Show HN: The new Windria – 2-5km precision wind forecasts | Hacker News Reader

Show HN: The new Windria – 2-5km precision wind forecasts

windria.net

63 pointstimedivers35 comments

Awesome site. Is there any sort of API ? I'd love to build a surf predictor/alert for Stockholm, which is very dependent on wind.

I noticed a UX bug here: http://snpy.in/E34hR9

Hi William

Thanks for reporting the bug. Yes there is an API, let's talk via email: info AT windria DOT net

Phenomenal work! I live in a city known for its sporadic gusts and high wind speeds (no surfing around here, unfortunately!), so this will be quite useful! I'm also interested in an API, so I'd like to get in touch with you as well.

Certainly, just shoot us an email to info at :)

Noticed the same bug using Chrome and MS Edge

Interesting. WRF is the model that RASP uses - so surface winds should be similar. It would be nice to have higher level winds there, too, although since that's a sailing site this is probably not a priority for the people behind the site.

We use the WRF-ARW core for some regions, it's a solid model. Haven't considered higher level winds, are you a pilot/paraglider? :)

Glider pilot here. The website looks awesome! Thank you for it! If you guys could add winds at higher levels, it would really be useful for all kinds of sky sailors as well.

Yes, would be useful for paragliding.

love the design, shall be recommending it to my gliding club.

RASP is the gold standard of guessing in which direction you're going to land out today, but my god it is ugly.

yes please, high level winds also!

When I was working for electrical utilities and tracking climate data over a service territory a service like this with an API would have been invaluable. I would have thrown money hand over fist at them. Particularly if I could predict accurately micro-conditions within the next 12-24 hours.

The microconditions are tricky, but yes there is business there. We've done a few proof of concepts using CV + ML/statistics in the area of storm tracking.

Do you have any insight into what the electrical utilities companies are doing these days in terms of weather forecasting?

At the time I was using the National Climatic Data Center's datasets from AWOS and ASOS and combining that with locally sourced airport weather stations, mobile micro-climate stations, and helicopter mounted climate collection systems. It was a lot of fun... :(

We were evaluating using a service provider such as Weather Underground or AWS Truepower. But those providers couldn't deliver the granularity or locality of collection that we needed for our obligations.

The utility industry required aerial LiDAR collection to be partnered with same time collection of local climate conditions for use in modeling the impact of wind/solar/temp on conductor line sag. We did a bunch of pioneering work in that area back in 2012.

Very impressive. Temperature seems a bit broken for extreme cold weathers though http://imgur.com/dRPFgs4

Yikes, 502C? I think most things we use would burst into flames by then.

I am wondering how wind directions are calculated? Is there a realistic model that uses height maps and fluid dynamics or just a simple one for visual purposes?

There are many different models, all of them taking into account terrain data, laws of physics and much more: https://en.wikipedia.org/wiki/Numerical_weather_prediction

Does that mean you can possibly also get indications of gusting in direction as well as speed from the terrain? That would be really useful for sailing, particularly inland. Or maybe that's a more local effect, and beyond the resolution of these models?

"Please signup"? No thanks.. There's very impressive alternatives that don't want my email.

eg: https://www.windyty.com/

I currently pay for iKiteSurf and this could be a viable alternative if it's accurate. iKSurf does actually place miniature weather stations at kite beaches though, so it might be hard to match that accuracy.

It's giving me what look like quite unlikely wind conditions (FF42/Linux) : https://imgur.com/QejZqTY

That looks like a rendering problem, we will look into it!

No worries - the only thing I did was slide the time forward by a couple of hours - it was rendering ok before that :-)

I didn't get a prompt asking for permission to use my location, yet the map was already focused over my location (though the loaded city text was wrong). Any idea what is going on here?

We use your IP to determine your approximate location. Not very precise, but useful because it puts at least in the right country/state, so you don't have to search for it :)

As a kitesurfer - this is awesome! Really well done.

You might want to use some color scales that are perceptually uniform. Right now you are using the bad kind of rainbow for many.

Any pointers where we could look? (we are not designers..)

If you google matplotlib viridis there is a discussion of the issue. Using perceptually uniform colormaps(not to mention ones that translate to gray scale correctly) is normally the correct thing to do, but not always. Sometimes you want to highlight particular conditions. For instance, jet with radar data tends to nicely separate different atmospheric conditions if set up right. The same perceptual colormap version feels somehow lacking.

http://www.windguru.cz/int/help_index.php

Use the standard implemented on WindFinder and WindGuru.

What's the source of your data? I get quite different predictions on WindFinder and WindGuru.

Depends on the area. Windfinder and Windguru are using the GFS model which has a resolution of only 50/100km.

For example for winds, in the US we are using NOAAs HIRESW with a resolution of 5km and in Europe we are using mostly WRF models from national agencies with 2-10km resolution depending on area.

Windfinder's Superforecast has a 12km resolution:

http://www.windfinder.com/help#superforecast

Extremely well done. Good job guys.

great!

Reply on news.ycombinator.com