Estimating ungulate abundance while accounting for multiple sources of observation error

Abstract

Whether a species is rare or overabundant, accurate estimates of population abundance are essential for the development and assessment of conservation plans and management goals. Aerial surveys are commonly used to estimate population abundance and a variety of methods have been used to account for recognized biases associated with imperfect detection. Rarely addressed, however, is the possibility of recording duplicate observations and the influence of duplicate observations on estimates of abundance. Using data provided by Global Positioning System (GPS)‐collared bison (Bison bison) and helicopter survey paths conducted during 2011–2013 in the Henry Mountains, Utah, USA, we determined whether GPS‐collared bison were observed, missed, or observed multiple times. Using these data, covariates hypothesized to influence observation errors, and generalized linear models, we found the probability of detecting a GPS‐collared bison (study‐wide mean of 95%) was consistently greater for individuals in larger groups and less rugged terrain. We also found that the probability of recording a duplicate observation (study‐wide mean of 4.4%) increased with time since the prior survey observation. By incorporating both types of observation error into a modified Horvitz–Thompson estimator, more accurate estimates of bison abundance can be attained, particularly in years when animals segregate into small groups, utilize terrain that makes them difficult to see, or have ample time to move into areas not yet surveyed and be counted again. Our approach of accounting for multiple sources of observation error can easily be applied to aerial surveys for other species in other systems, and has the potential to improve the rigor of wildlife abundance estimation from aerial surveys. © 2016 The Wildlife Society.

Modelling wildlife species abundance using automated detections from drone surveillance

“…The abundance estimate and lower confidence interval obtained for the majority of surveys was greater than the actual number of target animals on site, and this result is similar to those of Terletzky and Koons (2016). However, unlike the results reported in that study the difference between the abundance estimates and true counts was not due to inflation of raw counts, but from a failure to reduce the initial number of detections (Terletzky & Koons 2016). This suggests that accounting for duplicate detections alone may not be sufficient to eliminate spurious detections and obtain accurate abundance estimates from automated counts.…”

“…Following the methodology of Terletzky and Koons (2016), generalized linear models (GLMs) with a binomial distribution and a logit link function were used to examine the impact of covariates on probability of detection (y = 1 for a successful detection, y = 0 for a miss), and of a detection being a duplicate (y = 1 when a detection is a duplicate, y = 0 for a unique detection). For probability of detection these covariates included visibility score, tree height and koala height in tree ( Table 1).…”

“…For probability of detection these covariates included visibility score, tree height and koala height in tree ( Table 1). Group size and movement covariates suggested by Terletzky and Koons (2016) were not included as koalas tend to be solitary and remain stationary throughout surveying (Dique et al 2003). However, the field of view of the drone led to overlap in images along and between transects, which could have resulted in the same koala being seen and detected multiple times during surveys.…”

“…Therefore time since previous detection was considered a potential covariate for duplicate detection, as proposed by Terletzky and Koons (2016) ( Table 1). Time since the beginning of the survey was not applicable to koalas as this was originally proposed to account for movement of animals during surveying which does not occur with this target species (Terletzky & Koons 2016;Dique et al 2003) Table 1. List of covariates used in models of probability of detection and duplicate detection for automated detections of koalas in RPAS-derived thermal imagery.…”

“…Unlike conventional aerial surveys for which the biases affecting abundance estimates have been extensively enumerated, the unique biases inherent in RPAS surveying of wildlife have rarely been investigated or accounted for in previous models of population numbers (Brack et al 2018). The same is true of the biases uniquely affecting automated detection methods, despite many studies demonstrating that computer vision is not impacted by the same constraints as the human eye (Conn et al 2014) The aim of this paper is to investigate the feasibility of integrating automated detections of koalas in RPASderived thermal imaging into an existing method of abundance estimation for wildlife developed by Terletzky and Koons (2016). Their approach was determined to come the closest of any existing modelling technique to addressing the errors inherent in the method used to survey koalas in a study by Corcoran et al (2019), as the number of missed animals and duplicate detections in surveys were calculated using telemetry to determine the presence and location of radio-collared animals within the survey area in a similar method to Terletzky & Koons (2016).…”