Dispformer Datasets: Synthetic and Real-World datasets for Surface Wave Dispersion Curve Inversion

Dispformer: Synthetic and Real-World datasets for Surface Wave Dispersion Curve Inversion

This dataset is designed for surface wave dispersion curve inversion, particularly suited for deep learning-based inversion studies. It provides both synthetic and real-world datasets, including global and local models, enabling the evaluation of various inversion methods such as zero-shot and few-shot strategies. Researchers can refer to the DispFormer framework for details on the model and the corresponding code used for inversion tasks. The related paper for DispFormer can be found here.

For further details on how to use the datasets and the associated neural network models, visit the GitHub repository: https://github.com/liufeng2317

Datasets Overview

The LITHO1.0 global synthetic dataset is primarily used for pre-training the model. The Central and Western US Dataset (CWD) and Continental China Dataset (CCD) are used to validate the effectiveness of zero-shot and few-shot strategies. Finally, the datasets retrieved from the China Seismological Reference Model (CSRM) are used to test the model's performance on real-world data.

Data Files

• depth_vp_vs_rho.npz: Contains a 1-D velocity model, including depth, P-wave velocity, S-wave velocity, and density.
• depth_vs.npz: Contains a 1-D velocity model with depth and S-wave velocity (used for training the DispFormer).
• lon_lat.npz, lat_glat_lon.npz: Contain the station locations for the observed data.
• period_phase_group.npz: Contains synthetic and observed dispersion curves, including period, phase velocity, and group velocity.
• Folders train_data, valid_data, and test_data: Contain data directly used for training, validating, and testing the model

Example of Loading Data Using Python

Here is an example of how to load the data using Python:

import numpy as np
data_path = ""  # Specify your data path
all_disp_loc = np.load(data_path)["data"]  # Load data from the file

For more details on how to use the data, please refer to the DispFormer GitHub repository.

References

1. Liu, F., Deng, B., Su, R., Bai, L. & Ouyang, W.. DispFormer: Pretrained Transformer for Flexible Dispersion Curve Inversion from Global Synthesis to Regional Applications[J]. arXiv preprint arXiv:2501.04366, 2025.
2. W. Shen, M. H. Ritzwoller, and V. Schulte‐Pelkum, “A 3‐D model of the crust and uppermost mantle beneath the Central and Western US by joint inversion of receiver functions and surface wave dispersion,” JGR Solid Earth, vol. 118, no. 1, pp. 262–276, Jan. 2013, doi: 10.1029/2012JB009602.
3. S. C. Griffiths, B. R. Cox, E. M. Rathje, and D. P. Teague, “Surface-wave dispersion approach for evaluating statistical models that account for shear-wave velocity uncertainty,” J. Geotech. Geoenviron. Eng., vol. 142, no. 11, p. 4016061, Nov. 2016, doi: 10.1061/(ASCE)GT.1943-5606.0001552.
4. M. E. Pasyanos, T. G. Masters, G. Laske, and Z. Ma, “LITHO1.0: An updated crust and lithospheric model of the Earth,” JGR Solid Earth, vol. 119, no. 3, pp. 2153–2173, Mar. 2014, doi: 10.1002/2013JB010626.
5. X. Xiao et al., “CSRM‐1.0: A China Seismological Reference Model,” JGR Solid Earth, vol. 129, no. 9, p. e2024JB029520, Sep. 2024, doi: 10.1029/2024JB029520.