Contains the files that are different from the original WRF v4.3 files and used for implementing the WRF-ML coupler. To run the WRF coupled with ML-based mskf cumulus schemes, you need to add files that did not exist or overwrite the existing WRF files.

module_cu_mskf_dl_replace.F: subroutine infer_init_cu: used to initialize the ML-based mskf cumulus scheme, this is called in main/wrf.F，the dl model to be called is in the test/em_real/model subrountine infer_run_cu: used to run the the ML-based mskf cumulus scheme subroutine save_fortran_array: used to save wrf variables into npy or npz files

module_cu_mskf.F: difference from the original module_cu_mskf.F is that it outputs one 2 dimensional variable pblh, one 3 dimensional layer variable u, and one 3 dimensional level variable w, Yu Xing will help to check whether these variables are consistent with the npz files that he provided before using this method.

module_cumulus_driver.F: difference from the original module_cumulus_driver.F is that the same as changes in module_cu_mskf.F

Makefile: used to add module_cu_mskf_dl_replace.o

wrf.F: call infer_init_cu to initialize ML models

depend.common: add dependencies for new files such as module_cu_mskf_dl_replace.o

Contains the files used for building the library in dl-inference-plugin, and the python script to run ML models within WRF_DL

build_wrf_dl_plugin_debug_so.py: build python functions and generate the library in dl-inference-plugin

mskf_data_preprocess_short.py: used to preprocess features, such as data normalization

mskf_data_postprocess.py: used to postprocess variables, such as output data denormalization

cu_utils.py: contain function used to calculate some auxiliary features

config_wrf.py: wrf configurations

config_norm.py: contains the norm_mapping used for data normalization

Contains the onnx model, and the python script to convert model to onnx.

Contains the python scripts for training and evaluating ML models, and for generating figures.