Story: (Too) scientific approach to GPS accuracy

When I bought a new GPS device, Garmin Gpsmap 64s, I also started to think how I can compare it to my Garmin Forerunner 225. And because I am a scientist, I ended up to think that what I can really measure.

Determining the accuracy is quite simple – in principle. If you have a device you like to analyze, you must compare its results to the those given by a device that is more accurate. And calculate how much the results will deviate.

But the problem is that there is not always that more precise device available. For example, I have these two GPS devices, but before checking their accuracy I can not say that one is more precise than the other. The price or the manual or the number of the digits in the screen are not reliable criteria.

So, I decided NOT to measure the actual accuracy but the repeatability of the measurement. That is a different measure, but can still give a hint how accurate the device is.

First, I have to assume that there are no systematic errors in the operation. After that, the deviations in the results between different measurements should contain only some random variations. And if I measure these variations, I can determine how well the results can repeated. Which is the measure of the repeatability.

Before making the actual comparison, I liked to the this measurement with my Forerunner device that I have had now over a year. Thus, I have already quite a lot of data available.

So, I need some data that is measured in an identical way multiple times. Fortunately I have such: I take very often the same walk route with our dog in the evenings. And I have recorded that track almost every time with my Forerunner.

I downloaded ten examples of the tracks from where I synchronize the data from my Forerunner. Then, I uploaded that data to and plotted those tracks into the map.

Of course, this is not a precise approach. For that I should use the GPS data directly and calculate the mean track and the deviation. But I am not SO much interested in the results. So, I just use visual observation to obtain a rough estimation of the deviation.

The first question is that if there is any systematic error. This is why I use those maps. I just make a quick visual check that the tracks follow at least approximately the road that I was walking along. No problems there.

Then, I know that the road is about 4 meters wide; I walk it in both directions. So, if the GPS would be very accurate, I should just see two sets of tracks – one in both directions on the different sides of the road.

Above is one part of the tracks – without the map plotted. So, in this part the accuracy is not that good but there are random variations. Based on the scale it seems that variations are about +/- 5m. This is just a rough estimation now for the repeatability of the measurement in the case of Garmin Forerunner 225.

The repeatability seems to vary along the track, which is understandable. The terrain varies and the strength of the signals from the satellites therefore is not stable. But here the terrain is not too difficult; I can expect more deviation when going to, e.g., dense forests. In this kind of simple terrain the repeatability is really fine. I my opinion the accuracy is good enough if I see the actual place from the position given by GPS. And 5 meters is really close enough always.

So, now I have a method to compare those two devices: I will check the repeatability of both devices with these walking tracks when I have enough data. Then, I will also test the repeatability of single mark points for Gpsmap. Those tests are now in progress. Interested to see the final results.

As a conclusion: Before making any measurements one needs to think what is really measured and what the results are actually representing.