About the Results of IFS

[fei88kk]

The values of IFS forecasting errors in Table 10 have larger than expected numbers compared to WeatherBench2: https://sites.research.google/weatherbench/deterministic-scores/.

I wonder how do you get the scores of IFS? Perhaps there is a difference in how the scores are being calculated.

[rejuvyesh]

The values of IFS forecasting errors in Table 10 have larger than expected numbers compared to WeatherBench2: https://sites.research.google/weatherbench/deterministic-scores/.

I wonder how do you get the scores of IFS? Perhaps there is a difference in how the scores are being calculated.

Those scores depend on what year you are looking and what is the "ground truth". We had been comparing against ERA5 and not IFS-Analysis.

[fei88kk]

Thank you for your answer! However, I found that even if ERA5 is replaced, there is still a large gap between the results in paper, especially the data results of 5.625 degrees.

I‘m also very confused. Also, I don't understand why weatherbench2's accuracy increases when you downscale (say, from 0.25 degrees to 1.4 degrees to 5.6 degrees), and it does so significantly that I don't know if the author has any idea. Thank you!!