[mod: Bender, cut out the name games]

http://climexp.knmi.nl/start.cgi?someone@somewhere

At this point all I can say is: Geert, we laypersons love you man.

With data in hand for the 1958-2000 period from KNMI, I was able to verify the wide ranging and localized trend data in the Kaufman Arctic are defined as 60N pole ward.

After some thought on the matter, it is my view the composite plus scale (CPS) method used by Kaufman et al assumes at the very start (standardizing the proxies from the various Arctic locations) that the proxies are all reacting to the same climate conditions, i.e. assumes the Arctic climate is uniform over the entire area.

I have also thought that the further the actual climate deviates from that assumption the more uncertainty that that development adds to the final reconstruction results. I am not aware of a method of calculating that added uncertainty.

Another assumption in K09, I trhink, is that the summer months (JJA) anomalies are perfectly correlated to the annual anomalies. I suppose for a purely scientific paper this assertion would not be necessary, but that the paper’s authors were probably aware of the policy implications and that requires being able to say something about the annual anomalies.

The K09 authors being keenly aware of the assumptions noted above attempt to show evidence to support them. They point to the correlations of summer month anomalies to the annual ones for all of the Arctic area. In my mind, the correlations should be broken down by some logical subdivisions of the Arctic. To that end I show these regional correlations in the table presented below.

The K09 authors do not address the Arctic local variations in trends other than to show the trend color coded map in their paper with no discussion of the import of that variation to the uncertainty it must add to the reconstruction. Below in the table noted above I added larger area regional trends for the Arctic.

The authors also point to the CPS reconstruction showing an almost straight line decline for 2000 years and then claim an upward swing in the HS blade for the final 50 years that would correspond with presumed AGW effects. I have shown in previous posts that breakpoints in the reconstruction and individual that contribute to the composite do not correspond to the authors claim for the most recent 50 years. I made a proviso that annually resolved data, unlike the decadal resolution presented in K09, would perhaps make end point break points easier to detect. With that in mind I found rather complete monthly data in KNMI using CRUTEM 3 data for the period of 1898-2008.

I did break point calculations in R as documented previously on that data for the spring (MAM), summer (JJA), autumn (SON) and the winter DJF) months and for the combined annual series. The graphs of the series with breakpoints are shown below. Note that while perhaps one of two break points found in most of the series could be construed to fall close to the 50 year before present time, the direction of the segment slopes are in the wrong directions. The instrument period does not appear to agree with the K09 thesis and I believe that Lucia has found this same inconsistency. I personally do not see the K09 conclusion consistent with their reconstruction results.

In this post I merely wanted to make my points about what I see as the K09 assumptions implicit in using CPS methodology about instrumentally measured temperature anomalies and the added uncertainty when analyzed regionally in the Arctic and not averaged over the entire Arctic as the K09 authors did. The break point inconsistency becomes clearer with the use of instrumental data that are resolved annually.

Currently I have no reason to doubt the map.

Neither do I. What I am questioning is the location of the proxies. For a strong calibration (where you know the proxy really is a proxy) you want your samples to be located in strongly contrasting areas in terms of instrumental temperature trend (whereupon you run the risk of falsely conlcuding that trends in the proxy that match the instrumental record are tempearture-related, but you have the advantage that if there is no match, then you truly don’t have a proxy.) If you locate the proxies in areas with no trend you have to hope that annual temperature fluctuations are highly variable and that your proxy responds instantaneously quickly to temperature.
.
Locating the samples in the red and blue pools maximizes your chance of falsifying the proxy’s validity at the risk of over-estimating its sensitivity. (At this point I think we’re far more concerned that these things just are not temperature proxies than we are about possibly inflated sensitivity coefficients.)

My preliminary calculations, using KNMI station data, show that, indeed, the summer temperature trends are significantly smaller than those for the entire year and that the summer months in the Arctic can have negative trends over the entire 1958-2000 period.

These calculations using the KNMI data are slow going. ERA-40 access would require much less effort. I will sample the 60N-70N data and then determine a fit with the map. Currently I have no reason to doubt the map.

Are you referring to a high resolution Digital Elevation Model?

Yes.

I have a difficult time accepting that the very wide ranging negative to positive and localized trends over 43 years is the proper interpretation of the map

I was going to ask: why?
.
But then I reviewed the map, knowing that you are pretty careful even in your casual observations. And I have to agree. It is interesting that the strongest instrumental trends occur in areas where there are no proxies. The centre of Hudson Bay, for example. It would be good to plot the weather stations on this map to confirm that the coloured blobs are indeed centred on those station locations. Curiously, in the areas where the proxies are located, the instrumental trends are very weak. #1, #17, #22 might be considered exceptions. For a proper calibration you want to examine the areas with the warmest vs. the coolest trends. This obviously was not done. How many of these “proxies” consistently show the correct, opposing pattern when they are chosen from those anomalous little blue patches?

I suspected that the only way to use spatially sparse data to obtain the detail of the map would require kriging. Are you referring to a high resolution Digital Elevation Model?

I wanted to look at the data used to construct the map because I have a difficult time accepting that the very wide ranging negative to positive and localized trends over 43 years is the proper interpretation of the map – even though it seems straight forward.

Re: bender (#87),

As I recall, in pondering the use of ERA-40 data, the 1958-2000 period was the available reanalysis period from that series. Evidently a legitimate scientist can obtain the ERA-40 data series free, while non-scientists have to pay for it. I’ll look at KNMI data in the meantime.