Examining the occupancy-density relationship for a low-density carnivore

A flexible framework for spatial capture-recapture with unknown identities

Camera traps or acoustic recorders are often used to sample wildlife populations. When animals can be individually identified, these data can be used with spatial capture-recapture (SCR) methods to assess populations. However, obtaining animal identities is often labor-intensive and not always possible for all detected animals. To address this problem, we formulate SCR, including acoustic SCR, as a marked Poisson process, comprising a single counting process for the detections of all animals and a mark distribution for what is observed (eg, animal identity, detector location). The counting process applies equally when it is animals appearing in front of camera traps and when vocalizations are captured by microphones, although the definition of a mark changes. When animals cannot be uniquely identified, the observed marks arise from a mixture of mark distributions defined by the animal activity centers and additional characteristics. Our method generalizes existing latent identity SCR models and provides an integrated framework that includes acoustic SCR. We apply our method to estimate density from a camera trap study of fisher (Pekania pennanti) and an acoustic survey of Cape Peninsula moss frog (Arthroleptella lightfooti). We also test it through simulation. We find latent identity SCR with additional marks such as sex or time of arrival to be a reliable method for estimating animal density.

Evaluating density-weighted connectivity of black bears (Ursus americanus) in Glacier National Park with spatial capture-recapture models

BACKGROUND

Improved understanding of wildlife population connectivity among protected area networks can support effective planning for the persistence of wildlife populations in the face of land use and climate change. Common approaches to estimating connectivity often rely on small samples of individuals without considering the spatial structure of populations, leading to limited understanding of how individual movement links to demography and population connectivity. Recently developed spatial capture-recapture (SCR) models provide a framework to formally connect inference about individual movement, connectivity, and population density, but few studies have applied this approach to empirical data to support connectivity planning.

METHODS

We used mark-recapture data collected from 924 genetic detections of 598 American black bears (Ursus americanus) in 2004 with SCR ecological distance models to simultaneously estimate density, landscape resistance to movement, and population connectivity in Glacier National Park northwest Montana, USA. We estimated density and movement parameters separately for males and females and used model estimates to calculate predicted density-weighted connectivity surfaces.

RESULTS

Model results indicated that landscape structure influences black bear density and space use in Glacier. The mean density estimate was 16.08 bears/100 km2 (95% CI 12.52-20.6) for females and 9.27 bears/100 km2 (95% CI 7.70-11.14) for males. Density increased with forest cover for both sexes. For male black bears, density decreased at higher grizzly bear (Ursus arctos) densities. Drainages, valley bottoms, and riparian vegetation decreased estimates of landscape resistance to movement for male and female bears. For males, forest cover also decreased estimated resistance to movement, but a transportation corridor bisecting the study area strongly increased resistance to movement presenting a barrier to connectivity.

CONCLUSIONS

Density-weighed connectivity surfaces highlighted areas important for population connectivity that were distinct from areas with high potential connectivity. For black bears in Glacier and surrounding landscapes, consideration of both vegetation and valley topography could inform the placement of underpasses along the transportation corridor in areas characterized by both high population density and potential connectivity. Our study demonstrates that the SCR ecological distance model can provide biologically realistic, spatially explicit predictions to support movement connectivity planning across large landscapes.

Site occupancy by American martens and fishers in temperate deciduous forests of Québec

Interspecific interactions can mediate site occupancy of sympatric species and can be a key factor in habitat use patterns. American martens (Martes americana) and Fishers (Pekania pennanti) are two sympatric mesocarnivores in eastern North American forests. Due to their larger size, fishers have a competitive advantage over martens. We investigated site occupancy of martens and fishers in temperate deciduous forests of Québec, an environment modified by forest management and climate change. We formulated hypotheses on the spatial distribution of the studied species based on the knowledge of local trappers and on the scientific literature regarding forest cover composition, habitat fragmentation, and competitive relationships. We used a network of 49 camera traps monitored over two fall seasons to document site occupancy by both species. We used two-species site occupancy models to assess habitat use and the influence of fishers on martens at spatial grains of different sizes. None of the habitat variables that we considered explained site occupancy by fishers. Availability of dense old coniferous stands explained the spatial distribution of martens both at the home range grain size and at the landscape grain size. We identified the characteristics of habitat hotspots based on the knowledge of trappers, which highlighted the importance of stand composition, height, age, and canopy closure. The characteristics of habitat hotspots for martens in temperate deciduous forests refine the habitat suitability model for American martens that was originally developed for boreal forests of Québec.

N-mixture model-based estimate of relative abundance of sloth bear (Melursus ursinus) in response to biotic and abiotic factors in a human-dominated landscape of central India

Reliable estimation of abundance is a prerequisite for a species' conservation planning in human-dominated landscapes, especially if the species is elusive and involved in conflicts. As a means of population estimation, the importance of camera traps has been recognized globally, although estimating the abundance of unmarked, cryptic species has always been a challenge to conservation biologists. This study explores the use of the N-mixture model with three probability distributions, i.e., Poisson, negative binomial (NB) and zero-inflated Poisson (ZIP), to estimate the relative abundance of sloth bears (Melursus ursinus) based on a camera trapping exercise in Sanjay Tiger Reserve, Madhya Pradesh from December 2016 to April 2017. We used environmental and anthropogenic covariates to model the variation in the abundance of sloth bears. We also compared null model estimates (mean site abundance) obtained from the N-mixture model to those of the Royle-Nichols abundance-induced heterogeneity model (RN model) to assess the application of similar site-structured models. Models with Poisson distributions produced ecologically realistic and more precise estimates of mean site abundance (λ = 2.60 ± 0.64) compared with other distributions, despite the relatively high Akaike Information Criterion value. Area of mixed and sal forest, the photographic capture rate of humans and distance to the nearest village predicted a higher relative abundance of sloth bears. Mean site abundance estimates of sloth bears obtained from the N-mixture model (Poisson distribution) and the RN model were comparable, indicating the overall utility of these models in this field. However, density estimates of sloth bears based on spatially explicit methods are essential for evaluating the efficacy of the relatively more cost-effective N-mixture model. Compared to commonly used index/encounter-based methods, the N-mixture model equipped with knowledge on governing biotic and abiotic factors provides better relative abundance estimates for a species like the sloth bear. In the absence of absolute abundance estimates, the present study could be insightful for the long-term conservation and management of sloth bears.

Occupancy winners in tropical protected forests: a pantropical analysis

The structure of forest mammal communities appears surprisingly consistent across the continental tropics, presumably due to convergent evolution in similar environments. Whether such consistency extends to mammal occupancy, despite variation in species characteristics and context, remains unclear. Here we ask whether we can predict occupancy patterns and, if so, whether these relationships are consistent across biogeographic regions. Specifically, we assessed how mammal feeding guild, body mass and ecological specialization relate to occupancy in protected forests across the tropics. We used standardized camera-trap data (1002 camera-trap locations and 2-10 years of data) and a hierarchical Bayesian occupancy model. We found that occupancy varied by regions, and certain species characteristics explained much of this variation. Herbivores consistently had the highest occupancy. However, only in the Neotropics did we detect a significant effect of body mass on occupancy: large mammals had lowest occupancy. Importantly, habitat specialists generally had higher occupancy than generalists, though this was reversed in the Indo-Malayan sites. We conclude that habitat specialization is key for understanding variation in mammal occupancy across regions, and that habitat specialists often benefit more from protected areas, than do generalists. The contrasting examples seen in the Indo-Malayan region probably reflect distinct anthropogenic pressures.

Precision and bias of spatial capture-recapture estimates: A multi-site, multi-year Utah black bear case study

Spatial capture-recapture (SCR) models are powerful analytical tools that have become the standard for estimating abundance and density of wild animal populations. When sampling populations to implement SCR, the number of unique individuals detected, total recaptures, and unique spatial relocations can be highly variable. These sample sizes influence the precision and accuracy of model parameter estimates. Testing the performance of SCR models with sparse empirical data sets typical of low-density, wide-ranging species can inform the threshold at which a more integrated modeling approach with additional data sources or additional years of monitoring may be required to achieve reliable, precise parameter estimates. Using a multi-site, multi-year Utah black bear (Ursus americanus) capture-recapture data set, we evaluated factors influencing the uncertainty of SCR structural parameter estimates, specifically density, detection, and the spatial scale parameter, sigma. We also provided some of the first SCR density estimates for Utah black bear populations, which ranged from 3.85 to 74.33 bears/100 km² . Increasing total detections decreased the uncertainty of density estimates, whereas an increasing number of total recaptures and individuals with recaptures decreased the uncertainty of detection and sigma estimates, respectively. In most cases, multiple years of data were required for precise density estimates (<0.2 coefficient of variation [CV]). Across study areas there was an average decline in CV of 0.07 with the addition of another year of data. One sampled population with very high estimated bear density had an atypically low number of spatial recaptures relative to total recaptures, apparently inflating density estimates. A complementary simulation study used to assess estimate bias suggested that when <30% of recaptured individuals were spatially recaptured, density estimates were unreliable and ranged widely, in some cases to >3 times the simulated density. Additional research could evaluate these requirements for other density scenarios. Large numbers of individuals detected, numbers of spatial recaptures, and precision alone may not be sufficient indicators of parameter estimate reliability. We provide an evaluation of simple summary statistics of capture-recapture data sets that can provide an early signal of the need to alter sampling design or collect auxiliary data before model implementation to improve estimate precision and accuracy.

Occupancy of wild southern pig-tailed macaques in intact and degraded forests in Peninsular Malaysia

Deforestation is a major threat to terrestrial tropical ecosystems, particularly in Southeast Asia where human activities have dramatic consequences for the survival of many species. However, responses of species to anthropogenic impact are highly variable. In order to establish effective conservation strategies, it is critical to determine a species’ ability to persist in degraded habitats. Here, we used camera trapping data to provide the first insights into the temporal and spatial distribution of southern pig-tailed macaques (Macaca nemestrina, listed as ‘Vulnerable’ by the IUCN) across intact and degraded forest habitats in Peninsular Malaysia, with a particular focus on the effects of clear-cutting and selective logging on macaque occupancy. Specifically, we found a 10% decline in macaque site occupancy in the highly degraded Pasoh Forest Reserve from 2013 to 2017. This may be strongly linked to the macaques’ sensitivity to intensive disturbance through clear-cutting, which significantly increased the probability that M. nemestrina became locally extinct at a previously occupied site. However, we found no clear relationship between moderate disturbance, i.e., selective logging, and the macaques’ local extinction probability or site occupancy in the Pasoh Forest Reserve and Belum-Temengor Forest Complex. Further, an identical age and sex structure of macaques in selectively logged and completely undisturbed habitat types within the Belum-Temengor Forest Complex indicated that the macaques did not show increased mortality or declining birth rates when exposed to selective logging. Overall, this suggests that low to moderately disturbed forests may still constitute valuable habitats that support viable populations of M. nemestrina, and thus need to be protected against further degradation. Our results emphasize the significance of population monitoring through camera trapping for understanding the ability of threatened species to cope with anthropogenic disturbance. This can inform species management plans and facilitate the development of effective conservation measures to protect biodiversity.

Indigenous-led camera-trap research on traditional territories informs conservation decisions for resource extraction

The resource extraction that powers global economies is often manifested in Indigenous Peoples’ territories. Indigenous Peoples living on the land are careful observers of resulting biodiversity changes, and Indigenous-led research can provide evidence to inform conservation decisions. In the Nearctic western boreal forest, landscape change from forest harvesting and petroleum extraction is intensive and extensive. A First Nations community in the Canadian oil sands co-created camera-trap research to explore observations of presumptive species declines, seeking to identify the relative contributions of different industrial sectors to changes in mammal distributions. Camera data were analyzed via generalized linear models in a model-selection approach. Multiple forestry and petroleum extraction features positively and negatively affected boreal mammal species. Pipelines had the greatest negative effect size (for wolves), whereas well sites had a large positive effect size for multiple species, suggesting the energy sector as a target for co-management. Co-created research reveals spatial relationships of disturbance, prey, and predators on Indigenous traditional territories. It provides hypotheses, tests, and interpretations unique to outside perspectives; Indigenous participation in conservation management of their territories scales up to benefit global biodiversity conservation.

Spatially Explicit Capture-Recapture Through Camera Trapping: A Review of Benchmark Analyses for Wildlife Density Estimation

Camera traps have become an important research tool for both conservation biologists and wildlife managers. Recent advances in spatially explicit capture-recapture (SECR) methods have increasingly put camera traps at the forefront of population monitoring programs. These methods allow for benchmark analysis of species density without the need for invasive fieldwork techniques. We conducted a review of SECR studies using camera traps to summarize the current focus of these investigations, as well as provide recommendations for future studies and identify areas in need of future investigation. Our analysis shows a strong bias in species preference, with a large proportion of studies focusing on large felids, many of which provide the only baseline estimates of population density for these species. Furthermore, we found that a majority of studies produced density estimates that may not be precise enough for long-term population monitoring. We recommend simulation and power analysis be conducted before initiating any particular study design and provide examples using readily available software. Furthermore, we show that precision can be increased by including a larger study area that will subsequently increase the number of individuals photo-captured. As many current studies lack the resources or manpower to accomplish such an increase in effort, we recommend that researchers incorporate new technologies such as machine-learning, web-based data entry, and online deployment management into their study design. We also cautiously recommend the potential of citizen science to help address these study design concerns. In addition, modifications in SECR model development to include species that have only a subset of individuals available for individual identification (often called mark-resight models), can extend the process of explicit density estimation through camera trapping to species not individually identifiable.

Spatial proximity moderates genotype uncertainty in genetic tagging studies

Accelerating declines of an increasing number of animal populations worldwide necessitate methods to reliably and efficiently estimate demographic parameters such as population density and trajectory. Standard methods for estimating demographic parameters from noninvasive genetic samples are inefficient because lower-quality samples cannot be used, and they assume individuals are identified without error. We introduce the genotype spatial partial identity model (gSPIM), which integrates a genetic classification model with a spatial population model to combine both spatial and genetic information, thus reducing genotype uncertainty and increasing the precision of demographic parameter estimates. We apply this model to data from a study of fishers (Pekania pennanti) in which 37% of hair samples were originally discarded because of uncertainty in individual identity. The gSPIM density estimate using all collected samples was 25% more precise than the original density estimate, and the model identified and corrected three errors in the original individual identity assignments. A simulation study demonstrated that our model increased the accuracy and precision of density estimates 63 and 42%, respectively, using three replicated assignments (e.g., PCRs for microsatellites) per genetic sample. Further, the simulations showed that the gSPIM model parameters are identifiable with only one replicated assignment per sample and that accuracy and precision are relatively insensitive to the number of replicated assignments for high-quality samples. Current genotyping protocols devote the majority of resources to replicating and confirming high-quality samples, but when using the gSPIM, genotyping protocols could be more efficient by devoting more resources to low-quality samples.

Fine scale characteristics of catfish aquaculture ponds influencing piscivorous avian species foraging use in the Mississippi Delta

Piscivorous avian species are the main source of catfish depredation at aquaculture facilities in Mississippi, resulting in the economic loss of millions of dollars every year. Most notable of these avian species are the double-crested cormorant (Phalacrocorax auritus), great blue heron (Ardea herodias), and great egret (A. alba). Understanding why these species select specific ponds can increase management efficiency directed at avian dispersal and provide insight into their decision making with respect to foraging behavior. We collected species presence data on catfish ponds by flying 35 surveys from October through April of 2015-2017, during which an average of 973 catfish ponds were observed each year. We collected data associated with each pond's physical surroundings and contents and used occupancy modeling to determine their influence on avian occupancy probability. We also collected data associated with stocking practices and catfish health on a subset of ponds, and constructed resource selection functions to model their influence on avian presence. Pond area was positively related to occupancy probability of each species. Cormorant occupancy increased as pond distance from forest cover and activity centers, such as workshops and offices, increased. Distance to nearest activity center was positively related to egret occupancy, while distance to nearest forested area was negative. Ponds containing diseased catfish had an increased probability of use by both herons and egrets. In general, cormorants and egrets showed greater probability of use on the periphery of pond clusters. The abundance of catfish was positively related to cormorant and heron presence. Specific pond contents and characteristics influenced presence of each avian species in different ways, including fish species cultured, production methods, pond systems, and fish types. Many pond selection relationships were species-specific, illustrating inherent differences in their foraging ecology. Consequently, specific management actions aimed to reduce avian presence will depend on the targeted species.

Estimating population density of insectivorous bats based on stationary acoustic detectors: A case study

Automated recording units are commonly used by consultants to assess environmental impacts and to monitor animal populations. Although estimating population density of bats using stationary acoustic detectors is key for evaluating environmental impacts, estimating densities from call activity data is only possible through recently developed numerical methods, as the recognition of calling individuals is impossible.We tested the applicability of generalized random encounter models (gREMs) for determining population densities of three bat species (Common pipistrelle Pipistrellus pipistrellus, Northern bat Eptesicus nilssonii, and Natterer's bat Myotis nattereri) based on passively collected acoustical data. To validate the results, we compared them to (a) density estimates from the literature and to (b) Royle-Nichols (RN) models of detection/nondetection data.Our estimates for M. nattereri matched both the published data and RN-model results. For E. nilssonii, the gREM yielded similar estimates to the RN-models, but the published estimates were more than twice as high. This discrepancy might be because the high-altitude flight of E. nilssonii is not accounted for in gREMs. Results of gREMs for P. pipistrellus were supported by published data but were ~10 times higher than those of RN-models. RN-models use detection/nondetection data, and this loss of information probably affected population estimates of very active species like P. pipistrellus.gREM models provided realistic estimates of bat population densities based on automatically recorded call activity data. However, the average flight altitude of species should be accounted for in future analyses. We suggest including flight altitude in the calculation of the detection range to assess the detection sphere more accurately and to obtain more precise density estimates.

Pre-eradication assessment of feral cat density and population size across Kangaroo Island, South Australia

Abstract Context Feral cats (Felis catus) are a significant threat to wildlife in Australia and globally. In Australia, densities of feral cats vary across the continent and also between the mainland and offshore islands. Densities on small islands may be at least an order of magnitude higher than those in adjacent mainland areas. To provide cat-free havens for biodiversity, cat-control and eradication programs are increasingly occurring on Australian offshore islands. However, planning such eradications is difficult, particularly on large islands where cat densities could vary considerably. Aims In the present study, we examined how feral cat densities vary among three habitats on Kangaroo Island, a large Australian offshore island for which feral cat eradication is planned. Methods Densities were compared among the following three broad habitat types: forest, forest–farmland boundaries and farmland. To detect cats, three remote-camera arrays were deployed in each habitat type, and density around each array was calculated using a spatially explicit capture–recapture framework. Key results The average feral cat density on Kangaroo Island (0.37 cats km−2) was slightly higher than that on the Australian mainland. Densities varied from 0.06 to 3.27 cats km−2 and were inconsistent within broad habitat types. Densities were highest on farms that had a high availability of macropod and sheep carcasses. The relationship between cat density and the proportion of cleared land in the surrounding area was weak. The total feral cat population of Kangaroo Island was estimated at 1629±661 (mean±s.e.) individuals. Conclusions Cat densities on Kangaroo Island are highly variable and may be locally affected by factors such as prey and carrion availability. Implications For cat eradication to be successful, resources must be sufficient to control at least the average cat density (0.37 cats km−2), with additional effort around areas of high carcass availability (where cats are likely to be at a higher density) potentially also being required.

An Evaluation of Systematic Versus Strategically-Placed Camera Traps for Monitoring Feral Cats in New Zealand

Simple Summary

Feral cats are detrimental to native biodiversity worldwide. In New Zealand, feral cats are well established across much of the pastoral landscape, including forested areas. Feral cats, like many carnivore species, are elusive in their nature, and often occur at low densities, making them difficult to detect. Camera traps are a useful, non-invasive monitoring device, capable of ‘capturing’ feral cats as they pass by. Although cameras provide a wealth of information about animals within their field of view; there remains much to be learned about optimal camera trap placement within a landscape, if maximizing detection probability is the objective. Here, we report the results of two methods of camera trap deployment within similar sites: (1) systematic deployment on a grid and (2) strategic deployment, predominantly favoring habitats with assumed higher cat activity. Using the Royle–Nichols abundance-induced heterogeneity model (RN), which assumes detection probability and animal abundance are linked, we found that more cats were detected by cameras at forest margins than in mixed scrub or open farmland (but only slightly more than in forest locations). If maximizing cat detections is the aim, we recommend that cameras should be placed at the edges of forests (including forest fragments) whenever feasible.

Abstract

We deploy camera traps to monitor feral cat (Felis catus) populations at two pastoral sites in Hawke’s Bay, North Island, New Zealand. At Site 1, cameras are deployed at pre-determined GPS points on a 500-m grid, and at Site 2, cameras are strategically deployed with a bias towards forest and forest margin habitat where possible. A portion of cameras are also deployed in open farmland habitat and mixed scrub. We then use the abundance-induced heterogeneity Royle–Nichols model to estimate mean animal abundance and detection probabilities for cameras in each habitat type. Model selection suggests that only cat abundance varies by habitat type. Mean cat abundance is highest at forest margin cameras for both deployment methods (3 cats [95% CI 1.9–4.5] Site 1, and 1.7 cats [95% CI 1.2–2.4] Site 2) but not substantially higher than in forest habitats (1.7 cats [95% CI 0.8–3.6] Site 1, and 1.5 cats [95% CI 1.1–2.0] Site 2). Model selection shows detection probabilities do not vary substantially by habitat (although they are also higher for cameras in forest margins and forest habitats) and are similar between sites (8.6% [95% CI 5.4–13.4] Site 1, and 8.3% [5.8–11.9] Site 2). Cat detections by camera traps are higher when placed in forests and forest margins; thus, strategic placement may be preferable when monitoring feral cats in a pastoral landscape.

Incorporating citizen science data in spatially explicit integrated population models

Information about population abundance, distribution, and demographic rates is critical for understanding a species' ecology and for effective conservation and management. To collect data over large spatial and temporal extents for such inferences, especially for species with low densities or wide distributions, citizen science can be an efficient approach. Integrated models have also emerged as an important methodology to estimate population parameters by combining multiple types of data, including citizen science data. We developed a spatially explicit integrated model that combines opportunistically collected presence-absence (PA) data, commonly collected in citizen science efforts, with systematically collected spatial capture-recapture (SCR) data, which are often limited to small spatial and temporal extents. We conducted single and multi-season simulations with parameters informed by North American black bear (Ursus americanus) populations, to evaluate the influence of varying amounts of opportunistic PA data collected at larger spatial and temporal extents on the estimation of population-level parameters. Integrating opportunistic PA data increased the precision and accuracy of posterior estimates of abundance, and survival and recruitment rates. In some cases, adding PA locations improved abundance estimates more than increasing PA detection probability. Posterior estimates were as precise and unbiased as when higher quality, but sparse, SCR data were available. We also applied the integrated model to SCR and citizen science PA data collected on black bears in New York, with results consistent with our simulations. Our findings indicate that citizen science in integrated models can be a cost-efficient way to improve estimates of population parameters and increase the spatiotemporal extent of inference. Continued developments with integrated models and citizen science data will offer additional ways to improve our understanding of population structure and demographics.

Genetic tagging in the Anthropocene: scaling ecology from alleles to ecosystems

The Anthropocene is an era of marked human impact on the world. Quantifying these impacts has become central to understanding the dynamics of coupled human-natural systems, resource-dependent livelihoods, and biodiversity conservation. Ecologists are facing growing pressure to quantify the size, distribution, and trajectory of wild populations in a cost-effective and socially acceptable manner. Genetic tagging, combined with modern computational and genetic analyses, is an under-utilized tool to meet this demand, especially for wide-ranging, elusive, sensitive, and low-density species. Genetic tagging studies are now revealing unprecedented insight into the mechanisms that control the density, trajectory, connectivity, and patterns of human-wildlife interaction for populations over vast spatial extents. Here, we outline the application of, and ecological inferences from, new analytical techniques applied to genetically tagged individuals, contrast this approach with conventional methods, and describe how genetic tagging can be better applied to address outstanding questions in ecology. We provide example analyses using a long-term genetic tagging dataset of grizzly bears in the Canadian Rockies. The genetic tagging toolbox is a powerful and overlooked ensemble that ecologists and conservation biologists can leverage to generate evidence and meet the challenges of the Anthropocene.

American marten and fisher do not segregate in space and time during winter in a mixed-forest system

Understanding the mechanisms of coexistence between ecologically similar species is an important issue in ecology. Carnivore coexistence may be facilitated by spatial segregation, temporal avoidance, and differential habitat selection. American martens Martes americana and fishers Pekania pennanti are medium-sized mustelids that occur sympatrically across portions of North America, yet mechanisms of coexistence between the two species are not fully understood. We assessed spatial and temporal partitioning in martens and fishers in the Upper Peninsula of Michigan, USA, using camera trap data collected during winter 2013-2015. To investigate spatial segregation, we used a dynamic occupancy model to estimate species' occupancy probabilities and probabilities of persistence and colonization as a function of covariates and yearly occupancy probability for the other species. Temporal segregation was assessed by estimating diel activity overlap between species. We found weak evidence of spatial or temporal niche partitioning of martens and fishers. There was high overlap in forest cover selection, and both marten and fisher occupancy were positively correlated with deciduous forests (excluding aspen [Populus tremuloides]). There was strong temporal overlap (Δ ^ 4 = 0.81 ; CI = 0.79-0.82) with both species exhibiting largely crepuscular activity patterns. Co-occurrence of martens and fishers appears to be facilitated by mechanisms not investigated in this study, such as partitioning of snow features or diet. Our results add additional insights into resource partitioning of mesocarnivores, but further research is required to enhance our understanding of mechanisms that facilitate marten and fisher coexistence.

Making inference with messy (citizen science) data: when are data accurate enough and how can they be improved?

Measurement or observation error is common in ecological data: as citizen scientists and automated algorithms play larger roles processing growing volumes of data to address problems at large scales, concerns about data quality and strategies for improving it have received greater focus. However, practical guidance pertaining to fundamental data quality questions for data users or managers-how accurate do data need to be and what is the best or most efficient way to improve it?-remains limited. We present a generalizable framework for evaluating data quality and identifying remediation practices, and demonstrate the framework using trail camera images classified using crowdsourcing to determine acceptable rates of misclassification and identify optimal remediation strategies for analysis using occupancy models. We used expert validation to estimate baseline classification accuracy and simulation to determine the sensitivity of two occupancy estimators (standard and false-positive extensions) to different empirical misclassification rates. We used regression techniques to identify important predictors of misclassification and prioritize remediation strategies. More than 93% of images were accurately classified, but simulation results suggested that most species were not identified accurately enough to permit distribution estimation at our predefined threshold for accuracy (<5% absolute bias). A model developed to screen incorrect classifications predicted misclassified images with >97% accuracy: enough to meet our accuracy threshold. Occupancy models that accounted for false-positive error provided even more accurate inference even at high rates of misclassification (30%). As simulation suggested occupancy models were less sensitive to additional false-negative error, screening models or fitting occupancy models accounting for false-positive error emerged as efficient data remediation solutions. Combining simulation-based sensitivity analysis with empirical estimation of baseline error and its variability allows users and managers of potentially error-prone data to identify and fix problematic data more efficiently. It may be particularly helpful for "big data" efforts dependent upon citizen scientists or automated classification algorithms with many downstream users, but given the ubiquity of observation or measurement error, even conventional studies may benefit from focusing more attention upon data quality.

Evidence of significantly higher island feral cat abundance compared with the adjacent mainland

Context Feral cats (Felis catus) impact the health and welfare of wildlife, livestock and humans worldwide. They are particularly damaging where they have been introduced into island countries such as Australia and New Zealand, where native prey species evolved without feline predators. Kangaroo Island, in South Australia, is Australia’s third largest island and supports several threatened and endemic species. Cat densities on Kangaroo Island are thought to be greater than those on the adjacent South Australian mainland, based on one cat density estimate on the island that is higher than most estimates from the mainland. The prevalence of cat-borne disease in cats and sheep is also higher on Kangaroo Island than the mainland, suggesting higher cat densities. A recent continental-scale spatial model of cat density predicted that cat density on Kangaroo Island should be about double that of the adjacent mainland. However, although cats are believed to have severe impacts on some native species on the island, other species that are generally considered vulnerable to cat predation have relatively secure populations on the island compared with the mainland. Aims The present study aimed to compare feral cat abundance between Kangaroo Island and the adjacent South Australian mainland using simultaneous standardised methods. Based on previous findings, we predicted that the relative abundance of feral cats on Kangaroo Island would be approximately double that on the South Australian mainland. Methods Standardised camera trap surveys were used to simultaneously estimate the relative abundance of feral cats on Kangaroo Island and the adjacent South Australian mainland. Survey data were analysed using the Royle–Nichols abundance-induced heterogeneity model to estimate feral cat relative abundance at each site. Key results Cat abundance on the island was estimated to be over 10 times greater than that on the adjacent mainland. Conclusions Consistent with predictions, cat abundance on the island was greater than on the adjacent mainland. However, the magnitude of this difference was much greater than expected. Implications The findings show that the actual densities of cats at local sites can vary substantially from predictions generated by continental-scale models. The study also demonstrates the value of estimating abundance or density simultaneously across sites using standardised methods.

Large-scale variation in density of an aquatic ecosystem indicator species

Monitoring indicator species is a pragmatic approach to natural resource assessments, especially when the link between the indicator species and ecosystem state is well justified. However, conducting ecosystem assessments over representative spatial scales that are insensitive to local heterogeneity is challenging. We examine the link between polychlorinated biphenyl (PCB) contamination and population density of an aquatic habitat specialist over a large spatial scale using non-invasive genetic spatial capture-recapture. Using American mink (Neovison vison), a predatory mammal and an indicator of aquatic ecosystems, we compared estimates of density in two major river systems, one with extremely high levels of PCB contamination (Hudson River), and a hydrologically independent river with lower PCB levels (Mohawk River). Our work supports the hypothesis that mink densities are substantially (1.64-1.67 times) lower in the contaminated river system. We demonstrate the value of coupling the indicator species concept with well-conceived and spatially representative monitoring protocols. PCBs have demonstrable detrimental effects on aquatic ecosystems, including mink, and these effects are likely to be profound and long-lasting, manifesting as population-level impacts. Through integrating non-invasive data collection, genetic analysis, and spatial capture-recapture methods, we present a monitoring framework for generating robust density estimates across large spatial scales.