Observer error in identifying species using indirect signs: analysis of a river otter track survey technique

• Main Author: Evans, Jonah Wy
• Other Authors: Packard, Jane M., Higginbotham, Billy J., Stronza, Amanda
• Format: Book
• Language: English
• Published: Texas A&M University 2007
• Subjects: Observer error; observer bias; river otter; Lontra canadensis; Lutra lutra; animal tracks; tracking; spoor; pugmark; presence/absence; occupancy; index of abundance; Lontra
• Online Access: https://hdl.handle.net/1969.1/5853

Indirect signs of species presence (e.g., tracks, scats, hairs) are frequently used to detect target species in occupancy, presence/absence, and other wildlife studies. Indirect signs are often more efficient than direct observation of elusive animals, making such signs well suited for long-term and broad-scale monitoring programs. However, error associated with misidentification of indirect signs can be high, and should be measured if meaningful inferences about population parameters are to be made. This study addressed the need for systematic approaches to estimate and minimize variation due to observer error in identifying indirect signs. I reanalyzed data from 4 replicates of a presence/absence survey of northern river otters (Lontra canadensis) that had been conducted by Texas Parks and Wildlife Department (1996-2003). Sixteen observers had recorded tracks at sample points under bridges (n = 250) distributed throughout 27 counties in the Piney-Woods ecoregion of east Texas. My objectives were to 1) determine if observers were a source of bias in the survey, 2) estimate the proportion of error associated with track identification skill, and 3) evaluate the use of an international certification procedure that measured observer tracking skill. The null hypothesis that observers had no effect on the variation in reported sign was rejected. Indeed, binary logistic regression tests indicated that observers were significantly associated with variation in reported track presence. Observers were not randomly distributed among bridge sites, and therefore were significantly correlated with 4 habitat variables that may have influenced heterogeneity in otter occupancy and probability of detection (watershed, vegetation-type, water-type, bridge-area). On average, experienced observers (n = 7) misidentified 44% of otter tracks, with a range of 0% to 100% correct detection. Also, 13% of the tracks of species determined to be 'otter-like' were misidentified as belonging to an otter. During the certification procedure, ...