How many data sets are too many?

A range of different data sets are necessary to answer the research questions we ask in VEMOP Foto: Petra Schneidhofer/VTFK




10.02.2023 kl.10:40

To answer our research questions, we need a lot of different, complementary data sets. With over 100 GPR surveys, preparing all of them for analysis is quite an undertaking.

Time is flying by and we are deep into preparing all the different data sets collected during the last two years. This includes the actual GPR data from each of the surveys we conducted across the four test sites, as well as the monitoring data gathered from the subsurface materials and results from the soil and sediment analyses. A tedious, but necessary step before we can start looking for answers to our research questions. We gathered valuable insights for this stage of VEMOP from working on the Borre Monitoring pilot study (If you want to know more about the Borre Monitoring Project, you’ll find all the details in the last blog). With so many GPR data sets, every step in this process takes a bit longer and we try to account for that with custom-written scripts and by dividing the work load between us.

traktor med georadar
GPR surveys were conducted with two measurement systems at each of the four test sites resulting in a large amount of data sets

One of the first tasks comprises of estimating the velocity with which the electromagnetic signal travelled through the ground individually for each of the surveys in order to make the visibility of the archaeological structures in the data sets more comparable. For this, we are using a method called hyperbola-fitting. How fast or slow the electromagnetic signal travels is dependent on the material properties of the soil and sediments in the ground, so directly connected to our research question. Once the velocity models are established, we can proceed to processing the GPR data sets using the specialised software ApRadar.

velocity model
Velocity model of a GPR data set collected in Heimdal during December 2021

We will also look through all GPR data sets and classify them based on their quality for both the multi-channel MIRA and single-channel GX device at each of the four test sites. Again, with over a hundred individual surveys, this is a rather time-consuming undertaking, which is why we will be bundling our experience in a hands-on session in Oslo in the coming week.

et lag i datasettet
One of the many, many depth-slices derived from the GPR surveys at Hovland

The monitoring data we collected through the monitoring stations included bulk electric conductivity, volumetric water content, ground temperature and precipitation. With measurements every 30 min, we now have over 25.000 individual data points per site, telling us exactly what was going on in the ground before and during each of the surveys.

One of the soil moisture sensors installed at the test site Hovland

And finally, laboratory-based analyses of soil and sediment properties together with the observations made during the set up of the monitoring stations will help in characterise each of the test sites and provide the necessary background information for our investigations.

Tegning av stolpehull på Hovland
Documentation of the posthole sectioned at the excavation in Hovland

One of the big challenges ahead will be the integration of these data sets in order to understand exactly which of the environmental parameters were influencing the quality of the GPR data, to what extent and why. More on that in the nest blog.