Application of a Novel Machine Learning Methodology to the Study of Dipodomys spp. Response to El Niño Southern Oscillation events Throughout the Holocene

Summary

This is an abstract from the "Novel Statistical Techniques in Archaeology II (QUANTARCH II)" session, at the 84th annual meeting of the Society for American Archaeology.

El Niño Southern Oscillation (ENSO) events influence climatic variation on a global scale, considerably impacting modern human and animal populations. There is, however, a dearth of literature regarding the long-term effects of ENSO variation on prehistoric vertebrate populations. Here we examine how kangaroo rat (Dipodomys spp.) species abundance from Abrigo de los Escorpiones, a trans-Holocene rockshelter located on the Pacific coast of Northern Baja California, vary as a function of wet El Niño events. This study uses a Random Forest algorithm machine learning methodology to establish species level identification of Dipodomys spp. remains based on variation in cranial morphometrics. This novel technique to zooarchaeological research significantly improves upon more traditional statistical techniques, such as linear discriminate function analysis and principal components analysis, by more accurately predicting species identifications and by interpolating missing data more effectively. Through the use of open source statistical software, this study demonstrates the utility of machine learning techniques to perform archaeofaunal species identifications, which may serve as an invaluable addition to the zooarchaeologist’s toolkit. Moreover, as future ENSO strength and frequency is projected to vary with changing global climatic regimes, this study has important implications for understanding vertebrate population responses to these changing climatic events.

This Resource is Part of the Following Collections

• Novel Statistical Techniques in Archaeology II (QUANTARCH II)   •
• Society for American Archaeology 84th Annual Meeting, Albuquerque, NM (2019)

Keywords

Record Identifiers

Abstract Id(s): 24696