The Impact of Spatial Sampling and Migration on the Interpretation of Complex Archaeological Ground‐penetrating Radar Data. Issue 2 (28th January 2015)

• Record Type: Journal Article
• Authors: Verdonck, Lieven; Taelman, Devi; Vermeulen, Frank; Docter, Roald
• Is Part Of: Archaeological prospection. Volume 22:Issue 2(2015)
• Page Start: 91
• Page End: 103
• Publication Date: 2015-01-28
• Subjects: Archaeology -- Field work -- Periodicals; Prospecting -- Periodicals; 930.1028
• DOI: 10.1002/arp.1501
• Languages: English
• ISSNs: 1075-2196

Abstract

<abstract abstract-type="main"> <title>Abstract</title> <p>In this paper, the impact of spatial sample density and three‐dimensional migration processing on the interpretation of archaeological ground‐penetrating radar (GPR) data is assessed. First, the question of how to determine the sample interval required to take full advantage of the spatial resolution capabilities of GPR without oversampling is addressed. To this end, we transform a test profile into the frequency–wavenumber (<italic>f–k</italic>) domain and estimate the required sample interval from the wavenumber values. For the presented data set, collected at the Roman town Ammaia (Portugal), this resulted in a transect spacing approximately three times the distance prescribed by the <italic>λ</italic><sub>min</sub>/4 criterion (where <italic>λ</italic><sub>min</sub> is the minimum observed wavelength). Second, the effect of three‐dimensional migration is assessed. The data set, sampled as prescribed by the analysis of the <italic>f–k</italic> plot, is migrated with two‐ and three‐dimensional phase‐shift algorithms, and the migrated results are compared with non‐migrated data. It is shown that certain subtle features are better resolved by three‐dimensional migration. Third, it is investigated whether three‐dimensional migration following the application of an interpolation algorithm such as Delaunay triangulation or interpolation based on τ‐<italic>p</italic> transform, can further relax spatial sampling requirements. For the GPR data shown in this article, it is demonstrated that interpolation and three‐dimensional migration of slightly aliased data, collected with a transect spacing equal to five times the outcome of the <italic>λ</italic><sub>min</sub>/4 criterion, still allow a faithful reconstruction of the original, non‐aliased time‐slices. Copyright © 2015 John Wiley & Sons, Ltd.</p> </abstract>