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

[Charerg]

So, I recently got myself QGIS and fiddled a bit with WorldClim's 1970-2000 dataset for average monthly temperatures and precipitations. Previously, I used the 1960-1990 dataset (and I believe Azelor used 1960-1990 as well).

I ran across some interesting stuff when looking at the distribution of "Deadly Cold" (below-38 C average) and "Severely Cold" (-25 C to -15 C). Take a look at this map of Eastern Siberia with the elevations overlayed over the January temperature zones:
Eastern Siberia.png

The areas surrounded with red represent a "Chill" (-3 C to 0 C) category that I added in for Cc climates. The random dots around the Siberian coast are probably some kind of flaw in the dataset (given the extreme temperature differences with surrounding areas).

However, the thing I wanted to talk about is how the "Deadly Cold" seems to appear in the valleys and the Lena river basin, but not really in the highlands (the Verkhoyansk range shows clearly warmer temperatures relative to the surrounding lowlands). Similarly, in the Altai-Sayan region, "Severely Cold" occurs mostly in the lowlands.

And a similar thing seems to be going on in Alaska:
Alaska.png

Again, the "Severely Cold" seems to concentrate in the lowlands: the Brooks range is clearly warmer than the Yukon river basin. Again, we see some weird dots of warmer-than-plausible areas along the coast. In addition, the highest peaks of the Alaska range are shown with progressively warmer temperatures the higher up you go. Some of them even show summer temperatures, which is clearly a flaw in the dataset.


I have to admit that I am thoroughly confused by this. Does anyone have a clue if the lowlands are actually colder during the Siberian and Alaskan winter than the highlands, or is the WorldClim 1970-2000 dataset simply messed up regarding these areas?

[Azélor]

Yes the red part are clearly artifacts, bad overlapping data or lack of consistent data. The sea is frozen in winter there.

For Siberia, I considered a lack of data.

But it's an inversion of temperature. It's apparently very common in polar regions in winter. I did not know about that.
I don't understand it very well but it's when the air mass at ground level can't rise.
This video gives a good explanation : https://www.youtube.com/watch?v=T_U3TXHBt-0

[Charerg]

But it's an inversion of temperature. It's apparently very common in polar regions in winter. I did not know about that.
I don't understand it very well but it's when the air mass at ground level can't rise.
This video gives a good explanation : https://www.youtube.com/watch?v=T_U3TXHBt-0

I did a bit of google-fu about this, and it seems there is indeed a genuine temperature inversion associated with cold-cored continental highs (namely the Siberian High). I googled up a chapter in Robin Mcllveen's "Fundamentals of Weather and Climate", this is what the book has to say about it:

In early winter, the cooling of a land surface is greatly accelerated by the first snow falls, which practically eliminate further solar warming but maintain the net loss by terrestrial radiation. The surface and the overlying air cools progressively day by day, as in a polar night. The whole mass of overlying air sinks on top of the shrinking layer of cooling air near the surface, and the air aloft converges faster than the air near the surface diverges, so that the surface pressure rises. After setting up, the high surface pressure is maintained by a dynamic equilibrium between convergence and divergence which is common to all types of anticyclone, though not easily explained. Despite persistent subsidence, the absence of an underlying convecting boundary layer prevents the appearance of anything like the elevated subsidence inversions of other anticyclones; instead a relatively smooth inversion can extend several kilometres from the surface.

It then goes on to show a graph from Yakutsk (Jan 28, 1958 ) with the 1000 hPa temperature at -50 C, rising in a more-or-less linear fashion to approximately -25 C at 850 hPa, and then continuing to rise until 700 hPa (maybe about -17 C, the graph is unclear). Then the temperature starts dropping again. Looking at a basic pressure-to-altitude converter, 850 hPa is approximately 1500 metres, and 700 hPa about 3000 metres. I guess the presence of the high pressure area would push these altitudes a bit lower. Either way, it does seem like there's a pretty substantial temperature increase with altitude, especially from 0->1500 metres.

Looking at the temperature zones of Alaska, I'd guess a similar phenomenon (though probably less extreme) must be occurring with the Canadian High as well.

Edit:
This seems to happen in Mongolia too (reading this site). Apparently the coldest temperatures occur in the valleys between the mountain ranges. I guess that would be those "Severely Cold" areas appearing in the Altai-Sayan region.

Edit2:
Found some further info, which is a bit more applicable. From "Encyclopedia of Climate and Weather" (Dr. Stephen H. Schneider):

Mean surface temperatures during the winter range from about -20 C near the Mongolian-Siberian border to about -45 C in the northeastern corner of Asia. The high density of the cold air and the overland cooling produce a persistent temperature inversion, such that the warmest air, which can be up to 20 C warmer than air near the surface, is typically found at an altitude of 600 to 1000 meters. Such inversions are thickest in the Yakutsk area.

With that in mind I guess we can conclude that the "climatically pedantic users" of the tutorial should probably raise the temperature category by one at 1000 metres, if they have the "Siberian High case" and Deadly Cold or Severely Cold occurring.

Edit3:

I did some further research, and the Alaskan and Mongolian cases seem about the same, although there the temperature rises about 10-15 C, again with the warmest air at about 1000-1500 metres above surface. At least this seems to be the case in the Alaskan interior and Northern Mongolia, the temperature inversion is much weaker in Southern Mongolia. I guess the big effect is that you essentially have to reach approx. 3 km altitudes before you're "about equal" with surface-level temperatures, so the mountains in most cases don't really have colder winter temperatures than the lowlands if an area is influenced by a cold-cored continental high.

[Charerg]

I made a few plots of what the temp inversion profiles look like. Since I think a 4th edit would be a bit overkill, I'll make this a separate post. Note that these are fairly rough, intended for the purpose of making rough estimates. I used the following sources:

Fairbanks, Alaska:
A Climate Perspective of Observed and Modeled Surface-based Temperature Inversions in Alaska (Stefanie M. Bourne, link to paper)

The above thesis contains actual elevation-to-temperature profiles for various months averaged from 1957 to 2008. So, the simplified January profile I made based on this should be quite accurate. The data caps at 4 km though, so the dashed line from 4 km to 5 km in my profile is a guess.

Mongolia, Ulaangom and Arvaikheer:
Air temperature distribution over Mongolia using dynamical downscaling and statistical correction (B. Gerelchuluun, link to paper)

The above contains pressure-to-temperature profiles averaged over 1981 to 2010 for Ulaangom and Arvaikheer. I was able to somewhat accurately correlate pressure to elevation by comparing the start point of the profiles to the elevation of the aforementioned places (Ulaangom 931 m, Arvaikheer 1817 m). Still, I can't claim the profiles I made are terribly accurate, but they should be okayish. Notably Ulaangom experiences a very strong temperature inversion (being located in a deep valley near the centre of the Siberian High), whereas the temperature inversion is very weak in Arvaikheer (probably due to the high elevation). Ulaangom apparently has the strongest temperature inversion in N. Mongolia, while Arvaikheer has the weakest.

Yakutsk:
I wasn't able to find anything super-reliable for Yakutsk. I ended up making an estimate based on the Fairbanks profile and the literary sources I quoted in my prior post. Mcllveen's "Fundamentals of Weather and Climate" has that extremely deep inversion profile, but that's only for Jan. 28, 1958, so probably not representative of an average.


Anyway, here are the profiles:

Jan Temp Inversions.jpeg

What is notable from this is that apparently the inversion depth is about the same in Ulaangom (931 m) and Fairbanks (136 m). So, it seems that the inversion is relative to the "base level" of the surface. Although it may also be the case that the temperature inversion is just stronger in the Siberian High. If my estimate is anywhere near accurate, Yakutsk has an even deeper inversion profile, and you'd probably have to get near 6 km altitude before "breaking even" with the surface-level temperature.

Edit:
I found out that the Avg. Jan. Temp is about -41 C in Yakutsk (source). Also, found an extra source comparing the temperature inversion between Oymyakon (~700 m) and the Suntar-Khayata Glacier (~2000 m). Apparently, the Jan temp at the glacier is indeed ~20 C warmer than Oymyakon (link to paper). I somewhat smoothed this out since my estimate is intended to depict a long-term average. So, the Yakutsk estimate has been updated, and it's a bit less extreme now.

Also, apparently the temperature inversion happens in polar areas too. I found out one table from January 19, 1949, that provided temperatures compared to altitude in the Eureka Sound. At sea level, the temperature was -45 C, rising to a maximum of -21.8 C at 2.44 km (8000 ft) and then dropping to the "start point" at about 6.1 km (20 000 ft) at -44.7 C. So, the magnitude of the inversion seems relative to surface temperature: the colder the surface, the more strong the temperature inversion is.

With that in mind, I guess the following "Counterlapse-lapse Zones" (Cl-L Zone) could be summed up (the elevation needed to rise to max temp and then drop back to starting point):

Fairbanks: ~3 km Cl-L Zone, dT ~8 C
Ulaangom: ~4 km Cl-L Zone, dT ~13 C (I probably underestimated the Cl-L Zone in the temp profile slightly)
Yakutsk: ~5 km Cl-L Zone, dT ~15 C (rough estimate)


Edit2:
Here's a similar altitude-to-temperature graph to the one Azelor made previously, taking the temperature inversion into account. Note that this is only for winter in arctic or subarctic areas that experience continental conditions. For the summer temperatures, the usual graph should be followed.

Temp-to-altitude for Arctic Winter.png

[Charerg]

Okay, I spent a good chunk of time yesterday cleaning up all those artefacts from the temperature maps. I'm 100% certain I missed some tiny island in the Pacific somewhere, but overall the temperature maps should be in serviceable condition. Equivalent maps of the temperature zones using WorldClim's 1960-1990 dataset can be found in my prior post, or in Azelor's real world data.

Note about temperatures in the Alaska Range:
Originally, the Alaska Range had some serious temperature anomalies in the Jan temp map (the 6 km tall Denali with 10+ C temperature in January!). I fixed those up using the Jan temperatures of the 1960-1990 dataset. I suspect the anomalies were a genuine flaw in the 1970-2000 dataset rather than just artifacts like those random dots along the coastline.

January temperature zones:
January Temp (1970-2000).png

Major differences between 1970-2000 and 1960-1990:

1. Winter Temperature Inversion in Continental Subarctic Regions
- The old 1960-1990 dataset had (incorrectly) basically all of the Altai-Sayan region and Alaska blanketed as "Severely Cold" (-38 C to -25 C), and most of E. Siberia as "Deadly Cold" (below -38 C). As previously discussed, these regions actually experience a temperature inversion, and the highland areas are noticeably warmer than this.

2. Deadly Cold in Central Greenland
- I suspect the 1960-1990 dataset had more limited data for Central Greenland. The 1970-2000 has C. Greenland as "Deadly Cold" in January.

3. Data for Antarctica
- Comparing the data to Azelor's sample maps (which have E. Antarctica as "Very Cold" in January), it seems the central regions of E. Antarctica maintain "Severely Cold" temperatures even during January.

July temperature zones:
July Temp (1970-2000).png

Major differences between 1970-2000 and 1960-1990:

4. Largely unchanged
- The July temperatures show only minor differences. The only notable one is that C. Greenland is again depicted as colder. Apparently it maintains "Very Cold" (-25 C to -10 C) temperature even in July. In Antarctica, the Antarctic Peninsula is a bit colder than was the case in Azelor's sample map.

I did the precipitation maps too, but they seem basically the same in both datasets (only minor differences).


Edit:
Oh, and I split the "Cold" category (originally -10 C to 0 C) into "Cold" and "Chill". The tutorial uses the 0 C isotherm as the boundary between C and D climates. However, most Köppen maps tend to use -3 C isotherm as the C/D boundary instead. The idea here is to basically save the -3 isotherms for Jan and July in order to later redraw the C/D boundary after the final climates have been defined. During the climate definition stage, the two categories are recombined into the old "Cold" category.

This basically makes Cc and Cb a bit more widespread, and eliminates most of the "weird" D climates in places like the Southern Andes, Iceland and coastal Norway and Canada. It's optional since it's up to personal preference whether the 0 C or -3 C isotherm works better as the C/D boundary.

[Azélor]

Yea, the numbers look ok.
I noticed that the temperature lapse above 4000m seems constant based on your sample and close to the average I used to make the initial model. Around 7 / 7,5 C by 1000m.

[Charerg]

Yea, the numbers look ok.
I noticed that the temperature lapse above 4000m seems constant based on your sample and close to the average I used to make the initial model. Around 7 / 7,5 C by 1000m.

My profiles are intentionally somewhat simplified. I think it also depends a lot on the case. But yeah, I guess it does sort of go back to normal adiabatic lapse rate past a certain point, so it's probably ok to assume a linear lapse rate after passing "the hump".

Here's a sort of theoretical profile for those who want to have that "crazily high mountain at the top of the world". If you had a 10 km tall Nunatak jutting out of an approx. 2 km thick glacier, this is how the temperatures might look like (in winter):

Polar Super-Everest.png

The tropopause is about 10 km high over the poles, once you reach that point, the temperature no longer really drops (or drops much more slowly). Although the temperature at the tropopause could be a bit warmer than -60 C. According to wikipedia's "temperatures in the troposphere", the polar tropopause is only -45 C. Here's the quote from wikipedia:

At middle latitudes, tropospheric temperatures decrease from an average of 15 °C at sea level to about −55 °C at the tropopause. At the poles, tropospheric temperature only decreases from an average of 0 °C at sea level to about −45 °C at the tropopause. At the equator, tropospheric temperatures decrease from an average of 20 °C at sea level to about −70 to −75 °C at the tropopause. The troposphere is thinner at the poles and thicker at the equator. The average thickness of the tropical tropopause is roughly 7 kilometers greater than the average tropopause thickness at the poles.

[Charerg]

Some time ago I made a post about generating an extra map that depicts summer length (aka length of the growing season), defined as the number of months with mean temperature above 10 C. The primary purpose for creating an extra map like this is to fix the distribution of a/b and c/d climates. The present tutorial tends to generate a lot of nonsensical transitions (like lowland Cc-> highland Db, to give an example), so there's definitely room for improvement in this regard.

For those who don't know, Köppen defines c/d climates as having 1-3 months above 10 C mean temp, while a/b have a minimum of 4 months above 10 C. So the primary difference between Cb and Cc (or Db and Dc) is the summer length, not the the maximum or minimum temperatures per se.

Previously I didn't have anything in the way of instructions to offer about how to create such a map for a fictional world though, as Pixie noted:

The point that Geoff's and Azelor's techinques only care about the extreme months is indeed a limitation. And you might be into something to break out of that, Charerg, but you would need to come up with a way to predict that kind of map you produced for Earth, for any given arrangement of land/ocean (which we like to call conworld)... Any thoughts?

So, here is an attempt of sorts to provide something of a guideline. First off, I've split the summer length into five categories:

1. Eternal summer (12-10 Months above 10 C)
2. Long summer (9-7 Months above 10 C)
3. Mid-length summer (6-4 Months above 10 C)
4. Short summer (3-1 Months above 10 C)
5. Arctic summer (0 Months above 10 C)

Here's the above presented as a colour key:
Key.png



A. Creating a base map of summer length:

The most important factors that control summer length are latitude and elevation. I've created a graphic that can be used as a guideline to create a base map of summer length:

Latitude-Elevation Graph.png

Note that if you've already created a climate map (or just the temperature maps), then your "Arctic summer" category should already exist (the E climates are defined as "No month above 10 C"). You'll note that in tropical latitudes the transition from "Eternal summer" to "Arctic summer" is extremely abrupt. This is because the temperature varies very little during the year. So, either the elevation is enough to push the monthly mean temperature of all months below 10 C, or it is not. In actuality, there is a very narrow transition zone, but it's too marginal to be depicted in a world map of climates.

Here's a sample map of Earth I created using the above instructions:
Sample 01.png

Note that the above map is intentionally very rough, as I've "followed my instructions to the letter". For an improved map, I suggest utilising the techniques described by Azelor in the "temperature placement" section of the tutorial.


B. Creating a map about local influences:

However, the above doesn't take into account local influences. For the summer length map, you need to consider the "annual net influence" of various factors. In some cases, continental influence can cause a longer summer, sometimes shorter. The areas adjacent to oceanic high pressure centres tend to have a "hot continental" effect because of the clear and sunny weather they cause. Also, western Eurasia is significantly warmer than expected due to the combined effect of the Gulf Stream reducing arctic influence, and also the warm desert regions causing a strong "hot continental" effect. Eastern Eurasia, on the other hand, is colder than expected, with a noticeable "cold continental" effect due to the development of the Siberian High and the freezing of the Arctic Sea in winter. Northern Canada is similarly open to arctic influences since the sea freezes in winter and there is no moderating effect in play.

Here's an example map I made for Earth:
Influences.png

Note that you can largely ignore tropical areas when creating the influence map, those areas are always going to be "eternal summer". When adjusting the summer length map with the influences, maritime influences should generally only effect 0-1000 m areas, as higher elevations tend to effectively block oceanic influence. A rule of thumb is to move the "summer length zone" about 5 degrees equatorwards if a cold influence is in play, or 5 degrees polewards if a warm influence is at play. It's not an exact science though, and in practice you need to make judgement calls. For example, "cold continental" doesn't usually reduce summer length to 0 months (unless the "cold continental" effect is caused by glaciers), but "cold maritime" could, as in the coasts of the Labrador peninsula or the southern tip of South America.


C. Final map of summer length:

Here's my Earth map after adjustments based on influence:
Sample 02.png

Note that my map isn't super-good or anything, as I did this fairly quickly and sloppily. Nevertheless, even a basic map like this could be used to make a major improvement with the distribution of a/b and c/d climates and eliminate all those nonsensical c->b transitions that the tutorial generates.

Finally, here's the equivalent "summer length" map generated from WorldClim's 1970-2000 dataset. It's a good reference, and you can check to see the difference between my sloppier version and the actual data:

 

1970-2000 Months Above 10.png

That's it for this bonus map. I realise my instructions are a bit short, but since the factors that affect summer length are largely the same that affect temperature in general, I figured there was no need to repeat instructions already covered by Azelor in the "temperature placement" section. Any comments, questions and suggestions are welcome and I hope you found this interesting and/or useful!

[Pixie]

Previously I didn't have anything in the way of instructions to offer about how to create such a map for a fictional world though, as Pixie noted:

You certainly have now. I will have to find the time to try this but, at first sight, this is an awesome add-on to our effort.
Charerg, I said this before, I think, and I'll say it again: I like the way you think

[Azélor]

Chareng, are you interested in having the original files? I have documents made with Photoshop, Illustrator and Excel. But some of them are really messy.

The last post was pretty good.