Unveiling buried aeolian landscapes: reconstructing a late Holocene dune environment using 3D ground-penetrating radar

Across the UK, sandy beaches and dunes protect coastal infrastructure from waves and extreme water levels during large-scale storms, while providing important habitats and recreational opportunities. Understanding their long-term evolution is vital in managing their condition in a changing climate. Recently, ground-penetrating radar (GPR) methods have grown in popularity in geomorphological applications, yielding centimetre-scale resolution images of near-surface stratigraphy and structure, thus allowing landscape evolution to be reconstructed. Additionally, abrupt changes in palaeo-environments can be visualized in three dimensions. Although often complemented by core data, GPR allows interpretations to be extended into areas with minimal ground-truth control. Nonetheless, GPR data interpretation can be non-intuitive and ambiguous, and radargrams may not initially resemble the expected subsurface geometry. Interpretation can be made yet more onerous when handling the large 3D data volumes that are facilitated with modern GPR technology. Here we describe the development of novel semi-automated GPR feature-extraction tools, based on 'edge detection' and 'thresholding' methods, which detect regions of increased GPR reflectivity which can be applied to aid in the reconstruction of a range Quaternary landscapes. Since reflectivity can be related to lithological and/or pore fluid changes, the 3D architecture of the palaeo-landscape can be reconstructed from the features extracted from a geophysical dataset. We present 500 MHz GPR data collected over a buried Holocene coastal dune system in North Wales, UK, now reclaimed for use as an airfield. Core data from the site, reaching a maximum depth 2 m, suggest rapid vertical changes from sand to silty-organic units, and GPR profiles suggest similar lateral complexity. By applying thresholding methods to GPR depth slices, these lateral complexities are effectively and automatically mapped. Furthermore, automatic extraction of the local reflection power yields a strong correlation with the depth variation of organic content, suggesting it is a cause of reflectivity contrast. GPR-interpolated analyses away from core control thus offer a powerful proxy for parameters derived from invasive core logging. The GPR data collected at Llanbedr airfield highlight a complex dune system to a depth of 2.8 m, probably deposited in several phases over similar to 700 years, similar to elsewhere in North Wales. (C) 2021 The Authors. Journal of Quaternary Science Published by John Wiley & Sons Ltd.

Automated summary

This study introduces a novel semi-automated GPR feature extraction tool for reconstructing Quaternary landscapes. Field testing was conducted using 500 MHz GPR data on a buried Holocene coastal dune system in North Wales, demonstrating the method's effectiveness in automatically mapping lateral complexities in a complex geological environment.