While having a look at temperature data for Harvey in Western Australia, approx 140km south of Perth, (BoM station Number: 9812) – I noticed the daily maximum readings for 2011 had a large number of even-number readings – readings with no decimal point – integer I think is the term. Looking closer I noticed the readings to a decimal point seemed to involve a large number of .5’s. Anyway – on downloading these and counting using Excel I get the following very strange distribution.
Does anybody know how widespread this is ? I will check the minimums when I have time.
At first glance it looks like
Spot on to line
just above line
halfway between the 2 lines
just under the top line.
That’s assuming a mercury thermometer is involved. If recording is electronic (thermistor) then something is fishy in Denmark (and surrounding areas).
If it is a mercury thermometer, then I like Graeme’s explanation. It is something I noticed when analysing other measurements (size, in mm, I think) recorded by an assistant.
Surely it can not be due to reading a mercury thermometer – at this BoM page there is a pdf download available under the map – the site is an – Automatic Weather Station. And Harvey is linked on the BoM Observations map. So I assume they would not be updating that at 9am and 3pm from observers contacting them.
Warwick,
the minimums are worse: .0 – 146, .2 – 33, .3 – 1, .4 – 3, .5 – 172, .6 – 2, .7 – 0, .8 – 59, .9 – 0 and 11 temperatures not recorded for the 2011 year.
Sorry, there are not that many days in a year, unfortunately. The number for 0.5 should be 99
Benford’s Law doesn’t seem to apply to either data set.
As an automatic station, you would expect the temperature to be read to 0.1 C. There are nil totals for .1 and .9, so obviously those results have been rounded up or down to give a whole number, but why the shortage of .3, .4, .6, and above all no 0.7’s? All they do is confirm that the meter reads in 0.1 increments.
Warwick,
have you considered that it might be a code? To people smugglers perhaps?
Last digit =0 Patrol boats at sea
Last digit =5 Patrol boats about to put to sea
Last digit =2 Start run into landing
Last digit =8 No more runs into landing
Digits 3,4,5 and 7 probably refer to other information. For example the total number of 7’s could refer to the chances of this Government coming up with a sensible and workable policy.
.
Snow Alert! With February being traditionally the hottest month and only a few days away, snow falls near Lake Mahinerangi, 35 kms from Dunedin NZ. Christchurch had a chilly 4 deg overnight low and people in usually much warmer Blenheim at the top of the S. Is. woke this morning to only 9 deg. This unseasonal “cold snap” now travelling North with all vigour and due to arrive in Auckland Friday late pm or early evening.
And Canada has just ikenovd its legal right and pulled out of Kyoto. Are you listening Mr Key? Why are you hampering NZ business?Julian
To Previous, All Mr Key has done is “water down” and delay re The ETS – at this stage. With parts of QLD & NSW flooded & Kiwis grumbling about January being cold & unseasonal (only 16.4deg national avg.), we should spare a thought for parts of (esp Eastern) Europe where people are actually DYING of the bitterly cold conditions in the coldest Winter temps since records began 100 years ago. With -32deg and the Black Sea freezing over, unusual snow falling on Croation Is in the Adriatric Sea, Helicopters deployed to evacuate people from snow-choked areas/drop supplies. Plus Northern Cities of W. Europe have been informed to brace themselves for the coldest Winter in 27 years. British forecasters warning that temps could plunge as low as -7deg at night.
They’re probably just rounding up, you know, like they did with Yasi being a category 5 ( winds in excess of 280kph), in spite of an engineer’s expert opinion that, based on the evidence of the damage caused, that the winds were not in excess of 220 kph!
Oh, BTW, its a year since Yasi and BOM still havent corrected the mis-classification.
I did see a reference to a “marginal category 5” in one of their reports on Yasi.
Maybe thats their idea of a correction?
Looks like they are faking thermometer/instrument precision to tenth of degree, when most of their equipment does not have it.
First of all, start with some background on rounding strategies. If rounding-to-closest uses a uni-directional scheme for ties (values exactly halfway between rounding points), then the rounding process will introduce a bias into the data set. For example if you have a data set of 1.5, 2.5, 3.5, and 4.5, and always rounded up for .5 to whole integers, then rounding will have added +.5 to every value of the data set. To prevent the bias, there’s a half-half strategies, most common of which is rounding to the nearest even value (1.5 to 2 and 2.5 to 2).
What .0 .2 .5 .8 looks like is 1/4 degree precision rounded to 1/10 precision. The other values are probably the odd thermometers that had 1/10 precision and there weren’t many of them. Some thermometers may be limited to 1/2 degrees or to a full degree.