2017 ESA Annual Meeting (August 6 -- 11)

COS 69-8 - Camera-fishing: A novel method to estimate density from remote underwater video

Tuesday, August 8, 2017: 4:00 PM
B114, Oregon Convention Center
Juan Vargas1,2, Rowshyra A. Castañeda1,3, Nicholas E. Mandrak1,3 and Péter K. Molnár1,3, (1)Department of Ecology and Evolutionary Biology, University of Toronto, Toronto, ON, Canada, (2)Biological Sciences, University of Toronto Scarborough, Toronto, ON, Canada, (3)Department of Biological Sciences, University of Toronto Scarborough, Toronto, ON, Canada
Background/Question/Methods

Remote detection methods like camera traps and acoustic detectors are one of the best ways to obtain important demographic data about species that are difficult to sample otherwise. Mark-recapture methods can only be applied in camera-trap studies to a few species for which individuals can be identified from spot or line patterns. Rowcliffe et al. (2008, J. Appl. Ecol. 45) proposed a different approach to estimate population density based on a random encounter model, where the encounter rate depends on population density and movement speed. The derivation of the model therefore provides a way to estimate the unknown density from the detection frequency of a species (i.e. photographs per unit time). The same principles could potentially be applied to estimate population densities in aquatic systems. An important factor to consider, however, is that movement in aquatic systems occurs in three dimensions, which could break down the assumptions of the random encounter model as it has been used in terrestrial studies. We propose here a new method of analysis for remote underwater video to estimate population density of species moving in a three-dimensional environment.

Results/Conclusions

We used an encounter rate model for movement in three-dimensional space to characterize the relationship between the frequency with which individuals encounter cameras and population density (individuals per unit volume). This relationship is determined by animal movement speed and movement patterns, as well as the cameras’ detection angle and range. While the former should stay constant for a given camera model, the latter varies with environmental conditions, mostly turbidity. Two variants of the model are developed – one assuming completely random movement of all individuals, in analogy with Rowcliffe et al.’s terrestrial model, and one that accounts for differing probabilities for different directions in space (e.g. animals are more likely to swim horizontally than vertically). Our methods were developed for estimating fish densities from underwater video, but could also be applied in terrestrial studies of arboreal species, or with other technologies such as ultrasonic detectors.