Processed dataset used for developing tsunami inundation emulators(Part-I Training Dataset and Model Weights)

This dataset contains some of the processed training datasets covering tsunami inputs and outputs parameters (for offshore waveforms, local deformation fields and inundation depths used in developing a machine learning emulator discussed in the preprint article - Towards Using Machine Learning Emulation for Probabilistic Inundation Mapping: Multiple Earthquake Sources and Near-field Effects with project repo - https://github.com/naveenragur/ML4SicilyTsunami/tree/ptha_emulators.

####Code####

Source code from the above github project :  ML4SicilyTsunami-NPJ2024.tar.gz

The post-processed numpy dat files for the two test location of Catania(CT) and Siracusa(SR) are provided for each variable retrieved from data/processed folder of the project as compressed gzip files(.gz):

####Model Checkpoints#### ML Model checkpoints from all training experiments.- outSR.tar.gz (Siracusa)- outCT.tar.gz (Catania)

####Training Datasets####- t.tar.gz: Offshore waveforms for gauge sites in a 3D numpy file- dZ.tar.gz: Local deformation fields in 2D numpy files or reduced 1D flat files on flooded grids- d.tar.gz: Maximum inundation depth in 2D numpy files or reduced 1D flat files

####Flood Mask####- Identifies floodable and non-floodable areas in the domain. - zero_mask_SR_961.npy.gz - zero_mask_CT_892.npy.gz

####Flood Indices####Links between 2D and 1D numpy files:- Indices connecting 2D (events x m x n) numpy files to flattened 1D (events x m) numpy files:  - zero_indices_SR_961.npy.gz  - zero_indices_CT_892.npy.gz- Indices connecting latitude-longitude coordinates to 1D (events x m) numpy files:  - lat_lon_idx_CT_892.npy.gz  - lat_lon_idx_SR_961.npy.gz

The filenames follow nomenclature representing {parameter}_{site}_{size}

The first var represents the parameter of the file.

d - 2 dimensional file for inundation depth (events x m x n)

dZ - 2 dimensional file for local deformation( events x m x n)

dflat - 1 dimensional flat file for inundation ( events x locations)

dZflat - 1 dimensional flat file for local deformation( events x locations)

t - 3 dimension offshore waveform for (events x gauges x timesteps)

lat_lon_idx - index file for matching lat long coordinate with location indices(locations x lat x lon)

The second var represents site of the file.

CT - Catania

SR - Siracusa

The third var represents number of events in the file or the selection size.

892 - represents the training events for the file, of 4 different sizes for Catania and Siracusa each

More information on the attached readme.