This is Robert Grumbine's Typepad Profile.
Join Typepad and start following Robert Grumbine's activity
Join Now!
Already a member? Sign In
Robert Grumbine
Recent Activity
Neven, I was talking with Jim Overland this morning about the SIO. Didn't quite ask your question, but am pretty sure that the answer is more in the vein of 'the more the merrier' than 'professionals only'. I'll try to get him tomorrow about it. Generally, I think the SIO is more interested in method than in answers. For that reason, a popularity poll is less interesting than a logistic regression (my approach for simple-minded statistics) or Gompertz regression (wasn't that mentioned here?). More interesting would be a method that makes some use of model outputs -- say something that you developed that used the PIOMAS or PIPS outputs to provide your estimates. Or, as done by Wanqiu Wang, the CFS at NCEP (Coupled Forecast System). (Wanqiu, Xingren Wu, and I discussed some hows for that estimate; Xingren is still working on runs for his December 2010 estimate. And we've got a paper or two in mind to come from this and related runs, not least, the reason it's a December 2010 estimate. More, one of these days, over at my blog when I get back to blogging.) My statistical method, I'm pretty confident now is biased low. So I think it's relatively unlikely that the cover will below the 4.4 +/- 0.5 of that estimate. Wang's model/statistical method (it isn't straight model output; there's a statistical adjustment -- how to do it is what we were discussing) is probably biased high, at 5.0 +/- 0.5. What I think is likely to be the best is the model runs plus different adjustment method that Xingren is currently working on. My intuition tells me that'll come in around 4.7. One comment of mine that Overland is taking under advisement is to make it more prominent what we're trying to guess -- the September monthly average extent, as computed by NSIDC.
Robert Grumbine is now following The Typepad Team
Jun 20, 2011