Guest Essay by Kip Hansen
While researching for a future essay tentatively titled “Whither Original Measurement Error?”, I have been reading up on the origins of the modern meteorological thermometer. Fascinating stuff, those early scientific instrument makers and their creativity and engineering skills.
I came across an interesting little [e]book that was just the sort of thing I was looking for, written by John Henry Belville in 1850, who started work at the Royal Observatory, Greenwich, Kent, England, in 1811 as a meteorologist and was still at it 35 years later. Here I reproduce the Title Page and Preface from his book:
Included in this little volume is the following chart, which I offer here without comment for those interested in the fascinating study of the long-term Central England temperature record. This thirty-five year average, day by day, is fairly well guaranteed not to have been adjusted or modified in any way since its publication in 1850 and might have some use for comparison purposes.
* Concerning the decrease of the mean daily temperature from the 12th to the 14th of May, see Humboldt’s ‘Cosmos,’, vol. i. page 121. Bohn’s edition.
The small note under the chart was included in the book. It refers to something in the great tome: Humboldt’s COSMOS; or a Sketch of a Physical Description of the Universe. Copies are available online, but I was not able to trace the exact reference.
The full Belville book is available to read online free – albeit through Google Play’s eBook app — at:
http://books.google.com/ebooks/app#reader/9L0ZAAAAYAAJ
It is a rather stiffly worded, but enjoyable, trip into the scientific past.
# # #
Moderation Note: I would appreciate links from [any reader] to good sources for historical sources of information on expected measurement errors of meteorological thermometers in use from 1850 to present, including narrative sources of “operator error”. (Example: Several years ago, I did a Surface Station Project interview in Santo Domingo, Dominican Republic, in Spanish, on how the Stevenson Screen thermometers were read there. Acceptable expected error according to the Chief Meteorologist? +/- 1 °C)
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
Reply to BioBob ==> “only valid means allowing scientists to reveal the magnitude of error” We are obvious speaking different languages. I believe you are speaking “statistics” and I am speaking simple physical measurement — taking a temperature with a thermometer today (or any ‘today’ in the past), reading a thermometer, how to tell if a thermometer is accurately calibrated. If this explanation doesn’t clear up our disagreement, then we’ll just have to let it lie for now.
Reply to Robtin ==> No worries, mate! Glad we cleared it up.
Reply to rgb ==> Thank you for your informative (and entertaining) contribution on the effects of random and non-random errors introduced into temperature records. In the case of the Santo Domingo station temperature records, the errors depended greatly on the weather (if it was raining, the reading was done very haphazardly, a glance only) and on the height of the man on duty (as explained by the Chief Meteorologist – who was perfectly aware of the variables) which changed the angle at which the thermometer was actually seen by the recorder (many Dominican’s are quite short). The Chief tried to get the short ones to stand on a concrete block (which was to hand) to take the readings, but he was aware that they resisted this ‘demeaning’ solution.
Reply to Chris Martin ==> I’ll admit I haven’t read the whole little book yet. I will and I’ll see if he mentions which type of thermometer is in use at the Royal at the time. Sixe’s was pretty common as a Max-Min recording thermometer, and as I mentioned in another reply, as used on the great HMS Challenger voyage in 1874. Whenever I get to it, I’ll post the information here. As has come up from helpful readers, thermometers in this time period were fairly easily calibrated for freezing and boiling temperatures, but the in-between temps were less precise due to the unevenness of the size of capillary tubes for the mercury.
richard says:
February 23, 2014 at 8:10 am
bit chilly in those days.
I checked the average means in the CET for those years and compared them with 2013. Here are the results for each month. Six months of the year were colder in 2013. (-2.8 isn’t an error)
0.9 -0.9 -2.8 -0.4 -1.0 -1.0 2.6 1.5 0.7 2.7 -0.0 2.2
Reply to Kelvin Vaughan ==> Very Interesting. Many other readers had noted that they though the early 1800s appeared chilly, and that Global Warming had warmed up Central England a bit, about 0.5°C — but now you come along and ….hmmm… I was going to say dispel that …. but I’m getting out my little calculator…..== +0.375°C to a rough-back-of-envelope 12-moth-average is a wee-bit warmer, with some months showing exceptional Coolth (especially Spring and early Summer) and Warmth (July, Oct and Dec) , comparatively. Have I got that right?
Kip Hansen
For records in Canada, the HBC Hudson’s Bay Company (some people call it Here Before Christ), kept temperature records going way back. In 1994 they donated their records to the Archives of Manitoba. Some of these can be searched on line.
http://www.gov.mb.ca/chc/archives/index.html
rgb – your contributions here always raise the quality average by a statistically significant extent and introduce a long term upward bias in quality potential as a result of emulative, educational, and inspirational effects. I’m currently working on comparisons between the quantifiable results produced by your data x logic x argument + (clarity of expression squared)/reader numbers, and the actual anomalies of such in order that a clearer approximation can be gained of the approximate proximity of useful proxies for determining the utter uselessness of much so called scientific thinking/writing these days. You are saluted again sir . . . lang may yer lum reek . .
Reply to ES ==> Thank you, very helpful. [My wife and I have a red Hudson Bay Blanket, three stripes, for frosty nights. In ten years, we’ve needed it once. 🙂
Not only are there random reading errors, but thermometer bulbs shrink in the first few years:
http://tinyurl.com/pcaqd2k
The temperature data needs to be adjusted before use.
How prescient is that book;
rgbatduke on February 24, 2014 at 6:29 am
Thanks for taking the time to write that.
I have spent much of a 33 year career in underwater acoustics sweating absolute sound pressure level calibrations and your post clearly speaks of “been there, done that, got the tshirt.”
I have spent much of a 33 year career in underwater acoustics sweating absolute sound pressure level calibrations and your post clearly speaks of “been there, done that, got the tshirt.”
No, I avoided doing that — I’m a theorist. We reason in a cozy ivory tower where we presume that we know exactly what’s going on in a vastly oversimplified model that we hope catches the essential dynamics and works at least approximately to describe nature in some specific neighborhood. The only errors I worry about are things like cumulative roundoff error and algorithmic stability, averaging over carefully specified stochastic noise, etc.
But I teach classes with an experimental component, I have many experimental friends, and I’m not an idiot. Also, I probably know a lot more than most (even in math/science) about randomness, probability and statistics, as I’ve done decades of work in Monte Carlo, Langevin dynamical models in optics and magnetism, and uber-advanced predictive modelling. So while I feel your pain, I do it from a comfortable armchair where — no, I really don’t:-)
I do, however, appreciate the fact that “empirical truth” is a just as much an oxymoron as “theoretical truth”. They are “truths” in a precisely describable (but not easily quantifiable) statistical sense, not in the sense people use when doing boolean/aristotelian first order logic.
rgb
Reply to Allensworth and rgb ==> The ever running distinctions between the armchair and the field tent. I admit to having done a bit of both but tend towards the armchair (well, deck chair) now. At present, I am concerning myself, in this particular effort, at discovering facts about the physical errors in the day to day temperature taking and recording in historical climate science. I appreciate your contributions.
Reply to Chris Martin ==> John Henry Belville says in his book “A Self-registering Thermometer of simpler construction, by Dr. John Rutherford of Edinburgh, is now in general use.” This implies, but does not definitely state, that the Rutherford registering thermometer set was used at the Royal. Found on Page 7 of Belville’s book..
That’s interesting. I didn’t know that. The other – perhaps significant point – is that the definitive book by Brazell of London Weather, published in 1968 only published detailed temperature figures from the Greenwich Observatory from 1841 onwards, so that might imply that temperature figures before then may have been from non-standard instruments and/or exposures.
Further to Chris Martin ==> If you have an interest in the exposure issue, there is discussion of this in Belville’s book. It’s free online and the chapters are listed. The whole thing is only sixty pages and can be read in half an hour or so. I recommend using the Chrome browser and the Google Play eBook app therein, a slightly better interface (not my usual browser either) for this purpose.
Further to Kip Hansen===> OK. I may have a look at Belville’s book. Historical weather records have always been an amateur interest of mine – I used to live in the same county of England as Thomas Barker , who made detailed daily weather observations from 1733 to 1795, including temperature. This in a location about 100 miles north of London. This was a key source for Manley when he put together the Central England Temperature Series. It’s when you look at Barker’s register you realise there is a whole bunch of issues prior to the 19thC around type of thermometer (in Thomas Barker’s case a spirit in glass thermometer made by John Patrick of London with zero at 90F and 76 at 32F!!!!!); exposure (often they were kept in an unheated room with open windows) and observation (often it was two temperature measurements a day, before max-min thermometers were invented). Makes me admire Manley’s work even more that he was able to make something out of these and other records. I do sometimes wonder whether some of the higher extreme temperatures reported nowadays are also partly down to the speed with which electronic sensors react to temperature compared to the old spirit or mercury thermometers – and especially if many max temperatures before about 1800 were, I suspect, based on a single observation at 3-5 pm!
Kip Hansen says:
February 24, 2014 at 10:28 am
Reply to Kelvin Vaughan ==> Very Interesting. Many other readers had noted that they though the early 1800s appeared chilly, and that Global Warming had warmed up Central England a bit, about 0.5°C — but now you come along and ….hmmm… I was going to say dispel that …. but I’m getting out my little calculator…..== +0.375°C to a rough-back-of-envelope 12-moth-average is a wee-bit warmer, with some months showing exceptional Coolth (especially Spring and early Summer) and Warmth (July, Oct and Dec) , comparatively. Have I got that right?
Yes, it has got mild again and has been mild so far this year as the jet stream has got stuck in mild mode. Over this decade It’s more a case of we are loosing the cold end of the spectrum rather than the hot end climbing up.