Code and Data for "Representing Subgrid-Scale Cloud Effects in a Radiation Parameterization using Machine Learning: MLe-radiation v1.0 "

# Code and Data for: Representing Subgrid-Scale Cloud Effects in a Radiation Parameterization using Machine Learning: MLe-radiation v1.0 The content of this repository:- icon-a: exact ICON version used in this work. It is based on v2.6.4 and is slightly modified to output variables after the dynamical core. The version also includes two bug-fixes regarding vertical mixing that were included in a later version. For the most recent ICON version please visit [icon-model.org](https://www.icon-model.org) - data_*: contains input/output data for neural neutwork training and pyRTE. generated using icon-2024.10. data is split into training, validation and test set- predictions: NN output/prediction using the test set as input- runscripts: runscripts to run ICON and postprocessing script for coarse graining- trained_models: trained neural networks using training and validation data from the data folder. The trained models produce data in the predictions folder- statistics_2004_mig_qs.pickle: contains statisitics for coarse-scale ICON shown in Figure 2 and 3- statistics_coarse_grained_njaj.pickle: contains statisitics for coarse-grained ICON shown in Figure 2 and 3. Data to produce statistics is saved in data.- icon_output_coarse_scale.nc: contains data to produce statistics_2004_mig_qs.pickle (v2.6.4)- pyrte_output: output of pyRTE based on coarse-grained test data in data folder The code to train the NNs and reproduce all plots is published here: [MLe_radiation](https://github.com/EyringMLClimateGroup/hafner25GMD_MLe_radiation) Reference:ICON partnership (DWD; MPI-M; DKRZ; KIT; C2SM) (2024). ICON release 2024.01. World Data Center for Climate (WDCC) at DKRZ. https://doi.org/10.35089/WDCC/IconRelease01