Integrating telemetry data at several scales with spatial capture–recapture to improve density estimates

AbstractMany applications in ecology depend on unbiased and precise estimates of animal population density. Spatial capture recapture models and their variants have become the preferred tool for estimating densities of carnivores. Within the spatial capture-recapture family are variants that require individual identification of all encounters (spatial capture-recapture), individual identification of a subset of a population (spatial mark-resight), or no individual identification (spatial count). In addition, these models can incorporate telemetry data (all models) and the marking process (spatial mark-resight). However, the consistency of results among methods and the relative precision of estimates in a real-world setting are unknown. Consequently, it is unclear how much and what type of data are needed to achieve satisfactory density estimates. We tested a suite of models to estimate population densities of black bears (Ursus americanus), bobcats (Lynx rufus), cougars (Puma concolor), and coyotes (Canis latrans). For each species we genotyped fecal DNA collected with detection dogs. A subset of individuals from each species were affixed with GPS collars bearing unique markings to be resighted by remote cameras set on a 1 km grid. We fit 10 models for each species ranging from those requiring no animals to be individually recognizable to others that necessitate full individual recognition. We then assessed the contribution of incorporating telemetry data to each model and the marking process to the mark-resight model. Finally, we developed an integrated hybrid model that combines camera, physical capture, genetic, and GPS data into a single hierarchical model. Importantly, we find that spatial count models that do not individually identify animals fail in all cases whether or not telemetry data are included. Results improved as models contained more information on individual identity. Models where a subset of individuals were identifiable yielded qualitatively similar results, but can produce quantitatively divergent estimates, suggesting that long-term population monitoring should use a consistent method across years. Incorporation of telemetry data and the marking process can produce more accurate and precise density estimates. Our results can be used to guide future study designs to efficiently estimate carnivore densities for better understanding of population dynamics, predator-prey relationships, and community assemblages.

Download Full-text

Spatial Mark-Resight for Categorically Marked Populations with an Application to Genetic Capture-Recapture

10.1101/299982 ◽

2018 ◽

Cited By ~ 2

Author(s):

Ben C. Augustine ◽

Frances E. C. Stewart ◽

J. Andrew Royle ◽

Jason T. Fisher ◽

Marcella J. Kelly

Keyword(s):

Individual Heterogeneity ◽

Positive Bias ◽

Telemetry Data ◽

Detection Function ◽

Noninvasive Methods ◽

Data Set ◽

Density Estimates ◽

Mark Recapture ◽

Capture Recapture ◽

Scr Model

AbstractThe estimation of animal population density is a fundamental goal in wildlife ecology and management, commonly met using mark recapture or spatial mark recapture (SCR) study designs and statistical methods. Mark-recapture methods require the identification of individuals; however, for many species and sampling methods, particularly noninvasive methods, no individuals or only a subset of individuals are individually identifiable. The unmarked SCR model, theoretically, can estimate the density of unmarked populations; however, it produces biased and imprecise density estimates in many sampling scenarios typically encountered. Spatial mark-resight (SMR) models extend the unmarked SCR model in three ways: 1) by introducing a subset of individuals that are marked and individually identifiable, 2) introducing the possibility of individual-linked telemetry data, and 3) introducing the possibility that the capture-recapture data from the survey used to deploy the marks can be used in a joint model, all improving the reliability of density estimates. The categorical spatial partial identity model (SPIM) improves the reliability of density estimates over unmarked SCR along another dimension, by adding categorical identity covariates that improve the probabilistic association of the latent identity samples. Here, we combine these two models into a “categorical SMR” model to exploit the benefits of both models simultaneously. We demonstrate using simulations that SMR alone can produce biased and imprecise density estimates with sparse data and/or when few individuals are marked. Then, using a fisher (Pekania pennanti) genetic capture-recapture data set, we show how categorical identity covariates, marked individuals, telemetry data, and jointly modeling the capture survey used to deploy marks with the resighting survey all combine to improve inference over the unmarked SCR model. As previously seen in an application of the categorical SPIM to a real-world data set, the fisher data set demonstrates that individual heterogeneity in detection function parameters, especially the spatial scale parameter σ, introduces positive bias into latent identity SCR models (e.g., unmarked SCR, SMR), but the categorical SMR model provides more tools to reduce this positive bias than SMR or the categorical SPIM alone. We introduce the possibility of detection functions that vary by identity category level, which will remove individual heterogeneity in detection function parameters than is explained by categorical covariates, such as individual sex. Finally, we provide efficient SMR algorithms that accommodate all SMR sample types, interspersed marking and sighting periods, and any number of identity covariates using the 2-dimensional individual by trap data in conjunction with precomputed constraint matrices, rather than the 3-dimensional individual by trap by occasion data used in SMR algorithms to date.

Download Full-text

Does the punishment fit the crime? Consequences and diagnosis of misspecified detection functions in Bayesian spatial capture-recapture modelling

10.1101/2021.01.12.426298 ◽

2021 ◽

Author(s):

Soumen Dey ◽

Richard Bischof ◽

Pierre P. A. Dupont ◽

Cyril Milleret

Keyword(s):

Home Range ◽

Space Use ◽

Relative Bias ◽

Diagnostic Tools ◽

List Type ◽

Detection Function ◽

Density Estimates ◽

Use Patterns ◽

Wide Range ◽

Capture Recapture

AbstractSpatial capture-recapture (SCR) is now used widely to estimate wildlife densities. At the core of SCR models lies the detection function, linking individual detection probability to the distance from its latent activity center. The most common function (half-normal) assumes a bivariate normal space use and consequently detection pattern. This is likely an oversimplification and misrepresentation of real-life animal space use patterns, but studies have reported that density estimates are relatively robust to misspecified detection functions. However, information about consequences of such misspecification on space use parameters (e.g. home range area), as well as diagnostic tools to reveal it are lacking.We simulated SCR data under six different detection functions, including the half-normal, to represent a wide range of space use patterns. We then fit three different SCR models, with the three simplest detection functions (half-normal, exponential and half-normal plateau) to each simulated data set. We evaluated the consequences of misspecification in terms of bias, precision and coverage probability of density and home range area estimates. We also calculated Bayesian p-values with respect to different discrepancy metrics to assess whether these can help identify misspecifications of the detection function.We corroborate previous findings that density estimates are robust to misspecifications of the detection function. However, estimates of home range area are prone to bias when the detection function is misspecified. When fitted with the half-normal model, average relative bias of 95% kernel home range area estimates ranged between −25% and 26% depending on the misspecification. In contrast, the half-normal plateau model (an extension of the half-normal) returned average relative bias that ranged between −26% and −4%. Additionally, we found useful heuristic patterns in Bayesian p-values to diagnose the misspecification in detection function.Our analytical framework and diagnostic tools may help users select a detection function when analyzing empirical data, especially when space use parameters (such as home range area) are of interest. We urge development of additional custom goodness of fit diagnostics for Bayesian SCR models to help practitioners identify a wider range of model misspecifications.

Download Full-text

Density Estimates of Unmarked Large Mammals at Camera Traps Vary among Models, Species, and Years, Signalling Importance of Model Assumptions

10.22541/au.163768980.02146968/v1 ◽

2021 ◽

Author(s):

Jason Fisher ◽

Joanna Burgar ◽

Melanie Dickie ◽

Cole Burton ◽

Rob Serrouya

Keyword(s):

Density Estimation ◽

Black Bear ◽

Camera Traps ◽

Mammal Species ◽

Density Estimates ◽

Movement Rates ◽

Social Species ◽

Marked Variability ◽

Capture Recapture ◽

Best Fitting

Density estimation is a key goal in ecology but accurate estimates remain elusive, especially for unmarked animals. Data from camera-trap networks combined with new density estimation models can bridge this gap but recent research has shown marked variability in accuracy, precision, and concordance among estimators. We extend this work by comparing estimates from two different classes of models: unmarked spatial capture-recapture (spatial count, SC) models, and Time In Front of Camera (TIFC) models, a class of random encounter model. We estimated density for four large mammal species with different movement rates, behaviours, and sociality, as these traits directly relate to model assumptions. TIFC density estimates were typically higher than SC model estimates for all species. Black bear TIFC estimates were ~ 10-fold greater than SC estimates. Caribou TIFC estimates were 2-10 fold greater than SC estimates. White-tailed deer TIFC estimates were up to 100-fold greater than SC estimates. Differences of 2-5 fold were common for other species in other years. SC estimates were annually stable except for one social species; TIFC estimates were highly annually variable in some cases and consistent in others. Tests against densities obtained from DNA surveys and aerial surveys also showed variable concordance and divergence. For gregarious animals TIFC may outperform SC due to the latter model’s assumption of independent activity centres. For curious animals likely to investigate camera traps, SC may outperform TIFC, which assumes animal behavior is unaffected by cameras. Unmarked models offer great possibilities, but a pragmatic approach employs multiple estimators where possible, considers the ecological plausibility of assumptions, and uses an informed multi-inference approach to seek estimates from models with assumptions best fitting a species’ biology.

Download Full-text

Evaluating Live trapping and Camera-based Indices of Small Mammal Density

Canadian Journal of Zoology ◽

10.1139/cjz-2020-0298 ◽

2021 ◽

Author(s):

Mitchell Alan Parsons ◽

ALISHIA ORLOFF ◽

Laura Prugh

Keyword(s):

Small Mammal ◽

Abundant Species ◽

Monitoring Methods ◽

Density Estimates ◽

Full Dataset ◽

Live Trapping ◽

Minimum Number ◽

Capture Recapture ◽

Two Indices ◽

Best Fit

Density estimates are integral to wildlife management, but they can be costly to obtain. Indices of density may provide efficient alternatives, but calibration is needed to ensure the indices accurately reflect density. We evaluated several indices of small mammal density using live trapping and motion-activated cameras in Washington’s Cascade Mountains. We used linear regression to compare spatially-explicit capture recapture density estimates of mice, voles, and chipmunks to four indices. Two indices were based on live trapping (minimum number alive and number of captures per 100 trap nights) and two indices were based on photos from motion-activated cameras (proportion of cameras detecting a species and the number of independent detections). We evaluated how the accuracy of trap-based indices increased with trapping effort using subsets of the full dataset (n = 7 capture occasions per site). Most indices provided reliable indicators of small mammal density, and live trapping indices (R2=0.64 – 0.98) outperformed camera-based indices (R2=0.24 – 0.86). All indices performed better for more abundant species. The effort required to estimate each index varied, and indices that required more effort performed better. These findings should help managers, conservation practitioners, and researchers select small mammal monitoring methods that best fit their needs.

Download Full-text

Bringing clarity to the clouded leopard Neofelis diardi: first density estimates from Sumatra

Oryx ◽

10.1017/s003060531400043x ◽

2014 ◽

Vol 48 (4) ◽

pp. 536-539 ◽

Cited By ~ 20

Author(s):

Rahel Sollmann ◽

Matthew Linkie ◽

Iding A. Haidir ◽

David W. Macdonald

Keyword(s):

Mixed Forest ◽

Camera Trap ◽

Forest Loss ◽

Data Sets ◽

Panthera Tigris ◽

Density Estimates ◽

Clouded Leopard ◽

Data Limitations ◽

Capture Recapture ◽

Study Sites

AbstractWe use data from camera-trap surveys for tigers Panthera tigris in combination with spatial capture–recapture models to provide the first density estimates for the Sunda clouded leopard Neofelis diardi on Sumatra. Surveys took place during 2004–2007 in the Kerinci landscape. Densities were 0.385–1.278 per 100 km2. We found no statistically significant differences in density among four study sites or between primary and mixed forest. Because the data sets are too small to account for differences in detection parameters between sexes, density is probably underestimated. Estimates are comparable to previous estimates of 1–2 per 100 km2 from the lowlands of central Sabah, on Borneo. Data limitations suggest that camera-trap surveys for Sunda clouded leopards require traps spaced more closely, to increase the chance of recaptures at different traps. Nevertheless, these first density estimates for clouded leopards on Sumatra provide a benchmark for measuring future conservation impact on an island that is undergoing rapid forest loss.

Download Full-text

Density of the Vulnerable Sunda clouded leopard Neofelis diardi in two commercial forest reserves in Sabah, Malaysian Borneo

Oryx ◽

10.1017/s0030605311001694 ◽

2012 ◽

Vol 46 (3) ◽

pp. 423-426 ◽

Cited By ~ 29

Author(s):

Andreas Wilting ◽

Azlan Mohamed ◽

Laurentius N. Ambu ◽

Peter Lagan ◽

Sam Mannan ◽

...

Keyword(s):

Management Practices ◽

Iucn Red List ◽

Separate Species ◽

Density Estimates ◽

Clouded Leopard ◽

Forest Reserves ◽

Protected Forests ◽

Capture Recapture ◽

Commercial Forest ◽

Malaysian Borneo

AbstractRecently the Sunda clouded leopard Neofelis diardi was recognized as a separate species distinct from the clouded leopard Neofelis nebulosa of mainland Asia. Both species are categorized as Vulnerable on the IUCN Red List. Little is known about the newly identified species and, in particular, information from forests outside protected areas is scarce. Here we present one of the first density estimates calculated with spatial capture–recapture models using camera-trap data. In two commercial forest reserves in Sabah (both certified for their sustainable management practices) the density of the Sunda clouded leopard was estimated to be c. 1 per 100 km2 (0.84±SE 0.42 and 1.04±SE 0.58). The presence of the Sunda clouded leopard in such forests is encouraging for its conservation but additional studies from other areas, including protected forests, are needed to compare and evaluate these densities.

Download Full-text

Black and Brown Bear Density Estimates Using Modified Capture-Recapture Techniques in Alaska

Bears Their Biology and Management ◽

10.2307/3872604 ◽

1987 ◽

Vol 7 ◽

pp. 23

Author(s):

Sterling D. Miller ◽

Earl F. Becker ◽

Warren B. Ballard

Keyword(s):

Brown Bear ◽

Density Estimates ◽

Capture Recapture

Download Full-text

How Does Spatial Study Design Influence Density Estimates from Spatial Capture-Recapture Models?

PLoS ONE ◽

10.1371/journal.pone.0034575 ◽

2012 ◽

Vol 7 (4) ◽

pp. e34575 ◽

Cited By ~ 93

Author(s):

Rahel Sollmann ◽

Beth Gardner ◽

Jerrold L. Belant

Keyword(s):

Study Design ◽

Density Estimates ◽

Capture Recapture

Download Full-text

Assumptions about fence permeability influence density estimates for brown hyaenas across South Africa

Scientific Reports ◽

10.1038/s41598-020-77188-7 ◽

2021 ◽

Vol 11 (1) ◽

Author(s):

Kathryn S. Williams ◽

Samual T. Williams ◽

Rebecca J. Welch ◽

Courtney J. Marneweck ◽

Gareth K. H. Mann ◽

...

Keyword(s):

South Africa ◽

Population Density ◽

Wildlife Management ◽

Management Decisions ◽

Density Estimates ◽

Future Studies ◽

Dominant Feature ◽

Capture Recapture ◽

Number Of Individuals ◽

Density Results

AbstractWildlife population density estimates provide information on the number of individuals in an area and influence conservation management decisions. Thus, accuracy is vital. A dominant feature in many landscapes globally is fencing, yet the implications of fence permeability on density estimation using spatial capture-recapture modelling are seldom considered. We used camera trap data from 15 fenced reserves across South Africa to examine the density of brown hyaenas (Parahyaena brunnea). We estimated density and modelled its relationship with a suite of covariates when fenced reserve boundaries were assumed to be permeable or impermeable to hyaena movements. The best performing models were those that included only the influence of study site on both hyaena density and detection probability, regardless of assumptions of fence permeability. When fences were considered impermeable, densities ranged from 2.55 to 15.06 animals per 100 km2, but when fences were considered permeable, density estimates were on average 9.52 times lower (from 0.17 to 1.59 animals per 100 km2). Fence permeability should therefore be an essential consideration when estimating density, especially since density results can considerably influence wildlife management decisions. In the absence of strong evidence to the contrary, future studies in fenced areas should assume some degree of permeability in order to avoid overestimating population density.

Download Full-text