← Back to context

Comment by aleciffo

10 hours ago

Does anyone know what are we talking about in practice in terms of weather forecast prediction improvement? Like MAE/RMSE

9 comments

aleciffo

Reply
lauri_jo  8 hours ago

This data is from the third generation of Meteosats, which are the European meteorological satellites. A lot like GOES in the north-America. The main improvement is really significant improvement in resolution. The resolution is, depending on the channel, 9 times better than in the second generation. The main improvement in forecasting comes due to better information in the initial condition of the numerical weather prediction, but it is hard to quantify in advance. I'd be surprised if MAE, over the 15 days the prediction spans, would improve more than 0.1 C, if we talk about the raw prediction. There are plenty of things that this data is used for, but I would say that improved nowcast of cloud coverage, and energy production related parameters are likely to benefit most from the improvement in resolution.

  • mturmon  41 minutes ago

    They also feature that the IR hyperspectral measurement is new -- 1700 channels in IR for a telescope in GEO seems new to me, but I'm not sure what exists now in this space.

    They say they hope to retrieve trace gases at that global scale (seemingly with 30 minute cadence), which I think would be new. Also, they seem to say that this spectral resolution would enable them to retrieve temperature and humidity as a function of height -- not just surface temperature and column-integrated water content ("humidity").

    Aha, here's a nice link (https://www.ssec.wisc.edu/geo-ir-sounder/) on exactly this question, pointing out the NASA IR sounders that have existed for many years (AIRS). These instruments get vertically-resolved atmospheric information, but they are not at GEO so their coverage is different. This makes them less useful for NWP.

  • mjanx123  8 hours ago

    I use windy.com for its 'compare models', the models can differ by ~2C sometimes

    • Beretta_Vexee  5 hours ago

      The difference in wind speeds can be quite significant, which greatly alters some forecasts.

      The question is often when it will rain, not if it will rain.

peyton  9 hours ago

I was curious but it’s surprisingly hard to find info. These guys [1] are pretty stoked about “nowcasting”—which seems to be on sub-10-minute timescales to issue local severe weather warnings and such. It appears current sounders don’t scan as often.

This project ppt from 2011 [2] references different requirements for different areas/teams and shows the instrument spits out readings at 150 Mbit/s, which seems like a good clip. Overall it sounds like a lot of local knowledge is involved in turning this output into forecasts. Maybe there’s not a precise answer to your question.

Somebody else must know more.

[1]: https://www.eumetsat.int/features/think-global-act-local

[2]: https://www.researchgate.net/profile/Donny-Aminou/publicatio...

atoav  9 hours ago

This is an improvement as it provides better data and has nothing to do with the models that are used in a separate step to forcast anything. But that is said in the article as well, with the satellite being the first hyperspectral view on Europe and North Africa.

I am not sure what to make if your question.