The red-brown of South Africa (and I guess Argentina and Australia too) are basically midway points of their lush green regions and dry desert regions.
South Africa is as green as neighbouring Eswatini (formerly Swaziland) in the east and south coast and almost as arid as the Namib desert in the far north west.
I suppose that's stating the obvious but still interesting. And could probably add some comment about how this shows the dangers of averages of any heterogeneous data.
One interesting fact is that the arid regions of Australia look much more red than most other deserts (which look more yellow, look e.g. at the Sahara or the Arabian Peninsula), this is due to the high concentration of iron oxide in the soil of many inland areas of Australia - in other words our sand basically contains a little bit of rust :-D
Ah, thanks. Then I misremembered my high school chemistry from 25y ago. I thought rust (the verb) is oxidation and that when you add an acid, it would be reduction (to get back the iron).
It seems FTA that the ‘average’ color is found by resizing the color information within the bounds in question to 1:1 and then using that color to fill the bounds. I’m not sure but that seems to be an important algorithmic element of this process that is almost completely delegated to a side effect of an image library and not really discussed in the context of this write up.
I also routinely make visualizations of geographic and weather data (including satellite imagery) as part of my job and it doesn't surprise me.
Once you've 1) acquired the data you need (in this case, we're talking about images of countries without clouds and corrected for exposure across the Earth) and 2) fixed your data (because it's never in the shape you want, or the map projection you need), the core of the algorithm can seem pretty trivial sometimes, especially because R and GDAL come with a very large set of ready-to-use algorithms (and I'll never thank the creators of their respective libs enough for that).
There will be a considerable amount of averaging and re sampling before the final resize. Depending on shape of the region and how samples outside it are handled you may see considerable variation in the results.
What I would be interested in is how much the colour would differ if another colour space were used. I wouldn’t expect a huge change in the imital re projection as adjacent samples are likely to be close in whatever colour space is used, but I think it might change some of the average colours.
A lot of the complexity of the python code is hidden by the Geo* methods you have here; a more representative comparison would show how those methods are implemented (eg with the shape files).
There is a lot of complexity hidden in the R functions as well. The reason of using higher level languages with comprehensive standard libraries is exactly to abstract these things away. So, I think it's fair to say that, for this particular problem, wolfram language is better suited.
Stephen Wolfram gave the keynote at re:Clojure a few days ago, with an amazing demonstration of the Wolfram language. It's incredible how much data and functionality they've made available with it, and with Stephen's mastery it is bordering on wizardry.
A nature lover's tip for Google Maps; old growth primary forest appears a far darker shade of green than secondary forest. I've found this very useful for finding good hiking close to urban areas.
This is such a great tip. Do you have any others? I personally use Strava to find places to walk when I go to new cities. And when I'm looking for a neighborhood to stay in a new cities, I plot out all moderately expensive restaurants to use as a heuristic for safe but not touristy neighborhoods.
The other Google Maps tips I have is a bit less general; when searching for an anchorage for you boat you can often see the direction and magnitude of swell in the satellite photos.
Author forgot the Asian part of Russia, it's included neither to Europe, nor to Asia. But it's still on the general map of the world.
Anyway it would be green.
Agreed - I think it would be interesting to see it by state/province (or even better some kernel smoothing) instead of country. Country size is kind of arbitrary, as some countries are very large and others are relatively small.
I was talking whole continents. I did notice some other bits went missing. "I guess people only pay attention to what they want to find" - not sure exactly what you're trying to say there, but it didn't feel good.
I’ve always wondered what the average colour of the web is. Probably close to light gray. Maybe Facebook gives it a blue tinge. Dark modes have complicated the question.
It'd be particularly fun to see that change over time - probably very much light grey early on (black text on white or near white or patterned whiteish as was popular) but darker/more colour as 'webapps' became more prominent and took over the whole screen more.
Pretty much. Remember how close they are geographically. IANAG but the gulf stream is probably the only significant difference. Also, they probably should say GB not UK.
The chart doesn't capture ... something. Luminous intensity?
I (and others) have noticed that when flying into NZ (Auckland or Wellington), the green is so intense that I feel I have to put on sunglasses. I haven't felt that anywhere else, but I haven't been to Ireland. I imagine Ireland is the same in this respect.
Interesting how Turkey and Azerbaijan have the same brown color, while Georgia and Armenia which are in the same geographical location, have a very different deep green color.
These images don't make any sense to me. This map looks all black except for the sahara desert. But the world map before averaging looks mostly greenish.
One can see the same thing as in regular satellite maps: for example that Spain is the odd color out in Europe. Much more arid, even if it's not a desert.
Grass is also green, lots of agricultural land come up as green as well. Actually, Europe is still fairly forested (45%) on average, so I'm not sure what other color it could be seen as.
South Africa is as green as neighbouring Eswatini (formerly Swaziland) in the east and south coast and almost as arid as the Namib desert in the far north west.
I suppose that's stating the obvious but still interesting. And could probably add some comment about how this shows the dangers of averages of any heterogeneous data.