Auto-monitoring wildlife recovery
ABSTRACT: Wildlife monitoring during forest restoration addresses such questions as: What species re-colonize or disappear from restored areas? How many individuals are present? What are the population trajectories? In this review, we focus on issues related to automating surveys of forest birds and mammals, particularly bats. For both birds and mammals, the need to automate data collection and analysis is clear, but several constraints must be overcome, before such automation becomes practical, compared with labour-intensive, conventional methods. Currently, wildlife species can be recognized and their abundance estimated by using audio recording and photography. However, species recognition software, using audio data, generally performs poorly, compared with humans, particularly under field conditions, where such systems fail to distinguish multiple overlapping calls and separate them from interfering background noises. Similarly, for images, highly variable lighting and lack of clarity of camera-trap images often confuse auto-recognition software. Nevertheless, automated systems continue to improve, and it is likely that they will achieve parity with humans in the foreseeable future. In the near-term, they will have the ability to save considerable amounts of time, by searching through large numbers of files, to narrow searches for particular species and transmitting such files wirelessly over networks. Furthermore, outside of cellular network coverage, drones can be used to collect image or audio data from wireless devices in the field. Thus, while these techniques are currently far from being highly accurate, inexpensive and practical for broad-scale surveys, it is not difficult to imagine a future when assessments of the wildlife recovery that is expected to occur with forest restoration will become increasingly more automated.