Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

WeatherNext 2: Our most advanced weather forecasting model

blog.google

Post reply on HN
Print view View original

81–90 of 140 posts

Re: WeatherNext 2: Our most advanced weather forecasting model

#81

Post by counters »

rytill wrote: post #32
lysecret wrote: post #30

Im pretty deep into this topic and what might be interesting to an outsider is that the leading models like neuralgcm/weathernext 1 before as well as this model now are all trained with a "crps" objective which I haven't seen at all outside of ml weather prediction. Essentially you add random noise to the inputs and train by minimizing the regular loss (like l1) and at the same time maximizing the difference between…

What is the goal of doing that vs using L2 loss?

To add to the existing answers - L2 losses induce a "blurring" effect when you autoregressively roll out these models. That means you not only lose import spatial features, you also truncate the extrema of the predictions - in other terms, you can't forecast high-impact extreme weather with these models at moderate lead times.

Re: WeatherNext 2: Our most advanced weather forecasting model

#82

Post by tekkk »

empath75 wrote: post #65
lampiaio wrote: post #54

Reminds me of a funny WWII story: Kenneth Arrow and his statisticians found that their long-range forecasts were no better than numbers pulled out of a hat. The forecasters agreed and asked their superiors to be relieved of this duty. The reply was: "The Commanding General is well aware that the forecasts are no good. However he needs them for planning purposes."

There is a fairly compelling argument that divination in the ancient world was not a useless waste of time, as is commonly assumed, but that having either a process or a person that can make essentially random choices for them allowed people to make hard, consequential decisions where they might otherwise be paralyzed, especially when the penalty for not acting was worse than making a mistake.

Never thought of that. Probably a bit too generous given that it could be just as well waste of time and resources, nevermind the bias of the voodoo doctor. Most of it was just weirdly provided therapy I suppose to relieve stress.

But it is funny that humans put a great lot of weight on social contracts and being given explicit orders, maybe even publicly, must help pursuing action instead of rumination. Especially in a world where things seemed to happen randomly anyway.

Re: WeatherNext 2: Our most advanced weather forecasting model

#83

Post by mmooss »

NoiseBert69 wrote: post #6
binsquare wrote: post #4

I find it interesting that they quantify the improvement on speed and number of forecast-ed scenarios but lack details on how it results in improved accuracy of the forecast per: ``` WeatherNext 2 can generate forecasts 8x faster and with resolution up to 1-hour. This breakthrough is enabled by a new model that can provide hundreds of possible scenarios. ``` As an end user, all I care is that there's one accurate for…

As a end user I also want to see the variance to get a feeling of the uncertainty. Quite a lot of weather sites offer this data in an easily eatable visual format.

That would be great - do you recommend any sites?

Re: WeatherNext 2: Our most advanced weather forecasting model

#84

Post by mmooss »

bilsbie wrote: post #38

Where can I use this? I’ve been trying to find hyperlocal forecasts like darksky used to be.

Weather Underground used to include large numbers of personal weather stations - you could connect yours to their network - and might have provided forcasts based on them (?). IBM bought them and things changed, but maybe that project is still alive.

Re: WeatherNext 2: Our most advanced weather forecasting model

#85

Post by CaliforniaKarl »

For folks who are interested, I suggest checking out "The weather machine: a journey inside the forecast" by Andrew Blum[0]. It's a great read into the history of weather forecasting pre-Covid.

[0]: https://search.worldcat.org/title/1153659005

Re: WeatherNext 2: Our most advanced weather forecasting model

#86

Post by mmooss »

On what geometric surfaces do weather models run? Spheriods? Spheres? Projections on planes? Geoids??

Weather is three-dimensional and I would guess that the difference between sphere and (appropriate) spheroid could impact predictions. It seems possible that, at least for local and hyperlocal forecasts, geoids would be worthwhile. But as you go from plane -> sphere -> spheroid -> geoid, computing resources must increase pretty quickly.

And even if a geoid is used, that doesn't mean the weather user sees a geoid or section of geoid. Every consumer weather application displays a plane, afaict. Maybe nautical or aeronatautical weather maps display spheres?

Re: WeatherNext 2: Our most advanced weather forecasting model

#87

Post by dopa42365 »

culi wrote: post #75
trillic wrote: post #51

Earlier quoted context omitted.

The HRRR is VERY good in my opinion. It updates hourly with a 15-minute resolution 18 hours out and hourly 48 hours out. https://rapidrefresh.noaa.gov/hrrr/

HRRR only works for the US though. Windy.com is great for comparing different models. (switcher is in the bottom right hand corner) https://www.windy.com/?hrrrConus Also checkout HRDPS model if you're in Canada/northern US https://www.windy.com/?canHrdps

https://www.ventusky.com/ has a ton of different models too.

Re: WeatherNext 2: Our most advanced weather forecasting model

#88

Post by mikestorrent »

cogman10 wrote: post #80
lampiaio wrote: post #54

Reminds me of a funny WWII story: Kenneth Arrow and his statisticians found that their long-range forecasts were no better than numbers pulled out of a hat. The forecasters agreed and asked their superiors to be relieved of this duty. The reply was: "The Commanding General is well aware that the forecasts are no good. However he needs them for planning purposes."

I think it was a stats class where I learned this, but as it turns out bad weather is less common than good weather. To be a fairly accurate weather person, you merely need to say "there will be no precipitation" and you'll be right like 90% of the time anywhere on earth. What makes that funny is that historically, weather forecasters have been less than 90% accurate. Now, I will say that today's weather models are p…

Your fairly accurate weather person is going to have to stay away from Vancouver / the Pacific Northwest ;-)

Re: WeatherNext 2: Our most advanced weather forecasting model

#89

Post by strangegecko »

cogman10 wrote: post #80
lampiaio wrote: post #54

Reminds me of a funny WWII story: Kenneth Arrow and his statisticians found that their long-range forecasts were no better than numbers pulled out of a hat. The forecasters agreed and asked their superiors to be relieved of this duty. The reply was: "The Commanding General is well aware that the forecasts are no good. However he needs them for planning purposes."

I think it was a stats class where I learned this, but as it turns out bad weather is less common than good weather. To be a fairly accurate weather person, you merely need to say "there will be no precipitation" and you'll be right like 90% of the time anywhere on earth. What makes that funny is that historically, weather forecasters have been less than 90% accurate. Now, I will say that today's weather models are p…

I regularly encounter days when today's forecast is wrong and even in conflict with the current situation.

E.g. the weather app tells me there's a drizzle all day and currently and yet it's entirely dry. The opposite happens too.

Days of rain often shift in increments of days one or two days before as well.

I'd say it's location specific how accurate predictions are.

Re: WeatherNext 2: Our most advanced weather forecasting model

#90

Post by jeffparsons »

empath75 wrote: post #65
lampiaio wrote: post #54

Reminds me of a funny WWII story: Kenneth Arrow and his statisticians found that their long-range forecasts were no better than numbers pulled out of a hat. The forecasters agreed and asked their superiors to be relieved of this duty. The reply was: "The Commanding General is well aware that the forecasts are no good. However he needs them for planning purposes."

There is a fairly compelling argument that divination in the ancient world was not a useless waste of time, as is commonly assumed, but that having either a process or a person that can make essentially random choices for them allowed people to make hard, consequential decisions where they might otherwise be paralyzed, especially when the penalty for not acting was worse than making a mistake.

Additionally, what has been the correct choice five years in a row might be catastrophically wrong in the sixth year. We need some randomness injected into our behaviour so that some people are always making "suboptimal" choices, to stop everyone from crowding into one local maximum and then getting swept away when the rare but inevitable flood comes along.
Post reply on HN