• Asian elephant;
  • automated identity;
  • detection probability;
  • morphological features;
  • population monitoring;
  • variable traits


Endangered, wide-ranging megafauna have many threats to contend with during their struggle for survival in an ever-increasing human dominance of the environment. Reliable monitoring of endangered large mammal populations is therefore a critical conservation requirement. Photographic capture–recapture (CR) techniques have opened up avenues for population monitoring of individually recognizable large mammal species. The efficient application of these techniques, however, can be constrained by challenges in reliably identifying individuals arising from the use of multiple, and potentially variable traits, as well as issues of temporal sampling of populations in the field. We address these key problems by describing an automated process of rapidly identifying individual Asian elephants (Elephas maximus) from photographs, and comparing resultant CR-based population parameter estimates with those obtained using supervised visual identification of individuals. In addition, we assess the temporal effort necessary for robust estimation of demographic parameters in our study population. Morphological traits that maintain constancy over time, including variations in tusk characteristics, and ear fold and lobe shape, proved the most reliable for individual identification and subsequent estimation of population parameters. The use of temporally variable traits contributed to high probabilities of misidentification and biased estimates of population size. We found a minimum of seven sampling occasions necessary for reliable population estimation. Our study contributes to design issues for CR studies by providing insights into optimality of sampling effort such that precision of parameter estimates are not compromised while minimizing survey costs. We demonstrate the importance of accurate individual identification in the context of such studies and recommend the use of fixed morphological traits as the optimal individual identification strategy for species where animals are distinguished on the basis of multiple attributes, including some that may be variable over time.