Update, 8 Feb 2017: as a result of this comparison of data sources, some data values used to compute the CRU NAO index have been updated:
The original sources for the Gibraltar and SW Iceland sea-level pressure (SLP) data used to calculate the CRU NAO index (post-2000 updates) are documented in the original reference below. Since its publication, the NAO series has been extended by appending monthly SLP observations for Gibraltar and Reykjavik sourced mostly from monthly CLIMAT reports, but occasionally from Met Services if the data for a particular month were not included in a CLIMAT report.
The Gibraltar and SW Iceland (including Reykjavik) SLP series assembled by CRU have been compared with data obtained from two other sources, all downloaded on 6 Feb 2017:
CRU data run from 1821 (with some missing values in the first four years; the early values have been compiled from a mixture of sources with corrections to compensate for differences between measurement series, see reference below). ICE data for Reykjavik run from 1949. Met Office CLIMAT data run from 2000.
The CRU data are compared with the other two sources for the full period of their overlap. The comparisons are listed in Comparison of Gibralatar and SW Iceland SLP sources.
Available values are listed for each month. When a source differs from the CRU value, the difference is also listed (other source minus CRU value) preceded by "=D". The upper part of the comparison file lists all those months where at least one other source differs by at least 0.3 hPa from the CRU value. The lower part of the comparison file lists all years and all months, regardless of any differences or the size of any differences.
There are 3 months since January 2000 when the CRU value differs from the CLIMAT value by 0.1 hPa but matches the Iceland Met Office value. These represent cases where the CLIMAT value was not available at the time and the ICE value was used to update the CRU series instead. There are 8 other months where the CLIMAT value is still not available and the CRU value matches the ICE value in each case. These values are left unchanged, as originally published by CRU. There are large differences of 7.4 and 7.3 hPa for October 2000 and August 2011, respectively, because the station-level pressure (985.2 and 1003.9 hPa, respectively) was incorrectly used instead of the sea-level pressure (992.6 and 1011.2 hPa, respectively, in both CLIMAT and ICE sources). The CRU values have now been updated to match the CLIMAT and ICE sea-level pressure values. A typographical error affecting one digit of the June 2010 SLP value has also been corrected.
Prior to 2000 when only Iceland Met Office data are available for comparison, there are occasional differences of 0.1 or 0.2 hPa and one diference (Jan 1957) that reaches 0.4 hPa. Since the CRU data for this period were originally obtained from the Iceland Met Office, it is likely that these small differences have arisen from subsequent re-processing of the observations or due to rounding differences. Since they are small compared with the standard deviation of monthly SLP at Reykjavik (>6 hPa for all months from Oct to Mar, > 3 hPa for all months from Apr to Sep), these values are left as originally published by CRU.
There are 9 months since January 2000 when the CRU value differs from the CLIMAT value. The difference is only 0.1 hPa in 3 cases, so these are left unchanged as they likely arise from rounding error. In 6 cases the error is larger and likely arose from inadvertent use of station-level rather than sea-level pressure (for Oct 2000, Nov 2006 and Mar 2011) or transciption error affecting one digit (for Nov 2006, Oct 2009, Feb 2010 and Oct 2011). In these 6 cases, the CRU value has now been updated to match the CLIMAT value.
The monthly SLP values updated following this comparison of sources are not in the standardisation period (1951-1980), so they do not affect the calculation of the NAO index overall. Only the six updated months are affected. The most significant changes are the corrections to SW Iceland SLP in Oct 2000, Jun 2010 and Aug 2011, which amount to just over 1 standard deviation (SD) for the first two and just over 2 SD for the third. The most significant change to Gibraltar SLP is about 0.5 SD (Oct 2009), with the other five changes being about 0.3 SD or less.
The resulting NAO index changes are:
Data files prior to this update are available for comparison: