Skip to content

NCDC Cooling The Past Again

June 8, 2012

By Paul Homewood

 

Climate At A Glance

November Temperature
Alabama

 

image

 

Yesterday, WUWT picked up the story of how the NCDC have been cooling past temperatures in the US, thereby creating a warming trend. Essentially the state temperatures now shown by NCDC are lower than the originally recorded ones, which are still available online.

The explanation given is that “they changed from the  ‘Traditional Climate Division Data Set’ (TCDD) to a new ‘Gridded Divisional Dataset’ (GrDD) that takes into account inconsistencies in the TCDD. “. (In simple terms, the old system simply averaged together all stations in a state, ignoring any potential weighting issues, where there was, for instance, a preponderance of stations in one particular area, or in warmer cities. The new system allows for the full area of each state to be adequately represented).

This sounds fine in theory, but how does it all work out in practice?

I have taken a detailed look at November temperatures in Alabama, comparing 1934 with 2011. Why Alabama – it was the first one on NCDC’s list! Why 1934? It was 1934 that Anthony’s article homed in on. Why November? November 2011 is the latest available monthly report available on NCDC’s website.

 

As illustrated on the above graph, according to NCDC, November temperatures in 1934 were 57.0F, and compare with 55.5F in 2011. (The full data is available here.) However the mean temperature originally published by the US Dept of Agriculture in 1934 is shown as 57.6F, as in Figure 1.

image

Figure 1

Furthermore the latest state report, before any adjustments, issued by NCDC for Nov 2011 indicate a statewide average of 55.3F. (This figure is arrived according to the “old method” and is available here.) In other words, NCDC show Nov 1934 as being 1.5F warmer than 2011, but the state reports show 1934 as being 2.3F warmer, a discrepancy of 0.8F. But are the 1934 and 2011 reports directly comparable?

In 1934, as Figure 1 shows, the statewide temperature was worked out by averaging 40 stations together. The current method, however, takes 132 stations, spread across 8 divisions. Each division is then individually averaged, and the average of each of these 8 divisions can then be amalgamated to arrive at the state figure. The divisions are each deemed to be climatically significant, rather than purely political boundaries. This, of course, should go a long way to addressing the problem of weighting.

So how do 2011 temperatures on the latest list of stations compare with their 1934 ones? These are listed in Figure 2 and sub totalled by division.

 

Division Station Temp
Nov 1934
Temp
Nov 2011
Diff
1 Decatur 55.8 54.0  
  Sub Total 55.8 54.0 1.8
2 St Bernard 52.6 51.4  
  Scottsboro 54.0 52.7  
  Valley Head 54.4 52.0  
  Sub Total 53.7 52.0 1.7
3 Centerville 56.3 53.4  
  Clanton 56.4 53.3  
  Sub Total 56.3 53.3 3.0
4 Gadsden 54.6 53.2  
  Talladega 56.5 53.3  
  Sub Total 55.5 53.2 2.3
5 No Stations      
6 Greensboro 58.5 56.9  
  Selma 58.4 56.6  
  Union Springs 59.0 55.0  
  Sub Total 58.6 56.2 2.4
7 Brewton 59.0 58.7  
  Greenville 59.9 55.3  
  Highland 60.7 53.0  
  Mobile 62.0 60.1  
  Thomasville 56.6 57.4  
  Sub Total 59.6 56.9 2.7
8 Bay Minette 62.0 60.6  
  Fairhope 61.7 59.8  
  Robertsdale 63.6 60.3  
  Sub Total 62.4 60.2 2.2
ALL DIVISIONS AVERAGE 57.4 55.1 2.3

                                                                                                   Figure 2

Bearing in mind, the NCDC figures show a difference of 1.5F, it is immediately apparent that in every division, the difference is much greater. Over the state, there appears to be a discrepancy of 0.8F. Remember, we are not concerned with absolute temperature, but relative changes, so the argument about weighting should not really factor. A number of points seem to raise themselves:-

  1. Out of the 132 stations now currently monitored, only 19 have records back to 1934. Equally, most of the stations used in the original 1934 report no longer exist. This must surely make any comparison with earlier decades highly problematic, and must involve considerable “guesswork”.
  2. Out of the 132 stations, in Nov 2011, 50 have no data at all. (This applies when more than 10 daily figures are missing, and can be seen here).
  3. In Division 5, Piedmont Plateau, there are no stations at all that report back to 1934, which again must call into question how accurate NCDC’s temperature comparisons can be.
  4. The overall average of 2.3F calculated from the above sample, is exactly same as the figure arrived at by comparing the 1934 and 2011 state reports, i.e 55.3F v 57.6F 57.6F v 55.3F. This indicates that the 1934 methodology was probably pretty sound.

It is abundantly clear that, in Alabama at least, the extra warming trend shown by NCDC cannot be explained by the change to a Gridded Divisional Dataset. Which, of course, brings us back to “Adjustments”.

The individual station temperatures, above, it should be remembered, are all “raw”. Rather than adding a cooling adjustment to allow for UHI, the NCDC mincing machine adds 0.8F of warming.

About these ads
7 Comments leave one →
  1. June 8, 2012 2:46 pm

    Reblogged this on Climate Ponderings.

  2. June 9, 2012 1:54 pm

    I’se confuzed. You start out with 1934 being about 57°F and 2011 about 55°F, and end up with them reversed. Wassupwiddat?

  3. June 16, 2012 2:21 pm

    Excellent work, Paul!

  4. lolwot permalink
    July 7, 2012 12:55 am

    “It is abundantly clear that, in Alabama at least, the extra warming trend shown by NCDC cannot be explained by the change to a Gridded Divisional Dataset.”

    I don’t see how you can conclude that from what you’ve done. The difference in simple average between 1934 and 2011 is 2.3F. NCDC report a difference in weighted average between 1934 and 2011 as 1.5F.

    You seem to be arguing that the two should be equal, that the difference in weighted average should be 2.3F too, and thus NCDC have adjusted the real figure down by 0.8F. But I don’t understand that. Why should they be equal? You say: “Remember, we are not concerned with absolute temperature, but relative changes, so the argument about weighting should not really factor.”

    I don’t understand what you mean there. As far as I can tell even relative change will be affected by the method. As much as 0.8F? I don’t see why not. I could imagine calculating the weighted average in 1934 and 2011 and finding that yes the difference is 0.8F less than found by simply calculating the averages.

    “Out of the 132 stations now currently monitored, only 19 have records back to 1934. Equally, most of the stations used in the original 1934 report no longer exist. This must surely make any comparison with earlier decades highly problematic, and must involve considerable “guesswork”.”

    I would have thought little guesswork is needed. It would be easy to test the uncertainty of having less stations back then by taking subsets of the 132 stations in 2011. Eg calculate the state figure for Nov 2011 with 132 stations. Then calculate it with just 19 randomly picked stations out of that 132. See what the difference is. To get a good idea of how less accurate 19 stations are than 132, keep doing the calculation with different sets of 19 randomly picked stations. I am sure this is the method, or a similar method they will use to determine the confidence in station differences. There are all sorts of tests along these lines to test the dropping and adding of stations. With 132 stations it’s easy to simulate the dropping and removal of stations at various points and test how that impacts the state figures.

    “In Division 5, Piedmont Plateau, there are no stations at all that report back to 1934, which again must call into question how accurate NCDC’s temperature comparisons can be.”

    Similar solution. Run the calculation removing random divisions and see how much that changes the result by.

    • July 7, 2012 11:29 am

      According to the “old” divisional system, Alabama was 2.3F warmer in 1934 than 2011. On the “new” gridded system, it shows only 1.5F warmer. My analysis (in theory at least!) removes the weighting issue, which Chris Burt referred to in his Arizona example, and suggests that after allowing for weighting changes, you still get a number of 2.3F.

      If this is so, the only other explanation for the discrepancy has to be “adjustments”, as NCDC admit:

      Finally, none of the TCDD’s station-based temperature records contain adjustments for historical changes in observation time, station location, or temperature instrumentation, inhomogeneities which further bias temporal trends

      It would, of course, help if NCDC showed the amount they adjust by state, then we would not need to have this conversation! We do know though that nationally they add warming adjustments of about 0.6F, pretty much in line with my Alabama analysis.

      http://www.ncdc.noaa.gov/img/climate/research/ushcn/ts.ushcn_anom25_diffs_urb-raw_pg.gif

      Please be clear. I am not arguing against the logic of these adjustments. I would just like people to be aware that adjustments are the main reason for the difference between the 2 datasets, and not the weighting issue Chris Burt wrote about.

      Paul

Trackbacks

  1. June 2012 – Not that hot « sunshine hours

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: