I didn't make my own, I used the first website that came up when I searched "<language> letter frequency".

Dutch would be a bit trickier. Here's my best shot.

Dutch letter frequency: e n t a o i r d s l g h m v b w k u p c ij z j f é x

The home keys should be e n t a o i r d. Is ST common in Dutch? If not, you can switch D and S. By the way, what's IJ? I found it here: http://www.cryptogram.org/cdb/words/frequency.html

B is in a bad place. It's too common. Switching it with F would be an improvement.

There are already a few things that I like better about Colemak in Dutch. I think in English, K is too rare for the spot it has. But in Dutch, it's just about perfect. Also C is rarer than in English, so that annoying spot down there is easier to reach.

Considering how much more common G is than P, a G-P switch would help.

q w b g p j l u y ;
a r d t s h n e i o
z x c v f k m , . /

I agree with Tomlu. Furthermore, nowadays more and more users are typing a substantial amount of English no matter where they're from. This creates a difficult situation for a would-be non-English optimizer.

I'm leaning towards making one good English layout, leaving a few keys open for adding national characters and praying that the result is good enough for almost all. If we could reduce the layout jungle somewhat I envision a lot of benefits coming from it. I shudder at the thought of the Chinese factories making separate keyboard batches for Norway, Denmark, Sweden, the Faeroes, Iceland and who knows whether they have a Greenlandish one as well...

"ij" is a common ligature in Dutch, sometimes even considered a single character (like in the website you referred to), and pretty hard to type on Colemak.  j itself is also quite common, mostly in combination with vowels: "jij" (you), "ja" (yes), "je" (you, but with less focus).  I think it would be beneficial to swap j and y, as y is very uncommon in Dutch.  But then again, I probably type as much (if not more) English that Dutch, cfr. DreymaR's remark.

I see a lot of potential in your suggested d-s swap though.  According to your source, d is slightly more frequent than s (eg. "de" means "the"), and "dt" is a common and annoying same-finger digraph on Colemak.  Also d and t are easily mixed up (quite impossible with Azerty or Qwerty) which leads to embarassing so-called "dt-errors" in Dutch (since d, t and dt are the most common suffixes in verb conjugation).

DreymaR: yes, I agree with "better have one good international layout (with perhaps minor local tweaks per language) than hundred ultra-optimised national layouts".  I was just looking for an indication on how "close to ideal" Colemak is for Dutch vs. for English. :-)

The visual representations of key frequencies you referred to are indeed very nice, although I find the original black/grey/white images a lot clearer than the ugly "blackbody" versions.  Btw, they also appear on the Spanish Wikipedia page on Colemak (which still exists).

Yes, that's a good argument!  Perhaps I should actually make this change to my layout.  Shai, what's your opinion on swapping j and y?

Definitely: List of languages by number of native speakers (Wikipedia).

Well actually, Y has surprisingly low same finger with most consonants. Y should not be paired with L, R, O, E, or A. <EDIT: or M. end edit> Everything else is fine. NY/YN is about as common as FT/TF or RL/LR. However, N has L on the same finger, and LY is to be avoided. Of course, these stats are for English. I'm not sure how it works out for other languages.

If jij is common in Dutch, switching the Y and the J without any further changed wouldn't be great, as it would be a multiple same-finger combo on a weak finger. Also the word "my" would be very uncomfortable to type.