Open idharssi2020 opened 8 months ago
Please can you confirm if the issue exists on the raw output from ai-models, i.e. before you split into smaller files.
I have also encountered issues related to precipitation amounts. It appears that your precipitation data also includes negative values, which I believe should be abnormal data, correct? Related Issue: https://github.com/ecmwf-lab/ai-models/issues/28
I have only recently started looking at precip output from GraphCast and noticed the forecasts are corrupted as are fields of vertical velocity and specific humidity.
I use the following command to run GraphCast
ai-models --input file --file dump.grib --date $1 --time 0600 graphcast --debug 2> log.txt
I use the following to write out the data into smaller files:
Recently, I noticed that the total precipitation fields are corrupt,
![image](https://github.com/ecmwf-lab/ai-models-graphcast/assets/80441682/aa5a23f9-3bd7-4598-a47e-8df573608aef)
However, when I look at the precipitation in the output.nc file it all looks OK (I used CDO
![image](https://github.com/ecmwf-lab/ai-models-graphcast/assets/80441682/ab803be6-acb1-4b2f-ab54-139d8ef37703)
cdo -select,name=total_precipitation_6hr output.nc test_tp.nc
), The time units are actually days (not hours).I would be much obliged if the EC ai-models team can investigate.
Thanks in advance I'm taking some leave soon so I don't have time to investigate further. Apologies if it turns out to be a bug in my code.