by Euan Mearns
UK blogger Paul Homewood and Telegraph columnist Christopher Booker have managed to stir public interest in the veracity of adjustments made to temperature records by the Global Historical Climatology Network (GHCN).
The focus lies in adjustments made to GHCN V2 data in the homogenised GHCN V3 data that was released 2011. Pair wise homogenisation is supposed to detect and remove non-climatic artefacts from the data caused by, for example, moving a station, a tree growing and providing shade or a change in the thermometer.
It is useful at this point to read what NASA GISS have to say on their FAQ page (it was Gavin Schmidt that pointed me to this information).
To recap, from 2001 to 2011, GISS based its analysis on NOAA/NDCD’s temperature collection GHCN v2, the unadjusted version. That collection contained for many locations several records, and GISS used an automatic procedure to combine them into a single record, provided the various pieces had a big enough overlap to estimate the respective offsets; non-overlapping pieces were combined if it did not create discontinuities. In cases of a documented station move, the appropriate offset was applied. No attempt was made to automatically detect and correct inhomogeneities, assuming that because of their random nature they would have little effect on the global mean.
Using the excellent web platform provided by NASA GISS it is possible to access GHCN V2 and GHCN v3 records, compare charts and download the data. It does not take long to find V3 records that appear totally different to V2 and I wanted to investigate this further. At this point I was advised that the way homogenisation works is to adjust records in such a way that a warming trend added in one station is compensated by cooling added to another. This didn’t sound remotely scientific to me but I clicked on Alice Springs in the middle of Australia and recovered 30 V2 and V3 records in a 1000 km radius and set about a systematic comparison of the two. The results are described in detail below.
In summary I found that while individual stations are subject to large and what often appears to be arbitrary and robotic adjustments in V3, the average outcome across all 30 stations is effectively zero. At the regional level, homogenisation does not appear to be responsible for adding warming in Australia. But the thing that truly astonished me was the fact that the mean temperature trend for these 30 stations, 1880 to 2011, was a completely flat line. There has been no recorded warming across a very large portion of the Australian continent.
Some final notes on nomenclature. NASA GISS refer to GHCN V2 as unadjusted while in fact NOAA say the V2 data have been subjected to adjustments, a fact borne out by my subsequent work on Iceland. In my charts and text I refer to GHCN V3.1 while in fact I’m unsure what the version is that I accessed via the NASA GISS web platform. It is the data used in GISS temp. And finally, those offended by my averaging of raw temperatures in the post below will find an anomaly plot at the end of the post.
For reference, Figure 1 shows a map of the region surrounding Alice Springs.
Figure 1 A 1000 km radius around Alice Springs. Many of the station names (Figure 2) can be found on the map.
In this comment, Sam Taylor pointed out that the way homogenisation works is to modify data in groups of stations and that to get a proper picture of its effect it is necessary to look at a regional group. So I clicked on the middle of Australia and got the list of stations below. This series of posts began with Roger Andrews in Alice Springs. This has been a lot of work. One of the main conclusions is that homogenisation has not biased this regional group of records.
Before proceeding, lets see how homogenisation is defined. First Wikipedia:
Homogenization in climate research means the removal of non-climatic changes. Next to changes in the climate itself, raw climate records also contain non-climatic jumps and changes for example due to relocations or changes in instrumentation. The most used principle to remove these inhomogeneities is the relative homogenization approach in which a candidate station is compared to a reference time series based on one or more neighboring stations. The candidate and reference station(s) experience about the same climate, non-climatic changes that happen only in one station can thus be identified and removed.
And this from the NASA GIS FAQ page.
UK Press reports in January 2015 erroneously claimed that differences between the raw GHCN v2 station data (archived here) and the current final GISTEMP adjusted data were due to unjustified positive adjustments made in the GISTEMP analysis. Rather, these differences are dominated by the inclusion of appropriate homogeneity corrections for non-climatic discontinuities made in GHCN v3.2 which span a range of negative and positive values depending on the regional analysis. The impact of all the adjustments can be substantial for some stations and regions, but is small in the global means. These changes occurred in 2011 and 2012 and were documented at that time.
Figure 2 The system allows you to select a reference station and provides a list of surrounding stations. This printout from GHCN V3.1 is the list of stations analysed down to Larimah.
- A comparison of raw temperature records (GHCN V2) and homogenised temperature records (adjusted records GHCN v3.1) is presented for 30 climate stations (Figure 2) within a 1000 km radius of Alice Springs, Australia. The adjusted records are subtracted from the raw records which illustrates the degree of adjustment for each station.
- 29 of the 30 stations have been adjusted to a greater or lesser extent. Only Farina has no adjustments.
- The size of the adjustments increases back in time and are occasionally large, up to ±1.5˚C. Temperature trends are adjusted by either warming or cooling the past.
- In 29 records, adjustments are near ubiquitous and are frequently exact decimal fractions, for example exactly 0.5˚C. For individual stations, it is usually very difficult to reconcile the pattern of adjustment made to any geographic or historic system. Homogenisation has also deleted at least 85 annual records that hinders comparison of the two data sets.
- In Alice Springs the raw record is flat and has no sign of warming. In the adjusted record, homogenistaion has added warming by significantly cooling the past. Five other stations inside the 1000 km ring have similarly long and similarly flat records – Boulia, Cloncurry, Farina, Burketown and Donors Hill. There can be no conceivable reason to presume that the flat raw Alice Springs record is somehow false and in need of adjustment.
- Six records show a significant mid-1970s cooling of about 3˚C (Alice Springs, Barrow Creek, Brunette Down, Cammoo Weal, Boulia and Windorah) that owing to its consistency appears to be a real signal. Homegisation has tended to remove this real temperature history.
- The average raw temperature record for all 30 stations is completely flat from 1906 (no area weighting applied). There has been no measurable warming across the greater part of Australia. The main discontinuity in the record, pre-1906, arises from there being only 3 operating stations that do not provide representative cover.
- The average temperature trend for the 30 adjusted records is also flat and not materially different to the raw record. Hence, wholesale adjustments have not significantly biased the regional record. This raises the serious question of why GHCN have adjusted individual records in a way that introduces trends that do not exist and removes trends that do at the individual station level? The individual GHCN V3.1 records are not temperature records but carry a coded temperature signal that only makes sense when amalgamated with similar code from neighbouring stations.
Figure 3 The chart summarises the adjustments made to the 30 station records showing V2 raw record minus V3.1 adjusted record. It shows clearly how adjustment is near ubiquitous although there are often segments of a record that are not adjusted. Note Farina (red) is the only station with no adjustment. Note also how the scale of adjustment tends to expand back in time.
Figure 4 Example of individual station record adjustment. The raw record for Barrow Creek was flat. Adjustments have cooled the past to create a warming trend. Note the style of flat line decimal fraction adjustments. Also note the significant adjustment to the mid 1970s data that tends to remove a real cooling event observed in several stations.
Figure 5 Somewhat surprisingly, since 1907 the raw temperature record for this large part of Australia is completely flat (Figure 6). There has been no warming. (note no area weighting). Pre 1907 there were only three operating stations and this imparts bias to the record. Mid 70s cooling is observable. There were no large volcanic eruptions at the time but VEI4 eruptions in 1973 (Tiatia), 1974 (Volcan de Fuego), 1975 (Tolbachik) and 1976 (Mount Agustine).
Figure 6 A regression through the post-1907 data is completely flat.
Figure 7 Averaging the dT records for 30 stations (Figure 3) shows that since 1906 no significant trend or bias is introduced. But the negative dip in the mid-1970s removes what is likely a real climatic signal. I’m unsure what impact the large pre-1906 bias may introduce but suspect that this may be removed by expanding the area that would increase the number of pre-1906 stations to a representative level.
Figure 8 Prior to 1906 there were only 1 to 3 operating stations. In 1907 that number increased to 7 and the temperature signal settled on a representative regional average. The number of stations then grew steadily to a maximum of 27 in 1972. Then in 1993, there was massive station closure, down to 6 is barely enough to provide representative regional cover.
Figure 9 Following from Figure 7, it is difficult to spot the differences between the raw and the adjusted record. There is less variance in the homogenised data which I guess is what homogenisation does but I suspect that real climate signal has been smoothed out, in particular the possible mid-1970s cooling event.
Figure 10 Six stations record a rather similar style of mid-1970s cooling that seems it could be a natural signal that homogenisation has removed (V2 unadjusted records).
Figure 11 Six stations with old records do not show warming. Notably Farina was the only record to have no adjustments made. There is no evidence for warming or cooling anywhere and therefore no justification to add warming or cooling artificially using homogenisation (V2 unadjusted records).
Homogenisation of climate records changes virtually everything and nothing at the same time. The objective of homogenisation is to remove non-climate artefacts. Wholesale re-writing of the temperature history everywhere is not consistent with the stated aims. Homogenisation appears to have added warming or cooling to records where neither existed. Homogenisation may also have removed real climate signal.
I find zero warming over such a large part of the Australian continent to be a surprise result that is consistent with Roger Andrew’s observation of no to little warming in the southern hemisphere, an observation that still requires more rigorous testing.
There is no evidence in this data set to support the more serious allegation that has been made for GHCN and NASA GISS adjusting records to manufacture global warming. Individually, the GHCN V3.1 records cannot be treated as climate records since each one contains fragments of code designed to create regional homogeneity.
It seemed prudent to have an anomaly chart, so here it is. Doesn’t change anything, the average temperature series are completely flat from 1880 to 2011.
Acknowledgements. I need to acknowledge the very substantial contribution made by my blogging partner, Roger Andrews, who many years ago compiled a large number of “raw records” that showed scant evidence for warming across the whole southern hemisphere. He sent me his spread sheet and Roger’s results are summarised in his recent post Homogenizing the World.
JC notes: This post was submitted via email.As with all guest posts, keep your comments on topic and civil. In terms of moderation, this post will be treated as a technical thread.