The official code of MDG625: A daily high-resolution meteorological dataset derived by geopotential-guided attention network in Asia (1940-2023)".
Any questions can be touched by [email protected]
-
The low resolution data ERA5 can be downloaded from ECMWF.
-
The downloaded data needs to be organized into the appropriate format. The order of the parameters is ['WIN', 'TMP', 'PRS', 'PRE']. Here are some examples in 2023-12.
-
The pre-trained model can be downloaded from here.
The configuration can be modified in the configs/config.yml
.
The file struct should organized as:
|--data
|--exclude
|--2023
|--12
|--1
|--era5.np.npy
|--2
|--era5.np.npy
...
...
|--era5_global_std_level.npy
|--era5_global_mean_level.npy
|--cldas_global_std_level.npy
|--cldas_global_mean_level.npy
|--geoan.pt
Execute sh test.sh
can run the test program to produce high-resolution results.
The results will be shown in the proc
folder if each step is right.
The DOI of the produced dataset MDG625 is 10.57760/sciencedb.17408.