Sysmic v8.0.0 — Fractal Tomography of Seismicity

Sysmic is an open-source Python framework for precision-calibrated Bayesian fractal tomography of seismic catalogs. It implements the Grassberger–Procaccia correlation integral with MCMC inference of latent dimensionality ($D_3$), resolving the dual paradoxes caused by finite location uncertainty.

Companion software to: Firmenich, Firmenich & Firmenich (2026). Fractal Tomography and the Fisher Information Barrier of Seismicity. JGR: Solid Earth. Under review.
Preprint: doi:10.31223/x5rf3v (EarthArXiv, Dec 2025)

---

Key Results (Empirical — Zero Synthetic Data)

Region	Network	$\sigma_h$ (km)	$D_2$	$D_3$	IQI Score
Noto, Japan	Hi-Net	0.5	2.12	2.82 ± 0.05	✅ 0.995
Tohoku, Japan	Hi-Net	0.7	2.09	2.83 ± 0.06	✅ 0.966
Tokachi, Japan	Hi-Net	0.6	2.15	2.86 ± 0.07	✅ 0.972
Valais, Switzerland	Swiss-SED	0.8–1.2	1.50	1.911 ± 0.066	✅ 0.988
SCSN baseline	SCSN	0.1	—	—	✅ 0.999
Cook Strait, NZ	GeoNet*	<1.0	2.24	2.53 ± 0.17	✅ >0.90
Bay of Plenty, NZ	GeoNet*	<1.0	1.91	2.20 ± 0.20	✅ >0.90
Sumatra	GEOFON	7.5	2.21	2.998 ± 0.002	❌ 0.000
Cascadia	USGS	6.1	2.21	→ 3.0	❌ 0.329

*High-precision local relocations (not national GeoNet catalog).

Fisher Information Barrier: $\sigma_c = 2.3 \pm 0.4$ km — empirically validated precision threshold above which Bayesian $D_3$ inference collapses to the embedding-dimension prior.

Tectonic Hierarchy ($D_2$): Rifting (1.50) → Transform (1.81) → Subduction (2.12) → Collision (2.24) → Deep Slab (1.26).

---

Repository Structure

sysmic/                     Core Python library (v8.0.0)
├── core.py                 Grassberger-Procaccia estimator (O(N log N), k-d tree)
├── bayesian_d3.py          MCMC D₃ inference (emcee, Uniform[1.5,3.0] prior)
├── statistics.py           Bootstrap, Theil-Sen, b-value (Aki 1965)
├── geometry.py             Steiner formula, deck-of-cards model (Q=2/3)
└── zaccagnino_stability.py Mmin-independence stability test (IQI criteria $S$)

data/                       22 empirical CSVs (all from real catalogs)
figures/                    14 manuscript figures (PNG 300 dpi)
figs_gen.py                 Reproducible figure generation (all 14 figs)
notebooks/
├── sysmic_colab.ipynb      Interactive Google Colab (no install required)
└── tutorial.ipynb          Step-by-step tutorial
app/
└── streamlit_app.py        Interactive web dashboard
joss/                       JOSS submission files (paper.md, paper.bib)
docs/
└── methods.md              Extended methods reference
tests/
└── test_core.py            Unit tests (6 test classes, 25+ assertions)
setup.py / pyproject.toml   pip-installable package
Dockerfile                  Reproducible containerized environment
CITATION.cff                Machine-readable citation metadata
.zenodo.json                Zenodo archive metadata

---

Quickstart

Option 1 — Google Colab (zero install)

Option 2 — Docker (fully reproducible)

docker build -t sysmic .
docker run -p 8501:8501 sysmic
# Open http://localhost:8501

Option 3 — Local install

pip install -r requirements.txt
python figs_gen.py                          # Reproduce all 14 figures
pytest tests/ -v                            # Run unit tests
streamlit run app/streamlit_app.py          # Launch interactive dashboard

Option 4 — pip (editable install)

pip install -e .

---

Reproducing the Manuscript Figures

python figs_gen.py

All 14 figures are generated deterministically from the empirical CSVs in data/. No internet connection required. Output: figures/*.pdf (600 dpi) and figures/*.png (300 dpi). Runtime: ~2–3 min on a standard laptop.

---

MCMC Configuration (per manuscript §3.3)

Parameter	Value
Prior	Uniform[1.5, 3.0]
Walkers	32
Steps	10 000
Burn-in	2 000
ESS target	> 5 000
Convergence	$\hat{R} < 1.01$

---

Data Sources

All data from publicly accessible catalogs:

Dataset	Source	Access
Hi-Net Japan (JUICE-relocated)	NIED doi:10.17598/NIED.0003	Registered researchers
USGS Pan-American	earthquake.usgs.gov	Open
GeoNet New Zealand	geonet.org.nz	Open
GEOFON Sumatra	geofon.gfz-potsdam.de	Open
Swiss-SED Valais	seismo.ethz.ch	Open
ISC-GEM v12.1	doi:10.31905/D808B825	Open

Data Source (primary): United States Geological Survey (USGS)

---

Citation

@article{Firmenich2026,
  author  = {Firmenich, Facundo and Firmenich, Pau and Firmenich, Le\'{o}n},
  title   = {Fractal Tomography and the Fisher Information Barrier of Seismicity:
             Resolving Dual Paradoxes via Precision-Calibrated Bayesian Inference},
  journal = {Journal of Geophysical Research: Solid Earth},
  year    = {2026},
  note    = {Under review},
  doi     = {10.31223/x5rf3v}
}

@software{Firmenich2026sysmic,
  author  = {Firmenich, Facundo and Firmenich, Pau and Firmenich, Le\'{o}n},
  title   = {Sysmic: Fractal Tomography of Seismicity},
  version = {8.0.0},
  year    = {2026},
  doi     = {10.5281/zenodo.18480821},
  url     = {https://github.com/cedesur/sysmic}
}

---

License

GPLv3 — see LICENSE. Free to use, study, modify, and share.

---

April 8, 2026 — Sysmic v8.0.0