Tadpole fingerprinting: Using tail venation patterns to photo‐identify tadpole individuals of a threatened frog

Abstract

Traditional methods for identifying individual amphibians in capture-markrecapture (CMR) studies have been primarily confined to post-metamorphic stages, using artificial markers that come with a variety of limitations. An alternative that may open CMR studies to earlier life stages involves the use of a species' natural external markers in photo-based identification. In this study, we investigated whether it was possible to distinguish tadpoles of the threatened green and golden bell frog (Litoria aurea) at the individual level based on tail venation patterns. We collected photographs of the tails of captive-raised tadpoles using a smartphone over a 4-week period. This photo-library was used to create an electronic survey where participants were asked to detect matches for query tadpoles from small image pools. We found that most participants agreed on a match for each query, with perfect consensus achieved for most queries (83%). We detected a 14% decline in perfect consensus when participants were asked to match images of tadpoles separated by longer time intervals, suggesting that it is more difficult to visually identify recapture events of L. aurea tadpoles over extended periods due to changes to tail appearance. However, consensus was obtained by participants for all queries, with all matches verified as being correct by the primary researcher. The strength of agreement among participants with no prior experience in matching tadpole tails suggests that there is sufficient inter-individual variation in this feature for individuals to be manually identified. We thus propose that photo-identification is likely to be a valid, non-invasive technique that can be used for short-term studies on tadpole populations that display tail venation. This offers an alternative to artificial markers that may not allow for individual identification, while also opening up tadpole monitoring programmes to citizen scientists who can be recruited online to process image data from home.

Performance of different automatic photographic identification software for larvae and adults of the European fire salamander

“…Another study compared the tail fin venation of amphibian larvae by untrained citizen scientist by eye. Here, they demonstrated that the consensus of all participants on matches of Litoria aurea tadpoles varied between 67% and 83% [ 45 ] and thus showing a lower Recognition Rate in comparison to the software indicating that the use of a software outcompetes by eye matching of (untrained) citizen scientists. Hence, both software reflect a reliable and beneficial tool in European fire salamander larvae research and conservation.…”

It's not easy being green: Comparing typical skin colouration among amphibians with colour abnormalities associated with chromatophore deficits

“…All case studies used in this study are derived from observations made on Kooragang Island, NSW, Australia (32.85837°S and 151.72480°E). This island is situated at the mouth of the Hunter River and has undergone widespread modification from agricultural and industrial activities, as well as recent remediation involving pond construction, resulting in the presence of semi‐natural and artificial freshwater ponds interspersed by mangrove creeks and saltmarsh (Beranek et al., 2020; Gould et al., 2024).…”

The importance of quantifying fitness-determining traits throughout life to assess the application of reproductive technologies for amphibian species recovery

“…A less invasive approach is to draw on a species' natural external markers for photo-based identification. For example, inter-individual variation in tail venation was recently shown to be effective at identifying green and golden bellfrog tadpoles (Gould et al, 2023) and ventral patterns have been successful in identifying individual brown toadlets (Byrne and Silla, 2023). Such biometric identification can be facilitated by the use of pattern recognition software (Ribeiro and Rebelo, 2011), with recent innovations in artificial intelligence and deep learning providing the opportunity to train software to identify individuals with extreme precision (Takaya et al, 2023).…”