How can we define a robust metric for "most discontinuous country" and then rank all states according to it?
How can we define a robust metric for "most discontinuous country" and then rank all states according to it?
More of a classification question, but I'm really curious about what the metric would look like if we try to be systematic about it.
For context, there's several countries that are more or less famous for being geographically discontinuous. Top of the mind nowadays is Azerbaijan, whose sizeable territory of Nakhchivan has no land connections with the rest of the country. There's also Equatorial Guinea, whose capital city is on island which is smaller than the continental territory. That's the same for Denmark, although we seem to think of it less, because of the much smaller distances and significantly more connectivity. Then you have Indonesia which I currently think might be the most discontinuous country, with territory spanning across at least 4 major landmasses but which are shared with other countries.
But then you have countries such as Greece, Japan, or even Sweden, which are more or less archipelagic countries but do not stand out in the way Indonesia or Azerbaijan does.
How can we define a measure of geographic discontinuity that gives us a reasonable ranking? I would imagine we start with some measure that looks how much of the whole territory is in one contagious unit (less prominent main landmass = more discontinuity) but perhaps we also introduce average distance between units.
I think it would ultimately depend on a use case for that metric, otherwise you're putting the cart before the horse. There are many measurements and calculations you could come up with, but no obvious (to me, anyway) interpretation of "most discontinuous": something is either in one piece or not. If you needed a metric like this for a practical purpose, your specific needs would be a starting point for designing one. If it's more of a shower thought, you sort of have "too much freedom" to be able to define anything that's necessarily meaningful.
Simple examples would be just "number of 'discrete parts'", "minimal area needed to span all territories" and things like that. Maybe you're more interested in "total distance from all satellites to wherever the capital is" or something, in a different context. The point is they'd all tell you radically different things, so it's important to know which one to ask for.
You could argue that something like Hawaii and Alaska's distance from the rest of the US makes the US score highly.
You could argue that any number of island nations score highly because after all, most of e.g. the US is in one part.
You could argue e.g. Norway's territories near both poles make it pretty high-scoring too.
You could argue that for whatever reason, distribution of area and population matter, and so on.Maybe we look at the ratio of country perimeter to area? Counting the number of exclaves could also be a factor. And maybe a ratio of the distance to cover all the exclaves divided by their area?
So if a country were a perfect circle it's perimeter to area ratio would be 2/r, it has zero exclaves and then it's width would be the diameter.
If a country were two perfect circles of the same diameter, separated by a distance of the same diameter, it's area ratio would be 2/r, exclaves would be one, and it's width would be three times it's diameter.
So now you can imagine a country like Chile, modeled as a really skinny rectangle, has a pretty large perimeter to area ratio, no exclaves, and a width roughly the length of the rectangle.
I guess you'd have to decide if archipelago nations are measured as the geometry of the sea they own, or as discrete islands.
I would do perimeter^2/area, to avoid biasing toward small countries. Divide one circular country into two circles with the same total area and p^2/A goes from 4 pi to 8 pi. Divide a square country in two and p^2/A goes from 4 to 6.
I did this all in my head. I think you are right. Point being, small simple algebraic expressions stacked as a a polynomial can be used to create relative scores for this type of analysis.
Thanks for the proposal. That gets us somewhere already, although only for non-landlocked countries. Using the perimeter also opens us up to the coastline paradox.
I guess you’d have to decide if archipelago nations are measured as the geometry of the sea they own, or as discrete islands.
I think that it might serve us better to consider them as distinct islands, to keep the measures comparable with landlocked countries.
Will you use the sea borders instead of the coastline to avoid the paradox?
You didn't really mention how this is all being tabulated. Just in theory? Do you have access to data sets or are you making them? The coastline paradox only really exists for smaller and smaller measuring devices. If for example your method is using satellite images of a given resolution, or another consistent measuring tool, the measurements should be comparable between country borders and therefore the error or the paradox should come out in the wash.
Step 1: Find the area of each chunk. The biggest chunk is your main chunk.
Step 2: Find the distance between the closest edge of main chunk and the center of each other chunk individually.
Step 3: Discontinuity of each chunk is area of chunk * distance from main chunk / total area.
Step 4: Total discontinuity is sum of each chunk's discontinuity.
Bolded parts are important. If you use the center of the main chunk, larger main chunk radii make other chunks seem more discontinuous than they should be. If you use the closest edge of other chunk's, you don't account for the entire area of the other chunk.
This will give you a number that is bigger when there are more and/or bigger pieces that are further away, and smaller when the opposite is true, normalized for the total area of the country so bigger countries aren't penalized just because they're bigger.
That seems to capture the intuitive idea of discontinuity for me, thanks!
Not a problem. =)