How We Monitor Earth's Pulse: Our 3-Station Network

Why Three Stations

A single monitoring station can tell you what's happening at that location. It cannot tell you what's happening globally.

A thunderstorm 300 km from a magnetometer produces a local signal that looks identical to a global geomagnetic event on a spectrogram. Industrial machinery, agricultural equipment, nearby power lines — all generate electromagnetic noise that a single station cannot distinguish from real Schumann Resonance activity.

SunGeo monitors three stations across two continents. When a signal appears at all three, it's almost certainly real. When it appears at only one, it's almost certainly local noise. This cross-validation is what separates our data from screenshots of a single spectrogram.

Station 1: Tomsk, Russia

The Space Observing System in Tomsk has been monitoring Schumann Resonance continuously for years. It's one of the most widely referenced stations in the world — the spectrogram you see on most Schumann websites originates here.

What it measures: Full electromagnetic spectrum in the Schumann band (0-40 Hz), displayed as a 24-hour rolling spectrogram. The horizontal axis is time, the vertical axis is frequency, and color intensity represents signal strength.

Why it matters: Tomsk sits in central Siberia, far from most industrial electromagnetic interference. The continental location provides a different lightning-distance profile than Mediterranean stations, meaning it "hears" a different mix of global thunderstorm activity.

How we use it: Tomsk is our primary data source. We fetch the spectrogram image every hour, analyze it with both pixel-level algorithms and AI vision models, and use it as the baseline for our status scores.

Station 2: ETNA Observatory, Sicily

The ETNA Radio Observatory operates a coil magnetometer on the slopes of Mount Etna. Yes, the volcano.

What it measures: Electromagnetic signals from 0-105 Hz — well beyond the standard Schumann harmonics. The spectrogram covers an 8-hour rolling window, updated approximately every 30 minutes. Resolution: 813 x 601 pixels.

Why it matters: The Mediterranean location gives ETNA a fundamentally different perspective on global lightning. African and Middle Eastern thunderstorm activity registers more strongly here than at Tomsk. The volcanic environment also produces occasional electromagnetic signatures from Etna's own geophysical activity.

Reading ETNA data: The color scale runs from dark (quiet) through green and yellow (moderate) to red and white (intense). Fixed red horizontal lines at certain frequencies are instrumental artifacts — ignore those. Look for broad spectral features that evolve over time.

Station 3: Cumiana, Italy

The Cumiana station near Turin operates a geomagnetic sensor focused on VLF (Very Low Frequency) detection.

What it measures: Geomagnetic pulsations in the Schumann band. The sensor type is different from ETNA's coil magnetometer — it's optimized for detecting magnetic field variations rather than electric field components. Resolution: 815 x 569 pixels, updated roughly every 30 minutes.

Why it matters: Different sensor types respond differently to the same signal. A genuine Schumann Resonance event will appear on both electric-field sensors (like ETNA's coil) and magnetic-field sensors (like Cumiana's geophone). Agreement between sensor types is strong evidence that a signal is real.

The complementary pair: ETNA and Cumiana are roughly 900 km apart — close enough to share the same general thunderstorm environment, but far enough that local noise sources rarely affect both. When both Italian stations agree but Tomsk disagrees, we know the signal is regional (European/Mediterranean). When all three agree, it's global.

The Analysis Pipeline

Raw spectrograms are just images. They need interpretation. Here's what happens between the image download and the status you see on the homepage.

Step 1: Pixel Analysis

Before any AI gets involved, we run pixel-level analysis on the Tomsk spectrogram. The PixelAnalyzer scans five frequency bands corresponding to the first five Schumann harmonics (7.83, 14.3, 20.8, 27.3, 33.8 Hz).

For each band, it calculates:

• Baseline brightness using the 25th percentile (P25) — this represents the quiet background
• Peak brightness from multiple sample windows across the most recent 2 hours
• Delta (peak minus baseline) — how far above background the signal is
• Band score weighted by proximity to the fundamental frequency

The pixel score becomes a floor for the AI analysis. The AI can rate activity higher than the pixels suggest, but never lower. This prevents the AI from hallucinating calm conditions when the spectrogram clearly shows high activity.

Step 2: AI Vision Analysis

We send the spectrogram to Google Gemini Flash (a vision-language model) with a structured prompt that includes:

• The pixel analysis results as context
• Current solar wind data from NOAA
• The Kp geomagnetic index
• Instructions to assess status (calm/elevated/active/storm), dominant frequency, amplitude, and notable events

The AI returns a structured JSON response with status, score (0-100), frequency analysis, and a natural-language summary.

Step 3: Multi-Source Cross-Validation

ETNA and Cumiana get their own independent AI analysis. Each station's spectrogram is analyzed separately with station-specific prompts (because the image formats and frequency ranges differ).

The confidence score you see on the dashboard reflects cross-source agreement:

• High confidence (3/3): All three stations report consistent activity levels
• Medium confidence (2/3): Two stations agree, one disagrees or is offline
• Low confidence (1/3): Only one station reporting — treat the data as indicative, not definitive

Step 4: Translation and Display

The AI generates summaries in English. For other languages, a second AI call translates the summary while preserving technical accuracy and natural voice.

Total cost per analysis cycle: approximately $0.04 (Gemini Flash for vision and interpretation). At 24 cycles per day, that's roughly $1 per month for continuous AI-powered monitoring.

Data Sources and Costs

| Component | Source | Update Frequency | Cost |

|-----------|--------|-----------------|------|

| Tomsk spectrogram | Space Observing System, Tomsk State University | Continuous | Free (public data) |

| ETNA spectrogram | ETNA Radio Observatory, Sicily | ~30 minutes | Free (public data) |

| Cumiana spectrogram | VLF.it Observatory, near Turin | ~30 minutes | Free (public data) |

| Solar wind data | NOAA DSCOVR satellite | Real-time | Free (public API) |

| Kp index | NOAA Space Weather Prediction Center, 13 observatories | Every 3 hours | Free (public API) |

| AI analysis | Google Gemini Flash (vision model) | Hourly | ~$0.04/cycle (~$1/month) |

What Makes This Different

Most Schumann Resonance websites display a single spectrogram image and let you interpret it yourself. That's useful if you know how to read spectrograms. Most people don't.

SunGeo adds three layers that others don't:

1. Multi-station cross-validation — so you know whether activity is global or local noise

2. AI interpretation — translating complex spectral data into plain language ("Earth's pulse is elevated, you might feel more alert")

3. Solar context integration — because Schumann Resonance doesn't exist in isolation; solar wind, Kp index, and geomagnetic conditions all affect what you see

The goal is making this data accessible without dumbing it down. The raw spectrograms are always there if you want them. But you shouldn't need a physics degree to understand what Earth is doing right now.