3Y0K Bouvet Island — Propagation Dataset, Notebook, and Interactive Demo
Greg Beam
We're sharing a propagation analysis package from the recent 3Y0K Bouvet Island DXpedition (March 1-14, 2026). Everything is open, free, and reproducible.
Dataset
A standalone SQLite file (1.5 MB) with six tables: 3,209 raw RBN CW skimmer spots, 548 RBN signatures, 6,416 PSKR digital mode signatures, solar conditions (SFI, Kp, DSCOVR Bz), DXpedition metadata, and prediction results from both IONIS V22-gamma and VOACAP. Ready for Pandas, R, or any SQLite client.
Jupyter Notebook
A pre-run 9-section analysis covering solar timeline, band activity, geographic reach, day/night classification, SNR distributions, historical context (none for this one), and a 3-way comparison of Observed vs IONIS vs VOACAP predictions. All charts render on GitHub — no setup needed to view.
View: https://github.com/IONIS-AI/ionis-jupyter/blob/main/notebooks/dxpedition-3y0k-bouvet-2026.ipynb
Interactive Demo
A web app for exploring the data visually — band filters, maps, activity charts, solar timeline, and the prediction comparison page. No install required.
Explore: https://dxpedition-demo.vercel.app/
Prediction Comparison
We tested IONIS V22-gamma (a neural network trained on 175M propagation signatures) and VOACAP against the observed RBN data. Bouvet at 54S is a low-exposure location for both systems. IONIS showed an overall bias of -1.9 dB across 548 RBN paths; VOACAP showed -11.3 dB. The full per-band breakdown and caveats (survivorship bias, architectural differences, sample size limitations) are documented in the notebook.
A pre-registered blind prediction study is planned for TX9W Marquesas (April 2026).
We welcome feedback is always welcome.
73,
Greg Beam, KI7MT
IONIS-AI Project — ionis-ai.com
Phil Karn
Cool, I didn't know you were doing this. Some of the FT8 spot data to PSK Reporter came via ka9q-radio/wsprdaemon sites where I keep raw logs. Some of the sites are KFS (4 log-periodic antennas), KPH (one log periodic), VY0ERC (Eureka, Nunivut, CA @ 80N) and DP0GVN (Neumayer III research station, 70S). I also have some audio recordings made from DP0GVN.
If anyone would like them, let me know.
Phil
--
Please follow the HamSCI Community Participation Guidelines at http://hamsci.org/hamsci-community-participation-guidelines.
---
You received this message because you are subscribed to the Google Groups "HamSCI" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hamsci+un...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/hamsci/CAKKcKUt7akrOtqdM9Q_vC20dZxtcHStmcJNWWT9qS%3De0k%2BEWFg%40mail.gmail.com.
Greg Beam
Hi Phil,
Yes, absolutely—we would love access to those raw logs and audio. Thank you for reaching out and offering them.
This data is incredibly valuable for two specific limitations we're currently tackling:
Polar Path Blind Spots: The Neumayer (DP0GVN) and Eureka (VY0ERC) nodes provide the exact high-latitude observational baselines we need to close the training gap for extreme polar geometries that we noted in the Bouvet study.
Bypassing the Decode Floor: Standard PSKR data can truncate at the ~-24 dB FT8 decode floor, which introduces survivorship bias. Your raw logs and audio will allow us to analyze sub-threshold signals and actual band closures to get a true picture of the noise floor.
Since IONIS is an entirely FOSS project, our plan is to process these logs and publish the resulting signatures to our SourceForge repository alongside the 3Y0K dataset so the wider community can use them.
How do you prefer to handle the transfer? We can pull from an endpoint or I can provide a secure drop location on our end—whatever is easiest for you.
To view this discussion visit https://groups.google.com/d/msgid/hamsci/cdff433e-ca1a-4980-87f7-a8e719ab0d36%40ka9q.net.