Signature projectSame days, two thermometers: what moving the instruments did to the readings
Met Éireann's rescued Phoenix Park records for 1842 to 1847, when the minimum and maximum thermometers were read inside the new observatory and outside it, compared day by day. Temperatures in degrees Celsius.
Phoenix Park 1842 to 1847: inside the observatory against outside| Measure | Days with both | Inside compared with outside |
|---|
| Daily minimum | 2,034 | 0.83 degrees higher on average |
| Daily maximum | 1,978 | 0.42 degrees lower on average |
| Minimum higher inside | 2,034 | On 92.4 percent of days |
Warmer nights inside
Inside the observatory the overnight minimum averaged 0.83 degrees above the outside reading, and it was higher on 92.4 percent of the 2,034 days read both ways. Shelter keeps the coldest part of the night away from the bulb.
Cooler days inside
The daytime maximum went the other way, 0.42 degrees lower inside across 1,978 paired days. Housing that softens both ends of the day narrows the daily range without any change in the weather.
Why pairing works
Each of those differences sets two readings from the same day against each other, so that day's weather cancels out. What remains is the effect of where the thermometer hung, which is exactly the question.
Compare the decades and the answer is muddled
Set the last ten years at the window, 1832 to 1841, against the first ten in the observatory, 1843 to 1852, and summer nights appear about 1.5 degrees colder after the move, averaging 11.99 before against 10.51 after, while winter nights hardly shift, 2.95 against 2.88. Part of that gap is the move and part is simply two different decades of weather, and nothing in those numbers can say how much is which. The paired years can. That is the lesson in one line: when a method changes, only measurements taken both ways at the same time separate the change of method from a change in the thing being measured.
The station history Met Éireann publishes with the data explains why a move would matter. Until 1 March 1842 the instruments hung in a window recess on the second floor, shut off from the room by glass folding doors, and the practice was to open the window for some minutes before reading them; the registers note each autumn that "Fires were commenced in the room for the Winter". It also records corrections: a minimum thermometer used from 4 March 1835 to 9 November 1836 was "stated to have been too low by 2", in the Fahrenheit degrees the record was kept in, and its readings were raised accordingly. For 1855 to 1857 both the raw and the corrected readings survive, and the corrected maxima average 0.36 degrees Celsius lower. A series that looks like one long column is really several instruments, several places and several sets of corrections.
The learner's program loads the main file and the outside-observatory file, joins them on the date, keeps only days where both carry a reading, and subtracts. The same program then runs the unpaired version, one decade against another, and prints both answers side by side with the number of days behind each. Seeing a 0.83-degree paired effect next to a 1.5-degree decade gap, for the same move, is the point.
Gaps belong in the answer
The 1830s carry minimum readings for 2,543 of 3,287 days, the registers noting interruptions "owing to breakages", and the station history finds no traceable daily registers at all for 1853 and 1854.
A frozen Liffey
The notes keep the observers' own remarks too. On 14 March 1845: "The frost was so intense that the River Liffey was frozen over, above King's-bridge".
From 1880, a standard screen
The history quotes the Irish Meteorological Service in 1983: since 1879 to 1880, when the thermometers were housed in a Stevenson screen, "the data may be considered to have a high level of accuracy".
