Camera trap constraints in focus
: assessing detectability and identification of small mammals in camera trap studies

  • Larissa Carmel Potter

    Student thesis: Other thesis - CDU

    Abstract

    Camera trapping is a powerful and increasingly popular tool for mammal studies, but like all survey methods, there are limitations. Given the expanding use of camera traps, and the increasing number of management and conservation decisions based on camera trap data,  understanding the constraints of this survey method is crucial. I examined two important limitations for using camera traps to survey small mammals in Australia: detectability and identification from camera trap images. Through employing a web-based survey of 60 camera trap images and 20 questions regarding observer experience, I examined the effects of two species’ attributes (body mass and uniqueness) and two observer variables (experience and confidence) on the accuracy of mammal identifications from camera trap images. I also examined detectability and identification of an elusive small marsupial, Butler’s dunnart (Sminthopsis butleri) on Melville Island (Northern Territory), as a field-based case study of the effectiveness of camera traps for surveying small mammals where related and morphologically-similar species, including the Red-cheeked dunnart (Sminthopsis* virginiae), co-occur. With the survey, I found accuracy and confidence in species identifications from camera traps was lowest (36%) for small or non-distinct species and highest (100%) for large or unique mammals. There was surprisingly little relationship between accuracy of mammal identifications and levels of experience, but identifications made with greater confidence were more likely to be accurate. In the case study of Butler’s dunnart, comparing live-trap to camera trap detections was particularly useful for distinguishing sympatric small mammal species. Nightly detectability of Butler’s dunnart (probability of detection given the species was present) was lower for camera trapping (1%) than pitfall trapping (8%), and a longer survey duration was required with cameras to reach 95% detectability (255 days). However, camera trapping was significantly more cost-effective than live trapping. A seasonal peak in detections was found in the late dry season (early October). The results highlight that the capacity of camera traps to detect small mammals, such as Butler’s dunnart, may be low on a nightly basis, but the cost-effectiveness, ease-of-use and animal ethics benefits of this approach, allow them to be deployed at large spatial and temporal scales. Camera trapping is therefore likely to be an effective survey method for studies reliant on presence/absence data, provided the species is easily distinguishable from camera trap images, as unreliability of identifications is a significant limitation to camera trap studies, particularly where small or morphologically-similar species are the focus.
    Date of AwardJun 2017
    Original languageEnglish
    SupervisorBrett Murphy (Supervisor) & Christopher Joseph Brady (Supervisor)

    Cite this

    '