Dataset for: Synthetic Lagrangian Turbulence by Generative Diffusion Models

This dataset forms the basis of our research paper "Synthetic Lagrangian Turbulence by Generative Diffusion Models" published in Nature Machine Intelligence. It is available for download from the open-access Smart-TURB portal: http://smart-turb.roma2.infn.it, in the TURB-Lagr repository (DOI: 10.48550/arXiv.2303.08662). Instructions on how to read the dataset can be found in the code repository at https://github.com/SmartTURB/diffusion-lagr.

The files "Lagr_u1c_diffusion.h5" and "Lagr_u3c_diffusion.h5" are the datasets used to train the DM-1c and DM-3c models respectively. Correspondingly, "Lagr_u1c_diffusion-demo.h5" and "Lagr_u3c_diffusion-demo.h5" serve as their minimum datasets for testing the code. The files "samples_358400x2000x1_DM-1c.npz" and "samples_358400x2000x3_DM-3c.npz" are generated outputs from the DM-1c and DM-3c models. They support all the analyses presented in the paper.