Efficacy of depletion models for estimating abundance of endangered fishes in streams

Estimating riverwide abundance of juvenile fish populations is challenging because detection probability is typically low and juveniles can be patchily distributed over large areas. We used a hierarchical Bayesian model to estimate the abundance of juvenile steelhead Oncorhynchus mykiss in two rivers in British Columbia over 3 years based on a multigear, two-phase sampling design. These estimates were used to drive a simulation model to evaluate how the precision of abundance estimates varied with the number of single-pass index and mark–recapture sites that were sampled, the proportion of shoreline sampled, and the mean and variation of detection probability and fish density across sites. The extent of variation in fish densities across index sites was the most important factor influencing the precision of river-wide abundance estimates, and increasing the number of index sites was the best approach to reduce variability in abundance estimates. River size, which controls the proportion of habitat sampled for a given level of sampling effort, had a moderate effect on precision, but only when the extent of site-to-site variation in fish density was high. Factors affecting detection probability, such as the number of mark–recapture sites, the mean detection probability, or the extent of variation in detection probability across sites, had much less influence on precision of abundance estimates unless the proportion of river sampled was high. Hierarchical Bayesian models are no substitute for collecting informative data, but they improve our understanding of variance structure, which is critical for providing realistic estimates of uncertainty and designing informative and efficient sampling programs. Received July 16, 2015; accepted October 7, 2015 Published online March 8, 2016

Although point counts are frequently used in ornithological studies, basic assumptions about detection probabilities often are untested. We apply a double-observer approach developed to estimate detection probabilities for aerial surveys (Cook and Jacobson 1979) to avian point counts. At each point count, a designated “primary” observer indicates to another (“secondary”) observer all birds detected. The secondary observer records all detections of the primary observer as well as any birds not detected by the primary observer. Observers alternate primary and secondary roles during the course of the survey. The approach permits estimation of observer-specific detection probabilities and bird abundance. We developed a set of models that incorporate different assumptions about sources of variation (e.g. observer, bird species) in detection probability. Seventeen field trials were conducted, and models were fit to the resulting data using program SURVIV. Single-observer point counts generally miss varying proportions of the birds actually present, and observer and bird species were found to be relevant sources of variation in detection probabilities. Overall detection probabilities (probability of being detected by at least one of the two observers) estimated using the double-observer approach were very high (>0.95), yielding precise estimates of avian abundance. We consider problems with the approach and recommend possible solutions, including restriction of the approach to fixed-radius counts to reduce the effect of variation in the effective radius of detection among various observers and to provide a basis for using spatial sampling to estimate bird abundance on large areas of interest. We believe that most questions meriting the effort required to carry out point counts also merit serious attempts to estimate detection probabilities associated with the counts. The double-observer approach is a method that can be used for this purpose.

Although point counts are frequently used in ornithological studies, basic assumptions about detection probabilities often are untested. We apply a double-observer approach developed to estimate detection probabilities for aerial surveys (Cook and Jacobson 1979) to avian point counts. At each point count, a designated “primary” observer indicates to another (“secondary”) observer all birds detected. The secondary observer records all detections of the primary observer as well as any birds not detected by the primary observer. Observers alternate primary and secondary roles during the course of the survey. The approach permits estimation of observer-specific detection probabilities and bird abundance. We developed a set of models that incorporate different assumptions about sources of variation (e.g. observer, bird species) in detection probability. Seventeen field trials were conducted, and models were fit to the resulting data using program SURVIV. Single-observer point counts generally miss varying proportions of the birds actually present, and observer and bird species were found to be relevant sources of variation in detection probabilities. Overall detection probabilities (probability of being detected by at least one of the two observers) estimated using the double-observer approach were very high (>0.95), yielding precise estimates of avian abundance. We consider problems with the approach and recommend possible solutions, including restriction of the approach to fixed-radius counts to reduce the effect of variation in the effective radius of detection among various observers and to provide a basis for using spatial sampling to estimate bird abundance on large areas of interest. We believe that most questions meriting the effort required to carry out point counts also merit serious attempts to estimate detection probabilities associated with the counts. The double-observer approach is a method that can be used for this purpose.

<i>Abstract</i>.—A common assumption in acoustic or radio telemetry studies is that tag transmission strength is homogeneous for a particular tag type, which in turn supports the assumption that detection ranges or mark–recapture detection probabilities are homogenous among tagged fish. Variation among tags in acoustic intensity could reduce precision in detection probability estimates that do not account for it, and therefore possibly in the precision of survival or abundance estimates. Simple methods are suggested for quantifying variation in tag strength prior to tagging fish and incorporating these measurements into mark–recapture models. At little extra effort to the researcher, these measurements could explain part of the variation in detection probability estimates and therefore could increase the precision of survival or abundance estimates of migrating fish. This potential source of variation in detection probabilities was investigated in a case study with migrating salmon smolts. An index of tag strength was quantified while coded acoustic tags were activated prior to tagging fish. Detection and survival probabilities were estimated with standard mark–recapture methods for the downstream and early ocean migration. A model that included the tag strength index as an additive covariate of detection probabilities had a reasonable level of support compared to a model without the index, suggesting that this source of variation should not be ignored.

Camera traps are an increasingly popular tool for monitoring medium to large mammals, but the influence of camera trap placement on the detection probabilities of different species has seldom been investigated. In this study we explore the influence of roads on the detection probability of medium to large mammals in three vegetation types in the Little Karoo, an arid biodiversity hotspot. We placed cameras in nine 100 m-long transects, running perpendicular from roads within a conservation area. The camera traps were spaced at ~25 m intervals, and were active for an average of 88 days each. Detection probabilities relative to distance from roads showed extensive variation between species and habitat types. There was no clear relationship between distance from the road and the detection probability of most species and guilds, although carnivore detection probability declined significantly as distance from roads increased in all vegetation types. Our results suggest that there is considerable inter-specific variation in detection probability that is significantly influenced by camera trap location relative to roads. Therefore studies that seek to maximise the detection rates of particular species or guilds (e.g. carnivores) by placing cameras on prominent roads and trails are unlikely to provide reliable estimates of the relative abundance of the broader range of sympatric species; a trend observed elsewhere but hitherto untested in arid environments. We recommend that future studies employ a mixed design of cameras located on- and off-roads to provide better estimates of biodiversity in general and predators specifically.

The pooling robustness property of distance sampling results in unbiased abundance estimation even when sources of variation in detection probability are not modeled. However, this property cannot be relied upon to produce unbiased subpopulation abundance estimates when using a single pooled detection function that ignores subpopulations. We investigate by simulation the effect of differences in subpopulation detectability upon bias in subpopulation abundance estimates. We contrast subpopulation abundance estimates using a pooled detection function with estimates derived using a detection function model employing a subpopulation covariate. Using point transect survey data from a multispecies songbird study, species-specific abundance estimates are compared using pooled detection functions with and without a small number of adjustment terms, and a detection function with species as a covariate. With simulation, we demonstrate the bias of subpopulation abundance estimates when a pooled detection function is employed. The magnitude of the bias is positively related to the magnitude of disparity between the subpopulation detection functions. However, the abundance estimate for the entire population remains unbiased except when there is extreme heterogeneity in detection functions. Inclusion of a detection function model with a subpopulation covariate essentially removes the bias of the subpopulation abundance estimates. The analysis of the songbird point count surveys shows some bias in species-specific abundance estimates when a pooled detection function is used. Pooling robustness is a unique property of distance sampling, producing unbiased abundance estimates at the level of the study area even in the presence of large differences in detectability between subpopulations. In situations where subpopulation abundance estimates are required for data-poor subpopulations and where the subpopulations can be identified, we recommend the use of subpopulation as a covariate to reduce bias induced in subpopulation abundance estimates.

Species occurring at low density can be difficult to detect and if not properly accounted for, imperfect detection will lead to inaccurate estimates of occupancy. Understanding sources of variation in detection probability and how they can be managed is a key part of monitoring. We used sightings data of a low-density and elusive raptor (white-headed vulture Trigonoceps occipitalis) in areas of known occupancy (breeding territories) in a likelihood-based modelling approach to calculate detection probability and the factors affecting it. Because occupancy was known a priori to be 100%, we fixed the model occupancy parameter to 1.0 and focused on identifying sources of variation in detection probability. Using detection histories from 359 territory visits, we assessed nine covariates in 29 candidate models. The model with the highest support indicated that observer speed during a survey, combined with temporal covariates such as time of year and length of time within a territory, had the highest influence on the detection probability. Averaged detection probability was 0.207 (s.e. 0.033) and based on this the mean number of visits required to determine within 95% confidence that white-headed vultures are absent from a breeding area is 13 (95% CI: 9–20). Topographical and habitat covariates contributed little to the best models and had little effect on detection probability. We highlight that low detection probabilities of some species means that emphasizing habitat covariates could lead to spurious results in occupancy models that do not also incorporate temporal components. While variation in detection probability is complex and influenced by effects at both temporal and spatial scales, temporal covariates can and should be controlled as part of robust survey methods. Our results emphasize the importance of accounting for detection probability in occupancy studies, particularly during presence/absence studies for species such as raptors that are widespread and occur at low densities.

Failure to estimate capture efficiency, defined as the probability of capturing individual fish, can introduce a systematic error or bias into estimates of fish abundance. We evaluated the efficacy of multipass electrofishing removal methods for estimating fish abundance by comparing estimates of capture efficiency from multipass removal estimates to capture efficiencies measured by the recapture of known numbers of marked individuals for bull trout Salvelinus confluentus and westslope cutthroat trout Oncorhynchus clarki lewisi. Electrofishing capture efficiency measured by the recapture of marked fish was greatest for westslope cutthroat trout and for the largest size-classes of both species. Capture efficiency measured by the recapture of marked fish also was low for the first electrofishing pass (mean, 28%) and decreased considerably (mean, 1.71 times lower) with successive passes, which suggested that fish were responding to the electrofishing procedures. On average, the removal methods overestimated three-pass capture efficiency by 39% and underestimated fish abundance by 88%, across both species and all size-classes. The overestimates of efficiency were positively related to the cross-sectional area of the stream and the amount of undercut banks and negatively related to the number of removal passes for bull trout, whereas for westslope cutthroat trout, the overestimates were positively related to the amount of cobble substrate. Three-pass capture efficiency measured by the recapture of marked fish was related to the same stream habitat characteristics that influenced (biased) the removal estimates and did not appear to be influenced by our sampling procedures, including fish marking. Simulation modeling confirmed our field observations and indicated that underestimates of fish abundance by the removal method were negatively related to first-pass sampling efficiency and the magnitude of the decrease in capture efficiency with successive passes. Our results, and those of other researchers, suggest that most electrofishing-removal-based estimates of fish abundance are likely to be biased and that these biases are related to stream characteristics, fish species, and size. We suggest that biologists regard electrofishing-removal-based estimates as biased indices and encourage them to measure and model the efficiency of their sampling methods to avoid introducing systematic errors into their data.

Abundance and density estimates are central to the field of ecology and are an important component of wildlife management. While many methods exist to estimate abundance from individually identifiable animals, it is much more difficult to estimate abundance of unmarked animals. One step toward noninvasive abundance estimation is the use of passive detectors such as remote cameras or acoustic recording devices. However, existing methods for estimating abundance from cameras for unmarked animals are limited by variable detection probability and have not taken full advantage of the information in camera trapping rate. We developed a time to event ( TTE ) model to estimate abundance from trapping rate. This estimate requires independent estimates of animal movement, so we collapsed the sampling occasions to create a space to event ( STE ) model that is not sensitive to movement rate. We further simplified the STE model into an instantaneous sampling ( IS ) estimator that applies fixed‐area counts to cameras. The STE and IS models utilize time‐lapse photographs to eliminate the variability in detection probability that comes with motion‐sensor photographs. We evaluated the three methods with simulations and performed a case study to estimate elk ( Cervus canadensis ) abundance from remote camera trap data in Idaho. Simulations demonstrated that the TTE model is sensitive to movement rate, but the STE and IS methods are unbiased regardless of movement. In our case study, elk abundance estimates were comparable to those from a recent aerial survey in the area, demonstrating that these new methods allow biologists to estimate abundance from unmarked populations without tracking individuals over time.

Summary Methods that aid in understanding the variation in the detection probability in acoustic telemetry can aid proper interpretations of data derived using such systems. In their recent paper, Huveneers et al. (Methods in Ecology and Evolution, 7, 825‐835, 2016) claim that a general detection probability model (Gjelland & Hedger, Methods in Ecology and Evolution, 4, 665–674, 2013) does not have the intended applicability across systems. However, Huveneer et al. applies predictions for a shallow freshwater application to results from a much deeper marine application, which is an invalid use of acoustic theory and the proposed general model. Users of acoustic telemetry are encouraged to acknowledge how environmentally induced variation in the acoustic attenuation coefficient influences the detection probability, because an understanding of this will aid prediction of how acoustic telemetry systems will work under various environmental conditions. Models used for predictions must be appropriately parameterized. Proper incorporation of spatiotemporal variation in acoustic detection probability may help reduce effects of environmentally induced biases in detection data. We challenge scientists to make further contributions to how acoustic theory can be incorporated in the modelling of detection probability in acoustic telemetry.

Population monitoring is fundamental for informing management decisions aimed at reducing the rapid rate of global biodiversity decline. Herpetofauna are experiencing declines worldwide and include species that are challenging to monitor. Raw counts and associated metrics such as richness indices are common for monitoring populations of herpetofauna; however, these methods are susceptible to bias as they fail to account for varying detection probabilities. Our goal was to develop a program for efficiently monitoring herpetofauna in southern Texas. Our objectives were to (1) estimate detection probabilities in an occupancy modeling framework using trap arrays for a diverse group of herpetofauna and (2) to evaluate the relative effectiveness of funnel traps, pitfall traps, and cover boards. We collected data with 36 arrays at 2 study sites in 2015 and 2016, for 2105 array-days resulting in 4839 detections of 51 species. We modeled occupancy for 21 species and found support for the hypothesis that detection probability varied over our sampling duration for 10 species and with rainfall for 10 species. For herpetofauna in our study, we found 14 and 12 species were most efficiently captured with funnel traps and pitfall traps, respectively, and no species were most efficiently captured with cover boards. Our results show that using methods that do not account for variations in detection probability are highly subject to bias unless the likelihood of false absences is minimized with exceptionally long capture durations. For monitoring herpetofauna in southern Texas, we recommend using arrays with funnel and pitfall traps and an analytical method such as occupancy modeling that accounts for variation in detection.

We describe an approach for estimating occupancy rate or the proportion of area occupied when heterogeneity in detection probability exists as a result of variation in abundance of the organism under study. The key feature of such problems, which we exploit, is that variation in abundance induces variation in detection probability. Thus, heterogeneity in abundance can be modeled as heterogeneity in detection probability. Moreover, this linkage between heterogeneity in abundance and heterogeneity in detection probability allows one to exploit a heterogeneous detection probability model to estimate the underlying distribution of abundances. Therefore, our method allows estimation of abundance from repeated observations of the presence or absence of animals without having to uniquely mark individuals in the population.

1. Monitoring programmes serve to track changes in the distribution and abundance of species. A major problem with most monitoring programmes is that species detection is imperfect and some populations are inevitably missed. Estimates of abundance and distribution are biased when detection probabilities are not taken into account. 2. Data were analysed from a large-scale volunteer-based amphibian monitoring programme using recently developed methods for estimating site occupancy. The analysis revealed that detection probabilities were much smaller than 100%, and varied among species and between years in an unpredictable way despite the fact that standardized field methods were used. Different environmental covariates best explained variation in detection probabilities among species and between years. It is therefore very difficult to standardize field methods in a way that eliminates such variation. 3. When detection probabilities were not taken into account, estimates of the proportion of sites where a species occurred and estimates of trends in site occupancy were biased. Bias was large and was related to detection probability. 4. Site occupancy models are a useful tool for monitoring the distribution of amphibians (and other animals and plants). These methods should be useful for assessing 'favourable conservation status' as required under the European Habitats Directive. Copyright © 2005 John Wiley & Sons, Ltd.

A multigear sampling method to quantify fish richness and abundance in Amazonian Terra firme streams was assessed. This method is based on the four‐pass removal method using the combined application of different nets. The efficiency of the method was explored over 10 replicated sites along three streams over day, night and seasons. Use of four successive passes allowed both abundant and rare species to be collected and the abundance of common species to be estimated. On average, a high proportion (41%) of rare species was collected per sample (only one or two individuals) after four fishing passes. The efficiency of the sampling method to detect species richness per sample between successive passes was estimated using an autosimilarity approach. Although species richness and abundance increased with successive passes, no major differences were obtained between the third and fourth pass. A single pass considerably underestimated the richness and abundance of species in these type of streams, and night sampling also increased beta‐diversity by at least 20%. Abundance estimates demonstrated high efficiency with an overall sampling error of only 8 and 11% for samples and single species, respectively. Capture efficiency differed among fish species exhibiting different ecological traits and showed significant differences among seasons for total samples. Results supported the robustness of the method and its suitability to quantify fish richness and abundance in small, wadeable Amazonian Terra firme streams inhabited by highly diverse fish assemblages.

This paper compares methods for modeling the probability of removal when variable amounts of removal effort are present. A hierarchical modeling framework can produce estimates of animal abundance and detection from replicated removal counts taken at different locations in a region of interest. A common method of specifying variation in detection probabilities across locations or replicates is with a logistic model that incorporates relevant detection covariates. As an alternative to this logistic model, we propose using a catch–effort (CE) model to account for heterogeneity in detection when a measure of removal effort is available for each removal count. This method models the probability of detection as a nonlinear function of removal effort and a removal probability parameter that can vary spatially. Simulation results demonstrate that the CE model can effectively estimate abundance and removal probabilities when average removal rates are large but both the CE and logistic models tend to produce biased estimates as average removal rates decrease. We also found that the CE model fits better than logistic models when estimating wild turkey abundance using harvest and hunter counts collected by the Minnesota Department of Natural Resources during the spring turkey hunting season.