Methodology

Every SunGeo forecast pulls from public scientific feeds. We never invent data — we aggregate, cross-check, and translate it into plain language.

• · NOAA SWPC — Kp index (planetary geomagnetic activity), 3-day forecast, solar wind, X-ray flares. swpc.noaa.gov
• · BGS Eskdalemuir — Eskdalemuir (Scotland) Schumann magnetometer — daily ELF spectrograms. bgs.ac.uk
• · INGV (ETNA) — Mt Etna (Sicily) ELF coil magnetometer — hourly spectrograms. ingv.it
• · HeartMath GCI — HeartMath Global Coherence Initiative — California and Alberta magnetometers. heartmath.org/gci
• · Tomsk State University — Siberian broadband ELF spectrogram (sos70.ru) — hourly. sosrff.tsu.ru

All readings are timestamped in UTC and stored locally for cross-validation. When a station is stale or offline, we flag it visually instead of silently substituting.

Each tool exposes a 0-100 score with a deterministic formula. No black-box scoring — every input is documented and reproducible.

Pet Anxiety Index (PAI)

Weighted sum of geomagnetic Kp factor (peak forecast next 24h), barometric pressure drop (last 6h), Schumann state, and lunar illumination — multiplied by a species coefficient.

PAI = (Kp_factor + pressure_factor + schumann_factor + moon_factor) × species_multiplier
species_multiplier: dog 1.00 · cat 0.85
buckets: ≤30 calm · ≤60 mild · ≤80 high · 80+ extreme

Sleep Quality Index (SQI)

Combines Kp forecast for the sleeping window, moon phase (full moon penalty), and Schumann amplitude. Outputs a 0-100 score with bucket interpretation (poor/fair/good/excellent).

Headache & Migraine Risk

Tracks barometric pressure rate of change (dP/dt). Drops greater than 6 hPa in 24h are correlated with migraine onset (Hoffmann 2015 — Cephalalgia). Kp ≥4 adds a geomagnetic component.

We cross-validate Schumann readings across 6 stations on 3 continents. A single station with anomalous values is flagged, not blindly trusted. The displayed status is the cross-station median.

Forecast accuracy is reviewed monthly by comparing predicted Kp peaks against NOAA actuals. Algorithms are versioned and documented in our public GitHub repository.

SunGeo uses two AI models. A vision model (Google Gemini 3.1 Flash Lite) analyzes raw spectrograms hourly to extract amplitude, frequency stability, and quality factor. A text model (Gemini 2.5 Flash Lite) translates the deterministic forecast into 8 languages.

AI does not invent forecast scores. Scores are computed by deterministic formulas (above). AI is used only for spectrogram interpretation and natural-language translation. Every reading is logged with cost, model version, and timestamp in our admin pipeline.

Health-related interpretations carry a visible disclaimer and link to peer-reviewed sources. We never present AI output as medical advice.

How this page is maintained

This page is updated whenever we change a formula, add a data source, or modify the AI pipeline. Material changes are logged in our public changelog and reflected in the "last reviewed" date below.

Last reviewed: 2026-04-28