In the case of Lincoln, the station Broadfield Ews is stipulated as the Reference Station, but there is only 1 full year in CliFlo.

This is the data returned from a query on 17603 Broadfield Ews

Lincoln, Broadfield Ews 2000 2 11.9

The other Open (but no longer being updated) station at Lincoln is Broadfield Edl

4882 H32642 01-Jun-1987 30-Apr-2000 100 Lincoln, Broadfield Edl

A 1971-2000 query of that station returns this data:-

Lincoln, Broadfield Edl 1988 2 12.3
Lincoln, Broadfield Edl 1989 2 12.1
Lincoln, Broadfield Edl 1990 2 12
Lincoln, Broadfield Edl 1993 2 10.8
Lincoln, Broadfield Edl 1994 2 11.3
Lincoln, Broadfield Edl 1995 2 11.8
Lincoln, Broadfield Edl 1996 2 11.6
Lincoln, Broadfield Edl 1997 2 11.6
Lincoln, Broadfield Edl 1998 2 12.5
Lincoln, Broadfield Edl 1999 2 12

And the Closed station 4881 H32641 01-Jan-1881 31-Dec-1987 100 Lincoln returns this data,

Lincoln 1971 2 12.1
Lincoln 1972 2 11.4
Lincoln 1973 2 11.9
Lincoln 1974 2 11.4
Lincoln 1975 2 11.6
Lincoln 1976 2 10.4
Lincoln 1977 2 10.8
Lincoln 1978 2 12.1
Lincoln 1979 2 11.6
Lincoln 1980 2 11.5
Lincoln 1981 2 11.9
Lincoln 1982 2 11.4
Lincoln 1983 2 11.2
Lincoln 1984 2 11.6
Lincoln 1985 2 11.9
Lincoln 1986 2 11.5
Lincoln 1987 2 12

If what you say is correct that “the default procedure is to bring all past temps into phase with the current “open” station” then the Reference data for Lincoln is 10 years of Broadfield Edl and 17 years Lincoln 4881 brought into phase with 1 year of Broadfield Ews.

So 3.33% of the baseline data comes from the last Open (and stipulated) station but 96.66% of the data comes from other stations. This does make sense because the 2 earlier stations are better representations of the climate 1971-2000.

The allocation of “Reference Station” to Broadfield Ews is something of a misnomer IMO, if I am understanding this correctly. A better stipulation would be Lincoln 4881 because it provides 56.66% of the data for the baseline.

I checked Lincoln, Broadfield Ews in CliFlo but it doesn’t have 30 years of data:-

17603 H32645 01-Jul-1999 31-Jan-2011 100 Lincoln, Broadfield Ews

I really don’t understand these “Reference Stations”. It seems to me that it is a separate dataset that should be published along with the series. How for example, did they compile 30 years 1971-2000 with just 1 1/2 years of data from CliFlo for Broadfield Ews?

There seems to be some other mystery data source.

Barry Brill briefly covers Reference Stations in the post but I’m not much wiser.

A similar thing is happening in the NZT7 series. Way up-thread I checked CliFlo for Masterton 2000-2011 but the years of data available was not contiguous although NZT7 was but the introduced data did not look right.. I don’t get it.

You use a bit of common sense about averages and it all gets backed up in NIWAs own procedures. What a strange world we live in.

Remind me again, what exactly is this strawman we are trying to destroy?

Not quite I think.

The standard (as I understand from Bob D – haven’t read it myself) is that the 3 year gap guideline is for calculations within one stations history (e.g. monthly values added to obtain annual averages), not across several stations (I could be wrong).

I do agree on the spatial issue however. I think it is far more important to retain spatial balance about the centre-of-area than to haggle over missing data. If say there were 3 SI stations and 1 NI station, then 2 SI stations should be dropped and the 1 that is retained should be diagonally opposite across the centre-of-area to the NI station. Otherwise the average of the 4 stations is really only representative of northern SI.

“However, when you come to calculate the average temperature (not the anomaly) they must add their 4-station anomaly obtained above to the average temperature for the whole 7-station series (1971-2000).”

The NZT7 method says this:-

3) The climatologies are specific to the “Reference” stations at each location, which are:
Auckland Aero (Auckland), East Taratahi AWS (Masterton), Kelburn (Wellington), Hokitika Aero (Hokitika),
Nelson Aero (Nelson), Lincoln Broadfield EWS (Lincoln), and Musselburgh EWS (Dunedin).

Also, the NZT7 station values for say Lincoln are not necessarily all from Broadfield EWS. The climatologies of the reference seem to be a separate exercise.

For an actuals composite, if all the 7 values are there its a simple average of the actuals. If say 1 value is missing then the average of the 6 anomalies is added to a reference to obtain actual temp. I’ve done a few checks on my understanding using NZT7 data but it would be better for you to do some some yourself to get a feel for it and you may even find I’m putting you crook (i.e. don’t believe everything you read on the internet)

Bob D has explained well but there is also this on the NIWA site:-
——————————————————————————————————————————
Technical note on the treatment of missing data in the 11-station series

http://www.niwa.co.nz/news-and-publications/news/all/2009/nz-temp-record/temperature-trends-from-raw-data/technical-note-on-the-treatment-of-missing-data

Missing months during the year

* Calculate the monthly anomalies by subtracting the climatology for that month (in this case the 1961-1990 average), then
* Average the monthly anomalies to obtain the annual anomaly, ignoring missing months.

Missing annual values or missing sites

The ‘eleven’-station average is simply the average of the individual station anomalies over those stations where the annual anomaly is available.

This is a very important point, often misunderstood. Averaging the anomalies (deviations from climatology at each site respectively), and not the actual temperatures, means that no bias is introduced if the number of cold versus warm sites fluctuates during the period of analysis. For example, over the period 1931-1937, Palmerston North is colder than Queenstown in ‘anomaly’ terms, even though Queenstown is colder by about 3 °C in absolute terms. Using the absolute numbers will bias the results; using the anomalies will not.
—————————————————————————————————————————–
A way to view this is that missing months is a longitudinal calculation but missing years is a cross-sectional calculation (think in terms of a spreadsheet)

This is reasonable to me but I’m not a fan of anomalies because it’s just a convention. The baseline is a standard 30 years that moves every 30 years (currently 1971-2000 I think) but I have seen 30 yr baselines placed arbitrarily at the end of a series for convenience. From what I can gather, the thinking with the convention is that climate changes every 30 years which may well be close to the truth given that 60 year cycles have been extracted statistically from global temp series.

The composition of the reference baseline becomes very important when using anomalies but I’d bet most people don’t give the baseline much thought. It’s really a form of data smoothing I think. So-what, if a warm or cool bias accrues, isn’t that the point of the exercise – to obtain a representation of what is happening to temperatures? But note this is just a personal opinion and to be taken with a grain of salt.

What happens is that as a series gets longer, the 30 yr baseline gets proportionately less than the series length so its OK when the series is not that much longer than the baseline but then the anomaly gets progressively irrelevant as the series lengthens. It also makes comparison of series with different baselines a real mission and in that case the best way is to convert to actuals.

As far as I’m concerned with temperature, the baseline is 0 K or 0 C and I much prefer to use actuals but I’m willing to conform to anomaly convention where appropriate.