The Köppen–Geiger climate classification made simpler (I hope so)

[AzureWings]

I meant binary morphological erosion - the opposite of morphological dilation. https://en.m.wikipedia.org/wiki/Erosion_(morphology), not terrain erosion from weathering processes.

Thanks for clarifying! The result of applying morphological erosion to landmasses in this case though seems like it would be fairly similar to a simple distance-from-ocean threshold (if somewhat smoother), and obtaining distances from nearest ocean is something that would already be worthwhile in and of itself.

Not just lines but fill the whole landmass. I know it's another step but it should not take long to do. Whether it's worth it or not depend on the complexity of the algorithm, which a have absolutely no idea.

It would save on running time, though the amount it would save by, while not insignificant, would still be marginal in that similar steps in terms of computational expense would still be necessary to construct other data (like distance from nearest ocean). It is an option though.

To be honest, I don't know what CUDA or OpenCL are.

They're frameworks for parallel programming on GPUs (they're a bit broader than that but that's what I take to be the suggestion for their use here). Since there are a lot of individual data units (e.g. pixels) with similar operations being performed on each, if and when we aim to make any automated system for climate generation more efficient it's potentially a very good candidate for such parallelism.

[Azélor]

It takes about 5-10 minutes to do the continent step with Earth. Could have taken less time but I was concerned about what to do with the islands like Japan, Newfoundland, Madagascar, South east Asia...
It probably does not matter which continent they are part (or if they are part of a continent to begin with) since they are likely to be considered coastal.
How would the algorithm manage archipelagos like the Canadian Arctic or Patagonia?
In winter, the Arctic would be considered like a single large landmass because of the ice sheet. At least that was the assumption in the tutorial. (although I did put Greenland as a continent)

Have you considered automating the precipitation part as well? I just want to know if you thought about it.

[AzureWings]

I'd figure you could have a configurable threshold for what latitude to approximate the start of an ice sheet in each season, and that could be taken into account by an algorithm for painting the temperature influence zones (i.e tend to paint even coastal land north of that line with continental zones unless they're under a hot/mild-current influence). If going into winds/pressure systems/other earlier data you might actually have the algorithm construct a 'mock landmass' out of the ice extent and treat it as a landmass.

If temperature generation is being automated, I figure it only makes sense to give precipitation a shot too, but I hadn't really looked at the task in any detail yet.

[Azélor]

I made a script for Photoshop to indicate the distance from the sea. I'm not sure if it's better than the default tool included (interior shadows, I think) since it has all these strange lines, but maybe it's supposed to be like that?
It does not take in account the map projection, and Antarctica is wrong since the script thinks there is a ocean south of the south pole.

[AzureWings]

While it looks visually a bit strange, I think there's a bit of visual trickery/optical illusion going on with how we see the darkness of the lines as compared to their surroundings, since the surroundings are varying too - all of the lines no matter how faint should essentially be 'ridges' where multiple coastal points are tied for nearest ocean pixel (moving off the line gets you closer to at least one of those points so it gets brighter, which is what forms the lines). So I think they're expected, just visually non-intuitive (especially since distinguishing relative brightness is something that is very sensitive to the surrounding visual context, such as with the checker shadow illusion).

One test you might give it is a perfect circle - outside of issues caused by aliasing artifacts there shouldn't be any striations there, since aside from the center of the circle any point within it should have only one nearest point on its perimeter.

[Azélor]

I guess it's normal.
The circle eventually becomes an octagon, though very gradually. It can't be a perfect circle since PS is a raster program.

While it might be useful to know the distance from the sea, it does not help figuring the distance between points. So, I guess using PS for this is not a really good idea.

I was also considering using Qgis to help with the precipitation step.

[AzureWings]

Unfortunately an averaging pass or two over the temperatures is not offering all that satisfactory of results overall (and causes the program to take several times longer to run). I'm taking a bit of a break from automation attempts to run through the hand process again with a bit more care on a new map - had the thought about previous comments regarding pack ice causing surrounded land and islands to act like continental zones in terms of temperature influence. Looking at where it's marked on that large, detailed current map you posted in the currents step, and on another one I found from the 1930's, permanent pack ice extent (at least, pre modern-era climate) comes down to about 70-75° latitude from the poles, while winter extent gets to about 60°, and from those baselines veering away from wherever warm/mild currents are present. Does that sound about right?

[Azélor]

Unfortunately an averaging pass or two over the temperatures is not offering all that satisfactory of results overall (and causes the program to take several times longer to run). I'm taking a bit of a break from automation attempts to run through the hand process again with a bit more care on a new map - had the thought about previous comments regarding pack ice causing surrounded land and islands to act like continental zones in terms of temperature influence. Looking at where it's marked on that large, detailed current map you posted in the currents step, and on another one I found from the 1930's, permanent pack ice extent (at least, pre modern-era climate) comes down to about 70-75° latitude from the poles, while winter extent gets to about 60°, and from those baselines veering away from wherever warm/mild currents are present. Does that sound about right?

Yes, the presence of an ice sheet is highly dependent on the water temperature.

[martinmike2]

Hey,

Im giving this tutorial a try and have just finished my heightmap. What do you guys think?

[AzureWings]

The high areas seem a bit... uniformly centered on the continents? If you're going for a height map that reflects some sort of tectonic basis, I might expect the mountains to reflect that in some way - right now they seem a little too much just 'located at the middle' everywhere.

A point I think Charerg mentioned in another thread is that (if you're going for a relatively Earthlike configuration) >4km elevations are actually rather uncommon, and most continents on Earth don't have much of it (and not all continents reach the same sort of elevations - most of them don't have the Andes or Himalayas, and there are quite a few mountainous regions on Earth that don't hit 4km at all); moreover >6km is extremely rare and confined to the very highest peaks. It may be a consequence of just defining general areas but unless your intent is for your mountains to in general reach higher extremes than Earth's your white areas look a bit large. Likewise, you mostly just have the lowest elevation category along the coastlines; Earth has some large regions that are mostly comprised of the lowest elevations. Again, this only is a big deal if you're going for a very Earthlike configuration, although it will affect climates you end up with.

Also, you've got a few of what look like awkward artifacts around the north-central island chain - some lines that look like unfinished island outlines. That does raise another thing about the islands - while having them located in chains is good the average island size seems a little bit high (or put another way, the island chains seem to be really dense in terms of the amount of land in a given area). This seems like a 'sense of scale' issue that's pretty easy to run into (I do it a lot too) but in terms of relative size on a map of Earth many islands are just a couple of pixels (even at a resolution like 40000x20000). That's not to say islands the size of yours don't exist - obviously they do - but they tend to (often though not quite always) be accompanied by a number of smaller islands too. Granted, putting those smaller islands there can get tedious (and is often a bit visually unsatisfying), so it's a question of how precise you want to go.