date: Thu, 15 Jul 2004 06:41:09 -0600 from: Tom Wigley subject: next to: Sarah Raper Sarah, I will cover the new points and add the refs. They are important in case we get some of these authors as reviewers. It is hard to add text specifically to knock the TAR method. One problem is that it should be a differential equation at the outset. My addition changes it to a differential equation -- but in a rather roundabout way. The reason that the TAR method is OK is that if the time scale is >> 100 years (as it seems to be) then the d/dt term drops out. Praps I can add a little bit about this at the end. What is annoying to me is that the TAR method is conceptually flawed and it happens to work not becoz of cleverness and forward thinking by Jonathan, but by a fluke. Oh well. Jane Leggett, who is funding some of my MAGICC work, wondered about sensitivity proportional to remaining V instead of A. To test this I can use proporional to V**n for different 'n'.. Then we could cite the J. Glac. paper too. This is only one line of code. What 'n' range? Here is what I said to Jane ..... ------------------ > The linear with volume vs linear with area is something I will look > into. There > is a paper that Sarah is first author on in J. Glaciology a few years > back where > we note that one can use A = V**n. I can put this in the code to see the > sensitivity to 'n'. My intuition says it will be small. To 2100, > volume has no > effect, so 'n' cannot be important. For large times the paths must > tend to the > initial volume, so 'n' can have no effect there either. Still, it is > easy to do the > full sums to check this out. Good point. ---------------------------- Tom.