Improving geographically extensive acoustic survey designs for modeling species occurrence with imperfect detection and misidentification

Abstract

Acoustic recording units (ARUs) enable geographically extensive surveys of sensitive and elusive species. However, a hidden cost of using ARU data for modeling species occupancy is that prohibitive amounts of human verification may be required to correct species identifications made from automated software. Bat acoustic studies exemplify this challenge because large volumes of echolocation calls could be recorded and automatically classified to species. The standard occupancy model requires aggregating verified recordings to construct confirmed detection/non‐detection datasets. The multistep data processing workflow is not necessarily transparent nor consistent among studies. We share a workflow diagramming strategy that could provide coherency among practitioners. A false‐positive occupancy model is explored that accounts for misclassification errors and enables potential reduction in the number of confirmed detections. Simulations informed by real data were used to evaluate how much confirmation effort could be reduced without sacrificing site occupancy and detection error estimator bias and precision. We found even under a 50% reduction in total confirmation effort, estimator properties were reasonable for our assumed survey design, species‐specific parameter values, and desired precision. For transferability, a fully documented r package, , for implementing a false‐positive occupancy model is provided. Practitioners can apply to optimize their own study design (required sample sizes, number of visits, and confirmation scenarios) for properly implementing a false‐positive occupancy model with bat or other wildlife acoustic data. Additionally, our work highlights the importance of clearly defining research objectives and data processing strategies at the outset to align the study design with desired statistical inferences.

Accounting for heterogeneity in false‐positive detection rate in southeastern beach mouse habitat occupancy models

“…Ferguson et al (2015) recommended that researchers always use occupancy models that allow falsepositive detections, to avoid bias in estimates of occupancy. We agree, but caution that in our simulations and in those of others (Ferguson et al 2015, Clement 2016, Banner et al 2018, Chambert et al 2018a), optimal Notes: SD, standard deviation. Ψ is the probability that site is occupied during the first season, φ is the probability that an occupied site stays occupied, c is the probability that an unoccupied site becomes occupied, and fp is the probability of false detection at an unoccupied site.…”

“…In conclusion, previous studies have shown, both empirically and with simulations, that even when including some level of certain detections, heterogeneity in detection and false occupancy rates can induce bias in estimates of habitat occupancy (Ferguson et al 2015, Banner et al 2018, Chambert et al 2018a). These studies also provide several methods to reduce this bias, including random sampling in study design and modeling the heterogeneity in detection parameters through covariates.…”

“…, Clement , Banner et al. , Chambert et al. ), optimal survey design involved subtle trade‐offs between bias and precision, and in some cases, detection parameters did not behave as expected.…”

“…It is well established that not accounting for false-positive detections (when present) will cause bias in estimates of dynamic habitat occupancy parameters (Miller et al 2013, Chambert et al 2015, Ferguson et al 2015. However, even when false-positive detection is included in occupancy models, heterogeneity in true and false detection rates can induce bias in estimates of habitat occupancy (Ferguson et al 2015, Banner et al 2018, Chambert et al 2018a. Our simulations showed that large bias can occur when heterogeneous falsepositive detections are ignored, and models including false-positive detections and accounting for heterogeneity among habitats fit our data much better than models that did not account for false-positive detections.…”

Temporally adaptive acoustic sampling to maximize detection across a suite of focal wildlife species

“…Automated detection methods may fail to detect sounds issued by species of interest (false negatives), or mistakenly detect false alarms not issued by the target species (false positives; Acevedo, Corrada‐Bravo, Corrada‐Bravo, Villanueva‐Rivera, & Aide, ; Balantic & Donovan, ; Buxton & Jones, ; Duan et al, ; Marques et al, ). Occupancy modeling frameworks are a well‐established approach for accommodating the detection mistakes that arise from remote acoustic monitoring and can deal both with false negatives (Cerqueira & Aide, ; Furnas & Callas, ; Rich, Beissinger, Brashares, & Furnas, ) and false positives (Balantic & Donovan, ; Banner et al, ; Chambert, Miller, & Nichols, ; Chambert, Waddle, Miller, Walls, & Nichols, ). However, false negatives due to suboptimal automated detection methodologies are distinct from false negatives that occur as a consequence of deficient audio sampling schedules.…”