The results from Darrowfield

First of all, apologies for the delay to this post. Life has been a bit busy, and the data processing got a bit complicated, but I think I have cracked it now.  Our principal aim was to complete the area using magnetometry.  This we did easily.  We also aimed to try and cover as much as possible using GPR and Earth Resistance.  We managed quite a bit of GPR, but not much resistance survey, partly due to manpower and time, and partly due to the stunningly uninteresting results.  Lastly, we did two resistivity pseudo-sections and two topographic surveys, one using the dGPS and one using a UAV and “structure from motion”.  So be prepared for a long post!


I am going to start by discussing the topography because this has an impact on the results and the interpretation.  We did two types of topographic surveying.  Firstly, I walked back and forth taking readings with the dGPS at six-pace intervals, or closer if there was a change of slope such as a bank.  Secondly, Peter Alley used his DJI Phantom UAV to take a series of aerial photographs which can be used to create a topographic model using a technique called “structure from motion”.

Fig. 1: Peter Alley landing his UAV after taking aerial images of the site.

The two techniques have their strengths and weaknesses.  The model using SfM can be very detailed, far more so than is easily possible with the dGPS, but the model is of the top of the vegetation.  The camera cannot see through the stinging nettles.  Mostly, this does not matter but there are areas where one has to be aware that the nettles are masking the topography.  I, however, could not easily get dGPS readings in the garden or the orchard because the trees blocked the satellite signals.  The UAV could, however, map these areas. Using the two methods side-by-side is thus a useful thing to do.  The other big advantage of the UAV is that the survey took about 20 minutes compared to almost a day using the dGPS!

Fig 2: Darrowfield paddock topographic model from the dGPS survey.

Fig. 3: Topographic map derived from the UAV survey.

Ignoring the areas outside the paddock / gardens in Fig. 3, we can see that the two surveys show pretty much the same thing.  The highest ground is the bank along the SW edge of the paddock with a steep slope along one edge.  The bank is less pronounced as it curves around through the garden and the orchard in the NW corner of the area.  I had initially thought this was a road line, but it doesn’t line up with the roads seen in the wider magnetometer surveys of the park.  It does seem to be mapped as a long thin enclosure on early OS maps. There is a second, much less pronounced bank along the NE side of the paddock, which fades out towards the east corner.  This bank is assumed to be the line of the Roman wall. Between the two higher areas is a lower lying area in the middle of the field.  In the east corner, the UAV seems to show a channel running west-east into the corner (Fig. 3, the dark blue line running through a lighter blue area).  This is a footpath running through an area of high stinging nettles.  On the dGPS survey (Fig. 2) we can see a low channel running north-south in the same area — shown in red — which is masked by the nettles.

The magnetometer survey

The main area of the paddock was surveyed on the normal grid we have been using at Verulamium.  The two small extra areas, on the lower lawn of Darrowfield House and the orchard, were surveyed using small floating grids set-up to fit the awkward areas better.  In the paddock, nettles, rabbit holes and badger burrows all made surveying a bit awkward.  Very many thanks to Pat Johnson of Foerster for bringing the repaired magnetometer down to site from Tamworth.

Fig. 4: Ruth Halliwell and Jim West surveying the Lower Lawn of Darrowfield House.

Given where we are in the Roman town, i.e., just behind the basilica in the centre of the the city, the results were surprisingly unexciting!

Fig 5: the magnetometry results.

The majority of the area seems to be dominated by a faint striping running SW-NE across the middle of the field.  I am guessing these are cultivation marks, and this interpretation is supported, as we will see below, by the GPR results.

Fig. 6: annotated magnetometry results.

In Fig. 6, at point A, we can see a curving “linear magnetic anomaly”.  This matches the curving bank seen in both topo surveys (Figs. 2 and 3).  The clearest indications of buildings are along the SW edge of the paddock, at B.  The light lines mark walls, and the darker areas magnetically enhanced materials between the walls. It is difficult to know if these are Roman or buildings associated with the farm.  If, however, we project the line of the roads from the wider survey (Fig. 7), it seems likely these are Roman buildings lining the SE–NW running road.

Fig. 7: rough projection of the Roman roads into Darrowfield Paddock.

At Fig. 6, C there appears to be a linear feature running across the lawn.  At the west end is a black dot with a bright white halo.  This is something ferrous, it has a range of -104nT to 2371nT!  The linear feature next to it only has readings from -5nT to +15nT and is much more likely to be something like a ditch.  Just inside this linear feature are three black “blobs” which have similar ranges and may be pits.

The steep bank shown on the topographic maps which runs parallel to the SW hedge has a linear low magnetic feature (Fig. 6, D).  This may just an artefact of the bank, but I am beginning to wonder if there is something holding the bank up, especially in the light of the GPR results below.

Lastly, At Fig. 6, E there are two faint linear low magnetic features.  Is this a hint of the wall line?

The GPR survey

Between the animal burrows and the nettles, the GPR survey could be quite a struggle…

Fig. 8: Jim versus the nettles at Darrowfield (image: Mike Smith).

We did, however, get quite a bit of the paddock done.  As usual, I have used Larry Conyers’ GPR Process program and Surfer v.8 to produce the time slices.  I have created five slices 3ns in thickness.  I’ll discuss them in order from the top down.

Fig. 9: time slice 3, 10.5–13.5ns.

The topmost time slice shows the bank parallel to the SW hedge very clearly indeed.  The clarity of this feature makes me think that this bank must be artificially reinforced to create the flat space between it and the hedge. Perhaps this was part of the entrance to the house, or something to make a view from it?  The other feature is the very clear line to the east which matches where the bank curves in the topo data.  I’m unsure what would make a line like this.  It is very close to the surface being in the top slice.

Fig. 10: time slice 4, 13.5–16.5ns.

In the next time slice (Fig. 10), the bank to the SW has turned into two linear features.  The bank to the NW shows as less clear linear features as well.  There are a series of lines running SE-NE across the lower part of the field.  These become clearer in the lower slices.  Finally, in the middle of the NE edge, cut into by the “dog tooth” pattern of the survey edges, is a broader linear feature running NW-SE.  This may be a hint of the town wall.

Fig. 11: time slice 5, 16.5–19.5ns.

In the next slice (Fig. 11), the parallel lines in the lower area are now very clear.  These seem most likely to be some form system of land drains.  To the south are some other linear features which may be the tops of surviving walls.

Fig. 12: time slice 6, 19.5–22.5ns.

In the next slice (Fig. 12) we are getting below the drains, but the signal is starting to attenuate.  There are, however, some linear features to the south and the west which are parts of buildings lying quite low.  A number of the “wiggly” features, particularly to the north, are badger setts.

Fig. 13: time slice 7, 22.5–25.5ns.

The last time slice (Fig. 13) finally shows the buildings that can be seen along the SW edge in the mag data.  The GPR cannot really push into the undergrowth as well as the mag and so the evidence is a little more bitty.  We can see, however, the road which runs from the SW to the NE, almost parallel to SE hedgeline, pretty much where I suggested it should be in Figure 7.

It would appear, therefore, that the archaeology is more deeply buried in most of this field than elsewhere at Verulamium.  Perhaps soil eroding down the slope built-up at the back of the Roman wall?  The wall itself, however, is conspicuously absent.  A massive structure like that should show quite clearly in GPR data.  The one thing that I ought to do is build the topography into the GPR processing, but I cannot do that as yet.  Watch this space!

Earth Resistance survey

Earth resistance came a poor third in the techniques we used on the site.  We managed one block 40m by 60m using the Institute of Archaeology’s RM85 at 50cm intervals.  We have also recently had the Welwyn Archaeological Society’s machine repaired (thanks Bob!), and did just one 20x20m square at 50cm.

Fig 14: Earth Resistance surveys.

After the excellent results from Gorhambury last summer and Durobrivae last October, these are more the sort of thing I expect in Hertfordshire, i.e., nothing very useful!  In this case, I suspect that, based on the GPR data, the interesting archaeology is deeper than the method with a 50cm probe spacing is looking.

Resistance Pseudosections

The West Essex Archaeological Group has kindly lent us their Resistance Pseudosection kit.  The idea of this is that one takes a series of readings in a line with a 1m spacing between the four probes.  One then repeats the operation with a 2m spacing, then again with 3m, 4m, 5m and 6m.  By increasing the spacing between the probes each time one repeats the line, one is looking deeper into the surface.  The data can then be plotted using a program called res2Dinv which creates a “pseudosection”, an image of a vertical slice through the ground.  We decided to try two of these: one across the road and one across the bank where the wall may be.  The idea was that maybe being able to look deeper into the ground we might be able to see something the other techniques missed.

Professional versions of this system use a switching box and are very expensive.  Bob Randall, however, created a useful budget version using his TRCIA resistance meter.  With just a couple of people this can be hard work moving all the probes between pegs, but if you have five people it can be quite quick. It does look, however, quite amusing as four people bob up-and-down changing probes in unison!

Fig. 15: the team undertaking a resistance psuedosection.

Figure 16 shows the location of the two lines we did overlain on the GPR data.

Fig. 16: the location of section A (red) and B (blue).

For both of the sections, I also ran the GPR along the same line so that we would have a direct comparison.  Unfortunately, the free version of res2Dinv will not undertake corrections for topography and will not save images for sections 40 probes (39m) long, although I can get a screen grab and use that, which is good enough for this blog.  The commercial version is extremely expensive.  I hope I may be able to process the data with the topographic correction in the future.

The first thing I did was to compare each section with its companion radargram.  The radargrams have been corrected for topography.

Fig. 17: Pseudosection A and a radargram along the same line.

There isn’t a huge correlation between the two for section A (Fig. 17).  The broad curving layer shown low on the left of the radargram does not have a similar signature in the resistivity data unless the large grey-green area of slightly higher resistivity is reflecting that.  The very high areas of resistivity shown to the right of the section in red similarly do not clearly match anything much in the radargram (although see below about the ‘circular thing’).

Fig. 18: Section B and its radargram.

The second section was remarkably featureless.  There is a large area of lower resistivity to the right which may represent the robbed out wall, but this doesn’t really show in the radargram. The large curved response in the radargram on the left doesn’t really show in the resistivity section.  All rather frustrating.  I decided to put the images onto the Google Earth images to see if that helps.

Fig. 19: Resistivity pseudosection overlain on the mag data. Note that the red line marks the position of the pseudosection and north is to the left.

In Fig. 19 the red line indicates the position of the pseudosection.  The area of medium resistivity on the left shown in a greeny colour matches where the section cuts across the linear area of low magnetism shown in light grey / white.  These are probably, therefore, caused by the same feature which would appear to be something associated with agricultural activity judging by the wider pattern in the mag data.

Fig. 20: The radargram for the same line, marked in red. North is to the left.

Comparison of the radargram to the mag data doesn’t help much.  The large curving feature to the left of the radargram (N), doesn’t really show in the mag data.  The area of low magnetism mentioned above only really correlates with half of that feature, if at all.

Fig. 21: Radargram overlain on the lower time slice (cf. Fig. 13).

Fig. 21 does show, however, that the radargram goes right over one of three clear sub-circular features which can be seen in the bottom time slice (cf. Fig. 13).  This feature shows very clearly in the radargram.  In Fig. 22 I have annotated it radargrams take some use to ‘reading’.

Fig. 22: Radargram showing the position of the subcircular feature.

I don’t know what these three features are, but there are three of them in a line at right-angles to the suggested line of the road.  Very intriguing!

Fig. 23: Pseudosection B on the mag data. The blue line marks the position of the section.

In Fig. 23 the area of slightly higher resistivity to the west matches very well with the bank of lower magnetic responses that run parallel to the bank from NW to SE.  The low resistivity block on the right (east) doesn’t seem to match anything in the mag data.

Fig. 24: Radargram along section line B. The actual line of the GPR transect is the blue line.

Fig. 24 shows the radargram on the mag data.  The curved feature low in the profile on the western side only partially matches the low magnetism band.  I cannot see anything that clearly matches the low resistivity feature seen in the pseudosection.


As I have often said, using more than one technique can be very powerful as the different methods detect different properties of the below-ground stratigraphy, and therefore different aspects of the archaeology.  The downside is the huge increase in complexity in comparing, contrasting and trying to interpret the various results.  This small survey is an excellent example of this,  There is much more that could be done, especially going through the GPR results in more detail and comparing and contrasting them with the magnetometry data.

For now, we can say the following:

  1. There is no clear evidence for the town wall.  There is a good possibility that it has been robbed in the this area as suggested by the fact it only shows in one time slice, and there is a low resistivity feature in the pseudosection in about the position one would expect the wall.
  2. The clearest evidence for buildings is along the SW hedge line.  These show best in the mag data, and parts show in the GPR but difficulties with nettles and burrows prevented the GPR getting closer into the hedge line.
  3. The lower part of the field in the centre mainly shows evidence for agriculture in both the mag and GPR data, including probable land drains.
  4. The deepest time slice does show some evidence of the road and Roman buildings.  If my estimates of the speed of the radar signal are about right, the tops of these remains lie between 90cm and 125cm below the surface.  The extra depth of soil compared to elsewhere may be due to soil creep down the slope building up behind the bank to the NE.
  5. Part of these building remains include three sub-circular features which have strong radar reflections.

Many thanks to everyone who came out to help with the survey.  Many thanks also to Pete and Flora Letanka for allowing us to survey in their paddock and garden, and to Stuart Gray for putting us in touch.  Also, thanks to WEAG for the use of their pseudosection equipment, and SEAHA for the loan of the GPR.  Lastly, thanks to Pat Johnson for bringing us the mag at the start of the survey.


1 thought on “The results from Darrowfield


    Fascinating. Pity all that effort – and all your work interpreting it – didn’t show anything more conclusive. Pauline


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 )

Google photo

You are commenting using your Google 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