Traditional and Geometric Morphometrics and Invariant Shape Descriptors of Catfish Otoliths in the Yucatán Peninsula: Tools for Species Identification and Individual Size Estimation at Maya Archaeological Sites

ABSTRACT This study investigates the use of otolith shape analysis for species identification and size estimation in Ariopsis felis and Bagre marinus , based on 181 modern otoliths obtained from a scientific collection and recent sampling in the coastal regions of Campeche and Yucatán, as well as 39 archaeological otoliths corresponding to the Early Classic and Late Classic periods of the Maya culture, from three pre‐Hispanic coastal settlements: Xcambó, Vista Alegre, and Isla Cerritos. Using geometric morphometrics, discriminant analysis with cross‐validation yielded classification accuracies of 94% for A. felis and 92% for B. marinus , and in the final classification, we achieved an accuracy classification of 100% for A. felis otoliths and 60% for B. marinus otoliths using K‐NN and Naive Bayes. Classical morphometric descriptors and invariant shape descriptors combined with machine learning yielded classification accuracies of 100% for A. felis and 60% for B. marinus with three algorithms (SVM, ANN, and RF). This allowed us to achieve an overall accuracy of 80%. Our analysis identified key morphological features, such as the lateral edges and rostrum, which distinguish between species. The study also explored the relationship between otolith size and fish size, finding species‐specific growth patterns. While simple models provided accurate estimates for A. felis , more complex machine‐learning models were necessary for B. marinus due to its variable growth. These results highlight the utility of integrating different methodological approaches to improve classification accuracy and efficiency and demonstrate the importance of archaeological otolith analysis in understanding historical fish populations and the effects of human activities on marine ecosystems.