Temperature Adjustments In The Canadian Arctic – some supplementary information for a comment

This post extends a comment posted at https://notalotofpeopleknowthat.wordpress.com/2015/02/08/temperature-adjustments-in-the-canadian-arctic/

While Paul Homewood’s post deals with GHCN adjustments, this comment examines the influence of one of these Canadian Arctic stations, Eureka, on the Gistemp gridded output. As station coverage in the northern Canadian Arctic is sparse, this one station has considerable influence on neighbouring (and not so closely neighbouring) grid cells.

Please remember that this comment illustrates the influence of one station in a single month, not on annual values.

Like it or not however, single month results such as shown below (provided of course that they match the appropriate beliefs – certainly not when they conflict)  are eagerly grasped by a certain class of credulous journalist or “specialist environmental writer and commentator”.

That last local reference may be lost on the general reader – for entertainment Google it to follow a trail of dubious “science”: the recent “27 million to one”  probability claim; increase in temperature in °C expressed as a percentage of initial temperature in °C; “IEA’s projected global average temperature increases of 3.6C by 2035” – a blunder in the Irish Times, substituting 2035 for 2200, which should have been rather obvious to a “specialist environmental writer and commentator”, but for the credulous blinkers imposed by an alarmist mindset (and do journalists and “specialist environmental writers and commentators” not bother with original sources these days?); “In the 20 years since the Earth Summit, Ireland’s average temperature has increased by 0.75C, exactly in line with a projected 4C calamity this century” – another blunder in the Irish Times, the first line of the report did indeed state this, if not using those exact words, but again a “specialist environmental writer and commentator” might be expected to read on and notice that this was not what the Irish Met Office was saying (and the increase derived from GISS gridded data would be roughly half that 0.75°C); …

Returning to the matter in hand, what follows is part of a comment I sent to Reto Reudy of GISS last autumn, prompted by an earlier comment in email from Reto that “even that extreme case of planting a midlatitude station into the tropics did not visibly affect any of our results or maps as a comparison of the original with the corrected results showed”. [I had in fact raised the issue of faulty location metadata only in the context of urban/rural misclassification with nightlight luminance and not in the context of gridded data]. It came as a surprise to me too to find that dropping a single station, and one which is correctly located at that, could cause a change which would be reflected in the gridded data, GISS maps, and even make it through to GLB.Ts.txt. As mentioned above, this is in the context of monthly rather than annual results, and relatively minor in the context of other problems with the Gistemp analysis, not just the wrong classification of hundreds of stations as a result of faulty metadata. (The WMO Volume A modification programme to accommodate station coordinates with higher precision alone provides corrections for hundreds of stations included in the GHCN inventory, although it should be born in mind that not all of these stations are currently reporting in GHCN, leaving the possibility that the GHCN metadata could still reflect the “last known location” at the time GHCN data stopped. I have lightly edited that comment to Reto below (removing references to you/your, etc.):

Out of curiosity I dropped 40371917000 EUREKA,N.W.T., and compared maps and results in GLB.Ts.txt:

2014    94   60   73   87  101   77   74   76   99***************     82****     78   87   76*****  2014
Year   Jan  Feb  Mar  Apr  May  Jun  Jul  Aug  Sep  Oct  Nov  Dec    J-D D-N    DJF  MAM  JJA  SON  Year


When Eureka is dropped:

2014    95   62   74   88  101   77   74   76   98***************     83****     79   88   76*****  2014
Year   Jan  Feb  Mar  Apr  May  Jun  Jul  Aug  Sep  Oct  Nov  Dec    J-D D-N    DJF  MAM  JJA  SON  Year

The September value is reduced from 99 to 98 (0.995 to 0.982), and the February value increases from 60 to 62 (0.597 to 0.618), and that large red shaded area in the Arctic reduces. I had expected the shading reduction, but not that the change would make it through to the global means.


And just for further information, not affecting global means:

Dropping the other station marked in green as well, 40371082000 ALERT,N.W.T., cleared the rest of that red shading, and increased the number of gridcells with missing data, but did not further change the global means.

Alert has data only to 2004, whereas Eureka has current data.


And a final warning regarding Canadian data in GHCN (Mould Bay is mentioned in a comment on Paul Homewood’s post):

40371826000 Nitchequon has crept back in to the Gistemp analysis – see the Gistemp Updates to Analysis:

April 12, 2010: Reports have been coming in from Nitchequon, a ghost town in Quebec, after a long gap from 1986-2006; however those data don’t seem to be consistent with the older records. Hence they are disregarded until further notice.

The newer reports are not in fact from Nitchequon, but from Pangnirtung. What has happened is that Canada has started to recycle WMO IDs. 71826 was Nitchequon previously but is now Pangnirtung. GHCN has however included both stations in one record, still calling it Nitchequon and with those coordinates. Pangnirtung is some 17 degrees further north. Fortunately Gistemp appears to use only the data up to 1985, which is Nitchequon. I checked the Environment Canada data for both stations about two years ago and verified that the GHCN record combines the two stations.

Other WMO IDs in the GHCN inventory which I know have been recycled are:

403 71072 000   MOULD BAY, N., which is now recycled as THUNDER BAY, ONT, which Gistemp uses up to 1997

71141 NORWAY HOUSE, which is now recycled as GATESHEAD ISLAND, NU (GHCN inventory has “non-WMO”  40371141001 NORWAY HOUSE,MA, which Gistemp uses up to 1945, so still safe even if this is the WMO station)

403 71187 000 BAIE COMEAU,Q, which is now recycled as HOPE, BC, which Gistemp u uses up to 1989

403 71196 000 BONAVISTA,NFL, which is now recycled as PEMBROKE CLIMATE, ONT, which Gistemp uses up to 1988

403 71874 000 LETHBRIDGE,AL, now recycled as BATHURST INLET, NU, which Gistemp uses up to 2001, and the GHCN data appears to be Lethbridge

So, so far Gistemp still uses only the data from the original locations, but if sufficient recent data is added by GHCN there is the risk of this new data, from different stations altogether, being used by Gistemp as if it came from the original locations.

This entry was posted in Uncategorized and tagged , , , . Bookmark the permalink.

2 Responses to Temperature Adjustments In The Canadian Arctic – some supplementary information for a comment

  1. A small technical explanation for anyone aware of rounding to integer values in tenths or hundredths of a degree at several steps in Gistemp, subsequently read and converted back as floating point values, and who may as a result raise an eyebrow on seeing a comparison of results with and without Eureka resulting in a reduction of the September value from 0.995 to 0.982:

    The effect of such multiple earlier roundings was mitigated by running the analyses with and without Eureka only from Gistemp Step 3, using the homogenized Step 2 station output available from GISS as v3.mean_GISS_homogenized.txt as input, rather than running the prior steps. In this way the homogenized integer values expressed as hundredths of a degree were reconverted to floating point only once, and the calculation of grid cell values in Step 3 is then carried out with floating point values. My Gistemp implementation also produces parallel versions of results files such as GLB.Ts.txt with fractional degree values rather than the rounded integer hundredths of the main files.

  2. Verity Jones says:

    Very thorough, as ever Peter

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