To create realistic benchmarks we would like to reproduce times and locations of known sources of inhomogeneity as best we can. Please can you help us. If you know of any regional/countrywide changes to the observing system over time please can you list them here or point us to some documentation/reference. Any information is valuable - even if its quite vague.
Ideally we'd like to know:
WHEN - specific date or month or year or even decade etc.
WHERE - a region, a country, an international GTS/WMO change etc.
WHAT - a change in shelter, thermometer type, automation, observing time/practice etc.
HOW - are there any estimates of the size/direction/nature of the effect of this change?
Please post here and encourage others to do so. We then hope to reward you with some realistic error-worlds to play with.
Kate
Ideally we'd like to know:
WHEN - specific date or month or year or even decade etc.
WHERE - a region, a country, an international GTS/WMO change etc.
WHAT - a change in shelter, thermometer type, automation, observing time/practice etc.
HOW - are there any estimates of the size/direction/nature of the effect of this change?
Please post here and encourage others to do so. We then hope to reward you with some realistic error-worlds to play with.
Kate
3 comments:
Dear participants, please be careful, on some computers the comments seem to be deleted when you publish or preview them. Thus please write the comment in your favourite text editor and copy them into the comment box. If it does not work, please mail them to Kate Willett,
kate.willett@metoffice.gov.uk or Victor Venema, Victor.Venema@uni-bonn.de.
A few examples for Australia, none of them particularly 'clean'. All of these are documented in a paper just published in IJC, and at more length in a technical report on the Australian homogenised temperature data set project (http://cawcr.gov.au/publications/technicalreports/CTR_049.pdf):
Instrument shelters: progressive introduction of Stevenson screen – mostly early-mid 1890s in Queensland, South Australia, Tasmania and Western Australia, but not until 1906-1908 in New South Wales and Victoria. Lack of standardisation prior to change so no simple adjustment, but an inhomogeneity in maximum temperature in the order of -1 to -2°C was typical (usually more in summer, less in winter). More discussion in a paper by Linden Ashcroft et al. last year in the Australian Meteorological and Oceanographic Journal (http://www.bom.gov.au/amoj/docs/2012/ashcroft.pdf).
Observation time: current standard for both maximum and minimum temperature is 24 hours ending at 0900 local time. During 1932-1963, midnight-midnight used at major stations (~50 Australia-wide?), but 0900-0900 for max/1500-0900 for min elsewhere. Overall impact looks to be in the order of +0.05-0.10°C on Tmin at nationally averaged scale but up to 0.5°C at individual stations near the south coast, and with a particularly marked impact on the frequency of extreme high minima (Melbourne had only one minimum above 25°C in the 32 years 1932-63 but has averaged about 0.6 per year since then). No measurable impact on maxima. Introduction of daylight saving in some states from 1972 onwards introduced effective 1-hour observation time shift in summer but no significant impact.
Metric conversion: observations changed from Fahrenheit to Celsius in September 1972 (new thermometers issued throughout network). No clear evidence of shift in mean temperatures though difficult to assess due to lack of reference data, as well as shift from strong El Niño to strong La Niña in early 1973 producing strong natural variability signal at this time. Marked decline in incidence of rounding temperatures to nearest 1 or 0.5 degree after changeover which may affect occurrence of days over/under thresholds.
Conversion from manual to automatic stations: progressive over last 20 years (sometimes with site moves, sometimes without), with numerous stations changing over on 1 November 1996 (the automatic probe became the primary instrument as of this date at stations which had both types of instruments). Unlike many countries, there was no change in screen design – automatic probes were exposed in traditional Stevenson screens. No evidence of any systematic impact on temperatures if no site move involved.
Site moves: progressive tendency over time for sites to move from in-town locations to out-of-town locations (mostly airports) – proportion of sites in the homogenised (ACORN-SAT) network in towns dropped from 66% in 1930 to 20% in 2012. Highest frequency of moves was in 1940s (as aviation expanded during World War 2) and 1990s (often automation was the pretext for site moves outside towns). Impact of each move depends on local geography, but this systematic shift probably accounts for much of the 0.1-0.2°C difference between raw and homogenised Australian temperatures (at national scale) before 1940.
Dear colleagues,
I do not know how Victor has calculated the statistics for comparing detected platforms between real and surrogated HOME-networks.
Here, detection results with the latest version of ACMANT (www.c3.urv.cat) are shown for networks of HOME benchmark. In the first five networks of the Real section, less frequency of breaks were detected than in the surrogated data. But the striking difference is in the magnitude distribution of the breaks: no break with at least 0.8°C magnitude was found for these networks, while in surrogated series their ratio to the all breaks is near to 40%.
In real network 6 (it is a large network, it consists of 30 Tmin and 30 Tmax series) the relation is opposite. In that network the ratio of large breaks is even higher than in the surrogated data, although the difference is not so big than in case of the first 5 real networks.
From HOME surrogated data 4 networks of 15 series were selected, they are from the version of 200-network base (namely: network 1, network 2, network 4 and network 5).
One can see that when we work with real data containing a large number of large-size breaks, then the detected frequency of platforms of max. 5yr duration is twice as high as for the HOME surrogated data (at least with ACMANT).
A B C D E
First five real networks together 142 14.3 0.0 14.8 6.3
Real network 6 283 69.1 48.0 26.1 17.0
4 HOME networks of 15series 265 53.6 38.5 18.9 8.7
A = total of detected change-points
B = ratio of change-points of at least 0.5°C magnitude (percentage of A)
C = ratio of change-points of at least 0.8°C magnitude (percentage of A)
D = ratio of pairs of shifts with opposite signs and max. 10yr distance (percentage of A)
E = ratio of pairs of shifts with opposite signs and max. 5yr distance (percentage of A)
Each break was sorted maximum to 1 platform. For instance, if shifts of varying signs were detected in 1948, 1952 and 1954, then the shift in 1952 could be sorted to two platforms (1948-1952 and 1952-1954), but in such cases only the shorter option was chosen.
The results can be controlled by anyone, since ACMANT and HOME dataset are accessible and the use of automatic software is straightforward.
Victor in his last mail to Homogenisation-list about short platforms did not react to all the relevant indications of likely high frequency of short platforms, for instance did not comment at all the findings of M. Rienzner. He confessed that he was skeptic and advised to keep the frequency of short platforms very low. With the method that Victor suggests the frequency of platforms of shorter than 5yr (together the ones of detectable size and the ones of invisible) is deeply less than 1 per 100yr. I reiterate that we have no access to quantify accurately the frequency of short platforms, so I, as a private person, tend to respect Victor’s skepticism. Human made global warming also has many-many indications, but there are still climatologists who are skeptic on that issue.
However, ISTI prepares to make an experiment that is intended to obtain widely applicable results. I suppose that the results are expected to serve as reference about observed climate change and data homogenization issues. At present, 10 scenarios are planned, and how the things stand now, each of them will reflect Victor’s skepticism about short platforms.
I have the strong feeling that if the things do not change, the benchmarking group is only 1 step before to make a serious mistake.
Best regards,
Peter Domonkos
Centre for Climate Change
Univ. Rovira i Virgili, Spain
e-mail: peter.domonkos@urv.cat
Post a Comment