The Herpetological Journal

https://doi.org/10.33256/31.2.99110

pp. 99-110

Authors: Manon Dalibard, Laëtitia Buisson, Olivier Calvez, Morgan Nguyen-Hong, Audrey Trochet & Pascal Laffaille

Abstract: Individual recognition of animal species is a prerequisite for capture-mark-recapture (CMR) studies. For amphibians, photoidentification of body pattern is a non-invasive and less expensive alternative than classical marking methods (e.g. passive integrated transponder). However, photo-identification is effective only if the patterns are (i) sufficiently variable between individuals, and (ii) stable over time. This method also depends on the observer’s judgment. In the present study, we assessed the effectiveness of an automatic algorithm (AmphIdent) to recognise ventral colour patterns of the Pyrenean brook newt (Calotrion asper), endemic to the Pyrenees Mountains of France. To assess the performance of the tested method, 113 individuals from two different streams were marked with passive integrated transponders (PIT-tags). We used false rejection rate (FRR), false acceptance rate (FAR) and true acceptance rate (TAR) as metrics to evaluate performances of photoidentification. Mean FRR was 7.3 %, FAR was 5.2 %, and TAR was 92 % across both streams, both sexes and all the observers. FAR was significantly different between sexes, while FRR and TAR were significantly influenced by the interaction between the sex and the stream. Despite these differences, our error rates are among the lowest values found in the literature for both amphibian and non-amphibian computer-assisted photo-identification. We found that poor-quality reference pictures could lead to an increasing difficulty to achieve a correct match when time since first capture rose. Consequently, individual photo-identification using AmphIdent software is a reliable tool to aid in the monitoring the Pyrenean brook newts, provided that pictures are taken with care, reference images are regularly updated and observers are properly trained to use the software and interpret images.

Keywords: AmphIdent, natural marking, computer recognition, amphibian monitoring, pattern changes