**********************************EXCERPT************************************
tim channon
The Antarctica AWS stations are not good. Stay with this a twist is coming.
The Harry data led me to start looking and what turned up was not good.
Turns out the AWS hardware is problematic.
Then I found what I am quite sure is a bus error and then lots of instances of hardware malfunctions. To make it worse, the radio links are awful, inadequate error detection. Try a 20C jump in a matter of minutes and then back.
Cutting that part of the story short, I had the high resolution data when BAS were fiddling around with their web published public version so I could back calculate, all highly amusing. Antarctica Treaty and BAS omitted to say they have to give the data to others and they make it available to the public.
Hence I discovered the hardware errors in the real data.
What’s this to do with the Arctic, surely they are poles apart?
Seeing that ancient mil spec A/D are likely or similar and being a former equipment designer I took a guess, grab calculator.
” The temperature jumps from 6.9 to 13.0 and back to 7.4 at 10:00 hrs and again 9.5 to 13.7 and back to 7.2 at 17:00 hrs. ”
Only takes condensation to do this, given it happens at room temperature.
13 – 6.9=6.1C not interesting. But this is american equipment.
6.1*9/5 = 10.98 and that looks likely, is basically internally F, the higher numeric resolution. Kind of things folks do.
13-7.4*9/5=1008
13.7-9.5*9/5=756
13.7-7.2*9/5=1170
Bet that unlike the BAS stuff which appears to use an old BCD interface this is binary and a duff bit, 1024, bit 9, 2^10
Is about right on resolution. Wouldn’t be surprised if 1.024 or 10.24 is the internal reference
Look for jumps around 5.7C
If a higher bit is dodgy, might get 22.8C jump.
However on looking at some of the examples, could be at least 3 bits playing games
11.4
11.4+5.7=17.1
One of the plots shown might have several instances where actually the temperature variation was rather boring.
It gets worse, the reading are probably averaged and faults are noisy so anything is possible.
Might not be the only pattern and so the entire record is suspect, other malfunctions are possible. The equipment housing probably has an air leak.
The above is guesswork so I might be very wrong.
*************************************EXCERPT*******************************************
I’m flabbergasted that temperature data collection can be prone to so many errors in so many dimensions from siting issues to human errors of data entry and misunderstanding the limits of statistics, though on to human analysis fabrications (intentional or otherwise). How can any of it be taken seriously?
Has anyone made a list or flow diagram showing all the steps and all the places that errors have been found to creep in?