Is GPR Interpretation Subjective?

I was in a meeting the other week and while discussing GPR one of the team said "Well, GPR interpretation is subjective." I had just started to disaggre when the conversation was dragged back on to the topic that we should have been on. It did, however, get me thinking.

I know that a lot of people agree with this and I even used to work with someone who liked to add in their reports "The analysis of GPR data is subjective therefore the results may not be correct." To them it was a given that two people viewing the same data could, or would, come out with a different answer. I can understand this view point, that two people with different backgrounds may look at results and draw different conclusions. However, were you to break out or dig where that data was collected there would only be one right answer.

Ground radar is an objective recorder. GPR will recorded all signals that come into the receiver and put them into the A-scans and B-scans. While there is noise in the data there is no biases, everything it records is from something. Granted it is not always from something we want to look at, a rock in the ground when looking for utilities or a stray EM signal from nowhere. If a survey was undertaken at the same place, in the same conditions, and with the same settings the data collected would always be the same, regardless of the surveyor.

This data is a reflection of the true nature of the subsurface. Therefore, could it not be argued that if the data is objective our interpretation of the data should be as well. Where there are differing opinions of what the data shows rather than saying it is subjective and passed off as "a draw back with GPR" we should be viewing it as differing levels of exactitude.

In the end there can only ever one correct interpretation and we are doing ourselves a disservice for not striving for it. Differing interpretations should be analysed and resolved not just written off and accepted.

Interpretation

Most people in the GPR community on LinkedIn have come across posts like the one below. Where one of our associates has come across some interesting or tricky data and asked for help on what’s going on, or just because they liked the radargram. Typically theses are a single scan, maybe two, and typically only has a basic description of where the scan is from.

When these pop up on our feeds they tend to elect a verity of interpretations from the data displayed. There is no doubt this ends up being subjective interpretation but not representative because it is based very limited information. No one would offer an explanation to the client based on one scan though.

What is interesting to see, in the discussions that follows these post, is that most people’s interpretation is biased towards what they are most used to looking at, or even where they last saw similar data. It is simple to do, when we don't have enough information we start drawing of our own experience. However, as we look deeper and get more information the options on what the data are showing reduce. In the end with all the facts there can only be one or two options that fit the data. Then its up to the surveyor to pick the most likely, while informing the client about the other possible options.

Subjective Interpretation

When I first started working with GPR we undertook a tree root mapping survey. Simple site survey on a half meter grid around the base of a couple of trees. Three of us undertook the analysis the data, our CAD and data manager, the surveyor, and me, the assistant. Low and behold we all produced totally different maps of roots, from the same set of data. At the time none of us had much knowledge of tree root structures and likely we were all totally wrong. This supports the assertion that GPR is subjective. What if, though, the data was processed by an expert in tree roots, who undertook this type of survey every day. His results would undouble be different and likely nearer the truth. With this type of survey the getting the interpretation of the data 100% right would be near impossible given the unpredictable nature roots. However, that does not mean the answer is not out there.

Objective interpration

Now let’s take another example; a GPR scan of a the third floor of a concrete office block. The floor is made up using clay pots supported on ribs running between transverse beams, in turn supported by a main spine beam and edge beams.

The data collected from here has a lot of information contained with numerus targets and a lot of noise. The data can still be interpreted subjectively, instead of clay pots you choose tendons or services, maybe a concrete precast unit. I suspect a utility surveyor would struggle to get all the information out of these scans if they are not familiar with different types of construction methods.

However, a trained and experienced surveyor, who is used to looking at data of this type of construction, will be able to look at one scan and get a good idea of the construction. Using many scans and site information, they would be able to determine all the key construction details (except bar size). This would then not be a subjective look but an objective look, based on experience.

In this example there is many ways the data could be read but only one would be correct. The data collected here will always be essentially the same and if you broke out the floor there would be a particular construction method.

If the surveyor presented results to the client, and said they were purely subjective, the client would not be pleased. Here we would have the problem of the client assuming the surveyor competent and that GPR is at fault. they would likely not use GPR again.

The question then; is GPR data interpretation subjective or is there right and wrong interpretation?

Should we then not be taking an easy route and saying "oh it’s all subjective; we can never really know." Rather we should be saying "there a lot of things this could be, I will use my knowledge and experience, and that of my colleagues and the industry at large, to determine the right answer or the most likely answer."

Yes there are always going to be questions and we cannot get it right all the time. We will always come across data that just blows our minds. However, we need to move away from saying subjective and towards a fuller understanding of what we see.