Archaeol Prospection 15 (2)

The efficacy of ground penetrating radar (GPR) methods is inhibited when surveying over a target that is structurally complex and/or hosted within attenuative media. Recent research has documented the ability of certain seismic methods to improve imaging using GPR. For imaging complex targets, three-dimensional acquisition and migration methods are applied. For attenuative sites, signal-to- noise ratio (SNR) may be boosted on acquisition of multi-offset data. The authors present results from an integrated three-dimensional multi-offset survey over a Romano-British villa at Groundwell Ridge, near Swindon. Data were acquired within a grid of dimension 21 m × 14 m, using a single-channel PulseEKKO GPR system equipped with common-offset (CO) 450 MHz antennas. To satisfy criteria for three-dimensional migration, the sample density over the grid was 0.05 × 0.05 m2. A smaller grid of three-dimensional multi-offset data was acquired, with fold-of-cover 2200%, targeting a low SNR section of data. The spatial resolution and SNR in the resulting images of the target are greatly improved compared with data acquired using a more conventional survey method. However, this improvement may not be justified by the greatly increased (some 10 times) fieldwork effort required to obtain three-dimensional multi-offset data. The authors therefore investigate a means of improving the efficiency of three-dimensional GPR surveying by applying a simple trace interpolation method to recover three-dimensional acquisition criteria. This trial suggests that, at this site, three-dimensional data can be simulated from a grid of pseudo-three-dimensional data, sampled at 0.05 × 0.25 m2. In this way, high quality images of an archaeological target can be obtained with minimal increase to survey effort.

It is argued that ground-penetrating radar (GPR) lags behind other archaeogeophysical methods in terms of speed, efficiency and ability to produce clean site-wide composites owing to complex and time-consuming data processing requirements. Two North American case studies illustrate problems that occur when survey is conducted over long and short periods of time. Some GPR defects have been blamed on differential solar heating of antennae and battery power levels but the authors present evidence to indicate that these effects are negligible. Major problems include gradual changes over time in ground moisture and low-level background noise, which can create discontinuities between adjacent survey blocks when data are collected at different times. These problems are often remedied by globally aligning traces using a stable trace position. Variations in ground moisture through time also cause differences in reflection amplitudes, necessitating different range gain curves to match amplitudes between survey blocks. In some cases changes in ground moisture cause noticeable differences in velocity between survey blocks requiring time-scales to be converted to depths to correctly match the data. These problems must be remedied before horizontal slicing can be considered. Subsequent image processing may also be necessary to generate a seamless mosaic and eliminate striping artefacts commonly seen in slice maps. The latter are probably caused by antenna lift and tilt and can be removed by a de-striping algorithm that uses a one-dimensional low-pass filter to characterize stripes followed by their subtraction from the data.