Replication Instructions

LINDA/Lagrangian Persistence nowcasts

Create a configuration folder for computing nowcasts named arbitrarily, e.g. arbitrary_folder_name under config.
Make a copy of config/example/prediction_builders/advection_prediction.yaml in that folder for each model to compute nowcasts for, each configuration separately, if many are run.
- For example: LINDA_config_1.yaml, LINDA_config_2.yaml, LP.yaml
modify each of these according to your needs, taking care that:
- datelist_path, data_source_name, input, preprocessing, postprocessing are matching for comparable results
- save parameters should most likely be the same, and set according to your choices.
- hdf5_path should be different for each model.
- nowcast parameters should differ for each model. Here the actual nowcast method and its parameters are defined.
Run python scripts/run_pysteps_swap_predictions.yaml {config_folder} where {config_folder} indicates your arbitrarily name folder name.
Prediction HDF5 archives should appear where you set them to be saved. Check for the integrity of the nowcasts.