I’ve been looking at speed distributions derived from GPS tracking data. When I started to see some strange peaks in the distribution, I investigated to see if they were being caused by the limited precision of my GPS coordinates (My east/west readings were on ~1m precision, but north/south coordinates to only ~11m.)
Take a look at this notebook, where I:
- I create a set of random single tracks in a 100 x 100 box (pairs of x, y, coordinates).
- Look at the distribution of their lengths
- Apply a truncation to limit the coordinate precision.
- Look at the new distribution of lengths.
- Attempt to reverse the effect of the limited precision, in order to reproduce the distribution original shape.
In conclusion I found this effect was responsible for the peaks!