Reader enragedparrot asks the rather sensible question, which appears to have been somewhat neglected in the vast war of words of 2035, 2350, and quite what is the source for what: if 2035 is badly wrong, what is the right date?
The answer, of course, is that I don’t know. But I may be able to tell you something useful along the way. If you’ve seen a better answer, please point me at it.
[Dragged from the comments: http://web.hwr.arizona.edu/~gleonard/2009Dec-FallAGU-Soot-PressConference-Backgrounder-Kargel.pdf is excellent -W]
So (forgive me, to clear more wrong stuff out of the way) 2035 is wrong, and there is a 2350 date bandied about. Unfortunately, the text of the report that provides the 2350 number is, as I’ve said before, utter junk. So that doesn’t help.
Lets go back to physical reality for a bit. The lapse rate is about 6.5 K per kilometer – which means, all else being equal and taking the rough with the smooth, that if you go vertically up 1 km it will get 6.5 K colder. So, again using a very broad brush, if a given patch of snow is going to just-not-melt all through the year at a given altitude, then about 200 years from now it will just-not-melt about 1 km higher up (this ignores enormous possible changes – precipitation is obviously very important, and shifts in the monsoon could have a massive effect on the glaciers – but that is yet more complexity). Since there is quite a lot of “higher up” available in the Himalayas, that pretty well guarantees there will be *some* glaciers there in 200 years time, barring absurd levels of GW. Attempting to predict more than 200 years ahead is a waste of time, so I won’t try (and I don’t say much in favour of 200, either).
To work out exactly how much, you’d need to run a super-spiffy hi-rez GCM, and we don’t have those. More plausibly, you could take a larger-scale warming prediction and feed it into a GIS, which would know the height and accumulation rate of all the existing glaciers, and then make a guess at their future response. As I recall, there are some Gregory / Oerlemans papers doing an even broader-brush version of this, but I’ve forgotten the details.
Was that vague enough for you?