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

[Charerg]

Temperature combinations left after merging similar ones: still 28 but the distribution is different. ( tundra and ice caps included)
I less D and one more C.

Attachment 104078

26*34+2=886

That's not too bad.

One thing I experimented with was splitting the Mild (18-10 °C) category in two in order to fix the Cc distribution. I found that making the "sub-Mild" category 10-14 °C still made Cc too extensive, but 10-13 °C seems to fairly closely match the distribution of Cc climates. Including that could be a viable alternative to making a separate map to sort out the b/c boundary. That way you'd only need the temp maps to get the Cc distribution right. It would however ramp up the temp possibilites from 100 to 121.

Here's the July map with the 10-13 °C sub-Mild category (in light blue). It also has -3 to 0 °C "Chill" in grey but that's not a necessary addition. If you compare the extent of the sub-Mild category to Cc/Dc distribution in British Columbia and the British Isles, I think it's a fairly close approximation. Seems to work pretty well for Norway, too. Though some areas become Dc instead if using the 0 °C isotherm (those maps use -3 °C).

Here's the map:
JulTemp Test.png

Also, I think maybe the Db/Dc distribution could potentially be improved by changing the Cold/V. Cold boundary from -10 °C. At least in Europe Dc is somewhat less extensive than what it should be in the present model. I haven't experimented with that yet, though.

Edit:

I experimented a bit with -8 °C and -6 °C Cold/VCold boundary. Tbh the results are much less promising here than with the Cc distribution. The Dc distribution in Scandinavia would be more-or-less correct with -6 °C, but then you end up with northwest Russia and Eastern Canada having too extensive Dc areas. I ended up with -8 °C as a compromise. The results might be slightly better than the original -10 °C boundary, but I'm not sure. With -8 °C the Dc areas in mountainous regions become even more extensive than previously, although in reality the Hindu Kush and other similar cases are mostly Db.

Climate zones with sub-Mild (13 to 10 °C) and -8 °C Cold/VCold boundary:
13to10 Sub-Mild and -8C Cold VCold boundary.png

In the end it might still be easier to create a separate map about the length of the growing season, since it doesn't look like the Db/Dc boundary can really be sorted with info from just two months.

[Azélor]

Your proposition is that we should add a new category 10-13?
It does increase the average temperature of the Mild climates.
It could look like that?

sheet31.png

Gave it the color yellow and named the other category below as Chilly. That looks good?
Are there any other changes I should make for the temperature?
You think we are good with 6 precipitation categories?
It would be easier to add new temperatures or changing them than adding new precipitation categories. So if the precipitations look right with 6 categories, we could add one temperature category.

I think the most problematic Db/Dc area is in Siberia and your changes did not improve anything in that region.
Also, I don't know what map you use as reference, hopefully the one Kottek because the one at Wikipedia seems pretty flawed. For instance, it shows a Dfb climate around the Bay James between Ontario and Quebec but that is clearly a Dfc area.

[Charerg]

Your proposition is that we should add a new category 10-13?
It does increase the average temperature of the Mild climates.
It could look like that?

sheet31.png

Gave it the color yellow and named the other category below as Chilly. That looks good?
Are there any other changes I should make for the temperature?
You think we are good with 6 precipitation categories?
It would be easier to add new temperatures or changing them than adding new precipitation categories. So if the precipitations look right with 6 categories, we could add one temperature category.

I think the most problematic Db/Dc area is in Siberia and your changes did not improve anything in that region.
Also, I don't know what map you use as reference, hopefully the one Kottek because the one at Wikipedia seems pretty flawed. For instance, it shows a Dfb climate around the Bay James between Ontario and Quebec but that is clearly a Dfc area.

For regional maps, I tend to use Adam Peterson's Köppen maps (they use Kottek's criteria to classify the climates though).

But yes, the Cc distribution is probably the biggest issue right now. Even with the 6-step precipitation system the climate distribution is fairly good, though using 8 would probably improve the results I don't know if it's a necessity as such. Although there are some cases like the Af-Am transitions where the 8-step system tends to work better.

However, I'm not sure if the Cc distribution is best handled with the extra category, because that still leaves Db/Dc boundaries off, so maybe it's still better to just create that extra map about summer length and use that to fix the Cb/Cc and Db/Dc boundaries.

Also, the difficulty with adding more temperature categories or modifying the boundaries between them is that then we'd have to alter the instructions for placing the temperature zones.


Edit:
Although I guess one possibility would be to split the tutorial into two: the "basic tutorial" using the present temperature and precipitation systems, and an "advanced tutorial" that uses a larger number of categories to provide more accurate results. But in the advanced version it would be more difficult to create the source maps, since you need to be a bit more precise with a greater number of relatively narrow categories. Though maybe that complicates things unnecessarily. I don't know if the "average user" would have any interest in an extra complicated version.

Edit2:
I guess we could try both 11 temp categories and 8 prec categories? I think it could still be handled if the total possibilities would end up in the 800-1200 range.

[Azélor]

How will you generate the summer length map?
It will solve most of our problems?

Adding the 10-13 category doesn't look too complicated. One would only have to reduce the old 10-18 by 30-40%.

Although I guess one possibility would be to split the tutorial into two: the "basic tutorial" using the present temperature and precipitation systems, and an "advanced tutorial" that uses a larger number of categories to provide more accurate results. But in the advanced version it would be more difficult to create the source maps, since you need to be a bit more precise with a greater number of relatively narrow categories. Though maybe that complicates things unnecessarily. I don't know if the "average user" would have any interest in an extra complicated version.

I just keep thinking about Mars and if the results of any of the 2 methods would make any sense. Precision would not matter at all since we would have no way to verify it.
Also, the geography of the planet is so different that all of our references are lost.


Assuming this is the version generated with the latest script: https://www.cartographersguild.com/a...9&d=1517397591


Areas at the tropic are too wet: Mexico (Northern half), Gujarat, Australia, South Africa
The Rockies in North America are also slightly too rainy.
These are things that could be improved but at the same time, changes might affect other ares and make them inaccurate instead.

[Charerg]

How will you generate the summer length map?
It will solve most of our problems?

Adding the 10-13 category doesn't look too complicated. One would only have to reduce the old 10-18 by 30-40%.

I just keep thinking about Mars and if the results of any of the 2 methods would make any sense. Precision would not matter at all since we would have no way to verify it.
Also, the geography of the planet is so different that all of our references are lost.

For the summer length map, there is that post I wrote (admittedly the instructions are fairly rough). It won't provide too precise results probably, but at least they will be consistent.

And you're right, precision is not a major concern with fictional worlds. I think the biggest problem is having "clean transitions" between different climate zones, and I do feel the 8-step system is a bit better in that regard. Since the jumps between the precision zones aren't so huge with 8, you end up with better transitions, generally speaking. So in that sense I'd definitely be interested in updating the 8-step script to work with the revised precipitation zones (for my own use if nothing else), since they did look fairly promising, at least in Excel.

Assuming this is the version generated with the latest script: https://www.cartographersguild.com/a...9&d=1517397591

That is the one, yes.

Areas at the tropic are too wet: Mexico (Northern half), Gujarat, Australia, South Africa
The Rockies in North America are also slightly too rainy.
These are things that could be improved but at the same time, changes might affect other ares and make them inaccurate instead.

Well, if the change we are speaking is going from 6 to 8 prec categories, I don't think it risks inaccuracy, since it just means a bit better resolution, so to speak. Also, adding that 10-13 °C category might also help since it would push the avg. temperature of "Mild" climates up, making some regions more prone to being arid. Since in general, cold steppes should extend further north in North America and Central Asia than they do in the model.



Edit: About that summer length map

One technique I have been considering is just creating a gradual surface-level map going from 0 to 12 months above 10 °C. Ideally it would be completely gradual, so 0->0.1->0.2.... and so on. I think that could be achieved if elevation could be ignored. Then the elevation could possibly be taken into account with a script that calculates the effect.

There would basically be three factors:

T_avg=Average Annual Temperature
dT=Temperature variance within the year
T_ele=Temperature drop due to elevation (this is subtracted from T_avg)

Normally, the start point for T_avg would be 10+dT (meaning 0 months below 10 °C). That is because this is intended to create an "adjustment layer" that is applied to that pre-existent surface-level map. So, for example, if the "base level" is 6M_a10 and the adjustment layer has 1M_b10 at 1 km, then the final number would be 5 M_a10 at that location. Although starting from tropical latitudes, there would also need to be a "buffer temperature" applied to push the below 12 M_a10 transition higher in altitude.

However, the big problem comes with dT. Based on latitude, we could assume that if Latitude= 90°, then dT=40 (just a random number in this case), and if Latitude=0°, then dT=2. And then assume a linear transition between the two extremes. That would describe the temperature variance between the seasons growing greater as you get closer to the pole. The problem comes because this ignores maritime and continental influences (which have a major effect on the temperature variance between the seasons), so I'm unsure if the results would still be acceptable.


Edit2:
Although with all that rambling said, I should add that I do think that the most "practical solution" right now is to simply accept that the Db/Dc distribution is going to be a bit off, and just fix what can be easily fixed (the Cb/Cc distribution). The summer length map should probably be considered a bonus map for those who have an interest in it in any case.

[Azélor]

Ok, I will finish updating the table later today and will send it to you when it's done.

[Azélor]

I haven't worked on the file yet but I have this!

iceland.png

[Charerg]

Does the "image-to-Excel-sheet" always produce the same shape? Since in our case you'd have to convert 4 maps from .png to Excel sheets. Either way, nice work.

[Azélor]

I haven't succeeded in converting a PNG.
I'm exporting the the geo tif from Qgis to Excel using an asc file.
It works so far but it's slow.

That was Iceland by the way.

[Azélor]

I've been busy doing a few thing, starting in no particular order.

Here, I've managed to put all the results from Hex climate (the 3600 combinations) in a single column
and then determining the climate, somehow

first it extract the temp and prec from the hex code
then search these values in 2 different tables to get the corresponding temperature, precipitation regime (s,f or w) and another one just for the A, and the threshold matching each temperature
Then we can calculate if the precipitation is greater than the threshold ans see if it is humid, steppe or desert
with the temp, we can figure out the 2 first letters and the aridity for the B climates
The two next colomn are just for A and B and use the data of other columns to figure out the missing letters, same for C and D on the last colomn
the climate columns puts everything together. it gives something like Ca__f but we can still understand
final colour reads the letters we just found and look into the colour table (classic koppen colours scheme) to give a single colour for each climate

So by reading the first and last colomn, it tells you : pick this colour and replace it with this


sheet32.png


This is the table assembled with links to the main tables for the precipitations:
It contain all the data used above and also a id colomn.
Excel need the data to be in a specific order to work properly. I included numbers in case we need to put back the data in the original order since it makes it easier to modify the table as the Hex code is gibberish.
You can notice that several hex have the same value. This is because we simplified the precipitations and temperatures. I think Excel always picks the first one, but it doesn't matter if the numbers would be different since all result in the same climates in the end. Also, including this redundant data means that if we change the categories, we won't have values missing from the tables.
But if we add new categories, we will have to expand the tables manually and adding the new links if necessary.

sheet 33.png

here is for the temperature, same logic

sheet 34.png

the last table looks like this:
You just have the relevant data left : input code, corresponding climate, and the final colours
ideally, we should copy the result in another layer for modification : ordering the results by climate and deleting the many missing values.
It is normal to have that many missing, they exist here only, not on the generated map.
we could also add the # to convert the hex to a colour, delete the text, and we have a colour key
the other use of this is to convert the data all using Excel and in that case, no colour key is needed.