kaszta, was originally published online on 12 January 2021 with Open Access under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
... The importance of incorporating spatial scale into modelling of species-habitat relationships is wellestablished (McGarigal and Cushman 2002;McGarigal et al. 2016). Failing to do so can undermine the performance of habitat selection models and their interpretation (e.g., Wasserman et al. 2012;Mateo-Sanchez et al. 2014;Shirk et al. 2014), potentially leading to errors of inference and application (McGarigal and Cushman 2002). Optimized multiple-scale approaches (see definitions by McGarigal et al. (2016)] have been used to model habitat selection, limiting factors, and threats for a number of species (e.g. ...
Context
Species habitat suitability models rarely incorporate multiple spatial scales or functional shapes of a species’ response to covariates. Optimizing models for these factors may produce more robust, reliable, and informative habitat suitability models, which can be beneficial for the conservation of rare and endangered species, such as tigers (Panthera tigris).
Objectives
We provide the first formal assessment of the relative impacts of scale-optimization and shape-optimization on model performance and habitat suitability predictions. We explored how optimization influences conclusions regarding habitat selection and mapped probability of occurrence.
Methods
We collated environmental variables expected to affect tiger occurrence, calculating focal statistics and landscape metrics at spatial scales ranging from 250 m to 16 km. We then constructed a set of presence–absence generalized linear models including: (1) single-scale optimized models (SSO); (2) a multi-scale optimized model (MSO); (3) single-scale shape-optimized models (SSSO) and (4) a multi-scale- and shape-optimized model (MSSO). We compared performance and resulting prediction maps for top performing models.
Results
The SSO (16 km), SSSO (16 km), MSO, and MSSO models performed equally well (AUC > 0.9). However, these differed substantially in prediction and mapped habitat suitability, leading to different ecological understanding and potentially divergent conservation recommendations. Habitat selection was highly scale-dependent and the strongest relationships with environmental variables were at the broadest scales analysed. Modelling approach had a substantial influence in variable importance among top models.
Conclusions
Our results suggest that optimization of the scale of resource selection is crucial in modelling tiger habitat selection. However, in this analysis, shape-optimization did not improve model performance.
... Second, both species distribution models and our regression analyses only incorporated remotely sensed data on climate and land cover variables. These variables relied on coarse cover type classifications, and previous research has shown that both martens and fishers also select for fine-scale habitat features (Buskirk and Powell 1994;Shirk et al. 2014;McCann et al. 2014). Moreover, our data captured the heterogeneity in land cover composition and configuration, but not the underlying structural complexity that is important for both martens and fisher habitat selection (Buskirk and Powell 1994;McCann et al. 2014). ...
ContextInterspecific competition can limit species distributions unless competitors partition niche space to enable coexistence. Landscape heterogeneity can facilitate niche partitioning and enable coexistence, but land-use change is restructuring terrestrial ecosystems globally with unknown consequences for species interactions.Objectives
We tested the relationship between landscape heterogeneity and carnivore co-occurrence in natural and human-dominated ecosystems to assess the landscape-mediated impacts of anthropogenic change on coexistence.Methods
We used boosted regression trees to model the distributions and co-occurrence of two competing forest carnivores, American martens and fishers, at two contrasting sites in the Great Lakes region, USA: a “natural” site largely devoid of human impacts and a “human-dominated” site with substantial development and a history of land-use change. We assessed the importance of climate and habitat variables for each species, measured spatial niche overlap, and quantified co-occurrence as a function of compositional (patch richness), configurational (landscape shape), and topographical (elevation range) heterogeneity per site.ResultsWe observed significant differences in the effect of heterogeneity on co-occurrence between sites. The natural landscape exhibited little niche overlap and co-occurrence had a significant, positive relationship with heterogeneity. Conversely, the human-dominated site exhibited high niche overlap with variable effects of heterogeneity on co-occurrence. Elevation, snowpack, and development had strong, contrasting effects on marten and fisher distributions, suggesting that differential use of habitat and anthropogenic features facilitates coexistence.Conclusions
Heterogeneity can facilitate coexistence, but too much may undermine carnivore coexistence in human-dominated landscapes where habitat and space are limited. Moreover, future climate change will likely erode niche partitioning among martens and fishers, with particularly strong consequences for coexistence in human-dominated landscapes and at range boundaries.
... Recall that the range distribution aims to predict where an animal might go rather than define where it went ( Figure 13.2). Historically, one of the most common uses of the range distribution was to define availability for third-order resource selection (Shirk et al. 2014). But because these models describe the process resulting in a particular home range, they allow researchers to ask questions about why animals move the way they do. ...
Estimating animal home ranges and understanding the processes that influence home range behavior are important components of ecological investigations. While ecologists have been interested in the concept of home range for many decades, turning this conceptinto a usable statistical model or quantifiable metric for scientific purposes has been quite challenging. Keys to overcoming this challenge are (i) letting well-defined questions, and a solid understanding of the study system point to a specific home range metric to be quantified;(ii) choosing an appropriate estimator of this target metric;(iii) collecting location data during appropriate time periods and at the right sampling frequency; and (iv) understanding the trade-off between simplicity and complexity in modeling ecological data. If researchers follow these recommendations, many of the traditional challenges to studying animal home ranges will be alleviated due to technological advances in tracking systems and a solid foundation of models for understanding space use. However, we should continue to strive for more fundamental approaches to understanding animal movements. In particular, we see great opportunity for the continued development of agent-based models (ABM) of animal movements that allow for home range behavior to be an emergent property of the model instead of an a priori structure imposed on the data.
... It is well known that ecological relationships and observable patterns vary across spatial scales (e.g., Wiens 1989;Levin 1992;Chave 2013). Analysis conducted at inappropriate scales can lead to weak or incorrect inferences regarding the relationships between species responses and the environment (Thompson and McGarigal 2002;Shirk et al. 2014;Wan et al. 2017). Yet ecological studies that do not address the effect of scale remain ubiquitous, and those that claim to have investigated multiple scales typically select scales without thorough analysis (Wheatley and Johnson 2009;Jackson and Fahrig 2015;McGarigal et al. 2016). ...
Background
Fire size and severity have increased in the western United States in recent decades, and are expected to continue to increase with warming climate. Habitats for many species are threatened by large and high-severity fires, but the effect of spatial scale on the relationship between fires and habitat modifications is poorly understood. We used the 2011 Wallow Fire—the largest wildfire in the state history of Arizona, USA—as a case study and assessed changes in predicted nesting habitat of the threatened Mexican spotted owl ( Strix occidentalis Xántus de Vésey) in the first three years following the fire. Our objective was to explore potential relationships between burn severity and changes in habitat suitability at different spatial scales. To accomplish this, we applied a multi-scale optimized habitat selection model to pre- and post-fire landscapes and compared the differences in predictions along a continuous scale gradient.
Results
Fire effects on habitat quality were spatially variable and the strength and direction of relationships were scale-dependent. Spatial patterns of burn-severity mosaic resembled the patterns of habitat suitability change. High burn severity reduced nesting habitat suitability and this relationship was strongest at broad scales. Pre-fire habitat suitability was positively related to burn severity, again at fairly broad scales, but the relationship was weak. Low-severity fires had little effect on habitat suitability.
Conclusions
Multi-scale analysis may influence the statistical measures of goodness of fit in assessing fire effects on species and their habitats. Future studies should explicitly address spatial scale when quantifying fire effects.
... However, insular Pacific martens, mink, and river otters overlap in the freshwater riparian and coastal marine interfaces and their diets may overlap. This is especially true for mink and otters, which exclusively exploit these aquatic environments, while in moist coastal environments martens are strongly associated with forests and prey in the streamside environments [66]. Raccoons commonly forage in streams and riparian areas and insular populations are frequent foragers in the intertidal and shallow subtidal marine interface [67], where they overlap with martens and river otters in HG and VI, and coastal mink on VI. ...
Amdoparvovirus and Protoparvovirus are monophyletic viral genera that infect carnivores. We performed surveillance for and sequence analyses of parvoviruses in mustelids in insular British Columbia to investigate parvoviral maintenance and cross-species transmission among wildlife. Overall, 19.1% (49/256) of the tested animals were parvovirus-positive. Aleutian mink disease virus (AMDV) was more prevalent in mink (41.6%, 32/77) than martens (3.1%, 4/130), feline panleukopenia virus (FPV) was more prevalent in otters (27.3%, 6/22) than mink (5.2%, 4/77) or martens (2.3%, 3/130), and canine parvovirus 2 (CPV-2) was found in one mink, one otter, and zero ermines (N = 27). Viruses were endemic and bottleneck events, founder effects, and genetic drift generated regional lineages. We identified two local closely related AMDV lineages, one CPV-2 lineage, and five FPV lineages. Highly similar viruses were identified in different hosts, demonstrating cross-species transmission. The likelihood for cross-species transmission differed among viruses and some species likely represented dead-end spillover hosts. We suggest that there are principal maintenance hosts (otters for FPV, raccoons for CPV-2/FPV, mink for AMDV) that enable viral persistence and serve as sources for other susceptible species. In this multi-host system, viral and host factors affect viral persistence and distribution, shaping parvoviral ecology and evolution, with implications for insular carnivore conservation.
The Humboldt marten (Martes caurina humboldtensis) has declined from over 95 % of its historic range and currently occurs in just four extant population areas (EPAs). Prior to their listing under the Endangered Species Act, a conservation strategy was developed to identify key conservation needs for this species. This assessment identified an area near the California–Oregon (CA–OR) border as the second EPA in California, yet little was known about the overall distribution or habitat used by this population. This prompted our investigation to provide the first systematic survey of the CA–OR EPA and to assess habitat use under an occupancy modeling framework. Between 2017–2018 we surveyed 51 survey units in and around the EPA and detected martens at 20 (39.2 %). We found that occupancy was most influenced by the spatial scale-specific amount of low-elevation late-seral old-growth forest habitat, riparian habitat, and mid-seral forest habitat. Occupancy by marten was greatest in low-elevation (< 800 m) habitat and was positively associated with late-seral forest habitat at the 1,170-m home range scale (Odds Ratio [OR] = 35.31, 95 % CI = 1.30–958.07), riparian habitat at the 1,170-m home range scale (OR = 3.20, 95 % CI = 1.01–10.1), and increased amounts of mid-seral forest habitat at the 50-m microhabitat scale (OR = 1.28, 95 % CI = 0.95–1.73). Our findings identified habitat types important for explaining the distribution of this understudied population, addressing two of the highest priority research needs identified in the Humboldt marten conservation strategy.
Recent research, has shown that species-environmental relationships and habitat model predictions are often nonstationary in space, time and ecological context. This calls into question modeling approaches that assume a global, stationary ecological realized niche and use predictive modeling to describe it. This paper explores this issue by comparing the performance of predictive models for wildcat hybrid occurrence based on (1) global pooled data across individuals, (2) geographically weighted aggregation of individual models, (3) ecologically weighted aggregation of individual models, and (4) combinations of global, geographical and ecological weighting. Our study system included GPS telemetry data from 14 individual wildcat hybrids across Scotland. We developed predictive models both using Generalized Linear Models (GLM)and Random Forest machine learning to compare the performance of these differing algorithms and how they compare in stationary and nonstationary analyses. We validated the predicted models in four different ways. First, we used independent hold-out data from the 14 collared wildcat hybrids. Second, we used data from 8additional GPS collared wildcat hybrids from a previous study that were not included in the training sample. Third, we used sightings data sent in by the public and researchers and validated by expert opinion. Fourth, we used data collected by camera trap surveys between 2012 – 2021 from various sources to produce a combined camera trap dataset showing where wildcats and wildcat hybrids had been detected. Our results show that validation using hold-out data from the individuals used to train the model provides highly biased assessment of true model performance in other locations, with Random Forest in particular appearing to perform exceptionally(and inaccurately) well when validated by data from the same individuals used to train the models. Very different results were obtained when the models were validated using independent data from the three other sources. Each of these three independent validation data sets gave a different result in terms of the best overall model. The average of independent validation across these three validation datasets suggested that the best overall model produced for potential wildcat occurrence and habitat suitability was obtained by an ensemble average of the global Generalized Linear Model (GLM) and Random Forest models with the ecologically weighted GLM and Random Forest models. This suggests that the debate over whether which of GLM vs machine learning approaches is superior or whether global vs aggregated nonstationary modeling is superior may be a false choice. The results presented here show that the best prediction applies a combination of all of these approaches in an ensemble modeling framework
Landscape genetics provides a valuable framework to understand how landscape features influence gene flow and to disentangle the factors that lead to discrete and/or clinal population structure. Here, we attempt to differentiate between these processes in a forest-dwelling small carnivore [European pine marten (Martes martes)]. Specifically , we used complementary analytical approaches to quantify the spatially explicit genetic structure and diversity and analyse patterns of gene flow for 140 individuals genotyped at 15 microsatellite loci. We first used spatially explicit and nonspatial Bayesian clustering algorithms to partition the sample into discrete clusters and evaluate hypotheses of 'isolation by barriers' (IBB). We further characterized the relationships between genetic distance and geographical ('isolation by distance', IBD) and ecological distances ('isolation by resistance', IBR) obtained from optimized landscape models. Using a reciprocal causal modelling approach, we competed the IBD, IBR and IBB hypotheses with each other to unravel factors driving population genetic structure. Additionally, we further assessed spatially explicit indices of genetic diversity using sGD across potentially overlapping genetic neighbourhoods that matched the inferred population structure. Our results revealed a complex spatial genetic cline that appears to be driven jointly by IBD and partial barriers to gene flow (IBB) associated with poor habitat and interspecific competition. Habitat loss and fragmentation, in synergy with past overharvesting and possible interspecific competition with sympatric stone marten (Martes foina), are likely the main factors responsible for the spatial genetic structure we observed. These results emphasize the need for a more thorough evaluation of discrete and clinal hypotheses governing gene flow in landscape genetic studies, and the potential influence of different limiting factors affecting genetic structure at different spatial scales.
Background
A suite of mammalian species have experienced range contractions following European settlement and post-settlement development of the North American continent. For example, while North American martens (American marten, Martes americana ; Pacific marten, M. caurina ) generally have a broad range across northern latitudes, local populations have experienced substantial reductions in distribution and some extant populations are small and geographically isolated. The Humboldt marten ( M. c. humboldtensis ), a subspecies of Pacific marten that occurs in coastal Oregon and northern California, was recently designated as federally threatened in part due to its reduced distribution. To inform strategic conservation actions, we assessed Humboldt marten occurrence by compiling all known records from their range.
Methods
We compiled Humboldt marten locations since their rediscover to present (1,692 marten locations, 1996-2020). We spatially-thinned locations to 500-m to assess correlations with variables across contemporary Humboldt marten distribution (n=384). Using maximum entropy modeling (Maxent), we created distribution models with variables optimized for spatial scale; pre-selected scales were associated with marten ecology (50 to 1170 m radius). Marten locations were most correlated with abiotic factors (e.g., precipitation), which are unalterable and therefore uninformative within the context of restoration or management actions. Thus, we created variables to focus on hypothesized marten habitat relationships, including understory conditions such as predicted suitability of shrub species.
Results
Humboldt marten locations were positively associated with increased shrub cover (salal ( Gautheria shallon ), mast producing trees), increased pine ( Pinus sp ) overstory cover and precipitation at home-range spatial scales, areas with low and high amounts of canopy cover and slope, and cooler August temperatures. Unlike other recent literature on the species, we found little evidence that Humboldt marten locations were associated with old growth structural indices, perhaps because of a potential mismatch in the association between this index and shrub cover. As with any species distribution model, there were gaps in predicted distribution where Humboldt martens have been located during more recent surveys, for instance the southeastern portion of Oregon’s coast range. Conservation efforts and our assessment of potential risks to Humboldt marten populations would benefit from additional information on range extent, population sizes, and fine-scale habitat use. Like many rare and lesser-known species, this case study provides an example of how limited information can provide differing interpretations, emphasizing the need for study-level replication in ecology.
Aim
To evaluate the role of seasonal and non‐seasonal productivity fluctuations in global patterns of species richness.
Location
Worldwide.
Time period
2000–2017.
Major taxa studied
Amphibians, birds, mammals.
Methods
We analysed time series of monthly variation of the Normalized Difference Vegetation Index (NDVI), a surrogate of primary productivity, within c. 100 km × 100 km cells across all continents, estimating the mean, periodic seasonal variation and aperiodic unpredictable fluctuations of the NDVI in these cells. We then explored the relationships between mean NDVI and the components of its temporal variation and evaluated the independent effects of the above‐mentioned variables on species richness in the three vertebrate groups by means of variation partitioning.
Results
There is a hump‐shaped relationship between mean productivity and variation in productivity, so that temporal variation in productivity is lowest in regions with minimum and maximum values of mean productivity. Although mean productivity is a strong determinant of species richness, both seasonal and non‐seasonal productivity variation significantly affect the species richness of all studied taxa when accounting for mean productivity. However, the direction of these effects differs between regions differing in the mean productivity level. High variation in productivity has a negative effect on species richness in regions with moderate to high productivity levels, whereas species richness is higher in arid regions with high variation in productivity.
Main conclusions
Species richness is affected by temporal variation in productivity, but these effects differ regionally. In productive areas, high environmental stochasticity may increase population extinction rates, whereas arid regions probably benefit from resource fluctuations that promote species coexistence. Our results indicate that contemporary changes in patterns of temporal resource fluctuations may affect future global patterns of biological diversity on Earth.
Purpose of Review
This review summarizes the characteristics of contemporary multi-scale studies (i.e., those that quantify ecological relationships over multiple spatial extents), outlines limitations associated with analyses typically used to model multi-scale ecological relationships, and highlights recent methodological progress overcoming these limits.
Recent Findings
The majority of recent studies investigating scales of effect in ecological relationships were conducted within a simplified model selection, information theoretic framework that suffers under the specters of collinearity, imperfect detection, and limited candidate scale space. The few studies that deviate from this framework offer greater flexibility in identifying or directly estimating relevant scales of effect by allowing predictors to have independent scales of effect, simultaneously evaluated, and explicitly consider the unique issue of collinearity in multi-scale studies, with extensions to hierarchical models that account for imperfect detection.
Summary
Accurate ecological inference requires that the scales of ecological processes, observations, and analyses align. Wider adoption of emerging analytical methods for identifying important scales of effect, though not traditional, and in some cases more computationally intensive, will broaden the scope for understanding the ecological factors operating on organisms at different scales. By estimating full models (i.e., not single-predictor models), which are not scale constrained (i.e., each predictor can have its own scale of effect), and that disentangle ecological processes from observational processes when relevant will provide a comprehensive base of estimates from which to explore underlying mechanisms governing scale dependence in ecology.
We used bivariate scaling and logistic regression to investigate multiple-scale habitat selection by American marten (Martes americana). Bivariate scaling reveals dramatic differences in the apparent nature and strength of relationships between marten occupancy and a number of habitat variables across a range of spatial scales. These differences include reversals in the direction of an observed association from positive to negative and frequent dramatic changes in the apparent importance of a habitat variable as a predictor of marten occurrence. Logistic regression on the optimally scaled input variables suggests that at the scale of home ranges, marten select landscapes with high average canopy closure and low fragmentation. Within these low fragmented landscapes, marten select foraging habitat at a fine scale within late-seral, middle-elevation mesic forests. In northern Idaho, optimum American marten habitat, therefore, consists of landscapes with low road density, low density of non-forest patches with high canopy closure, and large areas of middle-elevation, late successional mesic forest. Comparison of current landscape conditions to those expected under the historic range of variability indicates that road building and timber harvest in the past century may have substantially reduced the amount of suitable marten habitat in northern Idaho. Our results are generally consistent with previous research in the Rocky Mountains, with additional insights related to the relative importance, functional form, and scale at which each habitat variable has the largest influence on marten occurrence.
A review of the commonly measured habitat factors that affect American martens and fishers is presented. Preferences for cover types and for specific habitat features by the North American Martes are compared, and biological principles are identified that may explain habitat selection by these species.
Modern ecological research often involves the comparison of the usage of habitat types or food items to the availability of those resources to the animal. Widely used methods of determining preference from measurements of usage and availability depend critically on the array of components that the researcher, often with a degree of arbitrariness, deems available to the animal. This paper proposes a new method, based on ranks of components by usage and by availability. A virtue of the rank procedure is that it provides comparable results whether a questionable component is included or excluded from consideration. Statistical tests of significance are given for the method. The paper also offers a hierarchical ordering of selection processes. This hierarchy resolves certain inconsistencies among studies of selection and is compatible with the analytic technique offered in this paper.
A combination of discriminant function analysis and multiple regression was used to develop a linear model of American marten Martes americana winter habitat use within second-growth boreal forest in northeastern Ontario, Canada. Four structural variables significantly discriminated between sites that were used or not used by martens: the percentage of spruce or fir trees, tree height, the number of downed logs, and canopy closure. The model was tested against a second data set and was not invalidated. The results demonstrated that martens were using second-growth forests in Ontario, and that their response to structural characteristics was similar to responses described previously in uncut forests.
Population connectivity is mediated by the movement of organisms or propagules through landscapes. However, little is known about how variation in the pattern of landscape mosaics affects the detectability of landscape genetic relationships. The goal of this paper is to explore the impacts of limiting factors on landscape genetic processes using simulation modeling. We used spatially explicit, individual-based simulation modeling to quantify the effects of habitat area, fragmentation and the contrast in resistance between habitat and non-habitat on the apparent strength and statistical detectability of landscape genetic relationships. We found that landscape genetic effects are often not detectable when habitat is highly connected. In such situations landscape structure does not limit gene flow. We also found that contrast in resistance values between habitat and non-habitat interacts with habitat extensiveness and fragmentation to affect detectability of landscape genetic relationships. Thus, the influence of landscape features critical to providing connectivity may not be detectable if gene flow is not limited by spatial patterns or resistance contrast of these features. We developed regression equations that reliably predict whether or not isolation by resistance will be detected independently of isolation by distance as a function of habitat fragmentation and contrast in resistance between habitat and non-habitat.
Spatially explicit information on the species composition and structure of forest vegetation is needed at broad spatial scales for natural resource policy analysis and ecological research. We present a method for predictive vegetation mapping that applies direct gradient analysis and nearest-neighbor imputation to ascribe detailed ground at - tributes of vegetation to each pixel in a digital landscape map. The gradient nearest neighbor method integrates vegeta - tion measurements from regional grids of field plots, mapped environmental data, and Landsat Thematic Mapper (TM) imagery. In the Oregon coastal province, species gradients were most strongly associated with regional climate and geographic location, whereas variation in forest structure was best explained by Landsat TM variables. At the regional level, mapped predictions represented the range of variability in the sample data, and predicted area by vegetation type closely matched sample-based estimates. At the site level, mapped predictions maintained the covariance structure among multiple response variables. Prediction accuracy for tree species occurrence and several measures of vegetation structure and composition was good to moderate. Vegetation maps produced with the gradient nearest neighbor method are appropriately used for regional-level planning, policy analysis, and research, not to guide local management deci - sions.
A clear understanding of habitat associations of martens (Martes americana) is necessary to effectively manage and monitor populations. However, this information was lacking for martens in most of their southern range, particularly during the summer season. We studied the distribution and habitat correlates of martens from 2004 to 2006 in Rocky Mountain National Park (RMNP) across 3 spatial scales: site-specific, home-range, and landscape. We used remote-sensored cameras from early August through late October to inventory occurrence of martens and modeled occurrence as a function of habitat and landscape variables using binary response (BR) and binomial count (BC) logistic regression, and occupancy modeling (OM). We also assessed which was the most appropriate modeling technique for martens in RMNP. Of the 3 modeling techniques, OM appeared to be most appropriate given the explanatory power of derived models and its incorporation of detection probabilities, although the results from BR and BC provided corroborating evidence of important habitat correlates. Location of sites in the western portion of the park, riparian mixed-conifer stands, and mixed-conifer with aspen patches were most frequently positively correlated with occurrence of martens, whereas more xeric and open sites were avoided. Additionally, OM yielded unbiased occupancy values ranging from 91% to 100% and 20% to 30% for the western and eastern portions of RMNP, respectively.
Coarse woody debris (CWD) is an important component of temperate stream and forest ecosystems. This chapter reviews the rates at which CWD is added and removed from ecosystems, the biomass found in streams and forests, and many functions that CWD serves. CWD is added to ecosystems by numerous mechanisms, including wind, fire, insect attack, pathogens, competition, and geomorphic processes. Despite the many long-term studies on tree mortality, there are few published rates of CWD input on mass-area-1 time-1 basis. CWD is significantly transformed physically and chemically. Movement of CWD, especially in streams, is also an important but poorly documented mechanism whereby CWD is lost from ecosystems. Many factors control the rate at which CWD decomposes, including temperature, moisture, internal gas composition of CWD, substrate quality, size of CWD, and types of organisms involved. However, the importance of many of these factors has yet to be established in field experiments. CWD performs many functions in ecosystems, serving as autotrophic and heterotrophic habitat and strongly influencing geomorphic processes, especially in streams. It is also a major component of nutrient cycles in many ecosystems and is an important functional component of stream and forest ecosystems.
I systematically sampled American marten Martes americana presence in two large study areas in the Selkirk and Purcell Mountains of southwest Canada using hair removal traps and tracks in snow. Both study areas were mostly forested and contained a broad cross-section of stand ages including abundant early seral and mature forest. I extracted measures of forest structure and dominant tree species, climax ecosystem types and human use from digital resource databases and used multiple logistic regression to model habitat selection of martens. I summarized data in windows of 100 m to 10 km in radius around each sample location to investigate the effect of varying data resolution on habitat selection. Marten detection at hair sites was positively related to temperature and trap duration and negatively related to snowfall while the trap was set. Martens were detected in all habitats sampled including recently logged areas, regenerating stands, dry Douglas-fir Pseudosuga menziesii forest and subalpine parkland. Overall selection was mildly greater using mean habitat values in 100 m and 2 km radius windows for both study areas. Martens selected for greater crown closure and older stands at the finer resolution; no selection for forest structure was detected at the larger resolution except that martens selected against increased overstory heterogeneity as measured by the standard deviation of crown closure (within the window). Martens preferred coniferous stands over deciduous dominated stands and were more abundant in wetter than in dryer ecosystems. Selection for ecosystems and stand types was stronger in the larger window size. At the intensity sampled in this study, neither road density nor logging appeared to affect marten habitat selection when I accounted for variation in ecosystems and stand structure. This study examined habitat selection at relatively coarse scales; stronger associations with forest structure may be expected at finer scales. In addition, roads or logging may influence habitat selection below the scale of my analysis.
As the concepts of landscape ecology have been incorporated into otherdisciplines, the influence of spatial patterns on animal abundance anddistribution has attracted considerable attention. However, there remains asignificant gap in the application of landscape ecology theories and techniquesto wildlife research. By combining landscape ecology techniques withtraditionalwildlife habitat analysis methods, we defined an organism-centeredperspectivefor breeding bald eagles (Haliaeetus leucocephalus) alongthe Hudson River, New York, USA. We intensively monitored four pairs ofbreedingeagles during the 1999 and 2000 breeding seasons, and collected detailedinformation on perch and forage locations. Our analysis focused on threecritical habitat elements: available perch trees, access to foraging areas, andfreedom from human disturbance. We hypothesized that eagle habitat selectionrelative to each of these elementswould vary with the spatial scale of analysis, and that these scalingrelationships would vary among habitat elements. We investigated two elementsofspatial scale: grain and local extent. Grain was defined as the minimum mappingunit; local extent was defined by the size of an analysis window placed aroundeach focal point. For each habitat element, we quantified habitat use over arange of spatial scales. Eagles displayed scale-dependent patterns of habitatuse in relation to all habitat features, including multi-scale andthreshold-like patterns. This information supports the existence ofscale-dependant relationships in wildlife habitat use and allowed for a moreaccurate and biologically relevant evaluation of Hudson River breeding eagle habitat.
We used field surveys and Geographic Information System data to identify landscape-scale habitat associations of American
martens (Martes americana) and to develop a model to predict their occurrence in northeastern California. Systematic surveys using primarily enclosed
track plates, with 10-km spacing, were conducted across a 27,700km2 area of largely forested, mountain terrain. Martens were detected at 20/184 (10.8%) of the sample units, aggregated in three
distinct regions. We investigated habitat selection at multiple scales using circular assessment areas of 3, 20, and 80km2. The model for the largest assessment area best fit the data and included the following predictors: amount of reproductive
habitat, number of habitat patches and land ownership category. These results support the hypothesis that martens select habitat
based upon broad scale landscape conditions and that these conditions vary with ownership. We tested the model using an independent
set of data, collected primarily during the winter. Poor fit of the test data in some locations raised concerns that our model,
which was developed using data collected during the snow-free season, may not predict winter distribution well. We are investigating
possible causes for the seasonal variation and until they can be incorporated our model represents a conservative view of
marten habitat suitability based on summer occupancy. During the summer months, which is the reproductive season, martens
are predicted to occur largely in relatively undisturbed landscapes where high-elevation, late-successional forests are common.
Individual-based analyses relating landscape structure to genetic distances across complex landscapes enable rigorous evaluation
of multiple alternative hypotheses linking landscape structure to gene flow. We utilize two extensions to increase the rigor
of the individual-based causal modeling approach to inferring relationships between landscape patterns and gene flow processes.
First, we add a univariate scaling analysis to ensure that each landscape variable is represented in the functional form that
represents the optimal scale of its association with gene flow. Second, we use a two-step form of the causal modeling approach
to integrate model selection with null hypothesis testing in individual-based landscape genetic analysis. This series of causal
modeling indicated that gene flow in American marten in northern Idaho was primarily related to elevation, and that alternative
hypotheses involving isolation by distance, geographical barriers, effects of canopy closure, roads, tree size class and an
empirical habitat model were not supported. Gene flow in the Northern Idaho American marten population is therefore driven
by a gradient of landscape resistance that is a function of elevation, with minimum resistance to gene flow at 1500m.
KeywordsLandscape genetics-Scale dependency-Causal modeling-American marten-Population connectivity-Gene flow
In mobile animals, movement behavior can maximize fitness by optimizing access to critical resources and minimizing risk of
predation. We sought to evaluate several hypotheses regarding the effects of landscape structure on American marten foraging
path selection in a landscape experiencing forest perforation by patchcut logging. We hypothesized that in the uncut pre-treatment
landscape marten would choose foraging paths to maximize access to cover types that support the highest density of prey. In
contrast, in the post-treatment landscapes we hypothesized marten would choose paths primarily to avoid crossing openings,
and that this would limit their ability to optimally select paths to maximize foraging success. Our limiting factor analysis
shows that different resistant models may be supported under changing landscape conditions due to threshold effects, even
when a species’ response to landscape variables is constant. Our results support previous work showing forest harvest strongly
affects marten movement behavior. The most important result of our study, however, is that the influence of these features
changes dramatically depending on the degree to which timber harvest limits available movement paths. Marten choose foraging
paths in uncut landscapes to maximize time spent in cover types providing the highest density of prey species. In contrast,
following landscape perforation by patchcuts, marten strongly select paths to avoid crossing unforested areas. This strong
response to patch cutting reduces their ability to optimize foraging paths to vegetation type. Marten likely avoid non-forested
areas in fragmented landscapes to reduce risk of predation and to benefit thermoregulation in winter, but in doing so they
may suffer a secondary cost of decreased foraging efficiency.
KeywordsAmerican marten–
Martes americana
–Path-selection–Fragmentation–Limiting-factors
Kernel density estimators are becoming more widely used, particularly as home range estimators. Despite extensive interest in their theoretical properties, little empirical research has been done to investigate their performance as home range estimators. We used computer simulations to compare the area and shape of kernel density estimates to the true area and shape of multimodal two-dimensional distributions. The fixed kernel gave area estimates with very little bias when least squares cross validation was used to select the smoothing parameter. The cross-validated fixed kernel also gave surface estimates with the lowest error. The adaptive kernel overestimated the area of the distribution and had higher error associated with its surface estimate.
A species' range limits may shift in space either because of changes in ecological factors (e.g. climate, the abundances of interacting species) or because the traits which influence range limits (e.g. dispersal, niche characteristics) evolve by natural selection. In Geographical Ecology, Robert MacArthur (1972) provided a synthesis of the evolutionary factors that may drive range dynamics. In this paper, I revisit this theme in the light of recent theoretical studies of evolution in heterogeneous environments. These studies suggest that a rich range of evolutionary patterns in species' ranges may occur, including expansions or contractions, leading to dynamism in ranges even in epochs without strong directional environmental change.
Martes americana rested primarily in subnivean sites associated with coarse woody debris, including logs and stumps. Use of spruce (Picea spp.)-fir (Abies lasiocarpa) stands by adults was greater than expected and use of lodgepole pine Pinus contorta stands was less than expected on the basis of spatial availability. Juveniles used stand types in proportion to spatial availability. Fidelity to individual resting sites and to subnivean sites associated with coarse woody debris was highest among adults. Above-snow sites were used during the warmest weather; subnivean sites associated with coarse woody debris were used during the coldest weather. Marten rested for longer periods where coarse woody debris formed all or part of the resting site than they did at other sites. -from Authors
Mark-recapture was conducted from 1981 through 1983 in the Cascade Range of Oregon to examine the use of riparian and upland habitats by small mammals. Both number of individuals and species richness were greater in riparian than in upland areas. Because of the occurrence of several less-commonly captured species in riparian locations, species evenness was greatest in the upland. Among Insectivora, Sorex monticolus and S. trowbridgii were captured most often on riparian sites. Among insectivores adult males weighed more in riparian zones. Three rodents (Peromyscus maniculatus, Microtus oregoni, and Zapus trinotatus) were captured more frequently in riparian habitats; two rodents (Tamias townsendii and Clethrionomys californicus) were captured more frequently in upland habitats. Among all rodents, except Z. trinotatus, adult males weighed more in riparian areas. Mustela erminea was captured more frequently in riparian sites. Eight species had greater numbers of adults in breeding condition in riparian than in upland habitats. Conversely, six species had greater numbers of juveniles in the upland. Data from several species, including T. townsendii, P. maniculatus, Z. trinotatus, and Glaucomys sabrinus, indicated that riparian habitats act as a species source and upland areas act as a dispersal sink.
Food habits and habitat selection of marten (Martes americana) were studied over two winters in Yosemite National Park, California. Analysis of 91 droppings showed substantial differences in diet between the two winters. White-tailed jack rabbits (Lepus townsendii) and voles (Microtus spp.) were the principal food items in 1976-77 and 1979-80, respectively. Habitat use was determined from following 34.8 km of marten tracks. Travel routes occurred in all habitat types, but marten did not rest or hunt in meadows or on granite domes. Marten selected dense cover less than 3 m above snow level and used tree trunks and other objects for access to the subnivean zone. Logs served as den sites and hunting areas, and rocks were used for scent and urine marking. Mixed-aged forests were important in furnishing subnivean access and protective cover over a wide range of snow conditions.
Wildlife ecology and management studies have been widely criticized for deficiencies in design or analysis. Manipulative experiments-with controls, randomization, and replication in space and time-provide powerful ways of learning about natural systems and establishing causal relationships, but such studies are rare in our field. Observational studies and sample surveys are more common; they also require appropriate design and analysis. More important than the design and analysis of individual studies is metareplication: replication of entire studies. Similar conclusions obtained from studies of the same phenomenon conducted under widely differing conditions will give us greater confidence in the generality of those findings than would any single study, however well designed and executed.
American marten (Martes americana) are often associated with mature conifer-dominated forests; however, marten may require complex vertical and horizontal woody structure, rather than a coniferous overstory. To evaluate whether marten prefer forests with a mature conifer overstory, we investigated landscape- and stand-scale habitat selection of 38 (20 M, 18 F) resident (≥25 locations collected during ≥3 months), nonjuvenile (≥1 year [YR]) marten in a forest preserve in northcentral Maine where trapping and timber harvesting had been excluded for >35 years. At the landscape scale, nearly all of the available habitat was used by marten. Further, composition of forest types within home ranges of radiocollared resident marten was similar to areas outside of home ranges. At the stand scale, marten used forest types disproportionately (P = 0.03) from availability within their home ranges during summer (1 May-31 Oct), but exhibited no apparent selection (P = 0.83) during winter. During summer, stands with substantial spruce-budworm (Choristoneura fumiferana)-caused mortality (
Multivariate Graphical Displays Transforming to Normality Distributional Tests and Plots Robust Estimation Outlying Observations
FIA provides the highest quality, most comprehensive data available on forests in the United States. In the last few years, these data have been made more readily available to a variety of clients for managerial, policy, and scientific applications. Beyond providing data, FIA is interested in working with a variety of collaborators to make the best use of FIA data in diverse application areas. In particular, within the FIA program there is an increased emphasis on remote sensing and other geospatial data and analyses. In the uncertain fiscal environments faced by land management agencies, remote sensing data provide opportunities to develop information in a more cost effective manner. Providing forest inventory information spatially facilitates access by the general public, supports land managers, and will result in better understanding of our Nation's forest resources.
Marten (Martes americana) use of successional forest stages in north-central Washington was determined during winter by counting marten tracks along 65 transects in 6 different forest stages. Red squirrel (Tamiasciurus hudsonicus) tracks and middens were counted and small rodents were trapped to determine association of marten activity with successional stages used by these species. Marten frequented older aged (>82 years old) Engelmann spruce (Picea engelmannii)/subalpine fir (Abies lasiocarpa) and lodgepole pine (Pinus contorta) forests where voles (Microtus spp. and Clethrionomys spp.) and red squirrel middens were available. Younger-aged (≤43 years old) forests were used infrequently by marten. Mature and old-growth forests are important habitats for marten, providing voles for prey during critical winter periods. For these reasons, mature and old-growth stands should be managed for marten.
Introduction Multivariate Normal Distribution Wishart Distribution Hotelling's T2 Distribution Multivariate Beta Distributions Rao's Distribution Multivariate Skewness and Kurtosis
To better understand how a species known to thermoregulate behaviorally switches among microsites and habitats in response to weather and snow, we studied influences of weather and snow conditions on resting by the American marten, Martes americana. Vertical location of resting sites varied with air temperature and snowfall during the previous 24 h. Subnivean resting was most likely when air temperature was low and when recent snowfall had been heavy, and tended to be in stands dominated by spruce-fir. Supranivean resting tended to occur when weather was warmer and when recent snowfall had been light, and tended to occur in stands dominated by lodgepole pine (Pinus contorta), the predominant conifer in the study area. Fidelity of martens to resting sites varied with season; martens reused sites more in winter than in spring, hypothetically a result of trading-off increased energy savings accrued from using a few especially efficient sites against longer distances traveled to reach them. Such travel may not be rewarded in warm weather. Stand characteristics associated with resting in cold snowy winter periods are typical of low disturbance frequencies, including old-growth conditions in the Rocky Mountains.
The marten subspecies on the island of Newfoundland, Martes americana atrata, is threatened. Survey data suggest that most of the extant marten population lives in old uncut balsam fir (Abies balsamea) forests, but a very few live in adjacent 40- to 60-year-old second-growth stands of balsam fir. We compared habitat structure and composition and prey abundance in old forest and second-growth stands to test the hypotheses that either food abundance or habitat quality, or both, limit use of the 40- to 60-year-old forests by marten. Snowshoe hares (Lepus americana) were most abundant in 40-year-old forests and also occurred in old forests, but field voles (Microtus pennsylvanicus) were not found in second-growth stands. A multivariate discriminant model indicated that older, uncut forests contained more structure than younger forests al ground level, because there was more woody debris, more young balsam fir, less litter, more mosses, and more low shrubs. Canopy cover was similar in all forest types, and subnivean access did not differ among the three age-classes when snow was about 1 m deep. We suggest that marten did not use 40- or 60-year-old forest stands because of the lack of the meadow voles that form a necessary part of their diet. Meadow voles likely respond to ground-level forest structure in selecting habitat, and this structure is unavailable in young forests. We recommend a management strategy for marten that would preserve current old forests as long as possible and allow sufficient second-growth balsam fir forest to become old forest with the required characteristics to maintain a viable marten population.
Publisher Summary
This chapter reviews the rates at which Coarse Woody Debris (CWD) is added and removed from ecosystems, the biomass found in streams and forests, and many functions that CWD serves. CWD is an important component of temperate stream and forest ecosystems and is added to the ecosystem by numerous mechanisms, including wind, fire, insect attack, pathogens, competition, and geomorphic processes. Many factors control the rate at which CWD decomposes, including temperature, moisture, the internal gas composition of CWD, substrate quality, the size of the CWD, and the types of organisms involved. The mass of CWD in an ecosystem ideally represents the balance between addition and loss. In reality, slow decomposition rates and erratic variations in input of CWD cause the CWD mass to deviate markedly from steady-state projections. Many differences correspond to forest type, with deciduous-dominated systems having generally lower biomass than conifer-dominated systems. Stream size also influences CWD mass in lotic ecosystems, while successional stage dramatically influences CWD mass in boat aquatic and terrestrial settings. This chapter reviews many of these functions and concludes that CWD is an important functional component of stream and forest ecosystems. Better scientific understanding of these functions and the natural factors influencing CWD dynamics should lead to more enlightened management practices.
A technique for using kernel density estimates to investigate the number of modes in a population is described and discussed. The amount of smoothing is chosen automatically in a natural way.
Female yellow-headed blackbirds in eastern Washington State settle to nest at higher densities on marshes with higher emergence rates of odonates, the most important prey delivered to nestlings. However, settling densities of females were not correlated with odonate emergence rates on individual territories or on individual territories plus adjacent ones. Apparently, females assessed production of insects on breeding marshes at the time they settled, and they used this information when making settling decisions. However, they selected nest sites on the basis of vegetation density rather than food availability. The complexity of decision making by female yellowheads would not have been detected had our analysis been restricted to one spatial scale. Because interpretations of habitat selection behavior are scale-dependent, careful attention to scale and performing analyses on more than one spatial scale are essential in studies of habitat selection.
Home range and habitat utilization data for adult marten were gathered from May to September 1980 in northwestern Maine. Analysis of 455 radio locations of three postlactating females and two adult males showed that overall summer ranges averaged 2.9 km2 for females and 5.6 km2 for males, with females showing preferential use of softwood stands. The frequency distribution of activity radii differed during this period for females but not for males, while use of habitats did not change. Thirty-eight resting sites and dens were located; 6 of 21 sites used by females were identified as maternal dens. All resting sites used by males were in tree canopies, commonly in "witches brooms" (abnormal clumped growth of balsam fir branches caused by rust fungi). Den characteristics are discussed in relation to the presence and development of kits.
To better understand how a species known to thermoregulate behaviorally switches among microsites and habitats in response to weather and snow, we studied influences of weather and snow conditions on resting by the American marten, Martes americana. Vertical location of resting sites varied with air temperature and snowfall during the previous 24 h. Subnivean resting was most likely when air temperature was low and when recent snowfall had been heavy, and tended to be in stands dominated by spruce-fir. Supranivean resting tended to occur when weather was warmer and when recent snowfall had been light, and tended to occur in stands dominated by lodgepole pine (Pinus contorta), the predominant conifer in the study area. Fidelity of martens to resting sites varied with season; martens reused sites more in winter than in spring, hypothetically a result of trading-off increased energy savings accrued from using a few especially efficient sites against longer distances traveled to reach them. Such travel may not be rewarded in warm weather. Stand characteristics associated with resting in cold snowy winter periods are typical of low disturbance frequencies, including old-growth conditions in the Rocky Mountains.
Because marten (Martes americana) require subnivean access for cover, prey access, and homeothermic reasons, we developed a predictive model to explain their differential use of subnivean access holes in Yellowstone National Park. We included prey biomass and percent ground cover of coarse woody debris (CWD) as explanatory variables in a logistic regression model because of their biological importance to marten in winter. Taken singly, relative prey biomass yielded the best univariate predictive model (P = 0.001). However, we included CWD in a multivariate model because of its biological significance. Coarse woody debris provides structure that intercepts snowfall, creating subnivean tunnels, interstitial spaces, and access holes, and was found at used and unused access points. Mean prey biomass was 205.4 g/400 M2 (SE = 20.26) and 108.2 g/400 m2 (SE = 10.73) at used and unused points (P < 0.001), respectively, while mean percent ground cover of CWD was 24.7 (SE = 2.30) and 18.5% (SE = 1.18) at used and unused access points (P = 0.017), respectively. As CWD increased by 5%, the probability of use by marten increased 1.12 times, and for every 50 g increase in relative prey biomass, marten were 1.37 times more likely to use that access point. Prey biomass varied (P < 0.001) among subnivean access points, and marten chose between different access points primarily on the basis of prey abundance levels. Older growth forests with accumulated CWD will enable marten to forage effectively in winter.
Logistic regression is an important tool for wildlife habitat-selection studies, but the method frequently has been misapplied due to an inadequate understanding of the logistic model, its interpretation, and the influ- ence of sampling design. Topromote better use of this method, we review its application and interpretation under 3 sampling designs: random, case–control, and use–availability. Logistic regression is appropriate for habitat use–nonuse,studies employing,random,sampling,and can be used to directly model,the conditional,probability of use in such cases. Logistic regression also is appropriate for studies employing case–control sampling designs, but careful attention is required to interpret results correctly. Unless bias can be estimated or probability of use is small for all habitats, results of case–control studies should be interpreted as odds ratios, rather than probability of use orrelative probability of use. When data are gathered under a use–availability design, logistic regression can be used to estimate approximate odds ratios if probability of use is small, at least on average. More generally, howev- er, logistic regression is inappropriatefor modeling habitat selection in use–availability studies. In particular, using logistic regression to fit the exponential,model,of Manly et al. (2002:100) does not guarantee,maximum-likelihood estimates, valid probabilities, or valid likelihoods. We show that the resource selection function (RSF) commonly used for the exponential model is proportional to a logistic discriminant function. Thus, it may be used to rank habitats with respect to probability of use and to identify important habitat characteristics or their surrogates, but itis not guaranteed ,to be ,proportional ,to probability ,of use. Other problems ,associated with the exponential model,also are discussed. We describe,an alternative,model,based on Lancaster,and Imbens (1996) that offers a method for estimating conditional probability of use in use–availability studies. Although promising, this model fails to converge ,to a ,unique ,solution in some ,important ,situations. Further work ,is needed ,to obtain ,a robust method,that is broadly applicable to use–availability studies. JOURNAL OF WILDLIFE MANAGEMENT 68(4):774–789 Key words: bias, case–control, contaminated control, exponential model, habitat modeling, log-binomial model, logis-
American marten (Martes americana), a species sensitive to intensive logging, is often associated with old-growth coniferous forest. Recent results, however, question the specificity of this association. We studied habitat selection of marten at the southern fringe of the boreal forest, in mixed forest types. The presence of a white-tailed deer (Odocoileus virginianus) wintering area protected part of the study area from large-scale intensive logging [clear-cuts with protection of regeneration and soils (CPRS) and precommercial thinning (PCT)] but not from partial logging of mature coniferous stands. We radio-tracked 15 marten over 2 y and analyzed habitat selection at 2 scales: landscape and home range. Marten selected mature (> 60 y) coniferous forests at both scales, while they neither selected nor avoided Mature mixed forests. PCT forests (0-15 y old) were strongly avoided at the landscape scale (P < 0.001), as were Young forests (21-60 y old; P = 0.005). At the home range scale, marten avoided CPRS (0-20 y old; P < 0.001). Partial logging had no effect on selection at either scale. Female home ranges were smaller in the partially logged sector of the study area (2.6 ± 0.6 versus 7.4 ± 0.2 km2), while male home range averaged 5.5 ± 1.0 km2, resulting in a significant interaction between location of home ranges in the white-tailed deer wintering area and sex of individuals (F1,11 = 5.618, P = 0.037). Also, home ranges tended to be larger as the road density and proportion of light outbreak cover type increased. Our results showed that partial logging, CPRS, and PCT have different impacts on marten habitat selection and use of space. We conclude that partial logging rather than clear-cuts and precommercial thinning should be favored for conservation of American marten.
We investigated habitat selection and home-range characteristics of American martens (Martes americana) that occupied home ranges with partially harvested stands characterized by basal area of trees <18 m2/ha and canopy closure <30%. During the leaf-on season (1 May–31 Oct), martens selected second-growth (80–140-years-old, >9-m tree height) forest stands (deciduous, coniferous, and mixed coniferous-deciduous) and mixed stands that were partially harvested (x̄ = 13 m2/ha residual basal area, >9-m tree height), and they selected against forests regenerating after clearcutting (≤6-m tree height, cuts ≤24-years-old). Marten home ranges included a greater proportion of partially harvested stands during the leaf-on season (maximum = 73%) than during leaf-off (1 Nov–30 Apr; maximum = 34%). Higher use of partially harvested stands during the leaf-on season coincided with greater canopy closure, higher use of small mammals, and greater relative densities of small mammals. During the leaf-off season, martens exhibited reduced relative selection for partially harvested and regenerating stands and increased selection for second-growth forest types. Partially harvested and regenerating clearcut stands had canopy closure <30% and basal area of trees >9-m tall of <13 m2/ha; both were below published thresholds required by martens. Coincidentally, home-range areas of martens increased during the leaf-off season to include a greater proportion of second-growth forest and less partially harvested forest. Further, martens with partial harvesting in their home ranges used areas almost twice as large during the leaf-off season as martens with no partial harvesting. Snowshoe hares (Lepus americanus) were prevalent prey for martens during the leaf-off season, and partially harvested stands had the lowest density of hares among all forest overstory types. Our findings suggest that the combination of insufficient basal area and overhead canopy closure, subnivean behavior of small mammals, increased reliance on hares, and reduced density of snowshoe hares relative to second-growth forest types reduced habitat quality in partially harvested stands during the leaf-off season. We suggest land managers retain basal areas >18 m2/ha and canopy closure >30% during winter to maximize use by martens in stands where partial harvesting is practiced.
This is an R package (a piece of Software) to fit and do inference on mixed-effects models.
The package is Free Software (hence open-source) and the package and much documentation about it is freely available from CRAN at
https://cran.r-project.org/package=lme4
In recent years the use of species distribution models by ecologists and conservation managers has increased considerably, along with an awareness of the need to provide accuracy assessment for predictions of such models. The kappa statistic is the most widely used measure for the performance of models generating presence–absence predictions, but several studies have criticized it for being inherently dependent on prevalence, and argued that this dependency introduces statistical artefacts to estimates of predictive accuracy. This criticism has been supported recently by computer simulations showing that kappa responds to the prevalence of the modelled species in a unimodal fashion.
In this paper we provide a theoretical explanation for the observed dependence of kappa on prevalence, and introduce into ecology an alternative measure of accuracy, the true skill statistic (TSS), which corrects for this dependence while still keeping all the advantages of kappa. We also compare the responses of kappa and TSS to prevalence using empirical data, by modelling distribution patterns of 128 species of woody plant in Israel.
The theoretical analysis shows that kappa responds in a unimodal fashion to variation in prevalence and that the level of prevalence that maximizes kappa depends on the ratio between sensitivity (the proportion of correctly predicted presences) and specificity (the proportion of correctly predicted absences). In contrast, TSS is independent of prevalence.
When the two measures of accuracy were compared using empirical data, kappa showed a unimodal response to prevalence, in agreement with the theoretical analysis. TSS showed a decreasing linear response to prevalence, a result we interpret as reflecting true ecological phenomena rather than a statistical artefact. This interpretation is supported by the fact that a similar pattern was found for the area under the ROC curve, a measure known to be independent of prevalence.
Synthesis and applications . Our results provide theoretical and empirical evidence that kappa, one of the most widely used measures of model performance in ecology, has serious limitations that make it unsuitable for such applications. The alternative we suggest, TSS, compensates for the shortcomings of kappa while keeping all of its advantages. We therefore recommend the TSS as a simple and intuitive measure for the performance of species distribution models when predictions are expressed as presence–absence maps.
Few studies have examined the potential for clearcutting to fragment habitat of area-sensitive, forest-dependent mammals such as American marten (Martes americana). We examined relationships among measures of landscape pattern and spatial use of habitat by 33 resident and 32 nonresident, adult marten that were radio-monitored in an extensively logged landscape. Size and distribution of forest patches (trees over 6 m in height) were associated with patch use by marten. Patches of forest used by resident marten (median = 27 ha, n = 12) were larger ( p < 0.003) than patches with no observed use (median = 1.5 ha, n = 128). Further, patches used by residents were closer to the nearest patch larger than 2.7 ha (38 m versus 55 m;p = 0.057) and to an adjacent forest preserve (2.5 km versus 3.5 km;p = 0.075) than patches with no observed use. At four spatial scales (10, 65, 125, and 250 ha), grid cells used by resident marten comprised a greater percentage of residual forest over 6 m in height ( p≤ 0.008) and intersected forest patches of greater area ( p≤ 0.006) than cells with no observed use. Edge indices were not different ( p≥ 0.490) between used grid cells and cells with no observed use at any of the four spatial scales. Analyses of forest edge associations indicated that marten did not avoid residual-regenerating forest edge within home ranges or within the study area. Home ranges ( n = 27) of all resident, adult marten were composed of more than 60% forest cover over 6 m in height; median values were 78–80% for both sexes. The median size of the largest forest patch in marten home ranges was 150 ha for females and 247 ha for males. Our results suggest that reducing fragmentation by consolidating clearcuts and retaining large residual patches would help to maintain resident marten in extensively logged landscapes.
1. We investigated the effects of forest fragmentation on American martens (Martes americana Rhoads) by evaluating differences in marten capture rates (excluding recaptures) in 18 study sites with different levels of fragmentation resulting from timber harvest clearcuts and natural openings. We focused on low levels of fragmentation, where forest connectivity was maintained and non-forest cover ranged from 2% to 42%.
2. Martens appeared to respond negatively to low levels of habitat fragmentation, based on the significant decrease in capture rates within the series of increasingly fragmented landscapes. Martens were nearly absent from landscapes having > 25% non-forest cover, even though forest connectivity was still present.
3. Marten capture rates were negatively correlated with increasing proximity of open areas and increasing extent of high-contrast edges. Forested landscapes appeared unsuitable for martens when the average nearest-neighbour distance between open (non-forested) patches was <100 m. In these landscapes, the proximity of open areas created strips of forest edge and eliminated nearly all forest interior.
4. Small mammal densities were significantly higher in clearcuts than in forests, but marten captures were not correlated with prey abundance or biomass associated with clearcuts.
5. Conservation efforts for the marten must consider not only the structural aspects of mature forests, but the landscape pattern in which the forest occurs. We recommend that the combination of timber harvests and natural openings comprise <25% of landscapes ≥9 km2 in size.
6. The spatial pattern of open areas is important as well, because small, dispersed openings result in less forest interior habitat than one large opening at the same percentage of fragmentation. Progressive cutting from a single patch would retain the largest amount of interior forest habitat.