6 ms·

Possibly off-topic: the variety of color scales used here is wild, and interesting. It looks like of them may have pre-established meaning within the community

Possibly off-topic: the variety of color scales used here is wild, and interesting. It looks like of them may have pre-established meaning within the community of expertise (like the rainbow-ish one for the image after "Overall, no sign of any influence from below."), but many of them are versions of a double-ended scale (white in middle, at 0), but there's a wide variety of double-ended scales. I wonder if this less disorienting to the domain experts?

anigbrowl 4y ago
I'm kinda used to them, there's a 'library' of color scales used in scientific visualization that aims to balance contrast with complexity and also be accessible to color-blind people. Matplotlib docs have good reference material on this iirc.
- an1sotropy 4y ago
  Right, but with those goals in mind a lot people say "so viridis it is!" and leave it at that; but there's nothing like viridis on that page, which is intriguing.
  - anigbrowl 4y ago
    I mean why would you ever want anything else XD I think the most interesting scale on that weather page is the red-blue one where the extremes of both then veer into other colors - often the same colors, but that's OK because they will only appear within 'moats' of red or blue covering the typical temperature distributions. So you have increasingly deep shades of blue or red to indicate very cold or hot weather, and then purple to white within those areas for record-setting-breaking extremes. I guess you could just do normal vs deviant scales, but given our sensitivity to the dichotomies of weather people have an inherent need to know what sort of extreme it is.
    - an1sotropy 4y ago
      Yes I noticed that too, and I agree with your guess about why it’s ok. Fundamentally the functions here are smoothly varying, which avoids ambiguities and some some contrast effects that affect chloropleth maps. Long path length in colorspace == more opportunities for nearby discrimination, even if with overlapping hues

anigbrowl 4y ago

I'm kinda used to them, there's a 'library' of color scales used in scientific visualization that aims to balance contrast with complexity and also be accessible to color-blind people. Matplotlib docs have good reference material on this iirc.

an1sotropy 4y ago
Right, but with those goals in mind a lot people say "so viridis it is!" and leave it at that; but there's nothing like viridis on that page, which is intriguing.
- anigbrowl 4y ago
  I mean why would you ever want anything else XD I think the most interesting scale on that weather page is the red-blue one where the extremes of both then veer into other colors - often the same colors, but that's OK because they will only appear within 'moats' of red or blue covering the typical temperature distributions. So you have increasingly deep shades of blue or red to indicate very cold or hot weather, and then purple to white within those areas for record-setting-breaking extremes. I guess you could just do normal vs deviant scales, but given our sensitivity to the dichotomies of weather people have an inherent need to know what sort of extreme it is.
  - an1sotropy 4y ago
    Yes I noticed that too, and I agree with your guess about why it’s ok. Fundamentally the functions here are smoothly varying, which avoids ambiguities and some some contrast effects that affect chloropleth maps. Long path length in colorspace == more opportunities for nearby discrimination, even if with overlapping hues