cc: Phil Jones , "Willebrand, Juergen " , Nathan Bindoff , Mote Philip , "Solomon, susan" , Martin Manning , John Kennedy , "Parker, David (Met Office)" date: Fri, 05 Jan 2007 10:15:08 -0700 from: Kevin Trenberth subject: Re: SPM-3 Error estimates to: Peter Lemke Peter I am not sure this is right. Here is the series of emails about this: You will see that we were aware that the error bars were supposed to become comparable and ultimately this was left in the hands of John Kennedy and David Parker. I'll cc them. Kevin Hi all There has been a flurry of emails on this. I believe the action now is with John Kenedy and David Parker who are the only ones(?) with all the data sets and in a position to try to make the plots and error bars comparable. So let's see what issues arise from them. Thanks Kevin Phil Jones wrote: Nathan, See the other email trying to get the snow plot to look similar, i.e with error ranges. The temperature plots includes all the measurement errors and biases at the decadal scale. Also, if you look very closely at the temp plot you'll see that the error band is very slightly asymmetric, as the urban error is one sided. Look closely at the present - it is just about visible! Also all our errors are the known errors. There is no assessment of unknown errors. So just like Rumsfeld we haven't made estimates of these, although I think for his case they could have made some assessment of the unknowns. We can't as they really are unknowns. This range is 5/95%, so 2*SE times 0.8225. Cheers Phil At 00:44 20/07/2006, Nathan Bindoff wrote: G'day Kevin The graph looks fine. I would like to confirm that the error bars have been increased from standard errors to 90% confidence intervals by scaling by ~1.66. A note of caution, as we revise our chapter we may well increase the confidence interval to include uncertainty associated with unresolved decadal variability. The current confidence estimates are really estimates that come from the processing of the tide gauge data and the method and donot take into account decadal variability in climate processes. Also to confirm, the shaded area on the global air temperature is a confidence interval representing what?, instrumental noise, decadal variability and other unresolved process? Cheers Nathan Kevin Trenberth wrote: Hi All Dennis Shea reprocessed the sea level record to 1) Rezero the values relative to 1961-90 2) Replace the wisker plots with an error range more that the yelloe one in the sfc T plot. This does not have a decadal filter run through it though. Kevin Phil Jones wrote: David, I guess we could but it is just one more thing to remember to alter in January. So let's not do that. Cheers Phil At 12:27 19/07/2006, david.parker@metoffice.gov.uk wrote: Phil Thanks. I don't remember a newer version but John Kennedy will confirm. Were we supposed to add 2006 to the top panel in due course? David On Wed, 2006-07-19 at 11:53, Phil Jones wrote: > John, David (and Nathan), > I'm not sure where we are on this figure for the SPM. I'm > reattaching the latest version I had emailed to me. Anny has already > sent some > sea level data. Not sure how these in the zip file compare with those used? > I was only acting as a go-between on this figure, hoping that John > Kennedy would be able to draw it. I think it is John who produced the > attached. > John - can you check whether these files differ? > > This means that the sea level is now OK. So what about the snow area? > Has that got uncertainty estimates, Richard? > > Apologies if I've misunderstood anything here. Put it down to the > stifling heat here today. Supposed to be in mid-30s C somewhere in > the UK. Ch 3 has been discussing extremes. Today is HOT !!!!!!!!! > > Cheers > Phil > > > Cheers > Phil > > > > At 01:40 19/07/2006, Nathan Bindoff wrote: > >G'day Kevin > > > >Kevin Trenberth wrote: > >> > >>------------------------------------------------------------------- ----- > >> > >>Subject: > >>[Fwd: Re: IPCC diagrams FAQ fig] > >>From: > >>Kevin Trenberth > >>Date: > >>Wed, 12 Jul 2006 14:23:39 -0600 > >>To: > >>Nathaniel Lee Bindoff , Richard Alley > >> > >> > >>To: > >>Nathaniel Lee Bindoff , Richard Alley > >> > >>CC: > >>Phil Jones , david parker , > >>Martin Manning , Susan Solomon > >> > >> > >>Nate, Richard > >> > >>I am not sure who is in charge of this? > >Actually I thought Phil Jones was in the sense of creating the plot > >(with John Kennedy). Richard and I were to provide the data. The aim was > >to get the same look and feel for all three data. > > > >I have attached the same data that was used to make figure SPM3. > > > >The challenge for us was figure out the of adding the 5-90% shading > >representing the decadal noise to match the same figure in the upper panel. > >Is it true that Phil (with John) created that panel, and also the shading > >represening the decadal noise. And if so could he do it for the Sea-Level > >Data I have attached? Phil? > > > >In which case we have all the data necessary to complete the sea-level > >figure. Note that there is an arbitary level that needs to be added to > >each time series to get them to overly each other. The reference period > >of 1961-1990 can be calculated for both of the Holgate and Church time > >series. The altimeter data should be just adjusted to lie on top of the > >other two time series. > >>Please see the discourse below and the attached figures. The second one > >>is from our FAQ and has the global T with now 2 scales: anomalies from > >>1961-90 left and absolute values at right > >>It seems desirable to have the zero correspond to 1961-90 for all 3 on > >>the left scale and maybe (or maybe not) add absoluite values on right > >>(for sea level this would be 3,800 m or so, i.e. 3,800,000 mm??) > >I wouldn't bother with the absolute scale on the right side of the panel > >for sea-level. > >> > >>Anyway if you can provide the value on the left axis that corresponds to > >>the 1961-90 value we could do the offset. > >>Opinions? Should the zero line be drawn in each case? > > > >See above > > > >Is this okay Phil? > > > >Over to Richard > > > >Cheers Nathan > >> > >>Kevin > >> > >>-------- Original Message -------- > >>Subject: Re: IPCC diagrams FAQ fig > >>Date: Sun, 9 Jul 2006 09:17:19 -0600 > >>From: Susan Solomon > >>To: Kevin E Trenberth > >>CC: Phil Jones , david.parker@metoffice.gov.uk, > >>Martin.Manning@noaa.gov > >>References: < > >> > >> > >> > >>Dear Kevin et al., > >>Thanks for the info on this. I understood that there would be some > >>changes in the sea level and snow data presentation by those chapters, > >>for consistency with the atmospheric data normalization period, error > >>bars, etc. So representatives of those chapters will need to be in the > >>loop for this figure to reach final form. Please contact Bindoff and > >>Alley, respectively. > >> > >>Thanks, > >>Susan > >> > >> > >>At 1:15 PM -0600 7/8/06, Kevin E Trenberth wrote: > >> >Susan, David etc > >> > > >> >David thank John Kennedy for the 3 panel figure for the SPM. You sent me > >> >an eps and I have converted to a png: this is the second one labeled > >> >Figure... > >> >I have modified it a bit, changing axes and labels in the other version. > >> > > >> >Kevin Peter Lemke wrote: > Dear Friends, > I hope that you all had a good start into the New year and I wish you > all the best for the remaining part of it. > Concerning the error estimates in the figure SPM-3 the comment SPM-249 > is correct. I got the following comment from Phil Mote about the error > of the snow data: > > "As I recall, I calculated the standard deviation of the residuals of > the whole > time series after subtracting the smoothed curve, so it's a little > different from decadal means. We had to do this because we have no > *error* estimate of the underlying data. So the reviewer is correct in > that the snow panel is not strictly consistent with the temperature > panel." > > I think we HAVE TO unify the error shadings. Given that the errors of > single data points are all different and - at least in the case of the > snow data - not known, I wonder if it still makes sense that somebody > (Kevin, Phil J.) takes all time series and calculates the errors in > the same way. If we stick to decadal data (so that the widths of the > error range is changing with time) then there are only 10 data points > for the snow series to compute the variance from. This is not much, > but acceptable in this case, I think. > > Has this already been addressed by somebody? I am on vacation until > Monday and I am not really up-to-date. > Best regards, > Peter > -- **************** Kevin E. Trenberth e-mail: trenbert@ucar.edu Climate Analysis Section, www.cgd.ucar.edu/cas/trenbert.html NCAR P. O. Box 3000, (303) 497 1318 Boulder, CO 80307 (303) 497 1333 (fax) Street address: 1850 Table Mesa Drive, Boulder, CO 80305