Judith Curry writes: we are obviously interested in the implications of this SST issue for hurricanes.
Obviously, the first thing that needs to be done is develop the new time series of SSTs. Doing so will likely make the arguments over historical storm counts look simple and tame.
The figure at this link provided by Steve:
… suggests that some think the ocean temps will get bumped up for a brief period. Others have said that the ugly observational bump above the model spread in the IPCC reports will get pushed down. Steve makes a strong case that there is a longer term influence that needs to be considered (a convincing case IMO). It will be easy for people to pick the adjustment that best serves whatever argument they want to make — hurricanes, global temps, etc etc.
Thus, what is most important is that the discussion on the dataset take place first, and Steve’s blog provides a valuable public forum for a discussion.
Hadley Centre needs to open up their database and its “adjustments” to public scrutiny. Steve’s further suggestion that temperature databases should be maintained outside of the team is also important.
Thanks, Steve/Judith for this thread, it’s an interesting issue. I plan to prepare several plots this weekend (when I return home) incorporating the Hadley assumption and the Carl Smith assumption, to see what the tropical Atlantic SST would look like if the those adjustments indeed apply on a regional basis. One question I have is whether any adjustment should end by 1980, when satellite data began to be incorporated into the global SST estimates.
Kenneth or Bob K, if you have time and want to incorporate the possible adjustments into MDR and other tropical Atlantic SST subsets to see what they look like, that’d be great for people at CA to see.
ooops, there goes a correlation, or two, or three. Or maybe a big bunch of them. You’ve got to watch out for those trivial local curve fits to large-scale, meta-functional data using fancy distributions and stats and stuff. They’ll always come back to bite you in the behind. Another important area that could use some causality.
Full Disclosure: I am not a Certified Climatologist, don’t know much about hurricanes except when their names are in the daily news, and know even less about stats.
Steve McI can snip at will, of course.
David, since I have convinced myself, for the time being at least, that when the TC/hurricane count and ACE index for the NATL long time series are adjusted for detection capability changes and cyclical components, like AMM, the trends that could be attributed to SST changes are not evident, I do not have much motivation to look at SST corrections — unless it turns out that the SST has been flat over the long term.
I am interested in what the “adjusted” SSTs would look like. I have a link below that has a breakpoint for the global mean temperature (GMT)around 1945 and I believe various methods of measuring breakpoints show one at that approximate time. These breakpoints have been discussed here at CA previously with constant reminders from Steven Mosher. Breakpoints are seen additionally near 1900 and 1980 for the GMT.
Is there any data set for periods prior to the 1980’s that is capable of generating the level of certainty we need? Is there currently a data set in development that will in the future give us the level of certainty we only wish we had now? What are we asking of a data set?
re 5. Thanks Kenneth.
Another thing to note when looking at model hindcasts is to note that
they arent very skillful at modeling “non volcanic” cooling periods.
That’s probably a comment for a different thread.
On Unthreaded, Dan Hughes links to the latest in tardy ignorance from the New York Times linking AGW and worsening tornadoes.
I adjusted Kaplan SST for the MDR. The adjustment started at +0.3C and linearly declined to zero at 1980. This is sort of a pseudo – Carl Smith adjustment.
This adjustment improved the correlation ( r ) between Atlantic PDI and SST from 0.50 to 0.52 Not much of an impact.
The adjustment to Kaplan SST improved the correlation between western Atlantic sea-level pressure and SST to -0.59, from -0.42.
(Carl, by the way, I hope you don’t mind me naming the adjustment after you. I think you’re the reader mentioned by Steve M. in the earlier SST post.)
Time for a few plots. Here’s the unmodified Smith-Reynolds Extended Reconstruction SST for the Atlantic MDR:
and a modified plot, incorporating a warming of +0.3C in 1945 which linearly declines to zero by 1980:
Here’s a plot comparing unmodified Kaplan and Smith-Reynolds SST for the Atlantic MDR:
That’s a pretty significant difference between the two. If I’d used Kaplan instead of Smith-Reynolds I’d get a 1950s peak pretty close to the 2000s peak.
Which is more appropriate for tropical Atlantic SST? I don’t know but I’d sure like to see the detailed derivation of the SST differences for random example years, say 1970, 1985 and 2000, just so I could understand the mechanics of the derivations.
Kaplan uses ship reports only while other sets like Smith-Reynolds use buoy and satellite data. Alexy Kaplan gives a pretty good comparison of his vs others here . His comment #6 is interesting
There is also an important issue of SST trends. Jim Hurrell and Kevin Trenberth in their 1999 paper suggested that the analysis which conservatively assumes the stationarity of the mean SST climatology (like ours) might be underestimating the long-term changes in the SST, and argued that it is better to figure out the long-term variability beforehand and prescribe it a priori (e.g. like it is done in the HadISST1). Their criticism regarding long-term trend underestimation is certainly valid, as our preliminary study has shown; however, prescribing the global trend a priori might result in overestimation of it by the analysis — not a good outcome either (Kaplan et al. 2001).
Looks like a topic for another time.
Kaplan offers a cool/neat/hip/awesome plotter, located here , for a sample month. Simply plug in a latitude and longitude box and date, then hit plot. The result is a map of ship SST observation densities (# of observations per month in a 5 degree by 5 degree box).
An example for the tropical Atlantic is below. The red box is the MDR:
Other samples for September are below:
Note the lack of SST observations (and probably weather observations in general) in 1945.
The Hurrell Trenberth paper mentioned by Kaplan is here . The magnitude of assumptions and technique differences used to create SST time series is remarkable.