Deep Learning Parameterization of Tropical Cyclone Boundary Layer

These are the data used in this publication.

The raw data of input and output (data_in and data_out), the corresponding standard deviations and averages (Per-level calculated: large_std, large_avg, turflux_std, turflux_avg; Global: large_std_all, large_avg_all, turflux_std_all, turflux_avg_all) are uploaded.

The averaged resolved variables and diagnosed turbulent fluxes for old LES (avg_large_old_les, avg_flux_all_old_les) and new LES (avg_large_new_les, avg_flux_all_new_les) are uploaded.

The trained models are uploaded: DeepBL trained on original data (model_original_b, b ranging from 1-5 for model id), DeepBL trained on nonlinearly transformed data (model_nonli_a_b, a ranging from 0.3-0.9 for power number, b ranging from 1-5 for model id), FC-NN-NODE (model_same_nodes_b, b ranging from 1-5 for model id) and FC-NN-COMP (model_same_compute_b, b ranging from 1-5 for model id)

Github url for this project is https://github.com/lywang2237/DeepBL