اوپن سٹریٹ میپ
OpenWeatherService: Why doesn't this exist yet?
ایہہ 3؍February 2009ء English وچ «Mark Williamson» لیکھ چھپیا گیا سی۔I've been thinking about weather forecasting and wondering if wiki-like principles could be applied to it.
It seems to me that by leveraging a wide set of data sources (which will typically be user-contributors uploading local data but may also include public domain data where appropriate) and aggregating them into one database you could get a pretty accurate "current weather" view for large areas of the world reasonably easily. On top of that, a very-near-future forecast based on e.g. prevailing winds and observations of rainclouds seems like it should also be possible with reasonable accuracy.
I'm aware that "proper" meteorology is a complex science and uses vast data sampling and computational power, so I'm not proposing a replacement for that. Maybe one day it would be feasible to do some distributed computations in order to improve forecasts from an "OpenWeatherService" without requiring a supercomputer. However I think big gains could be made with only current and short range forecast data, plus a historical archive.
Consider:
* In remote areas, an OpenWeatherService could provide more relevant data than otherwise available (as mentioned OpenWeather-related links below).
* In highly populated areas with a number of OWS contributors, finer granularity data may be possible.
* A friendly-licensed historical archive might be useful for research purposes?
* Sites like OSM have shown that just having friendly-licensed data available can be great for encouraging unexpected innovation.
Imagine if YourNavigation.org could immediately warn you "there is rain on this route". Or how about if your MID or phone could detect through GPS that you're walking out the house and beep "Take an umbrella, it's raining within 10km of here". How about city articles on Wikitravel or Wikipedia being able to show CC-licensed temperature graphs of previous years' seasonal weather. Imagine just being able to overlay weather data onto the main OSM map.
I've investigated this sort of thing in the in the past, hoping to find existing projects. I found a couple of relevant links connected with the Open Weather project and the associated WxMonitor device:
* http://wxmonitor.sourceforge.net/
* http://www.catnip.co.uk/wx/
At the time I'd been hoping to find some suitable computer-attached weather station which could be used to update a hypothetical OpenWeatherService database". When I thought about it some more I decided that the contributor base could usefully be much broader than merely those who can afford a PC weather station - although if it's available then we would want an OWS server to be able to take data from it. Amateur weathermen could just as easily fill in temperature data manually using web forms, twitter, e-mail, whatever - they could read the data themselves from something as simple as a mercury thermometer or homemade wind meter, if they wanted.
Aggregating this information in one place and geolocating would be "all" we'd need to get this data usefully stored in a database. An open API for requesting data about a given area would be needed in order to give web developers the data they needed.
Summing up: Explicitly openly licensed data is better for innovation and use by other projects - weather data is (to me!) an obvious hole in current community geodata projects. User contributed data can potentially give better data density in populous areas and make data collection possible in remote areas not served by "official" weather services. The barrier to entry by individual contributors can be made very low, although it seems likely that automated weather stations and existing public domain data would also form a major component of this service.
I think this sort of service could be pretty cool, particularly if the API was made easily accessible. Has anybody thought about doing something along these lines? Is there any code out there...
Discussion
ایہہ 3؍February 2009ء 21:02 تے «es_ka» ٹپݨی کیتی گئی سی۔
No comments so far on such a big idea?
I think it's just great! The world needs those revolutionary idealistic thoughts.
ایہہ 3؍February 2009ء 21:55 تے «kaerast» ٹپݨی کیتی گئی سی۔
Couple of things worth mentioning - cumul.us aggregates weather forecasts from commercial providers and mixes in user-provided forecasts (including what people are wearing as a simplistic way of looking at current weather)
Also pachube, I haven't really played with it but their description of "a service that enables you to connect, tag and share real time sensor data from objects, devices, buildings and environments around the world. The key aim is to facilitate interaction between remote environments, both physical and virtual." seems accurate.
This reminds me of the old days when I'd hook a shortwave radio up to my 486 and monitor the incoming weather data from around the world. After an entire morning of receiving data I'd have a pretty accurate map of weather systems over the Atlantic and could forecast the afternoon's weather. Of course I could also have just looked out the window, which brings us back to cumul.us.
ایہہ 4؍February 2009ء 01:06 تے «iandees» ٹپݨی کیتی گئی سی۔
The United States does a great job giving all of its forecasting data out for free: http://www.weather.gov/
In fact, most of the weather forecasts I've seen from local weather stations just pull data from this website.
ایہہ 4؍February 2009ء 07:17 تے «Baloo Uriza» ٹپݨی کیتی گئی سی۔
This might be something that could be more useful for some areas than others. It would be interesting if an OpenWeatherService could integrate data from public sources required to diseminate this information without obligation (such as weather.gov), and supplement this information with data from third-party weather stations. This would be useful as some states, such as Oregon, have supplemental weather stations of particular interest to travellers (since most if not all of ODOTs weather stations are located on light posts or sign bridges located on highway medians or on the shoulder of the right of way (whichever is likely to give the most representative reading and is physically possible).
I'm interested to see where this goes.
ایہہ 4؍February 2009ء 07:41 تے «Chaos99» ٹپݨی کیتی گئی سی۔
I think the userbase for this would be slightly smaller than that for OSM, because it misses the 'adventure' part that is involved with going out and do some mapping. Crowdsourcing large datasets has more to do with social engineering than anything else. You could get a problem in motivating people to join in. But don"t let that stop you! Who knows ...
Here in Germany nearly all weather stations are commercial. Some big weather services build there own net of stations and than share data (for a fee of course). Just lately (ok, some 7 years ago) a new service turned up and said it can do a better job that the old services by drastically increasing the number of stations instead of using just better simulation moddels on existing data.
(This was after old services failed to correctly forecast the massive flooding of the river Oder in 2002).
Under these circumstances it is unlikely that you will get free (like in open licensed) data here in Germany. There is to much dispute about it in the comercial weather services.
But, said again, don't let that stop you.
ایہہ 4؍February 2009ء 12:11 تے «POHB» ٹپݨی کیتی گئی سی۔
Weather Underground has lots of data from personal weather stations e.g.
http://www.wunderground.com/weatherstation/WXDailyHistory.asp?ID=IMIDDLES2
ایہہ 10؍February 2009ء 00:32 تے «LivingWithDragons» ٹپݨی کیتی گئی سی۔
I thought it would nicer to have more access to weather data when I worked on some stuff for 'OpenSantaMap' over the holidays. I wanted to get all locations in an area if it is snowing. The Yahoo Weather API can tell you about 14 different types of snow fall, but you need to search for and know each station. The other thing is it gives snowfall, but not snow depth which I don't think would be hard for amateurs to measure. I'm sure being able to query a historic database would be very good too (and users/developers can process the data how they want).
Another problem I have with Yahoo and my Linux(using weather.com) forecast widget, is that the closest location to me is apparently Newcastle. Despite Durham being a city(small, but we have a cathedral!) and having a university that must have a weather station somewhere (I think the physics and geography departments are highly ranked). If a system like you propose was set up then I'd like to think it would be easy to get universities to automatically send some data. Or failing that, student societies/groups could be set up at the unis to read the equipment and enter into the project. This could go down to several levels, I remember my junior school getting a little weather gadget fixed outside the tiny computer room. Once we got to have a look at the values (temperature, pressure, wind chill, it was smaller than a brick). I imagine this being a lot easier for current teachers to set up/run.
My concern from general user contributions would be the continued reliability. Mapping is something that can be done once, and then have little updates (even by passing mappers or remotly reading planning notices). With the weather you hope that the number of stations your happy with remain online. If someone suddenly goes on holiday for two weeks, how will you know wether to put your coat on before leaving the house or not?