Reverse engineering a 1970s geophysics survey of Lake Farm Roman Fort, Wimborne, Dorset

During the 1970s and 1980s, the Ancient Monuments Laboratory undertook geophysics surveys with pioneering fluxgate magnetometer equipment. This was before there was access to personal computers, and results were plotted directly onto line trace plots. For the first time, buried archaeological remains could be surveyed in a non-invasive way, using a technique that was to revolutionise archaeological survey. However, the early trace plots were in some cases difficult to interpret, with broad patterns being clear but more subtle detail hidden in the mass of lines. Early attempts at interpretation consisted of annotation of the trace plot, as seen below.

Early magnetometer survey trace plot, copyright Historic England

A couple of weeks ago, a number of these trace plots were shown by Paul Cheetham of Bournemouth University during an excellent talk on Lake Farm Roman Fort, located near Wimborne, Dorset. This site has been subject to a recent magnetometer survey, and Paul showed a side by side view of an annotated 1970s plot, and the modern survey results. He commented how the 1970s interpretation was surprisingly accurate given the limitations of presentation of the results. It was however clear that the modern computer generated greyscale plot contained a huge amount of extra detail and subtle features. To me, this prompted a number of questions: how did the equipment used in the 1970s compare to the modern Bartington gradiometers? What would the results have looked like if there had been access to modern computers? What if we could reverse engineer the original readings from the trace plots, display them in greyscale, and for the first time directly compare the results?

I had tried the technique of reverse engineering trace plot results previously, for a long barrow survey near Old Sarum, with excellent results. For that site, however, there had been no modern survey, and so no comparison with modern techniques was possible.

The first step was to choose a suitable plot for Lake Farm. All the reports for the 1970s surveys are available on the Historic England website:

From those surveys, I chose a relatively small block of squares covering part of the southern section of the fort. I chose this plot because of the quality of the scan, with the lines mainly showing as clear black lines.

Plot chosen for reverse engineering (copyright Historic England)

This area is shown by the eastern block of survey squares on the overview plan of the fort:

Plan of survey squares (copyright Historic England)

A process was then followed to reverse engineer the original readings. This involved vectorising the lines from the image into line geometries, represented by thousands of XY coordinates. These line geometries needed some manual editing to connect broken lines, and to join lines where trace lines crossed. This process was particularly challenging where there was significant magnetic disturbance, as shown in the centre of the plot below. This plot shows the vectorised lines after manual editing, each plot line being represented with a random colour:

By using GIS software, it was then possible to treat each plot line as its own line graph, and read off XY values at intersection points with a set of vertical lines, the readings varying as the line meandered up and down (see below).

Intersect points between vertical lines and trace lines

The result was a set of readings for a set of row and column positions, matching the format of data produced by modern gradiometer surveys. As a result, this data could be loaded into Snuffler, a free software package for analysing geophysics data. This data could be displayed in a greyscale image and hence directly compared with the modern survey results. The result is shown below:

Greyscale plot of reverse engineered readings from 1970s trace plot

The quality of the greyscale plot is astonishing, and shows an incredible amount of detail that was held in the original line plot. Comparison can be made with the modern survey. The image below shows the reversed engineered plot alongside the original line plots (with and without annotation), and the same approximate area from the recent Bournemouth University survey:

The results are amazing. The vast majority of the features in the modern survey are visible on the greyscale plot of the reverse engineered 1970s data, including details of the internal rectangular linear features and significant anomalies. This underlines just how sensitive the 1970s equipment was. In addition, the 1970s survey fills in some narrow gaps that could not be surveyed due to the presence of modern fence lines.

1970s survey area chosen for reverse engineering shown as red box on Bournemouth University survey (copyright Bournemouth University)

It is clear that the 1970s and 1980s line plots encode an incredible amount of detail, this detail not being visible or capable of precise interpretation by the human eye, but recoverable through computer processing. This technique could be particularly valuable where previously surveyed sites are no longer accessible for survey, perhaps through destruction, contamination, or landowner permission.

This entry was posted in Technical and tagged , , , . Bookmark the permalink.

3 Responses to Reverse engineering a 1970s geophysics survey of Lake Farm Roman Fort, Wimborne, Dorset

  1. ajmc3 says:

    Fantastic work, Mike. Looks painstaking but it should open up a wealth of new opportunity.


  2. Kayt says:

    This is brilliant Mike. How did you assign values to the z axis of your line plots? Can they at all be approximated back to the in the field nT values, or are they an abstraction?

    • Mike Gill says:

      Hi Kayt, glad you liked the post, I have to admit I was very pleased with the result. The page has had so many hits today, more than we get in months for the whole site 🙂 . There is more detail in another blog post linked from the article – where I detail the steps at the end, including technologies. I used a tool called FME, which you can get a free license for if using just for personal use, and which is a powerful geoprocessing tool. You could probably do the same thing using some code in QGIS. I basically created lots of vertical lines, intersected with the trace lines, then read a spatial y-value from the intersection location. I worked out a standard separation between each trace line, then used simple maths to offset each value based on the index value of the line in the trace plot. So line 0 had no offset, line 10 had an offset of -10 * the standard separation etc. This allowed me to treat each trace line as a separate line graph. I think the values are an abstraction, but correct in their relative values. If the paremeters of the trace plot were known, you would be able to convert back to the absolute nT values.. Happy to provide any more info.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s