Examination of CRU data suggests no statistically significant warming

UPDATE: The StataSphere server can’t handle the load of interest, I’ve take the images offline from this article, and disabled the link to it. Once he gets the server up and running again I’ll put them back – Anthony

Readers may recall this quote from Dr. Phil Jones of CRU, by the BBC:

Q: Do you agree that from 1995 to the present there has been no statistically-significant global warming

A: Yes, but only just. I also calculated the trend for the period 1995 to 2009. This trend (0.12C per decade) is positive, but not significant at the 95% significance level. The positive trend is quite close to the significance level. Achieving statistical significance in scientific terms is much more likely for longer periods, and much less likely for shorter periods.

A.J. Strata has done some significance tests:

CRU Raw Temp Data Shows No Significant Warming Over Most Of The World

Bottom Line – Using two back-of-the-envelope tests for significance against the CRU global temperature data I have discovered:

• 75% of the globe has not seen significant peak warming or cooling changes between the period prior to 1960 and the 2000′s which rise above a 0.5°C threshold, which is well within the CRU’s own stated measurement uncertainties o +/- 1°C or worse.
• Assuming a peak to peak change (pre 1960 vs 2000′s) should represent a change greater than 20% of the measured temperature range (i.e., if the measured temp range is 10° then a peak-to-peak change of greater than 2° would be considered ‘significant’) 87% the Earth has not experienced significant temperature changes between pre 1960 period and the 2000′s.

So how did I come to this conclusion? If you have the time you can find out by reading below the fold.

I have been working on this post for about a week now, testing a hypothesis I have regarding the raw temp data vs the overly processed CRU, GISS, NCDC, IPCC results (the processed data shows dramatic global warming in the last century). I have been of the opinion the raw temp data tells a different, cooler story than the processed data. My theory is alarmists’ results do not track well with the raw data, and require the merging of unproven and extremely inaccurate proxy data to open the error bars and move the trend lines to produce the desired result. We have a clear isolated example from New Zealand where cherry picked data and time windows have resulted in a ridiculous ‘data merging’ that completely obliterates the raw data.

To pull this deception off on a global scale, as I have mentioned before, requires the alarmists to deal with two inconvenient truths:

1. The warm periods in the 1930′s and 1940′s which were about the same as today
2. The current decline in temperature, just when the alarmists require a dramatic increase to match the rising CO2 levels.

What is needed out the back end of this alarmist process is a graph like we have from NCDC, where the 1930′s-1940′s warm periods are pushed colder and the current temps are pushed higher.

[image offline]

People have found actual CRU code that does this, and it does it by smearing good temp data with inaccurate proxy data (in this case the tree rings) or hard coded adjustments. The second method used by alarmists is to just drop those inconvenient current temps showing global cooling, which has also been clearly discovered in the CRU data dump.

I have been attempting to compensate for the lack of raw temperature data by using the country-by-country graphs dumped with data from University of East Anglia’s Climate Research Unit (CRU). The file is named idl_cruts3_2005_vs_2008b.pdf, which tells me this is the latest version of the CRU raw temp data run in prep for a new release of the latest data (the PDF file was created in July 2009).

I am very confident this data is prior to the heavy handed corrections employed by CRU and its cohorts. The fact is you can see a lot of interesting and telling detail in the graphs. Much of the Pacific Ocean data has been flipped since 2005 trying to correct prior errors and you can see the 2008 data trend way downward in most of the graphs. In addition, the 1930′s-1940′s warm periods have not been squelched yet. The alarmists have not had a chance to ‘clean up’ this data for the general public (which is one reason I think it was in the dump).

Before we get to actual examples and my detailed (and way too lengthy) analysis, I need to explain the graphs and how I used them (click to enlarge).

[image offline]

In this graph we see the primary data we have available from CRU. This is a comparison of the 2005 runs in black and 2008 runs in light purple/red. At CRU all the data is blocked into quarters. This graph is MAM, which stand for March-April-May, for Argentina.

The love of trend lines and averaging by CRU and other alarmists is quite telling here. The ‘raw’ quarterly data is noted with the blue arrows, It is the highly variable lines from which the (much less accurate) trend lines are generated. I point this out to note that fact that to create a quarterly value for a country for a given year means the raw daily temp data has disappeared under a mountain of averaging already. Day/Night temps must be combined into quarterly temps by location and then combined into a country wide figure. Even with all this inaccuracy added in the ‘raw’ data is quite dynamic, which makes me wonder how dynamic the true sensor data is. CRU and others believe the trend lines mean something significant – but really all the do is mask the true dynamics of nature.

Anyway, now let me explain how I derived (by eye – ugh!) the two primary pieces of data I used to test my hypothesis that the 2000′s are not significantly warmer or cooler than the pre 1960 period (when CO2 levels were drastically lower). Here is how I measured the Peak-to-Peak change in each of the graphs (click to enlarge):

[image offline]

I simply find the highest pre 1960′s peak and the highest point in the 2000′s and subtract. I know this is subjective and error prone, but it is good enough for a ‘reasonableness test’. I would have preferred to use actual data and define min/max points for each time period and compare. But this is what happens when you don’t share the raw data, as true science demands.

Note I am using the 2005 trend line. I have noticed many graphs where the 2008 would given my hypothesis more strength, and maybe some day I will compute that version. I also know there were higher peaks prior to 2000 (especially around 1998). In fact I found myself averaging the slide from 1998 into the 2000′ many times. I tried to err on the alarmists’ side (my hypothesis to prove after all). Also please note that the ‘raw’ yearly data bounces around well beyond all trend line peaks – so I am not too concerned with fact some peaks are skipped. The next calculation will better explain why.

The P2P data is captured in my results file [offline] as shown (click to enlarge):

[image offline]

Note: I am trying to find a way to get a clean spreadsheet up so folks can copy out the data.

Anyway, what I did was compute the P2P value for each quarter for each country, and then averaged those over the full ‘year’. Then I applied three significance tests to see if the P2P value is (1) less than -0.5°C, (2) within the +/- range of 0.5°C or (3) greater than +0.5°C.

I decided used this significance test because of another file dumped with the CRU data which clearly showed where CRU stated its measurement accuracy was typically 1°C or greater. Here is the CRU report from 2005 containing their accuracy claims, along with their own global graph of temperature accuracy:

In my original post on these files I went into great detail on the aspect of measurement accuracy (or error bars) regarding alarmists claims. I will not repeat that information here, but I feel I am being generous giving the data a +/- 0.5°C margin of error on a trend line (which contains multiple layers of averaging error incorporated in it). Most of the CRU uncertainty data, as mapped on the globe, is above the 1°C uncertainty level.

What that really means is detecting a global warming increment of 0.8°C is not statistically possible. If I had used their numbers none of the raw temps would have been significant, which is why people do these back-of-the-envelope tests to determine if we have sufficiently accurate data to test our conclusions or hypothesis.

===========

Read the conclusion here: CRU Raw Temp Data Shows No Significant Warming Over Most Of The World

h/t to Joe D’Aleo