Have a look at the UTF-Grid
As I am playing with the raster-bases semantical generalisation via UTF-Grid…I use the opportunity to make a quick visual comparison.
This is the geometry/object that I am relying on. The german states…a dataset provided by german BKG and adjusted to scales smaller 1:1Mio (detailled information).
Let’s have a look at the corresponding UTF-Grid!
Each pixel of the 256×256 matrix of the tiled json file corresponds with one UTF-character. That is why the basic shape of the german states is, for a scale of 1, still very good recognizable.
More schematised for but still recognizable shape for a scale = 2. That means…one character represents an area of 4px² (2×2 Pixel)… Have a look a the repesentatives of Berlin (‘]’) and Hamburg (‘$’)…they are represented by only on character!
So…now one character represents an area of 16px² (4×4 Pixel) wow…this is heavily abstracted…a basic shape of germany is barely recognizable…but the states!?! I can find Hamburg but not Berlin.
I think this is the scale that is used by the guys of mapbox…I haven’t found detailled infos on that, so far, but in this visualisation…it looks like scale=4 ?!?
At the end…the decision on which scale you should choose…is a weighing up between ‘High performance’ & “Interaction granularity”. Do you want the requested *.json to be as small as possible (–> scale=4), do you need a very correct interaction (–> scale=1) or do you need both (–> scale=2).
I have to do some real tests on these different resoluted UTF-Grids…I will publish that in another post.