2010-10-10 00:13:29Basic 132: DMI shows Arctic cooling?
Peter Hogarth

peter.hogarth@geoacoustics...
86.152.121.48
I guess the most basic version is "no it doesn't", but that's no fun. I got a bit disheartened after starting, thinking surely no-one believes the Arctic is cooling?  In the end I did an intermediate and advanced version that probably act as rebuttals for other points too and vindicates the GISS Arctic data.  Comments welcome.
2010-10-10 02:29:03wouldn't it be appropriate
nealjking

nealjking@gmail...
84.151.40.134

to say what "DMI" means somewhere?

 

and what does this mean:

"Firstly it uses Arctic Summer "above zero" temperatures in isolation. It is well documented that the Summer maximum temperatures above ice, and thus above 80 degrees North are constrained to just above “zero” by the ice melt temperature,"

 I'm not getting it.

2010-10-10 04:16:48
doug_bostrom

dbostrom@clearwire...
184.77.83.151

The usual comprehensive Hogarth job. 

Peter, the first sentence of the Intermediate treatment goes:

"Danish Meteorological Institute (DMI) Arctic air temperatures and the Goddard Institute of Space Science (GISS) Arctic Surface Temperature Anomalies."

Might be better to cast those explicitly as definitions; because they appear as though they're the first paragraph they're a bit confusing.  

Regarding Neal's point, I took away the idea that a thermometer sitting in a thin pool of cold air in contact with a mass of ice is going to remain close to 32 degrees until all the ice is gone. Same deal an iced drink.  Needs a mental picture so perhaps that's the comms issue there.

Can't say enough in praise of the wealth of resources in the Advanced version. That ought to serve as an ultimate resource for any discussion on this top. 

I'd give both versions a thumbs up, now.  

2010-10-10 08:42:10DMI
Peter Hogarth

peter.hogarth@geoacoustics...
86.169.134.192

Thanks Neal, will alter the basic version.  I agree it is confusing.  As the rising air temperatures reach ice melt temperature in early summer the thermal energy is used to melt the ice instead of increasing the air temperature any further.  The surface temperature (and the air above it) can not increase much above this point until the ice has all melted.  Does this make more sense?  I'll look through the many references to see if a simple analogy is given, like Dougs.  Most papers just take it as given that the temperature is limited to ice melt point in Summer.

2010-10-10 12:27:54
nealjking

nealjking@gmail...
84.151.40.134

Maybe something like this:

"Firstly it uses Arctic Summer "above zero" temperatures in isolation. It is well documented that the Summer maximum temperatures above ice, and thus above 80 degrees North are constrained to just above “zero” by the ice melt temperature,"

=>

"First, his analysis is based only on the "above zero" temperatures for the Summer Arctic (above 90 degrees North). However, as long as there is any ice present at all, the local temperature will not get much above zero; so this set of numbers alone can't give an indication of heating trend."

I'm not familiar with the whole situation, so likely this doesn't capture everything you're trying to convey. However, my basic point is that you should not try to pack too much information into a Basic explanation: it should be short, but it must be comprehensible. 8th-grade reading level should be a target.

2010-10-11 01:36:17
nealjking

nealjking@gmail...
84.151.44.161

Ok, I've had a chance to look at the Advanced version. I congratulate you that you have cut back considerably from the Advanced version, while expressing the essential point much more briefly. Actually, with appropriate (brief) information, one of the graphs, that demonstrates that there is clearly a trend for winter although there is not a trend for summer, is actually useful. I would include it here.

I think the language is still too sophisticated; I have suggested something above, and with increased confidence that I've gotten the point, I would streamline it even much further. Many of the qualifiers are not really to the point, and could be dropped by appropriate re-wording. 8th-grade reading level => fewer syllables/word + fewer words/sentence.

2010-10-11 05:53:52
Rob Painting
Rob
paintingskeri@vodafone.co...
118.93.197.192

Pete, excellent job. I'm familiar with this argument having dealt with it on another blog 2-3 months ago. Thumbs up from me for the intermediate and advanced.

As the rising air temperatures reach ice melt temperature in early summer the thermal energy is used to melt the ice instead of increasing the air temperature any further. The surface temperature (and the air above it) can not increase much above this point until the ice has all melted.

That explanation is far easier to follow than that in the basic rebuttal.  

 

2010-10-11 06:34:41
nealjking

nealjking@gmail...
84.151.44.161

The more straight-forward the sentence structure, the better.

To quote the nuclear physicist Rutherford again: "If you can't explain to a barmaid what you're doing, you don't understand it."

2010-10-11 07:19:06
Peter Hogarth

peter.hogarth@geoacoustics...
81.157.15.17

Exactly what was Rutherford doing with the barmaid?  I think I missed that in the official biography...

The basic version was written in far more haste than the other articles.  I decided to get all three versions out into plain view as they have already taken far too much time (sorry John). This should have been a quick (!) distraction from other posts I hope to finish (sometime before I retire) on deep Ocean warming and recent work on early 20th Century warming. These are perhaps more worthy topics? 

I have no emotional investment in any of the original articles.  I'll tinker with them some more.  Any more good thoughts on titles?

2010-10-11 08:04:07comment
Robert Way

robert_way19@hotmail...
142.176.11.129
I'm intrigued regarding the new work on early 20th century warming, any tidbits you could share?
2010-10-12 08:02:56Update of basic version
Peter Hogarth

peter.hogarth@geoacoustics...
81.153.44.253

All, I've tried to update the basic version as plainly as I can for now.  Is this better?

Robert, I'm afraid my tidbits seem to go on a bit...  I'll try to get some of the information I've collated into a semi-logical flow, but it might be unwise to post it anywhere in this state.  I can get a list of the most recent papers together with links as I'll need to do this soon anyway.  My usual posts start off looking very unstructured as I gather together the related references and important points or supporting data in "themes", and most of my time is spent hacking back to something with a start, middle and finish, hopefully with the most relevant or recent references left in. I'm sure there's a better way of approaching this task. I am a humble beginner at this "blogging" game.

I'll next make some suggested changes to the intermediate version, will have a think about the advanced version, and let people know when done. 

2010-10-12 09:38:46
nealjking

nealjking@gmail...
91.33.105.82

"If we look at the Danish Meteorological Institute (DMI) Arctic temperture records, we see that every year the Arctic Surface temperatures rise to a maximum of just above 0 degrees C in summer, before falling again to extremely cold temperatures in the long dark winters.  Obviously the ice surface temperature will stop rising once the the melting point of ice is reached in early summer. Any further thermal energy is used to melt the ice. The surface temperature (and the air just above it) can not increase above this point while any significant amount of ice remains.  At the moment there are still significant amounts of sea ice above 80 degrees North through the Summer melt season. It is well documented that Arctic surface temperatures have not varied much in the Summer over the entire instrumental record for this reason."

=> "If we look at the Danish Meteorological Institute (DMI) Arctic temperature records, we see that every year the Arctic Surface temperatures rise to a maximum of just above 0 degrees C in summer, before falling again to extremely cold temperatures in the long dark winters. Why does it always stop there? Because the ice surface temperature will stop rising once the the melting point of ice is reached, early in the summer: Any further thermal energy is used to melt the ice. The surface temperature (and the air just above it) cannot increase above this point while any significant amount of ice remains.  At the present, there are still significant amounts of sea ice above 80 degrees North through the Summer melt season."

I would omit the following sentence: "It is well documented that Arctic surface temperatures have not varied much in the Summer over the entire instrumental record for this reason." It seems anti-climactic: You've already given an excellent reason, this seems to suggest that you need reassurance: "It is clear that the egg matter must be taken out of the shell before being eaten; and it is well-documented that eggs must be broken to make omelets."

"entire officia dailyl DMI data set"

=> "entire official daily DMI data set"


2010-10-12 10:21:19
Peter Hogarth

peter.hogarth@geoacoustics...
81.153.44.253
Thanks for spotting the typos Neal, getting late here.  I agree with your suggestion of omitting the sentence given, and substitution of "present".  However the rest is sentence re-arrangement and I have altered the structure yet again, I think the result is preferable. I am reminded of your asymptotic curves in the positive feedback post.
2010-10-17 03:31:54
doug_bostrom

dbostrom@clearwire...
184.77.83.151
These appear to have been published?
2010-10-27 05:46:46DMI data is needed !
Klaus Flemløse

KlausFlemloese@mail.tele...
62.243.102.250
Peter Hogarth
I have tried to do a statistical analysis of the Frank Lanser's pixel version of the DMI data.

My conclusion is that there is a significant decrease in the temperature even if 3 out layers are excluded.

Even if the result  might be misused of the skeptics , it must be made public.
The questions are:
1) Is the Frank data systematically wrong?
2) Does the original DMI data show the same trend?
3) If it is correct, that there is a decrease in temperature, what is the explanation?
Therefore the original DMI data is needed.
Summary of  analysis:
Call:
lm(formula = temp ~ Year)
Residuals:
     Min       1Q   Median       3Q      Max
-0.38512 -0.09318 -0.01426  0.08774  0.47171
Coefficients:
                 Estimate     Std. Error   t value Pr(>|t|)  
(Intercept) 11.131432   3.236385    3.439  0.00117 **
Year           -0.005177   0.001631  -3.174  0.00255 **
Residual standard error: 0.1817 on 51 degrees of freedom
Multiple R-squared: 0.1649,     Adjusted R-squared: 0.1486
F-statistic: 10.07 on 1 and 51 DF,  p-value: 0.002549

Regards

Klæaus Flemløse

2010-11-01 04:08:07Statistical analysis of Frank Lanser’s pixel-reading of DMI’s graph of the melting temperatures 80N – 90N.
Klaus Flemløse

KlausFlemloese@mail.tele...
62.243.102.250
Statistical analysis of Frank Lanser’s pixel-reading of DMI’s graph of the melting temperatures 80N – 90N.
Frank Lanser, HidetheDecline, has created a graph of the summer temperatures for the area 80N-90N by pixel-reading of the DMI’s graphs. From the graph he asserts that the temperatures are decreasing.
It has been a laborious job to carry out the pixel-readings. Thanks to Frank Lanser for doing this job.
In the following analysis, I rely 100% on the validity of this pixel-reading. However, I should like to do a reconciliation of the pixel-reading to the official DMI data sooner or later.
As an actuary/statistician I have little knowledge about the climate science. Therefore this analysis represents a statistician point of views.
As stated at DMI homepage:
The daily mean temperature of the Arctic area north of the 80th northern parallel is estimated from the average of the 00z and 12z analysis for all model grid points inside that area. The ERA40 reanalysis data set from ECMWF, has been applied to calculate daily mean temperatures for the period from 1958 to 2002, from 2002 to 2006 data from the global NWP model T511 is used and from 2006 to present the T799 model data are used.”
This means that the daily temperature measurements are put together from several sources. One source for the years 1958 – 2001, and other sources for the years 2002 - 2010. When datasets are combined from several sources there will always be a question of calibration. It is therefore plausible, that there exists a jump in the year 2002. The challenge to join one or more datasets is similar to joining two pieces of lumber into one piece. There will always be at jump at the joining. It is therefore justified - as a hypothesis – to include a jump in the statistical model.  
We know in general that there has been a significant increase in temperature in the Arctic over the most recent years. In addition we know that this increase is not likely to be shown in the summer melting temperatures since the temperature in this time period and in this area for physical reasons is expected to be constant. Observation of an increase or decrease of the summer melting temperatures will therefore be a surprise. However, this must not prevent us from testing this hypothesis using new/other temperature measurements.
A change in the temperatures over few years like the years 1958 – 2010 is expected to be linear in years and not quadratic or of higher order.  
In order to investigate this, I have set up a simple statistical model for the relation between summer temperatures for the years 1958 - 2010.  This model is a linear regression model with a jump in 2002:
Temperature ~ a + b*Year + c*jump in 2002
This means that there is a linear trend in the temperature development with a jump in 2002.
Sub models
In this model there are two sub models. In the following table these models are shown in addition to the AIC Akaike’s Information Criteria. Models with the highest AIC should in most cases be preferred.
Model specification
AIC
Model 1
Temperature ~  a + b*Year + c*jump in 2002
192,20
Model 2
Temperature ~  a + b*Year
178,84
Model 3
Temperature ~  a + c*jump in 2002
194,19
This indicates that Model 2 should be preferred. However, it depends on the outcome of the estimated parameters and test of significance.
By using the R-programming software I have estimated the following values and test probabilities for the full model:       
Model 1
Value
Standard error
T-value
P(T>t)
a
  1.2553305
3.7140747    
0.338   
0.7367   
b
-0.0001726  
0.0018762 
-0.092   
0.9270   
c
-0.3134765  
0.0764403 
-4.101
0.000151
This means that there is a small negative linear trend in the temperatures. However this trend is not significant. In contrary there is a significant jump in the temperatures in 2002 around
-0.31 °C.
There are 3 outliers in the dataset i.e. the years 1991, 1993 and 2010. The results are not sensitive to include or exclude these data points in the analysis.
Using this information the model is reduced to Model 3:
                 Temperature ~a + c*jump 2002
This model shows that the temperature is constant over the years 1958 - 2001 and over the years 2002 - 2010.
In this model the following parameters are estimated:
Model 3
Temperature
Standard Error
T-value
P(T>|t|)
a
 0.91374
0.02369
38.567
0.00000
c
-0.31805
0.05749
-5.532
0.00000
This means that the parameters are significantly different from zero.
The average temperature for the period 1958 - 2001 is 0.91 °C and a drop in temperature in 2002 of 0.35 °C to 0.56 °C.
The confidence interval for the average temperature is found to be:
Year
2.5% confidence limit
97.5% confidence limit
1958 -2001
0.86 °C
0.96 °C
2002- 2010
0.45 °C
0.69 °C
Alternative model with no jump
If no jump is included, we have the Model 2.
Temperature ~ a + b*Year
The estimated parameters are shown in the following table.
Model 2 Parameter
Estimated value
Standard Error
T-Value
P(T>|t|)
a
11.131432
3.236385
3.439
0.00117
b
-0.00517
0.001631
-3.174
0.00255
Both parameters are significant.  If there was no jump in 2002 the conclusion will be that there is at significant negative trend in the temperature. The predicted temperatures in 1958 and in 2010 are 0.99 °C and 0.73 °C respectively.
Under the conditions that there is no jump in 2002, the assertions made by Frank Lanser is correct.
However, we know from the first part of the analysis, that there is a significant jump in temperatures in 2002. Therefore model 2 is not valid.
Conclusions:
It is likely that the decrease in temperatures in the area 80N-90N as asserted by Frank Lanser from his pixel-reading of the DMI temperature is not correct due to data errors. These errors originate from joining two datasets in 2002 – the ERA40 dataset with the T511 dataset.
It is likely that the ERA40-dataset shows a significant higher temperature than the subsequent datasets showing around 0.35 °C lower temperature.
Again thank you to Frank Lanser for opening this discussion.
2010-11-02 04:56:52Statistical analysis of Frank Lanser’s pixel-reading of DMI’s graph of the melting temperatures 80N – 90N./Correction
Klaus Flemløse

KlausFlemloese@mail.tele...
62.243.102.250
Sorry to inform you, that I have made a mistake in respect of the AIC in the previous post. Please look away from the remarks about the AIC. I have sent an updated memo to John Cook.