A Pulse Ekko PE1000 console was used to collected trial profiles with both 450MHz and 900MHz antennas. From this data the 450MHz antenna was selected as the most suitable centre frequency for obtaining the depth of penetration and lateral resolution required for the survey. A common mid-point (CMP) velocity analysis was subsequently conducted with this antenna and confirmed that the velocity of the radar wavefront in immediate topsoil was ~0.08m/nS. This latter velocity was then used to estimate the depth to reflection events in the recorded profiles. Individual profiles were subject to post-acquisition processing involving the adjustment of time-zero to coincide with the true ground surface, removal of any low frequency transient response (dewow) and the application of a spreading and exponential compensation (SEC) gain function to enhance late arrivals.

A total of 120 parallel 60m EW profiles separated by 0.5m were collected over the site at a sample interval 0.05m (Figure 1). Amplitude time slices were subsequently created from the entire data set by averaging data within successive 4nS (two-travel time) windows (David and Linford 2000, Pulse Ekko 1996). This data is illustrated as a series of false colour images in Plan A and as histogram equalised greyscale images in Plan B.

Conditions at the site were generally good for GPR survey and the short cropped grass provided a suitable surface for the transport of the antenna. Depth penetration was apparently good with significant reflections recorded to a maximum two-way travel time of ~40nS (~1.6m). From the CMP velocity estimate made at the site each successive amplitude time slice represents a vertical window of ~0.16m extending from the modern surface to a maximum depth of ~1.6m. However, the true depth to anomaly source targets may vary due to the inhomogeneous nature of the subsoil.

A graphical summary of significant anomalies discussed in the following text is included on Plan B together with an extract of the gradiometer data from the same area of the site.

The initial two amplitude time slices (0-4 and 4-8 nS) are dominated by near surface reflections due, no doubt, to the air/ground interface and the response from the surface topsoil layer. However, the 4-8 nS time slice contains a number of more significant anomalies that are replicated with greater clarity in the subsequent 8-12 nS window. The location of the temple [1] is evident as two concentric circular responses enclosed within a more diffuse, rectilinear anomaly [2] that may well represent the original Society of Antiquaries excavation trench. Note how the inner circular anomaly of the temple is incomplete within the earlier (near surface) time slices and does not form a closed, circular response until the deeper 24-28 nS (~0.96–1.12m) window.

A NS road way [3] bisects the survey area and would appear to be generated from a 4m wide metalled surface with a ditch or drain type anomaly [4] running parallel to its course. Further evidence for the Roman street plan is found at [5] where part of an EW roadway meets with [3]. An additional road type anomaly is found at [6] but it is unclear how this relates to the grid pattern of major roadways through the Roman town revealed by both AP evidence and the results of the magnetometer survey.

High amplitude rectilinear anomalies related to former building remains are found in the vicinity of the temple at [7] and [8]. Of these, [8] would appear to be closest to the surface producing a strong response in the 12-16 nS time slice (~0.48-0.64m) compared to the later reflections from [7] that do not occur until the 20-24 nS (~0.8-0.96m). Further, less intense building-type anomalies are found at [9] but these appear as shallower responses that do not extend beyond a two-way travel time greater than 20 nS (~0.8m).

Plan B shows an extract of the magnetometer data (Martin 2000) for comparison with the GPR survey. In this case, the presumed stone wall footings that produced high amplitude reflections in the radar data are replicated as negative anomalies in the magnetometer survey due to the >depletion= of local magnetic field strength over the non-magnetic flint foundations. It is of interest to note that the magnetometer has responded to the apparently near surface remains, such as the incomplete inner radius of the temple structure but has failed to detect anomaly [8]. This is curious as the apparently deeper lying anomaly [7] detected by the GPR survey has been recorded in the magnetic data. In addition, the more ephemeral near surface anomalies found at [9] are reproduced with more clarity in the magnetometer data together with a series of discrete, possibly thermoremanent magnetic responses. It seems likely that the concentration of occupation activity, including the deliberate use of fire, may well have increased the local magnetic susceptibility in the vicinity of [9] resulting in a greater contrast between the magnetic properties of the soil and the stone wall footings.

The GPR survey has successfully identified a range of anomalies related to the street plan, former buildings and the Romano-Celtic temple at the site. Whilst this survey provides little additional information to that provided by the AP, excavation and magnetometer data the GPR technique is obviously well suited to conditions at the site. It is of interest to note how the GPR data complements the magnetic survey by revealing responses from more deeply buried features allowing a more detailed interpretation in terms of the apparent vertical stratigraphy. This may prove to be particularly useful in areas of the site where multiple-phase occupation is suspected or the near-surface survival of features is at issue.

David, A. and Linford, N., 2000, Physics and Archaeology, Physics World, 13, No. 5, pp27-31.

Sensors and Software, 1996, pulseEKKO IV RUN User’s Guide, Technical Manual 20.