Unmarked Populations Research Articles

Modeling spatiotemporal variations in the density of unmarked populations using camera trap data: An application to free-roaming cats

Camera traps have revolutionized wildlife monitoring, providing reliable information on animal occupancy and abundance. In recent years, several models have been developed to estimate the absolute density of unmarked populations (i.e., those lacking individually recognizable markings), which was previously challenging. By refining these models, they may become efficient monitoring tools applicable to a wide range of species. In this study, we extended one such model, the Random Encounter and Staying Time (REST) model, within a state-space framework to explicitly account for spatiotemporal variations in animal density. We applied this model to free-roaming cats using data obtained from 209 camera traps deployed across a study area (836 km2) over a two-year period (June 2018–May 2020). The estimated cat density was generally <0.3 cats per km2, with higher densities observed near human settlements and occasional excursions into deeper forested areas (>500 m). Densities also exhibited seasonal variation, with higher values in winter and spring compared to summer. Our model estimates the “average” density of animal individuals available to camera traps over the temporal duration of interest. Although this metric differs from conventional population density measures and requires careful interpretation, it remains a valuable indicator of the ecological impact or functioning of animals, including free-roaming cats.

Evaluating camera‐based methods for estimating badger (Meles meles) density: Implications for wildlife management

Abstract Accurate and precise assessment of population density plays a critical role in effective wildlife management, but reliable estimates are often difficult to obtain. Camera traps have emerged as valuable noninvasive tools for studying elusive species, offering cost‐effective solutions for both marked and unmarked populations. We evaluated the consistency of badger (Meles meles) density estimates obtained from the random encounter model (REM) and camera trap distance sampling (CT‐DS) with independent estimates from spatial mark‐resight (SMR) models and quantified the bias in CT‐DS arising from animals reacting to camera traps. Six camera trap surveys were conducted in Cornwall, UK, in 2019 and 2021, and data were used to estimate badger density using the REM and CT‐DS. Four sites were included in a badger vaccination research project, providing an opportunity to mark badgers with uniquely identifiable fur clips to facilitate resighting within a SMR framework. We found consistency in the density estimates across all methods, but results had wide confidence intervals. Density estimates derived from CT‐DS tended to be higher than those from the REM and were sensitive to the exclusion of reactive sequences, resulting in a twofold decrease in the estimated density in one case. The REM tended to be the most precise method; however, where badger density was low, precision was low using all methods. Practical implication: our findings suggest animal density can be assessed from camera traps in the absence of individual identification; however, it is important to account for reactive behaviours, especially where such behaviour is prevalent. In these circumstances, we recommend utilising the REM which offers a clear methodology for addressing bias arising from reactive sequences. In addition, we emphasise the need for improved precision to ensure the effectiveness of these methods in the context of wildlife management. We offer practical considerations to facilitate the broader application of these methods, drawing upon the example of disease control through badger vaccination.

Animal Density Estimation for Large Unmarked Populations Using a Spatially Explicit Model

AbstractObtaining abundance and density estimates is a particularly important aspect within wildlife conservation and management. To monitor wildlife populations, the use of motion-sensor camera traps is becoming increasing popular due to its non-invasive nature. However, animal identification is not always feasible in practice due to poor quality images and/or individuals not having uniquely identifiable physical characteristics. Spatially explicit models for unmarked individuals permit the estimation of animal density when individuals cannot be uniquely identified. Due to the structure of these models, a Bayesian super-population (data augmentation) approach is often used to fit the models to data, which involves specifying some reasonably large upper limit for the population. However, this approach presents substantial computational challenges for larger populations, as demonstrated by the motivating dataset relating to barking deer (Muntiacus muntjak) collected in Ujung Kulon National Park, Indonesia (with a population size in the low thousands). We develop a new and computationally efficient Bayesian algorithm for fitting the models to data that does not require specifying an upper population limit a priori. We apply the new algorithm to the large barking deer dataset, where the standard super-population approach is computationally expensive, and demonstrate a substantial improvement in computational efficiency.Supplementary material to this paper is provided online.

Reducing bias in density estimates for unmarked populations that exhibit reactive behaviour towards camera traps

Abstract Density estimates guide wildlife management, and camera traps are commonly used to estimate sizes of unmarked populations. Unfortunately, animals often alter their natural behaviour in the presence of camera traps, which may bias subsequent density estimates. We simulated 100 populations of known density to test several new and existing methods that aimed to reduce bias in density estimates from camera trap distance sampling (CTDS) and the random encounter model (REM). Within our simulated populations, we introduced different behavioural reactions including attraction towards cameras, freezing when near cameras and fleeing from cameras. CTDS and REM provided density estimates with decent coverage of confidence intervals (CTDS = 94%, REM = 87%), mean coefficient of variation (CTDS = 0.121, REM = 0.071) and minimal bias (root‐mean squared error: CTDS = 1.336, REM = 0.913) for simulated populations with no reactive behaviour. However, failure to implement a method to account for reactive behaviour resulted in low coverage, large bias and potentially imprecise density estimates when 30% of the simulated population potentially reacted by attraction to or fleeing from camera traps. We identified a corrective strategy that enhanced confidence interval coverage, increased precision and reduced bias for every behavioural reaction except when individuals potentially fled from cameras. Synthesis and applications. We provide empirically tested methods for reducing bias of density estimates. Wildlife managers requiring population estimates of animals that exhibit reactive behaviour can use our methods to reduce inaccuracy. We encourage future studies to quantify behavioural responses to camera traps and to implement, test and possibly extend our methods to reduce bias through simulation.

Estimating effective survey duration in camera trap distance sampling surveys.

Among other approaches, camera trap distance sampling (CTDS) is used to estimate animal abundance from unmarked populations. It was formulated for videos and observation distances are measured at predetermined 'snapshot moments'. Surveys recording still images with passive infrared motion sensors suffer from frequent periods where animals are not photographed, either because of technical delays before the camera can be triggered again (i.e. 'camera recovery time') or because they remain stationary and do not immediately retrigger the camera following camera recovery time (i.e. 'retrigger delays'). These effects need to be considered when calculating temporal survey effort to avoid downwardly biased abundance estimates. Here, we extend the CTDS model for passive infrared motion sensor recording of single images or short photo series. We propose estimating 'mean time intervals between triggers' as combined mean camera recovery time and mean retrigger delays from the time interval distribution of pairs of consecutive pictures, using a Gamma and Exponential function, respectively. We apply the approach to survey data on red deer, roe deer and wild boar. Mean time intervals between triggers were very similar when estimated empirically and when derived from the model-based approach. Depending on truncation times (i.e. the time interval between consecutive pictures beyond which data are discarded) and species, we estimated mean time intervals between retriggers between 8.28 and 15.05 s. Using a predefined snapshot interval, not accounting for these intervals, would lead to underestimated density by up to 96% due to overestimated temporal survey effort. The proposed approach is applicable to any taxa surveyed with camera traps. As programming of cameras to record still images is often preferred over video recording due to reduced consumption of energy and memory, we expect this approach to find broad application, also for other camera trap methods than CTDS.

Assessing trends in population size of three unmarked species: A comparison of a multi-species N-mixture model and random encounter models.

Estimation of changes in abundances and densities is essential for the research, management, and conservation of animal populations. Recently, technological advances have facilitated the surveillance of animal populations through the adoption of passive sensors, such as camera traps (CT). Several methods, including the random encounter model (REM), have been developed for estimating densities of unmarked populations but require additional information. Hierarchical abundance models, such as the N-mixture model (NMM), can estimate abundances without performing additional fieldwork but do not explicitly estimate the area effectively sampled. This obscures the interpretation of its densities and requires its users to focus on relative measures of abundance instead. Hence, the main objective of our study is to evaluate if REM and NMM yield consistent results qualitatively. Therefore, we compare relative trends: (i) between species, (ii) between years and (iii) across years obtained from annual density/abundance estimates of three species (fox, wild boar and red deer) in central Spain monitored by a camera trapping network for five consecutive winter periods. We reveal that NMM and REM provided density estimates in the same order of magnitude for wild boar, but not for foxes and red deer. Assuming a Poisson detection process in the NMM was important to control for inflation of abundance estimates for frequently detected species. Both methods consistently ranked density/abundance across species (between species trend), but did not always agree on relative ranks of yearly estimates within a single population (between years trend), nor on its linear population trends across years (across years trend). Our results suggest that relative trends are generally consistent when the range of variability is large, but can become inconsistent when the range of variability is smaller.

Estimating mesocarnivore abundance on commercial farmland using distance sampling with camera traps

AbstractMesocarnivores are of particular interest in wildlife management. Their adaptability makes them a focus of public attention globally, as negative interactions with people occur regularly, but their importance to wider ecosystem function is increasingly apparent. Robust methods for estimating mesocarnivore densities are essential for long‐term management strategies. Estimating densities of unmarked populations remains challenging, but new methods, based on camera trapping, have recently become available and require field testing.We conducted two camera trap surveys over two 200 km2areas of commercial farmland in South Africa. One survey sampled 25 locations, while the second used a migrating grid to sample 59 locations; total sampling effort was similar across the two surveys. We applied distance sampling with camera traps (CTDS), developing a method to estimate animal distances by applying a distance measurement overlay grid to camera trap images.We aimed to establish meaningful density estimates of the mesocarnivore guild and evaluate CTDS's suitability for broader use with these types of species. We obtained density estimates for four carnivores, African civetCivettictis civetta, black‐backed jackalCanis mesomelas, brown hyenaHyaena brunneaand caracalCaracal caracal, providing valuable insight into their status in commercial farmland. Imprecision in the estimates was almost exclusively due to encounter rate variance, which was not reduced with the migrating camera grid.We explored the sensitivity of our results to assumptions determining the value of the ‘snapshot interval’, demonstrating that careful selection of this parameter is vital to ensuring reliable estimates when using rapid‐fire photo burst modes.CTDS can provide useful density estimates for mesocarnivores, but future studies should aim to maximize precision and reliability by increasing sampling locations. More studies are required in areas with known densities to promote confidence in accuracy.

Modelling density surfaces of intraspecific classes using camera trap distance sampling

AbstractSpatially explicit densities of wildlife are important for understanding environmental drivers of populations, and density surfaces of intraspecific classes allow exploration of links between demographic ratios and environmental conditions. Although spatially explicit densities and class densities are valuable, conventional design‐based estimators remain prevalent when using camera‐trapping methods for unmarked populations.We developed a density surface model that utilized camera trap distance sampling data within a hierarchical generalized additive modelling framework. We estimated density surfaces of intraspecific classes of a common ungulate, white‐tailed deerOdocoileus virginianus, across three large management regions in Indiana, United States. We then extended simple statistical theory to test for differences in two ratios of density.Deer density was influenced by landscape fragmentation, wetlands and anthropogenic development. We documented class‐specific responses of density to availability of concealment cover, and found strong evidence that increased recruitment of young was tied to increased resource availability from anthropogenic agricultural land use. The coefficients of variation of the total density estimates within the three regions we surveyed were 0.11, 0.10 and 0.06.Synthesis and applications. Our strategy extends camera trap distance sampling and enables managers to use camera traps to better understand spatial predictors of density. Our density estimates were more precise than previous estimates from camera trap distance sampling. Population managers can use our methods to detect finer spatiotemporal changes in density or ratios of intraspecific‐class densities. Such changes in density can be linked to land use, or to management regimes on habitat and harvest limits of game species.

A practical guide for estimating the density of unmarked populations using camera traps

A practical guide for estimating the density of unmarked populations using camera traps

A cautionary tale comparing spatial count and partial identity models for estimating densities of threatened and unmarked populations

Population monitoring is critical to wildlife conservation, but density estimation is difficult for wide-ranging, unmarked species inhabiting remote habitats. Furthermore, recent investigations into density estimation with camera trap data has revealed unmarked models to be potentially unreliable, prompting cautious application and continued model development. Two related approaches with increasing appeal include spatial count (SC), which infer latent identities from the spatial pattern of detections, and spatial partial identity models (SPIM), which additionally leverage partial identity covariates (e.g., sex, antler point count, and presence of GPS/radio collar). To assess the performance of unmarked density models, we applied SC and SPIM to camera trap data of threatened boreal caribou in Canada, which are declining but have few rigorous density estimates across their broad distribution to inform conservation efforts. In particular, we focused on two spatially proximate caribou ranges in northern Alberta, Canada that differ in estimated caribou demographic trends, disturbance histories, and abundances of caribou predators and apparent competitors. Estimates of caribou density varied over a 4 year period (2016 – 2019), and were higher in the region with more stable reported growth rates and less anthropogenic disturbance (mode SPIM estimates: 155 – 225/1000 km 2 vs. 19 – 96/1000 km 2 ). However, density estimates differed by modeling approach and had low and variable precision, hindering inferences about population status and trajectories. Simulations suggest that SPIM estimates may have been less biased and more precise. SC models likely underestimated density by mistaking detections of neighboring individuals as recaptures of a single individual, although SPIM may also have overestimated density by inflating assignment probabilities of detections to non-existent individuals. Findings highlight the need to explore how grouping dynamics and non-independent movement violate assumptions and reduce the ability to distinguish individuals. We advocate continued investigation into the accuracy of unmarked density estimation approaches, the ecological and sampling conditions appropriate for different unmarked density models, and coordination across sampling efforts and analyses to improve population inferences.

Spatial dynamic N‐mixture models with interspecific interactions

Abstract Interspecific interactions and movement are key factors that drive the coexistence of metapopulations in heterogenous landscapes. Yet, it is challenging to understand these factors because separating movement from local population processes relied on capture‐based data that are difficult to collect. Recent development of spatial dynamic N‐mixture models (SDNMs) made it possible to draw inferences on local population growth and movement using count data of unmarked populations. However, no SDNMs have been developed to account for interspecific interactions and double‐counting observation errors. In this study we further developed SDNMs to account for interspecific interactions and both false‐negative and double‐counting observation errors. We conducted simulation studies to evaluate the inferential performance of these models under different ecological systems (competition, predator–prey), observation processes (binomial, Poisson) and sampling situations including the number of surveyed sites, detection and the adoption of robust sampling design. We then illustrated the applications of these models with two case studies, one representing a competition system (mallard Anas platyrhynchos, northern pintail Anas acuta) with binomial observations and the other representing a predator–prey system (bobcat Lynx rufus, wild turkey Meleagris gallopavo) with Poisson observations. The results of the simulation studies showed that the models provide unbiased parameter estimates regardless of the ecological system, observation process and sampling situation. Case studies further demonstrated the capabilities of these models in revealing important ecological processes. More specifically, the first case study revealed a negative effect of the superior competitor, Mallard, on the population growth of the inferior competitor, pintail as well as different movement patterns between these species, whereas the second case study reveal a top‐down effect of bobcat abundance on wild turkey population growth and their differential habitat preferences. The models developed in this study could be used by themselves on count data or serve as population sub‐models in integrated modelling frameworks to provide insights about metapopulation dynamics of interacting species in heterogeneous landscapes. The flexible structures of these models made them highly adaptive and relevant in population and community ecology.

Evaluating unmarked abundance estimators using remote cameras and aerial surveys

AbstractReliable population abundance estimates are invaluable to wildlife management programs. Mark‐recapture techniques are regarded as the gold standard for estimating abundance but can be financially or logistically prohibitive. Recent developments in remote camera analytical approaches have provided alternative methods to estimate unmarked populations efficiently but have undergone limited field testing. In September and October 2020, I assessed 2 camera methods that use the same set of time‐lapse photographs; a space to event (STE) model and instantaneous sampling (IS), applied to a case study of feral sheep (Ovis aries) at the US Army's Pōhakuloa Training Area (PTA) on Hawai`i Island. Results of the camera methods were compared to an abundance estimate calculated from an aerial distance sampling survey completed concurrently with camera sampling. A comparison of confidence intervals indicated no statistical difference between estimates, but the relative imprecision of IS suggested the STE model to be a more reliable estimator. The cost of camera sampling was approximately 3 times that of the aerial survey, but camera sampling became more economical if the area was resurveyed 3 or more times. My study showed the STE model to be a cost effective, practical abundance estimator of feral sheep when compared to well‐established aerial distance sampling techniques, and therefore should offer considerable potential for the conservation and management of other easily identifiable species.

Effectiveness of a New Self-Marking Technique in Aedes aegypti under Laboratory Conditions.

Simple SummaryMarking techniques are generally used to differentiate colony mosquitoes from wild ones in the implementation of vector control programs using the Sterile Insect Technique. Different mosquito marking techniques have been developed in the past years but need improvement due to the extensive handling that can affect the quality of mosquitoes. We present here a self-marking technique that can reduce the damage associated with mosquito handling in mass rearing. The marking technique consists of adding fluorescent powder (DayGlo: A-17-N Saturn yellow) directly to the surface water of the receptacle containing Aedes aegypti male pupae. The marking efficacy, powder persistence and mosquito survival were assessed for male mosquito. We observed a high marking rate that increased with increasing amounts of fluorescent powder. Fluorescent powder lasted long on the mosquito body and did not induce a negative effect on their survival. This self-marking method reduces human intervention and mosquito handling, improving the quality of marked mosquitoes.In the implementation of mosquito control strategy programs using Sterile Insect Technique and other rear and release strategies, knowledge on the dispersion, competitiveness and survival of mosquitos is considered essential. To assess these parameters, marking techniques are generally used to differentiate colony mosquitoes from wild ones. Most of the existing mosquito marking methods require numerous manipulations that can impact their quality. In this study, we have developed a self-marking technique that can reduce the damage associated with mosquito handling. The marking technique consisted of adding fluorescent powder (DayGlo: A-17-N Saturn yellow) directly to the surface water of the receptacle containing Aedes aegypti male pupae. Different quantities of powder were used, and marking efficacy, powder persistence and mosquito survival were assessed. The results show a mean marking rate of 98 ± 1.61%, and the probability of marking increased significantly (p < 0.001) with increasing concentrations of fluorescent powder. Fluorescent powder persisted up to 20 days and did not induce a negative effect on mosquito survival (χ2 = 5.3, df = 7, p = 0.63). In addition, powder transfer did not occur between marked and unmarked populations. This marking method significantly reduces human intervention and mosquito handling during the marking process, improving the quality of marked mosquitoes used to assess SIT programs.

Applying and testing a novel method to estimate animal density from motion‐triggered cameras

AbstractEstimating animal abundance and density are fundamental goals of many wildlife monitoring programs. Camera trapping has become an increasingly popular tool to achieve these monitoring goals due to recent advances in modeling approaches and the capacity to simultaneously collect data on multiple species. However, estimating the density of unmarked populations continues to be problematic due to the difficulty in implementing complex modeling approaches, low precision of estimates, and absence of rigor in testing of model assumptions and their influence on results. Here, we describe a novel approach that uses still image camera traps to estimate animal density without the need for individual identification, based on the time spent in front of the camera (TIFC). Using results from a large‐scale multispecies monitoring program with nearly 3000 cameras deployed over 6 years in Alberta, Canada, we provide a reproducible methodology to estimate parameters and we test key assumptions of the TIFC model. We compare moose (Alces alces) density estimates from aerial surveys and TIFC, including incorporating correction factors for known TIFC assumption violations. The resulting corrected TIFC density estimates are comparable to aerial density estimates. We discuss the limitations of the TIFC method and areas needing further investigation, including the need for long‐term monitoring of assumption violations and the number of cameras necessary to provide precise estimates. Despite the challenges of assumption violations and high measurement error, cameras and the TIFC method can provide useful alternative or complementary animal density estimates for multispecies monitoring when compared to traditional monitoring methods.

Double-observer approach with camera traps can correct imperfect detection and improve the accuracy of density estimation of unmarked animal populations

Camera traps are a powerful tool for wildlife surveys. However, camera traps may not always detect animals passing in front. This constraint may create a substantial bias in estimating critical parameters such as the density of unmarked populations. We proposed the 'double-observer approach' with camera traps to counter the constraint, which involves setting up a paired camera trap at a station and correcting imperfect detection with a reformulated hierarchical capture-recapture model for stratified populations. We performed simulations to evaluate this approach's reliability and determine how to obtain desirable data for this approach. We then applied it to 12 mammals in Japan and Cameroon. The results showed that the model assuming a beta-binomial distribution as detection processes could correct imperfect detection as long as paired camera traps detect animals nearly independently (Correlation coefficient ≤ 0.2). Camera traps should be installed to monitor a predefined small focal area from different directions to satisfy this requirement. The field surveys showed that camera trap could miss animals by 3–40%, suggesting that current density estimation models relying on perfect detection may underestimate animal density by the same order of magnitude. We hope that our approach will be incorporated into existing density estimation models to improve their accuracy.

Deriving observation distances for camera trap distance sampling

AbstractCamera trap distance sampling (CTDS)o is a recently developed survey method to estimate animal abundance from camera trap data for unmarked populations. It requires the estimation of camera‐animal observation distances, which previously was done by comparing animal positions to reference labels at predefined intervals. Here, we test a photogrammetry approach to derive camera‐animal observation distances. We applied both, the reference label and photogrammetry approaches to five ungulate species varying widely in body size (Giraffa camelopardalis,Equus grevyi,Oryx dammah,Kobus megacerosandEudorcas thomsonii) and one ground‐dwelling bird species (Numida meleagris) inhabiting a large enclosure and estimated their density with CTDS. Both procedures provided highly correlated observation distances (ρ = 0.99,p &lt; 0.001). A paired t test revealed a minor but significantly higher mean of the photogrammetry approach (MD = 0.28 m,p &lt; 0.001). This, however, seems negligible as for analyses, distances were grouped in intervals of 2 to 5 metres. In general, estimated animal abundance was close to the true number of individuals in the enclosure for both approaches, with the exception of zebra, whose density was underestimated. The photogrammetry approach offers an alternative approach for deriving camera‐animal observation distances and is particularly useful for application in open habitats, with little occlusion of animals.

Tapirs in trouble: estimating Baird's tapir densities in the Sierra Madre de Chiapas, Mexico

AbstractSpatial capture–recapture models have been widely used to estimate densities of species where individuals can be uniquely identified, but alternatives have been developed for estimation of densities for unmarked populations. In this study we used camera-trap records from 2018 to estimate densities of a species that does not always have individually identifiable marks, Baird's tapir Tapirus bairdii, in the Sierra Madre de Chiapas, southern Mexico. We compared the performance of the spatial capture–recapture model with spatial mark–resight and random encounter models. The density of Baird's tapir did not differ significantly between the three models. The estimate of density was highest using the random encounter model (26/100 km2, 95% CI 12–41) and lowest using the capture–recapture model (8/100 km2, 95% CI 4–16). The estimate from the spatial mark–resight model was 10/100 km2 (95% CI 8–14), which had the lowest coefficient of variation, indicating a higher precision than with the other models. Using a second set of camera-trap data, collected in 2015–2016, we created occupancy models and extrapolated density to areas with potential occupancy of Baird's tapir, to generate a population estimate for the whole Sierra Madre de Chiapas. Our findings indicate the need to strengthen, and possibly expand, the protected areas of southern Mexico and to develop an action plan to ensure the conservation of Baird's tapir.

Evaluation of camera trap-based abundance estimators for unmarkedpopulations.

Estimates of species abundance are critical to understand population processes and to assess and select management actions. However, capturing and marking individuals for abundance estimation, while providing robust information, can be economically and logistically prohibitive, particularly for species with cryptic behavior. Camera traps can be used to collect data at temporal and spatial scales necessary for estimating abundance, but the use of camera traps comes with limitations when target species are not uniquely identifiable (i.e., "unmarked"). Abundance estimation is particularly useful in the management of invasive species, with herpetofauna being recognized as some of the most pervasive and detrimental invasive vertebrate species. However, the use of camera traps for these taxa presents additional challenges with relevancy across multiple taxa. It is often necessary to use lures to attract animals in order to obtain sufficient observations, yet lure attraction can influence species' landscape use and potentially induce bias in abundance estimators. We investigated these challenges and assessed the feasibility of obtaining reliable abundance estimates using camera-trapping data on a population of invasive brown treesnakes (Boiga irregularis) in Guam. Data were collected using camera traps in an enclosed area where snakes were subject to high-intensity capture-recapture effort, resulting in presumed abundance of 116 snakes (density=23/ha). We then applied spatial count, random encounter and staying time, space to event, and instantaneous sampling estimators to photo-capture data to estimate abundance and compared estimates to our presumed abundance. We found that all estimators for unmarked populations performed poorly, with inaccurate or imprecise abundance estimates that limit their usefulness for management in this system. We further investigated the sensitivity of these estimators to the use of lures (i.e., violating the assumption that animal behavior is unchanged by sampling) and camera density in a simulation study. Increasing the effective distances of a lure (i.e., lure attraction) and camera density both resulted in biased abundance estimates. Each estimator rarely recovered truth or suffered from convergence issues. Our results indicate that, when limited to unmarked estimators and the use of lures, camera traps alone are unlikely to produce abundance estimates with utility for brown treesnake management.

Assessment and novel application ofN‐mixture models for aerial surveys of wildlife

AbstractAerial surveys are a critical tool for inference of wildlife populations, yet are limited by current available methods to account for imperfect detection without marking animals.N‐mixture models use count data from replicate surveys for estimation of wildlife abundance while accounting for imperfect detection. This statistical framework was recently developed but is rarely applied to aerial surveys of terrestrial wildlife. We applied anN‐mixture model incorporating temporary emigration to a novel aerial survey design to estimate the availability, detection probability, and abundance of an unmarked population of white‐tailed deer (Odocoileus virginianus) in Michigan, USA, using surveys in 2014 and 2016. We assessed the selection of the sample unit size for theN‐mixture model in a post hoc sensitivity analysis and compared all estimates to a hierarchical distance sampling (HDS) approach. Hierarchical distance sampling is an expansion of conventional distance sampling (CDS), a standard method applied to aerial survey designs for wildlife, where HDS allows for modeling spatial variation in abundance and detection probability using environmental covariates and can account for changing population demographics. We foundN‐mixture models produced similar parameter estimates to HDS with both models inferring a population increase between years, which was consistent with independently available estimates for this population (2014N‐mixture = 402 [CI: 303.8–515.27], and HDS = 391 [CI: 297.67–504.63]; 2016N‐mixture = 732 [CI: 572.89–924.22], and HDS = 708 [CI: 552.45–893.89]). Estimates from both models were sensitive to the sample unit size selected in an open population framework. We recommend using ecological (e.g., home range size) and statistical (e.g., sample size) considerations when selecting a sample unit size that reflects the movement ecology of a species and evaluating the sensitivity of parameters to sample unit size. In our study,N‐mixture models and HDS models afforded a practical field approach and analytical methodology that resulted in precise estimation in a population of unmarked deer over time. When used with aerial surveys in terrestrial systems where distance sampling methods may be difficult to implement,N‐mixture models can address observer error and estimate populations for conservation and management goals without marking individual animals.

Assessing the robustness of time-to-event models for estimating unmarked wildlife abundance using remote cameras.

Recently developed methods, including time-to-event and space-to-event models, estimate the abundance of unmarked populations from encounter rates with camera trap arrays, addressing a gap in noninvasive wildlife monitoring. However, estimating abundance from encounter rates relies on assumptions that can be difficult to meet in the field, including random movement, population closure, and an accurate estimate of movement speed. Understanding how these models respond to violation of these assumptions will assist in making them more applicable in real-world settings. We used simulated walk models to test the effects of violating the assumptions of the time-to-event model under four scenarios: (1) incorrectly estimating movement speed, (2) violating closure, (3) individuals moving within simplified territories (i.e., movement restricted to partially overlapping circles), (4) and individuals clustering in preferred habitat. The time-to-event model was robust to closure violations, territoriality, and clustering when cameras were placed randomly. However, the model failed to estimate abundance accurately when movement speed was incorrectly estimated or cameras were placed nonrandomly with respect to habitat. We show that the time-to-event model can provide unbiased estimates of abundance when some assumptions that are commonly violated in wildlife studies are not met. Having a robust method for estimating the abundance of unmarked populations with remote cameras will allow practitioners to monitor a more diverse array of populations noninvasively. With the time-to-event model, placing cameras randomly with respect to animal movement and accurately estimating movement speed allows unbiased estimation of abundance. The model is robust to violating the other assumptions we tested.