Integration of multidimensional archaeogeophysical data using supervised and unsupervised classification
When multiple geophysical methods are used to survey an archaeological site, an integrated approach to interpreting the data is often pursued. The use of supervised and unsupervised classification methods are tested using ground‐penetrating radar, magnetometry and magnetic susceptibility data sets f...
Gespeichert in:
Veröffentlicht in: | Near surface geophysics (Online) 2009-06, Vol.7 (3), p.147-158 |
---|---|
1. Verfasser: | |
Format: | Artikel |
Sprache: | eng |
Online-Zugang: | Volltext |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | When multiple geophysical methods are used to survey an archaeological site, an integrated approach to interpreting the data is often pursued. The use of supervised and unsupervised classification methods are tested using ground‐penetrating radar, magnetometry and magnetic susceptibility data sets from a site in the American Southwest. Pueblo Escondido was a large prehistoric village associated with the Mogollon culture in southern New Mexico, with peak occupation during the transition between pithouse and pueblo architectural periods (ca. 1280–1290 AD). Image classification has the benefit of producing unambiguous discrete maps and capitalizes on the multivariate relationships between data sets. Theoretically, unsupervised classification could identify new archaeological classes that were not anticipated but no such classes were identified. The K‐means cluster analysis succeeded only in identifying weak, moderate and strong positive anomalies found in the original data sets. Supervised classification utilizing Mahalanobis distance produced much better results. Training sites based on archaeological excavations were used to classify all locations in the survey area, yielding a predictive model of archaeological features in three classes, plus a background class. The result shows features that were not easily identified in the original data sets but are made visible by the multivariate model. The model could be used for guiding future excavations and arguably leads to a better understanding of the site’s subsurface content and spatial organization. |
---|---|
ISSN: | 1569-4445 1873-0604 |
DOI: | 10.3997/1873-0604.2009004 |