CET Issues

Figure above: The Central England Temperature area

This post will cover issues related to the Central England Temperature (CET) series, created initially by Gordon Manley in 1953, and maintained in recent decades by the UK Met Office Hadley Centre.

Definition

CET is meant to be an absolute average temperature, presumably at some location roughly in the “middle” of the area containing the source station data.

It might have been better to have made CET an average of temperature variations.

My Version of Central England Temperatures

A reconstruction of monthly average mean temperature variations back to 1760, from around 30 long temperature records, is given at diymetanalysis:

https://diymetanalysis.wordpress.com/2018/03/03/central-england-tavg/

Philip Eden version

An alternative version was being created by Philip Eden (a distinguished British meteorologist, now deceased), and the following figure shows both HadCET (the official version) and the Eden version, and their difference, all as 12-month moving averages:

Sources

• Eden version: http://www.climate-uk.com/page5.html
• HadCET: see the page above

More to follow …

Advertisements

GHCNM Spatial Sampling

SCOPE

This post is about the current spatial sampling of the stations with monthly average rainfall data in GHCNM version 2, and temperature data in GHCNM versions 3 and 4, the current source data for many “official” reconstructions of the global land rainfall and surface air temperature history.

Preliminary results are included for GHCNM version 4, which has not yet been released officially.

A figure of around 8000 stations is often quoted for GHCNM version 3, which might appear to be an adequate spatial sampling. However, most of those 8000 stations are no longer providing updates, and there are questions about the adequacy of the current spatial coverage in two distinct areas:

• Spatial sampling of the varying climate around the globe
• Detection and correction of inhomogeneities in the currently reporting stations

These two questions will be discussed on a per-country basis, starting with Australia.

AUSTRALIA

The Australian Bureau of Meteorology has 112 stations in ACORN-SAT(2012), intended to describe the varying temperature histories around the country:

The currently reporting stations in GHCNMv3 (unadjusted), for monthly TAVG (average temperature) data is shown in the following list for Australia, generated by recording only those stations with data in 2018:

These 62 stations in Australia that are currently reporting monthly TAVG to GHCNMv3 are possibly adequate to represent the varying temperatures around the country, but only if those stations remain unchanged in equipment, environment and procedures, and are free of errors.

ACORN-SAT was constructed by homogenisation involving many hundreds of other stations. This is no longer possible in GHCNMv3, which only has current data for 62 stations in Australia.

To confirm that all of the many other Australian stations in GHCNMv3 are currently non-reporting, here is a table of all Australian TAVG data (qcu: “unadjusted”) for 2017:

The 4-digit numbers in the table above, with separate columns for Jan to Dec, are the average temperature in hundredths of a degree C, -9999 means missing data.

GHCNM version 4 (preliminary) results are as follows. A total of 101 Australian stations contributed monthly average TAVG (unadjusted) data for January 2018, shown below in two parts:

101 stations will be a significant improvement on 62, but falls short of the 300 that is my guesstimate for the minimum number required to have a good chance of detecting and correcting inhomogeneities, and infilling missing data.

GHCNM version 2 (now containing only monthly rainfall totals)

Australian stations with rainfall data in 2018 are shown in the following list:

The above list includes the rainfall totals for January 2018, in tenths of a mm.

The 29 stations in the list are insufficient to allow analysis of contemporary rainfall, and detection/correction/infilling of anomalous/missing data.

The entirety of the 2017 rainfall data for Australia is given in the following table, with missing data (-9999) shown in red:

GHCNMv2 is no longer fit-for-purpose, and the Australian BoM has questions to answer about why so much data is missing from these stations, many/most of which are at Meteorological Offices.

More to follow later, about other countries …

DIYMETANALYSIS

Author: Dr Michael Chase

You are invited to visit a new website that describes a relatively simple, but nevertheless effective method of reconstructing the regional average history of monthly average surface air temperature variations for a region from its weather station data, aided by any metadata that is available:

https://diymetanalysis.wordpress.com/

The new website will be kept small and focused on the methodology, to help navigation.

This blog will continue to cover results from the method, and comparisons with “official” temperature reconstructions.

The Rumble at Rutherglen

Author: Dr. Michael Chase

Photo above: A recent picture of the weather station at Rutherglen, Australia, from the BoM webpage cited below. Other photos are shown at the end of the post.

Post Summary and Conclusions

This post documents some analysis of changes in minimum temperatures (Tmin) at Rutherglen, a rural weather station in South East Australia. It is found that:

• Early 20th century Tmin measurements are around 1.0C higher (annual average) than those that would have been measured if the recording system/location/environment of today had been in place then. There is some variation between months that make up this annual average.
• The annual average ACORN-SAT(2012) correction of 1.7C for early data is therefore substantially too high
• The daily ACORN-SAT(2012) corrections for 1920/21/22 (the only years examined) show a nonphysical discontinuity between the end of November and the start of December

Background

Rutherglen (BoM id 82039) is a rural weather station at a research farm, with no nearby man-made structures, at least from 1975, as revealed by photos and descriptions from the BoM webpage given below:

http://www.bom.gov.au/climate/change/acorn-sat/rutherglen/rutherglen-station.shtml

The RAW Tmin data from Rutherglen, and from many nearby stations, show a net cooling over the last 100 years, as revealed in the following figure:

Questions have been asked about why the raw temperature trend of net cooling has been adjusted in ACORN-SAT to a net warming trend, and the BoM have responded with the webpage cited above.

ACORN-SAT Corrections

The dates and sizes (annual average) of Tmin corrections applied by ACORN-SAT(2012) are given in the following extract from its adjustment summary document:

A later (September 2014) summary from the BoM about Rutherglen does not mention the 1928 Tmin correction:

http://www.bom.gov.au/climate/change/acorn-sat/documents/station-adjustment-summary-Rutherglen.pdf

but it is unclear if that correction has been disowned (without saying so) or simply not mentioned. The original 2012 documentation is taken to be definitive, as it matches daily temperature data available in October 2017.

Data prior to the last-listed correction in 1928 is reduced, on average, by 1.7C, the sum of all corrections. The following figure shows the daily corrections for 1920/21/23 (the only years examined):

The corrections appear to change in jumps from month to month, in particular with a very large jump (marked A in the figure above) from November to December, surely an undesirable and erroneous artifact rather than a genuine weather phenomenon.

My Analysis

I have estimated the monthly average corrections that would be needed to be applied to raw Rutherglen Tmin data to remove non-climatic influences relative to those present in recent years. The methodology is being documented in a separate blog:

https://diymetanalysis.wordpress.com

The following figure shows the annual average correction needed for periods of data (the bold blue lines are the moving averages) deemed to be stable, tracking the regional average (in red) reasonably closely:

The required correction is the temperature difference between the bold blue and dashed red lines, which are respectively the 15-year moving average of raw Rutherglen Tmin data, and the 15-year moving average of the regional average temperature variations. The figure also shows the 12-month moving average of weather-corrected raw Tmin data at Rutherglen.

The key features of the data shown in the figure above are as follows:

• 1914 to 1926: The average correction needed for Tmin data in this early period of stability is around 1.0C, the ACORN-SAT(2012) correction of 1.7C is too much
• 1914: There was a step change in temperatures, probably associated with the station move in January 1914 (source: Torok thesis 1997), a move that fails to get a mention or a correction in ACORN-SAT(2012)
• 1928: There was a step change in temperatures around 1928, but they recovered around 1936. ACORN-SAT (2012) has the step down in 1928, but not the recovery in 1936, an example of errors caused in ACORN-SAT by transient perturbations.
• 1966: There was a large drop in temperatures
• 1974: There was another drop in temperatures, but note that this was the date of some heavy rainfall (see below), and the temperature drop looks a bit like the sharp edge of a sawtooth perturbation
• 1984: This marked the start of a long period of stable temperatures with a trend matching that of the regional average
• 1998 (29th January): This was the date of a switch to an AWS system, which does not appear to have had a significant impact on measured temperatures
• 2012: There was a drop in temperatures at that date, possibly associated with a period of heavy rainfall, more on that below

The regional moving average temperature history was derived by averaging periods of stable temperature (such as the ones shown above in bold for Rutherglen) across stations in the region.

Monthly Corrections

The following set of figures show eyeball-estimated corrections for each month, being the average temperature difference between the raw data (in black, red for its average) and the regional average (in blue/mauve):

The figures shown above confirm that the periods 1914-1966 and 1984 to 2012 were roughly stable in terms of non-climatic influence, justifying the use of these periods in obtaining the regional average temperature history. If a corrected (“homogenised”) version of Rutherglen Tmin data is required then early data (before 1966) must be reduced by around 1.0C, with some monthly variation in that figure.

Regional Average

The following figure shows more of the periods of data used to form the regional average temperature history:

The complete set of the data periods used in regional averaging at Rutherglen is shown here:

https://diymetanalysis.wordpress.com/2017/10/09/example-01-rutherglen-tmin/

Finally, the following figure shows a summary of the regional average Tmin and rainfall history back to 1885, indicating the heavy rain that may explain some of the anomalous changes in temperature around 1974 and 2012:

Conclusions: See the start of this post.

Photos

The following photo of the Rutherglen station is from the ACORN-SAT station catalogue:

Photos of the Rutherglen site from the BoM website cited above (click to enlarge):

 

 

 

Merge Errors in ACORN-SAT

Author: Dr. Michael Chase

Summary

I believe that it was James Hansen and NASA/NOAA co-workers who first pointed out in print the problem of historical non-climatic warming in temperature reconstructions, as depicted in the following diagram:

Source of the figure above: https://pubs.giss.nasa.gov/docs/2001/2001_Hansen_ha02300a.pdf

This post shows examples of the problem manifesting itself in station mergers in ACORN-SAT, and quantifies the sizes of the resulting errors.

Methodology

I construct regional average temperature histories, separately for each month, of monthly average daily maximum (Tmax) and minimum (Tmin) temperatures. Full details of the methodology are being documented in this blog:

https://diymetanalysis.wordpress.com

Normalising station data to the level of the most recent regional average reveals the time-history of non-climatic influence at the stations being merged, indicating the size of corrections that would be needed in the construction of composite station records.

Example 01: Wagga Wagga (NSW) Tmin

The following figure shows 12-month (and longer) moving averages of Tmin for Wagga Wagga Kooringal, and AMO, with consistent normalisation for both stations, i.e. allowing actual measured temperature differences to be seen:

 

The figure above also indicates typical sizes of temperature corrections that would be needed in the construction of a composite record. The composite corrections needed for early data (1910-25) are much lower than those applied in ACORN-SAT, which are shown in the following extract from its documentation:

Working backwards in time from the present, ACORN-SAT applies the following corrections to Tmin data:

• 1968 (-0.46C): There was indeed a step change in temperature of around this size at that date
• 1964 (-0.09C): The data are consistent with this small correction
• 1948 (-1.62C): It appears that this enormous correction is the consequence of the anomalous warming at Wagga Kooringal between around 1930 and 1950, a period that I mark as being too anomalous to include in the regional averaging process.
• 1928 (+0.43C): This correction achieves some damage limitation, but the net correction (the sum of all of them) applied to data before 1928 is -0.46 – 0.09 -1.62 + 0.43 = -1.74 C, which results in considerable over-correction

 

 

Regional Average Stations

The list of stations used in constructing the regional averages in the analysis above is as follows, indicating which ones were omitted:

id BoM-id From To
01 82039 1912 2017;… % RUTHERGLEN RESEARCH
02 74128 1867 2003;… % DENILIQUIN WILKINSON
03 74258 1997 2017;… % DENILIQUIN AWS
04 82001 1908 1986;… % BEECHWORTH COMPOSITE
05 82002 1903 2006;… % BENALLA
06 82170 2006 2017;… % BENALLA AIRPORT
07 72151 1871 1950;… % WAGGA WAGGA KOORINGAL
08 72150 1942 2017;… % WAGGA WAGGA AMO
09 80015 1881 2017;… % ECHUCA AERODROME
10 80023 1903 2017;… % KERANG
11 82053 1901 1987;… % WANGARATTA
12 80002 1907 1986;… % BOORT
13 77042 1899 1996;… % SWAN HILL PO
14 77094 1996 2017;… % SWAN HILL AERO
15 74034 1907 2014;… % COROWA AIRPORT
16 74110 1914 1975;… % URANA PO
17 72023 1922 2017;… % HUME RESERVOIR
18 74009 1907 1975;… % BERRIGAN PO
19 80043 1908 1977;… % NUMURKAH PO **** OMITTED ****
20 82016 1909 1976;… % EUROA
21 73127 1913 1975;… % WAGGA WAGGA AG
22 80091 1965 2017;… % KYABRAM
23 81084 1965 1985;… % LEMNOS **** OMITTED ****
24 80049 1940 1975;… % ROCHESTER **** OMITTED ****
25 72000 1907 1994;… % ADELONG PO **** OMITTED ****
26 74039 1947 1977;… % DENILIQUIN FALKINER
27 74069 1949 1969;… % MATHOURA STATE FOREST
28 75080 1914 1927;… % WANGANELLA
29 74106 1970 2017;… % TOCUMWAL AIRPORT
30 81049 1965 2017;… % TATURA
31 82138 1987 2017;… % WANGARATTA AERO
32 72097 1970 1986;… % ALBURY
33 82100 1968 1986;… % BONEGILLA
34 82056 1954 1968;… % WODONGA
35 82038 1903 1921;… % RUTHERGLEN PO **** OMITTED ****
36 82085 1903 1937;… % RUTHERGLEN VITI **** OMITTED ****
37 82034 1927 1969;… % MYRTLEFORD PO **** OMITTED ****
38 81057 1965 1975;… % YARRAWONGA PO **** OMITTED ****
39 81124 1993 2017;… % YARRAWONGA
40 75038 1940 1956]; % KOONDROOK **** OMITTED ****

More examples to follow later.

The Late 20th Century Climate Shift in SE Australia

Author: Dr. Michael Chase

Introduction

Monthly average surface air temperature data in South-East Australia (and probably in other regions) show a relatively sudden increase in maximum temperatures at the end of the 20th century. Unfortunately, this was also the time when the BoM introduced Automatic Weather Stations (AWS) at many of its sites. This post presents some data on temperature and rainfall changes around this “climate shift” and shows graphically that the calibration of the AWS systems in the area examined had close matches to those of the systems they replaced, at least at the level of monthly Tmax averages. The seasonal differences in temperature and rainfall variations may provide clues to the cause(s) of the climate shift.

Regional Average Temperatures and Rainfall

The figure below shows the climate shift in a region of NSW/VIC bounded by lines joining Mildura, Hillston, Wagga Wagga, Rutherglen, Echuca, Nhill and back to Mildura:

The data shown in the figure above represent estimates of the regional average temperature history, in this case for 6-monthly Tmax data. Details of how to estimate regional averages, detecting and correcting inhomogeneities, will be given in later posts.

I have examined the regional average temperature history for each separate month, and find that each month from September to February has a similar upward shift in Tmax near the end of the 20th century, so have averaged over this 6-month period to illustrate the phenomenon (red curves above). The other months all show a similar lack of anything special happening around that time (blue curves for the 6-month average).

There is normally a close association between Tmax fluctuations  and rainfall levels, but the following figure shows that there was no particular trend in rainfall around the time of the climate shift:

Are AWS Systems Involved?

Many stations in the region had AWS systems installed in the late 20th century, for example becoming the primary sensors in November 1996 at both Mildura Airport and Wagga Wagga AMO. Fortunately, many nearby stations retained their manual systems, and I have checked their temperature histories against those that switched to AWS.

The following figure shows the temperature history (12-month and 15-year moving averages, after subtraction of regional average temperature fluctuations) for 3 stations that switched to AWS, together with the regional average temperature history (black curves):

Note that there are no substantial deviations from the regional average when the AWS systems became the primary sensors. For comparison, the following two figures show the same data for 6 stations that did not get converted to AWS:

Conclusions

There may be calibration differences of a tenth or two degrees C at the level of monthly Tmax averages between the AWS and manual systems employed in the region, but not more than that. This conclusion is consistent with the absence of corrections for AWS installations in ACORN-SAT, the early one at Cape Otway being the only one that has a correction.

Later posts will look at how the climate shift varied around Australia, which may shed some light on cause(s).

 

 

 

Averaging Multiple Temperature Records

Author: Dr. Michael Chase

Introduction

This is the first in a series of posts with the general theme of “Do it yourself temperature homogenisation“. The full series of posts will outline a simple but effective procedure for turning raw instrumental temperature data (aided by any available station history and data on rainfall) into reconstructions of the background temperature history for any area with a “sufficient” number of weather stations, the sufficient number depending mostly on the quality and extent of the data.

The overall procedure involves visual detection of inhomogeneities, followed by averaging and integration of interannual temperature differences in which the inhomogeneous data is omitted. Effectively, large inhomogeneities allow visual detection, followed by removal from the data, small ones are suppressed by the averaging process used to obtain regional histories.

Averaging Multiple Records

This first post in the series deals with the final step of the procedure, the method by which multiple temperature records are combined to give a regional average temperature history. The following figure gives a schematic picture of what typical temperature data looks like:

The figure above illustrates the following:

• Black data: Historical urban warming, followed by a station move (typically to an airport or other out-of-town site), followed by a switch to automatic sensors
• Red Data: A good rural station, but with some missing data
• Blue data: A station with a transient perturbation, possibly of non-climatic origin, or possibly due to a period of localised heavy cloud/rainfall

It is assumed for the purposes of this post that the previous stage of the overall procedure has identified the inhomogeneities, marked all periods of “transition” within a computer program, and that the analyst has corrected the data for any large residual inhomogeneities at transition boundaries and at all ends (more on that last issue below and in subsequent posts). The computer program, under analyst control, then does the following:

• Either infills missing raw data or leaves gaps if there is doubt about the station history within a gap. Infilling can also be done manually.
• Computes all valid differences in temperature, separately for each month, between years N and (N-1). Valid differences are those that do not cross, or lie within transitions
• Extrapolate temperature differences for all stations with missing years using the average of valid temperature differences

The following figure illustrates the resulting full coverage of temperature differences:

The temperature differences can now be averaged again with the important feature that each station has a constant weight in the averaging process (1/3 for each station in the example shown in the figure above). Finally, the average temperature differences can be integrated forwards and backwards in time from any desired reference year/temperature to obtain average temperature histories for each month.

Error Analysis

The reason for extrapolating temperature differences for all stations can be seen from the example of the blue data in the figure above. In some cases the dip in the blue data will be deemed to be associated with a period of heavy rainfall, i.e. a genuine climatic effect. This genuine climatic effect will only produce the correct impact on the regional average (i.e. only within the period of rainfall) if the weighting of the blue data is constant in time, which is what results from the extrapolation of temperature differences for the red data.

Using interannual temperature differences avoids the need to estimate and correct most inhomogeneities, but there is a small price to pay for that substantial benefit, the problem of residual inhomogeneities at boundaries, illustrated in the following figure:

The figure above shows an example of a temperature record with perturbations from the regional average. The perturbation within the record is not really a problem, because the downward shift in temperature is matched by the exact reverse in later data, with the constant weighting of each record ensuring that the perturbation will not influence the end-to-end shift in average temperature.

The problem illustrated in the figure above arises from the inhomogeneous ends of the record (more generally from any boundary, including ones created by defining “transitions”), which can distort the end-to-end variation of average temperature if the inhomogeneities are not detected and corrected, which can be done manually (see a later post).