Ground-penetrating radar (GPR) is used in this study to delineate the extent and internal structure of a large late Holocene buried shell matrix site at Thundiy, Bentinck Island, northern Australia. Shell matrix sites comprise a key component of the coastal archaeological record. The extensive nature of many shell matrix sites presents challenges for archaeological sampling regimes. While large-scale excavation is undesirable and impractical, limited test pits often represent only a tiny fraction of large shell deposits and are rarely considered representative. This study transforms GPR data into three-dimensional models which form the basis of deposit volume estimates. Volume estimates are evaluated against excavation data to test their accuracy. Results demonstrate that this novel methodology can generate accurate three-dimensional representations of buried shell matrices and highly accurate volume estimations with error margins of 3.5%±7%. It is recommended, though, that more inclusive error margins of 19.5%±17% are used to account for potential error, especially where results cannot be verified. This greater understanding of the extent and structural variability of deposits can be utilised to create robust sampling strategies for excavation. The methodology could also be further employed to enhance comparative regional studies and to add to conservation and management practices of buried shell matrix sites. If applied more widely this methodology will not only benefit our understanding of shell matrix deposits but also the wider archaeological record of coastal regions worldwide. Highlights GPR data were used to create volume and 3D models of buried shell matrix deposits  Volume estimates were evaluated against excavation data  GPR volume estimates found to be accurate within 3.5%±7%

Sampling issues represent a persistent problem in shell matrix research, particularly for large shell deposits. When small samples are taken from large buried deposits it is almost impossible, under current research practices, to understand how representative that sample is of the overall deposit. This case study tests a novel method for creating a better understanding of the buried deposits from which excavated samples are taken, thereby allowing for improved sampling strategies and a better understanding of how representative those samples are of the overall site. The case study employs two geophysical survey methods, ground‐penetrating radar (GPR) and electrical resistivity, to investigate buried shell deposits under experimental conditions. The survey results were used to create volume estimations and three‐dimensional (3D) models of buried shell deposits. This method is novel to shell matrix research and the current case study was designed to test the viability of the method under differing conditions. As well as testing the two geophysical methods, surveys were conducted under different moisture levels, soil types and survey transect spacings. Results showed that the 3D models and volume estimates of the deposit were successful in creating a representative understanding of the nature of the buried deposit, but with varying degrees of accuracy. GPR results created more accurate volume estimates and 3D models than the electrical resistivity results. Both geophysical methods produced more accurate results under drier conditions, though the electrical resistivity produced more visually distinct results with higher moisture levels. Analysis of the volume results revealed an error margin (to a confidence level of 95%) of 9.5% ± 15.5% for the GPR, and 44.5% ± 31.5% to 56 ± 70.5% for the electrical resistivity, depending on the interpretation method used to create the models.

No abstract

No abstract