3 ms·
> The model was trained on four decades of temperature, wind speed and air pressure data from 1979 to 2018 and can produce a 15-day forecast in just eight minut
by gwbas1c 2y ago
> The model was trained on four decades of temperature, wind speed and air pressure data from 1979 to 2018 and can produce a 15-day forecast in just eight minutes—compared to the hours it currently takes.
Basically, they trained the model on old observations, not old predictions.
IE, imagine someone created a giant spreadsheet of every temperature observation, windspeed observation, precipitation observation, ect, and then told the computer to learn how to predict a column.
- mnky9800n 2y agoFrom the paper: > GenCast is trained on 40 years of best-estimate analysis from 1979 to 2018, taken from the publicly available ERA5 (fifth generation ECMWF reanalysis) reanalysis dataset https://www.nature.com/articles/s41586-024-08252-9 https://www.nature.com/articles/s41586-024-08252-9 They say they trained on era5 data. That is modeled data not only direct observations. They do it this way to have a complete global data set that is consistent in time and space. https://cds.climate.copernicus.eu/datasets/reanalysis-era5-single-levels?tab=overview https://cds.climate.copernicus.eu/datasets/reanalysis-era5-s...