14 Jan
2021
14 Jan
'21
3:45 p.m.
E. Liddell wrote:
we are not writing a PhD here - we want v1.0 v2.0 etc. so lets start simple.
I just don't feel like doing this myself.
If you know where the name belongs to it would be much easier - to me the
names do not say anything (not Canadian)
What can I say? Canada is the second-largest country
in the world. We
have vast tracts of thinly populated land (and vast tracts of largely
unpopulated land too), and weather stations are sprinkled all across that.
In a lot of cases, though, a breakdown by country is going to be
sufficient. A small African nation might have 20 stations or so, which
isn't too much for one list.
Yeah it looks like we need a TDE KWeather committee to rule out the division
of the world :D
There are 6702
in the file from which 2590 are already done because they
are are in the US.
A big portion of the rest is just countries (like Europe etc)
Africa, Russia, Mexico are good candidates.
I could easily map everything from Canada to one country Canada, but it
will be difficult to find the right station. What do you think
For the Canadian stations specifically, it should be possible to allocate
a lot of them to provinces based on the annotations at the end of the
"city" field, which are mostly the old conventional provincial
abbreviations that everyone except the post office actually uses:
N. S. = Nova Scotia
N. B. = New Brunswick
PEI = Prince Edward Island
Nfld = Newfoundland
Que. = Quebec
Ont. = Ontario
Man. = Manitoba
Sask. = Saskatchewan
Alta. = Alberta
B. C. = British Columbia
N. W. T. = North-West Territories
Y. T. = Yukon Territory
There's some variation in capitalization and punctuation, and locations in
Nunavut are likely to marked as being in the NWT for historical reasons,
but it's a start and should be able to assign most stations in the more
populated areas to a province using a few regular expressions for
matching.
Some of them can also be assigned automatically based on
latitude/longitude ranges (north of a certain latitude mostly places
things in the Territories, south of it in the provinces, and Québec and
provinces westward *almost* adhere to specific longitudes, although there
are some ambiguous zones—I'll see if I can get some numbers tonight).
The rest of them can be dropped in a bin labelled
"Canada - Unknown" or
the like and assigned whenever someone figures out where they are. (If we
can find a way of making these assignments easy for a non-technical person
to do, it might be helpful.)
This is mainly a concern for really large countries (Canada, U.S.,
Russia), where if you pick the wrong weather station by mistake, you could
be getting forecasts for the other side of the continent.
Yes, I was going to suggest for version 1 to map the stations to something
like unknown. For example Mexico is done like this Mexico -> Mexico. This
will become Mexico -> Unknown
I could of course preserve the already used associations.
thank you and regards