Tag Archives: TerraSurveyor

Flying the Flag

Anyone new to this blog or geophysics in archaeology is recommended to read the material on the “Geophysical survey in archaeology” page.

Before I start this post, I should apologise to those waiting for other surveys…  There has been some fiddly data processing (including software training over Skype from Hawaii!), as well as problems with coordinate transforms and other boring technical stuff.  I promise I’m working on them!

The group were asked to provide a day’s training on geophysical survey at Flag Fen for the volunteers from Vivacity who help run the site and others in the Peterborough area.  Many thanks to Peter Alley for setting this up.  We planned to survey two areas: a small strip in the public area of the site across the known line of the Roman road, and a second larger area to the west through which is known to pass the Bronze Age wooden causeway.  We managed to survey the small strip using all three techniques: mag, Earth resistance and GPR, but the howling gale resulted in only the mag being able to survey a little in the second area.

Flag Fen is rightly famous for its prehistoric, mainly Bronze Age, archaeology (Fig. 1) which includes a great deal of preserved timber including a causeway (Fig 2).  It was extremely unlikely we would detect any of those features because wet wood in wet mud has little contrast in any of the three techniques we use (Fig 3).  There is, however, a Roman road known to run across the site as well as a medieval road with a toll building, and those we hope we would find. (Note that the site photos were taken in 2018 on a Welwyn Archaeological Society coach outing to Flag Fen.  The weather was not so kind last weekend. Also note that if you click on the image you can see them in higher resolution.)

Figure 1: Reconstructed round house at Flag Fen.

Figure 2: preserved timbers in situ at Flag Fen.

Figure 3: Flag Fen.

Area A was a small strip just 14m wide and 57m long.  It does, however, cut across the known line of the Roman road.  We managed to complete all three techniques in this area.

For the Earth Resistance survey we were using the “1 plus 2” method.  When we do this, the machine takes three readings when we stick the mobile probes into the ground (Fig. 4).  The first reading uses the outer two probes which are 1m apart.  This means we are looking roughly 1m below the ground surface.  It then takes two further readings, the second using the left-hand probe and the centre probe, and the third using the centre probe and the right-hand one.  These pairs are 0.5m apart and we are therefore looking about 0.5m below the ground surface.  The deeper survey is at a lower resolution of two readings per meter square than the shallower survey which is at four readings per meter square.

Figure 4: the Earth Resistance meter in action.

The results were very encouraging.  Figure 5 shows the 0.5m mobile probe spacing survey.

Figure 5: the Earth Resistance survey (0.5m probe spacing).

In Figure 5 high resistance readings are dark, and low resistance readings are light.  High represents solid things like paths and roads.  In this image we can see one large high resistance feature in the middle, a thin linear one to the south, and a wider north-south high resistance feature in the northern third of the survey area.  I have labelled these in Figure 6.

Figure 6: labelled version of Figure 5.

The big high resistance feature matches the known location of the Roman road.  The thin linear feature matches the current path (Figure 7).

Figure 7: the path at Flag Fen shown clearly by the moss. Photo ©Peter Alley.

This leaves us with Mystery Features A and B, and the hole in the road.  The latter could be (a) robbing of the road for building materials (b) an old archaeological excavation or (c) levelling of a spot for the toll house.  My money is on (a) or (b)!  Mystery feature B can be solved by looking at the historical imagery available on Google Earth.  I have put the geophysics images over the photos from 11th September 2006, which I have reproduced below (Figure 8).

Figure 8: imagery from September 2006.

One nice thing about this image is that it shows an excavation in progress.  It also solves Mystery feature B: there is clearly a path or walkway running across the site at that spot (remember that the Google Earth images are not very accurately georeferenced).  There is also a curious bright square in the photo which I am guessing is a roof to protect where the section of the Roman Road could be seen in the edge of the dyke.

Mystery feature A, however, remains just that,  mystery.  Maybe it is one wall of the toll building, but it is very wide and so would just be a spread of rubble.  There is clearly a high resistance feature here, but exactly what it is I do not know.

Figure 9 shows the deeper Earth Resistance survey.

Figure 9: the Earth Resistance survey (1m mobile probe spacing).

This survey does not show much new but three things should be noted.  Firstly, the hole in the Roman road has gone and is, therefore, not right through all the layers of the road.  Secondly, mystery feature A is even clearer suggesting that this feature is more substantial at depth and is unlikely to be merely ground compaction.  Lastly, the modern path has vanished.  This is because the wider probe spacing means that we are now looking below the level of the superficial path.

The Ground Penetrating Radar survey was very popular (Fig. 10), partially because one can see things one screen!  (I should note that everyone did try the other techniques too!)

Figure 10: the GPR survey in action.

With GPR we collect data in vertical “radargrams” which are difficult to read until one has had some practice (I’m still learning!).  One then takes those vertical slices and stacks them together in the software, and then slice them horizontally to give map-like images at different depths called time-slices (or amplitude maps if one wants to be posh).  In Figure 11 I have plotted the top 8 slices with the shallowest slice in the top-left corner and the deepest in the bottom right.  In these slices blue represents weak reflections, i.e., little or no radar waves are bouncing back to the antenna. Red represents strong reflections, i.e., a great deal of the signal is bouncing back.

 

Figure 11: time slices from the GPR survey.

In the first slice, the modern path shows very clearly right on the surface.  Note that the tracks from their little cart also show along the western edge.  This is simply from soil compaction. By the third slice, the Roman Road,  and mystery features A and B are starting to show nicely.  By slice 6, pretty all that one can see is the Roman Road.  Note that one should take the depths with a pinch of salt.  Firstly, I should have built-in to the processing the topography, which I haven’t.  Secondly, the speed of the signal has not been calculated so this is just a rough guess. Figure 12 shows the fifth slice in place so that you can see how it matches up to the res and the site.

Figure 12: GPR slice 5.

As I mentioned before, understanding the radargrams (the original vertical slices) can be difficult. This is because what we are dealing with are reflections.  High amplitude reflections can show down the profile like echoes in a empty room. Also, a single point, like a stone, will show as a hyperbola (a curve with the middle at the top).  This is because when you are off to one side, some of the radar signal will bounce off that point but the distance is the diagonal.  As you move closer, than diagonal gets shorter until you are over the top and the reflection is directly below you.  As you move past, the distance increases once more.  The software I use allows one to create 3D images with both the time slices and the radargrams combined which I find helpful in understanding the latter.  Figures 13 to 15 show three examples.

Figure 13: 3D representation of the GPR data.

Figure 14: 3D representation of the GPR data.

Figure 15: 3D representation of the GPR data.

In Figure 13 I have picked a time-slice which shows the Roman Road well. Notice the strong reflections in the radargram (showing as dark black and bright white) matching up with the red in the time slice showing the strong reflections.

In Figure 14 I have used a radargram that cuts across mystery feature A.  Notice the very jagged and noise area of the radargram over the line of the mystery feature.  We are not getting very strong reflections like the road, but we have also gone along the feature rather than across it.  When surveying something linear, it pays to try and cut across it rather than go along it.  Our transects are 50cm apart, but we take a set of readings every three centimeters.  That is why we always survey Christian cemeteries north-south not east west.

In Figure 15 I have two time slices.  the top one shows the modern path clearly.  Note how the strong reflections start from the very top. The bottom slice show the reflections from the Roman Road, just clipping the edge of the hole.

Hopefully, these images will help to explain how to work with GPR data!

Last, but not least, the magnetometry survey.  Magnetometry is the mainstay of geophysical survey because (a) it is quick and (b) is often finds stuff!  It’s main weakness is in areas with lots of ferrous material around, like a public spot with paths, fences, old excavations and a yurt…  As a technique, it likes the wide open spaces (Figure 16)!

Figure 16: Mag survey at Flag Fen (Area 2).

Figure 17 shows the results from Area A.

Figure 17: Area A magnetometry survey.

The modern path shows very clearly with a strong magnetic response.  Remember that magnets have a negative and a positive pole, and these are plotted as white and black in the image.  Mid-grey is the background reading.  The path probably used a magnetic rock, like granite or basalt, in the gravel. Mystery feature B also shows very strongly.  I’m not sure what they used for that path.  Along the western edge to the south we have picked-up the modern fence line. Mystery feature A does not show in the data.  There are various other bits of old iron about. In the northern half of the site are some features which are less strong than the others, and have a very weak negative pole (whitish) and the stronger positive pole.  Some of these may be pits full of organics or burnt materials.

Area B was much bigger.  I had, ambitiously, laid out four 40x40m blocks but the howling gale (I don’t exaggerate), we could only manage two of those.  (Tapes stretched out across 40m don’t stay straight for long in a gale!)

Figure 18: the Area 2 mag survey.

The first thing to note is the stripyness running diagonally across the grids.  This is the remnants of ploughing showing in the data.  This is very common in mag surveys.  I have labelled some of the other features in Figure 19.

Figure 19: the Area 2 mag surveyed labelled.

The black blobs like the one labelled A are not that likely to be exciting.  They are probably spreads of organics on the surface, or possibly burnt areas from getting rid of the stubble.  There is a moderate amount of old iron in the field, as is always the case in agricultural fields.  I have labelled just one as B.  There are a few features which might be something more interesting.  For example, at C, is a strong magnetic feature where the negative pole is very weak compared to the positive.  That might be a pit with organics or burnt material in it.  Perhaps the most exciting thing is the hardest one to see.  At D there is a faint curving line.  Almost a complete semicircle can be seen in the data which is about 7.5m in diameter.  Have we found a round house?  It is definitely a possibility!

Hopefully, everyone who attended the training day enjoyed it and got something from it.  Many thanks to Peter Alley, Jim West, Mike Smith, Pauline Hey and Nigel Harper-Scott for helping with the day, and to Gill Benedikz and the rest of the staff at Flag Fen for making this possible.

 

 

Advertisements

And now for somewhere completely different

Although this isn’t CAGG related, or Hertfordshire, I thought members of the group might be interested in my latest geophysical adventures.

Some 15 years ago I undertook a survey in Alba Iulia, Romania, for a colleague.   The site was part of one of the Roman cities at Apulum which grew-up alongside the legionary fortress.  The results were pretty good, but I was only taking one reading per square meter.  Since getting the RM85 I have been wanting to return and re-do the survey at higher resolution.  Well, be careful what you wish for!  Last Saturday, I found myself on the way…

Fig. 1: On my way…

Yes, you did read the time correctly.  I flew to Cluj-Napoca via Munich.  Sadly, when I got to Cluj, my luggage was still in Munich.  Thankfully, they delivered it all safe-and-sound the next day but it did mean I lost a half day of survey.

Alba Iulia has changed quite a bit in the fifteen years.  The citadel, especially, has been restored beautifully and now has a series of bronze statues decorating the area.

Fig. 2: scrumping.

Having lost half a day, we got started in the afternoon.  Three whole grids and partial that day, seven whole grids a four partials (including one very silly small one) the next day, and eleven yesterday.

Fig 3: Wyatt helping with the Earth Resistance survey.

Yesterday, going along the first line seemed fine with the wind at my back.  Then I turned into the howling gale and snow…  The effect was like star-trails in a science fiction movie as the snow blew past me horizontally.  Thankfully, the weather got better during the day.

There were some software issues to begin with, but thanks to David Wilborn’s excellent customer support, those were quickly resolved.  The results look pretty good.  In the next image I have applied a high-pass filter to even out the big changes in range that occur in this data set.

Fig. 4: the Earth Resistance survey results at the end of Day 3.

I’ll update you all as I go along.  I have six more days to try and complete the whole survey.  I suspect I’ll be a little tired by the end.

Your foreign correspondent.

Processing mag data (Part 1)

I thought it would be useful to outline how I have processed the magnetometry data from our surveys.

The first stage is to download the data from the Foerster datalogger.  I use their simple and free program IFR Dataload to do this.  Foerster sell a more complex program for processing data from the Ferex, but I find TerraSurveyor easier to use, as well as having the convenience of being able to process data from the resistance meters and other magnetometers.  When I download the files, you can give them a suffix, and they are numbered in sequence.  Usually I use some like “gorday4_” standing for Gorhambury survey, day four. I then end up with a sequence of files: gorday4_1.fdl, gorday4_2.fdl and so on.  FDL is Foerster’s own file format. You can look at each square in Dataload.

Screen grab from Dataload (best seen full size).

Screen grab from Dataload (best seen full size).

Although one can play around with the image at this point, there is no reason to.  If the square is a partial, it is worth making a note of its dimensions.  Each grid square is then exported as a “Text table” which is a fairly simple ASCII file.  I just number each text file sequentially, e.g., gor001.txt, gor002.txt and so on.  If I know that two or more text files are eventually going to be combined into one grid in TerraSurveyor, e.g., when we have surveyed a partial grid square in bits, I use gor003a.txt, gor003b.txt and so on.  You can see the importance of keeping good field notes or processing the data very soon after the day’s survey.

The next stage is to import the text files into TerraSurveyor.  If it is a new site, we need to create it.

Creating a new site in TerraSurveyor.

Creating a new site in TerraSurveyor.

Having created the new site, we have to import the Foerster text files.  First click on the import button.

The download button is TS.

The download button is TS.

This will make the import window pop-up. A Foerster template is not currently part of the default installation, you’ll need to get it from David Wilbourn or myself. Once you have it, just select it from the list.

Selecting the Foester text table import template.

Selecting the Foerster text table import template.

In the next screen, you need to navigate to where you saved the text files exported from dataload and then choose which ones you want to import.  On the first day it will be all of them, but subsequently just the grids since the last time you processed the data.  After that, just keep clicking next and accept all the defaults until the data is imported. The TS grid files will take the name of the text files, so gor001.txt becomes gor001.xgd.

Choosing text files to import.

Choosing text files to import.

The next stage is to assemble the grid into a composite.  Either click on “Assemble grids” to create a new composite, or select the existing one and click on “Open Grid Assembly” (all in the navigation window).

Grid assembly (part 1).

Grid assembly (part 1).

In the Grid Assembly window you can see thumbnails of the individual grids.  These can be drag-and-dropped into the grid in the correct pattern.  The direction of first traverse is always from left to right, so if we are surveying south-to-north on the first line of the grid, north will be to the right.  To make the grid bigger, use the blue arrows on the right.

Grid assembly (part 2).

Grid assembly (part 2).

Having completed adding the grids, click on Save or Save As.

Grid assembly (part 3).

Grid assembly (part 3).

At first, the composite just looks uniform gray.

First look at a new composite.

First look at a new composite.

This is because the extreme values from pieces of iron in the ground are being plotted as black and white.  The majority of the values which are much closer to zero are squeezed into a small number of mid-grays.  To see the pattern in the majority of the data, we need to clip the display.

Clipping the display values.

Clipping the display values.

The clip button on the processes toolbar on the left is indicated above.  You can see from the graph in the pop-up window how the values are compressed.  I have entered values of -9 and 9 for the clipping.  Depending on the site, you may need to clip down to as far as +/- 1.5nT.

In the image above I have cheated a bit as the composite has already been clipped.  You can see the archaeology, but the image is very stripy.  This is caused by the sensors being imperfectly compensated, by walking in zig-zags, and by sensor drift.  It is perfectly normal in mag data and can be removed using the destripe command in TS.  Most processing packages call this zero mean traverse as the process alters each traverse (a line of data within a grid square) so that its mean is zero.  TS offers ZMT, but also zero median traverse which is often more robust.  TS, therefore, bundles all the options in the destripe window.

Zero median traverse.

Zero median traverse.

I have indicated the position of the destripe button in the above image.  Often, just accepting the default values is fine but sometimes large numbers of large values can mess this up, and so I set absolute values for the process.  In this case I have used +/- 10nT.

Destriped and clipped.

Destriped and clipped.

The resulting image is rather flat.  This is because we have performed the processes in the wrong order.  We should only clip the final image, not the data on which other processes are working.  I did this in the wrong order so that you can see what was happening at each stage, and also to demonstrate the Modify command.  TS, very sensibly, does not alter the base values.  This allows us to edit the processes via the Modify command indicated above.

Modifying and re-ordering processes.

Modifying and re-ordering processes.

In this case we just want to move the destriping command down to below the clipping command.  Clipping should always be the last command.

Interpolating data values.

Interpolating data values.

The resulting image is pretty good and for quite a while that is all I would do.  Jarrod, however, persuaded that better images could be obtained by interpolation and smoothing.

Interpolating data values.

Interpolating data values.

The first step is to interpolate some new values.  The Foerster collects readings 10cm apart along the traverse, and the traverses are 50cm apart.  This is a rather unbalanced grid.  The traverses are the y-axis in TS and so by selecting the Interpolate button (shown above) and choosing to double the values on the y-axis, we create a grid which is now 25cm by 10cm.

Applying a low pass filter.

Applying a low pass filter.

The next stage is to smooth the data a little. Obviously, we don’t want to smooth the data so much we get rid of the archaeology! We use a low pass filter.  Click on the button indicated above and select low pass in the window.  I use values of x=7 and y=3.  This is because x=7 is 70cm and y=3 is now 75cm, i.e., close to the ideal of a circular “window” for the filter.

The resulting image will look rubbish.  This is because the processes are again in the wrong order.  Using the modify command, the processes should be in the following order (from bottom to top!)

  1. Destripe
  2. Interpolate
  3. Low pass filter
  4. Clip

I have done things a bit backwards so that you can see the effect of each stage.  Normally, I would just do everything in the right order from the start.  If you are adding new squares to an existing composite, the processes will be automatically applied when you save it at the grid assembly stage.  Normally one would only do all this once a survey.

In the next posting I will explain how I deal with partial grid squares.

Apologies to anyone who looked at this soon after I posted it as something went wrong and all the work I did this morning vanished.  I had to re-write half of it again!