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

[Charerg]

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.

I thought you posted the Excel files in the thread somewhere? But yeah you can send them, I'll try to have a look through them at some point. Though I doubt there's anything too major I could offer in the way of suggestions, your original "temperature placement" instructions are way more detailed than the stuff I put together for that summer length map . The only thing that comes to mind is that an alternative elevation-temperature graphic could be made for those who have access to a more detailed elevation map. For example, the "tundra line" (as defined by Köppen) in tropical areas is generally about 3500 m rather than 4000 m.

[Charerg]

Talking about that altitude graphic, in between working on Aduhr's height map, I've been trying to piece together something of a graph about altitude-to-temperature.

Unfortunately it's kind of tricky, because apparently the lapse rate varies a lot regionally. As an example, here's the approximate altitude of the tundra line as defined by Köppen (warmest month's mean temperature below 10 °C):

Tibetan Plateau (~35° N): 3850 m (more accurate)
Ethiopian Plateau (~10° N): 3500 m (approx.)
Venezuelan Andes (~5° N): 3250 m (more accurate)
Peruvian Andes (~15 °S): 3600 m (more accurate)

Note that the above are approximate altitudes, not exact. It still gives an idea about how much the lapse rate varies. The Tibetan Plateau is a lot warmer in summer than what one would expect based on just latitude and elevation. And on the other hand, the Venezuelan Andes seem chillier than one would expect. I guess these differences are caused by Venezuela having a very maritime, humid climate, with frequent cloud cover lowering the temperatures, whereas the Tibetan Plateau is extremely continental, with much hotter summer temperatures than might be expected.

Still, if we go by the tutorial's temperature placement guide, the sea level temperature in all these areas falls into the same category (Hot: 22 to 28 °C). Although I guess Tibet is at least partially Very Hot: 28 to 35 °C, which does alleviate the problem somewhat. Still, it's pretty tricky to come up with a system that assumes equal lapse rates everywhere, but still delivers acceptably accurate results.

Edit:
Actually, I did a bit of mistake with the low tundra line of the Venezuelan Andes (comes with using relatively low res data and elevation maps split into zones). I checked with a DEM-generated elevation map split into 25 metre intervals, and actually the tundra line is closer to 3500 metres, even in Venezuela (about 3250 metres). Well, spotting that mistake definitely raises my hopes that something of a "reasonably accurate universal guideline" might actually be possible to create .

Edit2:
Updated more accurate altitudes for the Tibetan Plateau, and added data for the Peruvian Andes as well. It does seem that the tundra line is indeed considerably lower in the northern Andes.

[Azélor]

Problem number one: using categories makes things less precis.
The 22 to 28 degrees is a pretty big range of temperatures considering an average lapse of 6 or 7 per 1000m increase in altitude.

I mean, if the base temperature is 22 at the base it reaches 0 at 3142m, assuming a lapse of 7.
At 28 at the base, it reaches 0 at 4000m, also assuming a lapse of 7.

Thus all you examples are valid.

Problem number two, 7 is just an average number for general purposes. If my memory is good, it usually range from 5 to 9, or maybe more if it is really dry.

[Charerg]

Problem number one: using categories makes things less precis.
The 22 to 28 degrees is a pretty big range of temperatures considering an average lapse of 6 or 7 per 1000m increase in altitude.

I mean, if the base temperature is 22 at the base it reaches 0 at 3142m, assuming a lapse of 7.
At 28 at the base, it reaches 0 at 4000m, also assuming a lapse of 7.

Thus all you examples are valid.

Problem number two, 7 is just an average number for general purposes. If my memory is good, it usually range from 5 to 9, or maybe more if it is really dry.

Note that in my examples the "tundra line" is actually 10 °C (since a climate with maximum mean temp below 10 °C is classified as ET in Köppen), not 0 °C (that would be the "glacier line", or EF).

From what I've calculated, when it comes to the effect of altitude on monthly mean temperature, the ratio definitely tends to be below 7 degrees. To follow up on my example about the Tibetan plateau, it hits the "glacial line" (mean temp below 0 °C) at roughly 5350m. If we use only the "known intervals", the Hot (28-22 °C)-Warm (22-18 °C) line is about 1700 metres in the Tarim basin (so, 22 °C July mean temp at 1700m).

So, the monthly mean temp drops by 22 degrees in 3650m (5350m - 1700m). Assuming a linear lapse rate, that's about 166 metres per 1 °C, or about 6 °C per 1 km. And often the temperature drops even more slowly, especially in lower altitudes. The Brazilian Highlands in January have about 200 metres per 1 °C, so 5 °C per 1 km. There seems to be an overall tendency for the lapse rate to increase gradually with greater altitude (due to increasing dryness of air perhaps?).


Edit:
As another example, in the Venezuelan Andes in January (which is slightly warmer than July), the altitudes between the different zones are as follows:
Hot-Warm boundary (22 °C): 1100m
Warm-Mild boundary (18 °C): 1750m

This gives us an approx. 1°C per 163 m, or ~6 °C per 1 km lapse rate between 1,1 and 1,75 kilometres.

Mild-Cool boundary aka "ET line" (10 °C): 3250m

This gives us 1°C per 188 m, which is about 5.3 °C per 1 km between 1,75 and 3,25 kilometres. Strangely enough in this case the lapse rate was seemingly faster at the lower altitude. That said, these are somewhat approximate numbers.

Note that since the surface level is Hot (28-22 °C), we have at least 183 m per 1 °C (~5.45 °C per 1 km) from 0 to 1,1 km (and could be 1 °C per 275 m [~3.6 °C per 1 km], if the sea level temp is 26 °C, as an example).

[Gridello_Aspart]

i need help that tutorial make me so confuse i cant figure out where its dry or where its not pls HELP i m not even sure i did other parts right i really need help i read other tutorial i understand it well but it always crated some blanks so i tried that one but it makes me confuse

[Charerg]

I ended up a learning a bit of Script-Fu in GIMP in order to automate the climate classification process (saving myself and others a lot of work in the future, I hope). For now, I've used the 8-step precipitation system introduced in my prior modification to the tutorial. Although I might also a create a version of the script for the original 6-step precipitation system at some point.

Here are the instructions for using the script:

Installation:
Place the script in the appropriate folder (usually /User/gimp-2.8/scripts). If uncertain, you can check Edit->Preferences->Folders->Scripts to see where the scripts are stored. Once the script is in the right folder, the script should be availabe (you can use Filters->Script-Fu->Refresh Scripts so you don't have to restart GIMP). Oh, and remember to extract it from the .zip file before use (but you knew that, right? ). You should now have the script available under the Image tab:

Activating the script.png

Restrictions for using the script:
- This has been written for and tested in GIMP 2.8, it might work with other versions as well, but I can't guarantee that
- The image needs to be RGBA (RGB with an Alpha channel)

Layer naming restrictions:
The temperature/precipitation layers need to have exactly the following names (the script searches for them by name and duplicates them in order to work out the climates):

JanTemp
JulTemp

JanPrec
JulPrec

Layer colouring restrictions:
The temperature and precipitation categories need to have exactly the following colours:

Temperature zones:

 

Temp Category	R	G	B
Severely Hot	160	0	65
Very Hot	210	60	80
Hot	245	110	65
Warm	250	175	95
Mild	255	225	140
Cool	230	245	150
Cold	170	220	165
Very Cold	100	195	165
Severely Cold	50	135	190
Deadly Cold	95	80	160

The temperature zones in a slider:
TempSlider.png

Precipitation zones:

 

Prec Category	R	G	B
200+ mm	135	0	180
140-200 mm	130	60	200
70-140 mm	105	70	200
40-70 mm	60	60	180
20-40 mm	70	95	150
10-20 mm	55	85	100
5-10 mm	35	50	50
0-5 mm	20	20	20

The precipitation zones in a slider:
PrecSlider.png

Processing time:
I've somewhat simplified the climate generation process by merging many of the temperature combinations (all combos that result in Cb, for example), in order to speed up the script and also reduce unnecessary complexity. Still, I've only tested this on approx. 4000x2000 and it does take maybe 30-60 sec to process that. So it may take a while for GIMP to go through all the operations if you have a very high resolution map.

Sample map:
Here are some sample maps I used to test the script. These are generated from WorldClim's 1970-2000 dataset, although I haven't cleared any artifacts from them.

Prec samples:

 

JanPrec.png

JulPrec.png

Temp samples:

 

JanTemp.png

JulTemp.png

Generated climates:
ZonesOut.png

Right, that's it for this post. The script itself can be found in the attachments. Any feedback about using the script or the generated climate zones is welcome of course.

Edit:
There's an updated version of the script available in this post.

[Azélor]

I have attached the files I was talking about earlier if you want to have a look.
It's strange that Psd files become much smaller if you turn the layers off.

Chareng.zip


I have to say that I'm impressed by the compression ratio. Normally it's in the range of 5%, here it's around 75%.


Also, Unless I missed something, you added more precipitation categories but haven't explained the difference this creates when placing them.
I'm not sure that the cutting halfway technique give very good results. The progression from one category to the other is not clear, even in my tutorial.
I find it not too difficulty to know where it rain and where it does not, the extremes. Yet everything in between is unclear.

The problem comes from using odd numbered categories. The categories are not bad per see but they are not helpful to understand the precipitation spread.
We should do another map for reference using 5 or 10 ml categories. It will be easier to see the progression.

[Charerg]

I have attached the files I was talking about earlier if you want to have a look.
It's strange that Psd files become much smaller if you turn the layers off.

Chareng.zip


I have to say that I'm impressed by the compression ratio. Normally it's in the range of 5%, here it's around 75%.


Also, Unless I missed something, you added more precipitation categories but haven't explained the difference this creates when placing them.
I'm not sure that the cutting halfway technique give very good results. The progression from one category to the other is not clear, even in my tutorial.
I find it not too difficulty to know where it rain and where it does not, the extremes. Yet everything in between is unclear.

The problem comes from using odd numbered categories. The categories are not bad per see but they are not helpful to understand the precipitation spread.
We should do another map for reference using 5 or 10 ml categories. It will be easier to see the progression.

That is a very good idea. Actually I kind of think both your original (extremely detailed, certainly more so than my calculations which used purely averages) and my precipitation threshold calculations are to some extent meaningless because the precipitation categories we use are so broad. For example, one temp combo has a threshold of 800 mm and another has 700 mm, and they both have the same precipitation pattern. Does the difference matter if the annual precipitation could be anything between 500 to 1000 within that same precipitation pattern?

This is actually a big reason why I decided to merge many of the temperature combos (within the same climate class) in the scripted version for GIMP (besides making it faster to write and easier to modify): I felt there wasn't any advantage to treating them separately as opposed to lumping them together and just using an approximate average threshold to determine whether it's an arid climate or not.

As to placing the precipitation categories, I admit that I don't have a really great way to place them myself. What I personally do is a combination of guesswork and looking at the example maps of Earth's precipitations for reference. As well as taking into account the general precipitation patterns covered in your instructions. At the end of the day it's an approximation, if only because we're working with precipitation data from only two months. Although I'd say that using more precipitation categories does have one distinct advantage: the sheer volume of combinations gives you more room to create a transitional BS stage between BW climates and more humid ones.

But making an example precipitation map using linear intervals for reference is indeed a great idea. And should be easy to generate too with a gradient map. I'll try to post some during the weekend.

[Azélor]

Well I already started

It's been a while I haven't used Qgis.
Here's a basic map for July showing the area with precipitations above 500ml.

test.png

I was going to subdivided the precipitations in 5ml categories but the highest value is 2381 ml.
Yep almost 2,4 m of rain in a single month. Almost 4 times what London receive in a year.
The maximum in January is barely over 900 ml. July is much more rainy world wide.

I'm trying to find a threshold where I can stop adding new categories. Beyond a point where increase in precipitation don't have any impact. Stopping at 300 ml still give 62 categories. That might still be too much unless I increase the resolution.

[Azélor]

I made a custom gradient in Qgis for each 5 ml
Yes, the pure black area receives exactly 0 ml of rain.

It's fine until I reach 200 ml. Then I ran out of colours. 200 ml is the saturated pink-red and the colour of the other category following it is salmon.
Past 200ml I made the categories 20 ml each. Til 500 when I ran out of desaturated colors to use. Everything above 500 ml is radioactive green.
I'm not sure how it look (psychedelic?). I hope we can easily differentiate the categories. Any tips for improvements are welcome.

The problem with the in built gradient is that they appear to follow a linear shift from one colour to the other. But the human eye is less sensitive to some colour, therefore I could only count a third of the colour used by the program.

January. It look like you will might need to download the map to see it properly. I'm using Firefox and by default the zoom is not good enough. I also have a zoom extension that allow to zoom further but he makes everything blurry.
precipitation 2 jan.png

July
precipitation 2 jul.png

These were made with the 5 minutes resolution maps.


But here's a comparison of Korea at the 3 different resolutions

10m
korea 10m.png

5m
korea 5m.png

2,5m

korea 2m5.png

and I'm downloading the 978 mb file for the 30s resolution. Just to see how much more detailed it is.

korea 30s.png