Re: Weather bot
Milos Rancic <millosh <at> gmail.com>
2009-06-19 11:26:03 GMT
On Thu, Jun 18, 2009 at 5:07 PM, Brian
McNeil<brian.mcneil <at> wikinewsie.org> wrote:
> If you go up a page from Milos’ detailed list there’s a world map with
> temperatures. This looks very like the old WeatherChecker output but I think
> it’s really overlaid temp. numbers. The whole thing looks really good, but
> pictorial form is what most people are used to digesting weather data in.
The page above is not maintained by bot. Bot is in the completed phase
related to the algorithm, but far from completed in relation to all
possibilities which may be added.
> How hard is it to convert these detailed figures into a map system that uses
> image maps and can be drilled down into a more local form?
Configuration page for each place assumes geographical coordinates.
Because of using Sun-based and Moon-based icons, I had to calculate is
there a night or a daylight. That means that bot already has
everything needed for making a map.
About openstreetmap, which mentioned Bawolff: yes, it is fully
possible to connect it.
> I’d expect to be able to go from World to one of our defined geographical
> regions (eg Oceania or Europe) and from there down to a country level (eg
> UK).
>
> Part of doing this would be altering the maps in some way to roughly mark
> out the areas that can be drilled to. Take a look at the BBC news page’s
> regional map for how they’ve done this. At regional level you’d want maps
> with country outlines, hover text would then be good to see where the final
> drill-down would go.
The most useful method for dealing with weather would be to have one
page per information; let's say: Template:Weather/London/pressure. I
started with that approach, but it took ~45-60 minutes just for Serbia
just for Wikinews in Serbian. So, I had to cut number of pages.
Note that the main problem is not processor power or Internet
connection, but: the process of changing one page, as well as
Pywikipediabot's waiting before changing the page. (Some time ago, I
tried to see how it would work if I remove that waiting, and it
doesn't work well: without waiting process of adding data into wiki
may fail.)
I have to check the next issues:
* Which amount of memory bot consumes.
* Which amount of CPU bot consumes.
* Which amount of Internet bot consumes.
After that we'll have the picture what do we need to cover all of the
planet. If it is not possible, then we may make some priority:
Capitals of countries, cities with more than 1M of inhabitants and so
on. Wikinews in languages localized in [relatively] small (like
Serbian or Italian are) may cover those areas, too. Also, if someone
is willing to cover her or his own country, he or she may run bot(s)
for that country for all Wikinews editions.
> Is this do-able? Are these sources as free as we need? Is there scope for
> cross-project collaboration? For the latter, it’d seem reasonable that
> Wikipedia have the ability to look up temperature history and say which day
> in a year was the hottest/coldest. For Wikiversity the data is a historical
> record for any research purpose they can come up with, and we have a need
> for it as news – if we do build a long-term record then there’s our ability
> to analyse it for trends and such. Wikipedia with weather data is one
> feature of Wolfram Alpha knocked off. Instead of fancy algorithms that
> aren’t very good at parsing English queries you’d have lots of curious geeks
> writing programs to query the data.
Sources are not free in the sense of sites, but they are free in the
sense of data. Actually, I think that there is some protocol for
sharing meteorological data, so we may use that protocol instead of
harvesting sites. At the other side, I think that all sites will be
happy because of linking them.
> This sort of begs the question, where should the data be held? We’re going
> to want it cross-wikinews – do we need weather.wikinews.org? (Or even
> weather.wikimedia.org) If so, and we have a good bot for it, could a WMF
> server have the spare capacity to run the bot?
There are more efficient method for adding meteorological data than
putting it on the wiki. Even writing it into the file system is more
efficient than remotely adding data to the wiki. If we have WMF
support and one PHP developer, we may cover all of the world with just
one system.
Generally, I am for making templates.wikimedia.org which would be used
as Commons is used. So, [[Template:Weather/London]] would point
firstly to the local template and then to the
"templates.wikimedia.org/wiki/Template:Weather/London". It would allow
a number of benefits: from putting weather data at Wikipedia articles
to making generic country/pop star/chemical element/... templates for
all Wikipedias.
> Lastly, one of the things I always find useful in a weather map is isobars
> with front markings. Rarely seen on TV now, you can quickly tell what
> weather is going where and do your own half-day to day forecast. Do we have
> enough data to produce these at any particular level of detail?
Actually, yes. Some work should be done and I'll need help for drawing
maps, but it is possible:
* I may draw lines according to temperature, pressure, wind and conditions data.
* I may calculate hourly changes, too.
But, I think that it would be good to have one meteorologist or
student of meteorology to explain to us some basic methods.
Some other possible features and needs:
* I may add astronomical data for current day: what is Moon's phase,
but where are Ganymede and Betelgeuse, too.
* I would like to see more systematized icons. So, someone who knows
to draw SVGs may help, too.
* I need the list of possible weather conditions.
* ...
RSS Feed