RESTORE

3.0.0

Added

• Support for multiple RESTORE realizations, with descriptive statistics for individual realizations.
• Automatic selection of a representative realization, defined as the successful run that is closest to the central behavior of the ensemble.
• settings.txt file reporting the input parameters for each realization.
• metadata.json file reporting the RESTORE version, Python version, input file path and file provenance.
• restore command line through pyproject.toml.
• Map showing the full input catalog and the spatial region used for STAI analysis.

Changed

• Input magnitudes are now required to be binned at ΔM = 0.1. Any conversion to this magnitude binning is left to the user.
• If no b-value is provided, RESTORE now uses the default value b = 1.0.
• Magnitude of completeness is now estimated with the Lilliefors test using truncated exponential dithering (Stallone and Spassiani (2026)).
• Stochastic reproducibility is now controlled through explicit random seeds and default_rng generator.
• Improved messages, documentation, and function docstrings.

Fixed

• Spatial smoothing now clips numerical values to the valid domain before applying arccos.
• Spatial sampling now validates that smoothed rates are finite and non-zero.
• Fitted hypocentral depth density is now validated for finite, positive, and non-zero values before sampling.
• Beta hypocentral depth distribution is now fitted in [0, 1] coordinates and then converted back to physical depth units for sampling and plotting.
• Fixed minor bugs.