This repository contains the coded data products, trained mixed-effects regression models, and MATLAB functions developed for estimating earthquake source parameters from finite-fault models.
The framework provides statistically robust estimates of:
• Fault geometry parameters:
- Effective length (Leff)
- Effective width (Weff)
- Effective area (Aeff)
- Aspect ratio (Leff/Weff)
- Very-large asperity area (Avla)
- Large asperity area (Ala)
• Slip-distribution parameters:
- Mean slip (Dmean)
- Maximum slip (Dmax)
- Slip standard deviation (Dstd)
- Approximate static stress-drop proxy (Dmean/Weff)
The estimates are obtained using a hierarchical mixed-effects formulation that explicitly accounts for:
- Tectonic regime
- Fault mechanism
- Seismic region
- Inversion data modality
- Fault-plane spatial resolution of the finite-fault model (FFM)
- Intra-event variability arising from multiple FFMs for the same earthquake
Model FUnctional Form:
The primary MATLAB function, predictSP.m, enables users to obtain
source-parameter estimates using minimal inputs (e.g., earthquake magnitude
and location), while optionally including or excluding selected random
effects and returning uncertainty bounds.
-
predictSP.m
Primary MATLAB function for predicting source parameters using the trained mixed-effects regression models -
Model.mat
Encrypted MATLAB data containing the input metadata and trained mixed-effects models -
loadModel.m
MATLAB helper function for loading the encrypted model files -
Examples.m
Worked examples demonstrating typical usage scenarios
- MATLAB R2021b or later
- Statistics and Machine Learning Toolbox
If you use this framework, please cite the associated publication (details will be added upon acceptance).