-17.8 low temps (loads of them)!

[Vegit8]

Hi
My friend https://www.charlesdowding.co.uk/weather/is running MX on a Ubuntu laptop connected to a WLL device. Having moved MX from running on an iMac (where there were issues).
The laptop does nothing except for running MX
He is not running the absolute latest build of MX.

A few days ago a spurious ultra low temperature appeared (-17.8).
I went round to his house yesterday and edited all the records (using the webpages provided by MX) where -17.8 was visible (All time, Monthly, this month, this year.) I could not find a day file editor.
On checking this morning (yes its early) on the betelsummary pages there are loads of the blighters 12 in May and 1 in April....
Where have these come from?
Why are they all the same value?

Looking at the data on Weatherlink.com there are no instances of this low value at all.
I know I can force a full download of the data from WLink, but that will wipe the sunshine hour data, and will also not fix the problem from recurring.

Any ideas???

[beteljuice]

On checking this morning (yes its early) on the betelsummary pages there are loads of the blighters 12 in May and 1 in April....
Where have these come from?

The one place you didn't edit - the clue is in the real name of the utility "Yet Another Dayfile Reader"

What put the values there ?????? - who knows

[Vegit8]

Ahhh
I didnt know the 'reader' was an editor!

Ok - I have now found the dayfile and data editing facility under Data Logs...
So I can fix the strange values.
Would still love to know where -17.8 comes from though --- Mark?

[beteljuice]

I didnt know the 'reader' was an editor!

?? It isn't - but it's good for finding 'bad' data

... but glad you've found the errant entries.
Was there any other corruption / problems ?

[sfws]

I don't know the WLL, so this answer might not apply in your case. But I hope it is a good guess.

I do know that some weather station don't supply Cumulus with a temperature using a scale like Celsius or Fahrenheit. By the way, as you will see later, I'm guessing that the WLL does use Fahrenheit, but let me carry on with this general point to help people. Instead they supply something that has to be converted to get a temperature, using a formula like

output = multiplier X input + constant X input

Also if you are asking for an output in Fahrenheit, or Celsius; but the station supplies Celsius or Fahrenheit respectively, Cumulus again uses a formula like above to do conversion to units you want to see.

Now it just happens that 0 o F is -17.8 oC; so my guess is some sort of fault is making the WLL unable to supply the correct temperature, my guess is that the WLL uses the Fahrenheit scale and this electronic fault, or whatever it is, is forced to report zero. Cumulus reads this and converts it to ‪-17.77778‬ on the Celsius scale (I am guessing that is what you are quoting) and that is why you see what you report.

[Vegit8]

Thanks SWFS,a plausible explanation.

Now .... this is the totally daft bit.
When I originally posted about 5 hours ago there were 12 instances of -17.8.
At 09:00 there are only 4 instances. The rest have disappeared.
I have not been to the site, and I know for a fact that the user will not have edited the values out.

How can MX recalculate values on its own once they have been recorded?
As I said in post one, the underlying data posted to Weatherlink.com is accurate with no spurious values.

Also in a similar theme the sun hours had disappeared from a number of days in May earlier. These too have now returned.

Most odd

[sfws]

Now .... this is the totally daft bit.
When I originally posted about 5 hours ago there were 12 instances of -17.8.
At 09:00 there are only 4 instances. The rest have disappeared.
I have not been to the site, and I know for a fact that the user will not have edited the values out.

How can MX recalculate values on its own once they have been recorded?

Again, I can make a guess, but it might be wrong.

When MX does a rollover, it reinitialises today.ini contents as well as other tasks. Consequently, any spurious values it might have been working with the previous day will no longer be being processed. If the change happened at 9am, maybe that is the rollover time being used.


See https://cumuluswiki.org/a/FAQ#.28Invent ... in_figures for some text originally entered by Steve Loft about this sort of issue (although in that case talking about rain)

[beteljuice]

When I originally posted about 5 hours ago there were 12 instances of -17.8.
At 09:00 there are only 4 instances. The rest have disappeared.

If you were / are seeing these in the data summary (YADR), then they must be entries in the dayfile.txt
The only kind of updates that happen to this file are new additions.

So ...
If you edited the file yesterday and did not restart Cumulus or upload the file manually, then your edits wouldn't have been visible until after station rollover.
That also suggests you missed four

I'm sure somewhere recently there was comment on if shutdown / reboot was not done correctly causing problems, but I would suggest putting in latest version and checking all settings first.