Image Processing of Thigh Color Pattern Is an Effective Method for Identifying Individual Cope’s Gray Treefrogs, Hyla chrysoscelis
Photographic mark–recapture is a recently developed method that uses photographs of naturally occurring body patterns to identify individual animals. This method may be especially appropriate for amphibians because they are difficult to externally mark for individual recognition. However, the reliability of the method depends on whether or not the considered pattern differs consistently among individuals over time. We took pictures of the black and yellow thigh pattern of Cope’s Gray Treefrogs (Hyla chrysoscelis) across two years to test whether this trait was consistent enough that individuals could be recognized from photographs, using automated photo recognition in Wild-ID software. We tested the performance of this program using sample sets of photos taken at three different timescales: photos from the same night, different nights within a year, and two different years. We also tested whether the likelihood of correct identifications decreased with increasing sample size. Overall, photographic mark–recapture produced a very high percentage of correct identifications across all timescales and sample sizes. Thus, we conclude that the use of this inexpensive and non-invasive technique on the thigh color pattern of H. chrysoscelis is a highly effective method for individual recognition. El marcaje–recaptura fotográfico es un método de reciente desarrollo que emplea fotografías de patrones corporales distintivos que aparecen de forma natural para identificar a los animales individualmente. Este método puede ser particularmente apropiado en anfibios ya que son difíciles de marcar externamente para su reconocimiento individual. Sin embargo, la fiabilidad del método depende del potencial que tiene el patrón considerado para diferir de forma consistente entre individuos a lo largo del tiempo. Se tomaron fotografías del patrón negro-amarillo del muslo en la rana arbórea Hyla chrysoscelis a lo largo de dos años para testar si este rasgo era suficientemente consistente como para que los individuos fueran reconocidos a partir de fotografías empleando el programa automático de foto reconocimiento Wild-ID. Examinamos el rendimiento del programa usando series muestrales de fotografías tomadas a tres escalas temporales distintas: fotografías tomadas durante la misma noche, entre noches dentro del mismo año, y entre dos años. También se exploró si la probabilidad de identificaciones correctas se redujo al incrementar el tamaño muestral. En general, la captura–recaptura fotográfica produjo un alto porcentaje de identificaciones correctas en todas las escalas temporales y tamaños muestrales. Así, concluimos que el uso de esta técnica barata y no invasiva en el patrón de color del muslo de H. chrysoscelis es un método altamente efectivo para el reconocimiento individual.

The diversity of patterns on the inner thigh region of H. chrysoscelis. (A, B) Photographs of the same individual. Photograph A was taken on 4 June 2019, and Photograph B was taken on 25 June 2019. (C–F) Photographs of four other individuals.

The effect of sample size on success rate. Points show success rate as the percentage of correct matches in the sample. n = 8 replications per percentage. Points are jittered along the x-axis and rendered partially transparent for better visibility given the high overlap across replications.

Match scores across the three different timescales in which photos were compared. Points show individual match scores (values closer to one indicate greater similarity) for each correct match. Red line indicates the median. Points have been jittered and rendered partially transparent to reduce overlap.
Contributor Notes
Associate Editor: D. S. Siegel.