Spatiotemporal shifts in distribution of a recolonizing black bear population

Abstract

Following one of the most successful relocations of a large carnivore species, American black bears (Ursus americanus; hereafter bears) have expanded from Arkansas to the Ozark Highlands of southern Missouri, USA, where they were potentially extirpated in the early 1900s. Our first objective was to estimate spatial and temporal shifts in probability of bear presence to understand the population distribution in Missouri. Our second objective was to assess which factors might influence any detected shifts. We used public sightings of bears to develop statewide spatial distribution models for early (1991–2003) and late (2004–2015) bear presence probability in Missouri using maximum entropy modeling and determine the rate and direction of population expansion. We evaluated the effects of environmental variables (e.g., land cover, roads) on bear presence probability and predicted the total area of three levels of bear presence probabilities. The average annual density of bear sightings increased temporally by about 3.4/1000 km2 in Missouri. The Ozark Highlands ecoregion contained the highest average annual density of sightings during both time periods. The bear presence probability increased spatially about 29 km in a north–northwest direction between the two time periods and the distribution became more dispersed over time. Our results corroborate prior findings suggesting bear presence is greatest in forests and rural areas. However, the increased importance of cultivated crops over time combined with the consistently positive relationship with roads indicates little negative effects from anthropogenic features on the bear presence probability in Missouri.

Conservation implications of sex‐specific landscape suitability for a large generalist carnivore

“…Another potential problem from not differentiating between sexes is over‐ or underestimation of smaller scale presence–environment relationships. Previous species distribution models developed in Missouri from citizen reported black bear sightings found a strong positive effect of roads and negligible effect of forested areas (McFadden‐Hiller & Belant, ), which contrasts with our results. Missouri bear sightings are likely male‐biased due to males moving larger distances (Gantchoff et al, ) and more likely to occur in human‐modified areas (this study).…”

The effectiveness of opportunistic public reports versus professional data to estimate large carnivore distribution

“…This sampling bias is a result of suitability being overestimated in areas near roads (due to greater “sampling” intensity; McFadden‐Hiller & Belant, 2018; Warton et al, 2013). Similarly, a previous study modeled bear range distribution and expansion in this area also using public reports (1991–2015), but even after implementing different spatial correction methods, a consistent positive relationship with roads remained (McFadden‐Hiller & Belant, 2018). Other alternatives to account for sampling bias include incorporating road biases as a covariate in models (Warton et al, 2013) or aggregating records at coarser scales (Elith et al, 2011; Fourcade et al, 2014).…”

“…To test whether the simple step of excluding road covariates can increase the similarity between opportunistic data‐based models and professional data‐based models, each of the three report‐based data sets (unfiltered, filtered, and verified) and the telemetry data set were modeled twice, once including the two road covariates (distance to major and minor roads) and once without. Our modeling approach and bias correction methods were informed in part by a previous study in this area, which modeled black bear range using older public reports (1991–2015) and evaluated several methods to mitigate sampling bias (e.g., spatial filtering, restricting background locations, and road exclusion; McFadden‐Hiller & Belant, 2018).…”

“…We present a unique approach that compares mapping results from telemetry data and public reports for a large carnivore, American black bear ( Ursus americanus ) in Missouri (USA). Our objective was to investigate whether distribution models developed using opportunistic public‐gathered data are congruent with those developed using professional (i.e., GPS telemetry) data, while evaluating two strategies to remove common biases in opportunistic data and improve the similarity with the professional model: data filtering and covariate removal (McFadden‐Hiller & Belant, 2018; Steen et al, 2019). We assume results from the telemetry models to better approximate the “true” black bear distribution, as the locations are largely free from human‐derived observation biases.…”

Integrated SDM database: Enhancing the relevance and utility of species distribution models in conservation management

“…Reintroduced species could expand outside the protected areas where they were introduced (Smeraldo et al., 2017), requiring rangers and decision‐makers to engage with local communities to generate awareness and support (Karamanlidis et al., 2016). Natural and unnatural barriers, such as topographic changes or roads, could inhibit a reintroduced species' expansion (Engler et al., 2012; McFadden‐Hiller & Belant, 2018). Furthermore, mapping a species' potential range requires understanding and proper handling of predictive model limitations, for example, how much extrapolation to allow into new regions (Elith et al., 2010) or a model's inability to account for environmental conditions that are non‐existent in a species' current range but become important as the species expands its range (Leroux et al., 2017; Swinnen et al., 2017).…”