A new framework for estimating abundance of animals using a network of cameras

While many ecology studies require estimations of species abundance, doing so for mobile animals in an accurate, non-invasive manner remains a challenge. One popular stopgap method involves the use of remote video-based surveys using several cameras, but abundance estimates derived from this method are computed with conservative metrics (e.g., maxN computed as the maximum number of individuals seen simultaneously on a single video). We propose a novel methodological framework based on a remote-camera network characterized by known positions and non-overlapping field-of-views. This approach involves a temporal synchronization of videos and a maximal speed estimate for studied species. Such a design allows computing a new abundance metric called Synchronized maxN (SmaxN). We provide a proof-of-concept of this approach with a network of nine remote underwater cameras that recorded fish for three periods of 1 h on a fringing reef in Mayotte (Western Indian Ocean). We found that abundance estimation with SmaxN yielded up to four times higher values than maxN among the six fish species studied. SmaxN performed better with an increasing number of cameras or longer recordings. We also found that using a network of synchronized cameras for a short time period performed better than using a few cameras for a long duration. The SmaxN algorithm can be applied to many video-based approaches. We built an open-sourced R package to encourage its use by ecologists and managers using video-based censuses, as well as to allow for replicability with SmaxN metric.