The original premise of ClimaCell as I understood when hearing about a few years ago was to utilize cellular cells (hence ClimaCell) to acquire information about the levels of humidity in the air, thus providing a cellular cell level immediate forecast. I believe they have expanded that to satellite communications and more.
There is an enormous amount of latent weather data to be extracted from existing sources and they, I think, are doing a great job finding and using them.
I'm shocked there aren't more efforts in this space. I know IBM is using barometers in phones now, and I suppose Apple may try that post-Dark Sky acquisition, but I'm still surprised there aren't more!
I have to say what benefit do you think that having all of these phone pressure measurements will bring to society? Does it help forecasting? Ive heard that we could use them as initial conditions for forecasting. which is cool, but i'm not sure the whole 'crowd source surface pressure obs' is really that important or impactful.
This should be far better than Dark Sky. AFAIK Dark Sky was far more of a simpler statistics play than real meteorology and data processing. I was surprised Apple bought them, because I don't honestly believe they had any real groundbreaking tech, more slimy marketing than anything.
But ClimaCell is serious about new data being used, real forecast analysis, and verification, etc.
The future of weather APIs and services is much brighter IMO, now that Dark Sky is out of the picture.
DS has an old article saying how they use statistical/ML approach rather than physical modeling. That doesn't make it illegitimate though, using ML in weather modeling is active and fruitful research: https://ai.googleblog.com/2020/01/using-machine-learning-to-...
I use ML in weather research too. That's not the problem. The problem is Dark Sky used more simple radar prediction methods, like taking existing rainfall, using wind speed and 'nowcasting' the rain into the future.
There are severe limits to this technology. It cannot detect the initiation of new precipitation. They never publicly explained what they were doing with their barometer data, which they claimed to collect for forecasting. But in such an active area of basic science research, they did not contribute data or algorithms or papers or any discussion at all. I believe it is because they did not overcome the challenges of using the noisy phone data to produce useful weather forecasts. But they lead people to believe they do. I am suspicious.
Dark Sky didn't use ML AFAIK when they started, it was straight statistics. Maybe they did change to add ML, I don't know, but for my knowledge of the space, they always used core math and statistics while ignoring physical science and not participating in the research community, or any community really.
I would love to know more about how Dark Sky worked. My gut feeling is that it is largely a scam, because they never published any of their own reanalysis data, or verification data for 1-minute forecast accuracy, etc.
They were so closed-off in an industry that is traditionally hugely open, and they made large claims without backing any of it up publicly.
Aha, I want a lot from weather apps. I am a weather app developer, for what that's worth. I expect a lot because the current state of weather data communication and accuracy is a disaster. There is a lot of work to do to improve forecasts, and when a company steps up to say they will do it, I expect a lot out of them.
I also take major issue with the claim that they provided reasonable forecasts. I think their forecast accuracy was extremely low in many cases and fares very poorly when put against competitors: https://theoutline.com/post/3826/crap-weather-apps-meteorolg...
Edit: "The scam" is that they exaggerated accuracy claims without backing it up.
Correct, at least here in Europe Dark Sky was super low-res in terms of its weather maps and predictions. As a result it was way off the mark. Especially the rain map here in Spain. I think it was really focused on the US.
At the moment I use openweathermap but interested in better ones.
No pricing is an immediate turn-off. Best they seem to have is another "Contact Us (and tell us how thick your wallet is)" button: https://www.meteomatics.com/en/eshop/
In southwest USA, if Dark Sky said "rain in 23 minutes", it was rarely off by more than a minute. That was really nice, in the "oh shit I should give the dog a pee break before the rain" sense.
This sounds as though it could be a nice replacement for the Dark Sky API that Apple is going to lock us out of.
The ClimaCell API docs say the weather API offers forecasts for 1) realtime, 2) hourly, 3) nowcast, 4) daily
The free-tier pricing plan [1] apparently offers limited access to 1-3 but no access to 4. That's too bad, but then no paid-tier plan offers 4, either. So, I have to wonder, Am I missing something or are they?
I actually started to use the free api user and I was able to get access to: 1. real-time, 2. hourly 3. nowcast, 4. daily and replace completely my DarkSky subscription.
All four endpoints are available on the free plan
Interesting concept, worth exploring. Except the documentation isn't very good. In so far as there are examples provided, with my API key, that don't work. If you're going to provide examples... test to see if they work. I don't care if it provides me weather in Timbuktu, but it should work.
Thanks! I am not paying for a plan, so the app is subject to their free quota. That line of source you found is simply going through a little proxy I set up to add my API key to the request.
It's a hassle to sign up though; at least they could show me an example weather forecast for some region so I could compare it to others before bothering to play with their API... of course I won't bother if their forecasts are no good.
Oh nice, I have something else I can use for my terminal status bar now that dark sky is dead. The free tier looks plenty generous for such a use case; polling once every two minutes is just about right. Much appreciated!
I used the ClimaCell api about a year ago as part of a POC and it was pretty nice.
The biggest feature for my use case was their granularity. The ability to query the service using a point (lat/long) and have returned the specific weather for that location was nice.
All other services required two requests, one to find the nearest weather station, another to query said station for data. ClimaCells model predicts weather for locations where no forecast data exists.
I would recommend ground-truthing the forecasted vs actuals for the modeled areas.
Beyond their api, they have a pretty killer dashboard where you can plot your assets and set geofences for weather alerts.
I've used Dark Sky API to plot temperature against COVID-19 cases[1], I found its coverage to be better than other historical weather API.
But, there's need gap in all these weather API's; they lack actual recorded historical weather data and provide only forecast data. Actual recorded weather data is important for any research conducted with historical weather data.
AFAIK actual recorded weather data is available only with individual country's met dept and few international agencies like NOAA; but I wasn't able to get recent global data from NOAA. Few private weather API provider like Weather Underground which uses crowdfunded data using actual sensors but obviously their global coverage is small.
I tried to use WU for my COVID data research and it didn't have data for the Chinese cities.
48 comments
[ 5.3 ms ] story [ 80.7 ms ] threadFor historical data, can you get historical forecasts as well as actuals?
Where do you get the data? How is it different from https://openweathermap.org/api ?
Are there APIs demo call that we developers can try?
Is there a cap limit, if so how much?
What are the pricing plans?
I am also working on similar efforts, focused on using labelled photos of the sky with ML to extract weather data from outdoor photos, and using barometers in phones: https://play.google.com/store/apps/details?id=com.allclearwe...
I'm shocked there aren't more efforts in this space. I know IBM is using barometers in phones now, and I suppose Apple may try that post-Dark Sky acquisition, but I'm still surprised there aren't more!
But ClimaCell is serious about new data being used, real forecast analysis, and verification, etc.
The future of weather APIs and services is much brighter IMO, now that Dark Sky is out of the picture.
There are severe limits to this technology. It cannot detect the initiation of new precipitation. They never publicly explained what they were doing with their barometer data, which they claimed to collect for forecasting. But in such an active area of basic science research, they did not contribute data or algorithms or papers or any discussion at all. I believe it is because they did not overcome the challenges of using the noisy phone data to produce useful weather forecasts. But they lead people to believe they do. I am suspicious.
Dark Sky didn't use ML AFAIK when they started, it was straight statistics. Maybe they did change to add ML, I don't know, but for my knowledge of the space, they always used core math and statistics while ignoring physical science and not participating in the research community, or any community really.
I would love to know more about how Dark Sky worked. My gut feeling is that it is largely a scam, because they never published any of their own reanalysis data, or verification data for 1-minute forecast accuracy, etc.
They were so closed-off in an industry that is traditionally hugely open, and they made large claims without backing any of it up publicly.
I also take major issue with the claim that they provided reasonable forecasts. I think their forecast accuracy was extremely low in many cases and fares very poorly when put against competitors: https://theoutline.com/post/3826/crap-weather-apps-meteorolg...
Edit: "The scam" is that they exaggerated accuracy claims without backing it up.
At the moment I use openweathermap but interested in better ones.
Available models can be found here: https://www.meteomatics.com/en/api/request/#model_selection
The ClimaCell API docs say the weather API offers forecasts for 1) realtime, 2) hourly, 3) nowcast, 4) daily
The free-tier pricing plan [1] apparently offers limited access to 1-3 but no access to 4. That's too bad, but then no paid-tier plan offers 4, either. So, I have to wonder, Am I missing something or are they?
[1] https://www.climacell.co/weather-api/pricing/
Error 400 - "location is not defined correctly".
Select it from the "Source" section in the options.
[1] https://github.com/ersimont/weather/blob/master/src/app/sour...
Not that I'm their target customer anyway...
I can compare it with 3 other APIs. I haven't been rigorous about it, but so far I feel like it has been the most accurate for near-term forecasts.
But, there's need gap in all these weather API's; they lack actual recorded historical weather data and provide only forecast data. Actual recorded weather data is important for any research conducted with historical weather data.
[1]https://abishekmuthian.com/covid-19-temperature-correlation/
Is there no way to just download an archive of previous history of weather data for a location?
I tried to use WU for my COVID data research and it didn't have data for the Chinese cities.