In May 2024, Earth was hit by the strongest geomagnetic storm of Solar Cycle 25.
The event — now known as the Gannon G5 storm — was triggered by multiple X-class solar flares erupting from solar active region AR3664. Within hours, Earth’s upper atmosphere began expanding dramatically as solar radiation dumped energy into the thermosphere.
For most people, the storm produced beautiful auroras.
For satellite operators, it created chaos.
Thousands of satellites in Low Earth Orbit suddenly experienced elevated atmospheric drag. Orbital trajectories became less predictable, conjunction warnings surged, and satellite operators were forced into large-scale orbital correction maneuvers just to maintain stability.
For SpaceX’s Starlink constellation, this became one of the largest real-world stress tests ever experienced by a mega-constellation.
I wanted to know something very specific:
How much did the storm actually affect orbital decay?
Not theoretically.
Not through simulations.
But through real orbital tracking data.
So I built a Python pipeline combining:
- NASA DONKI solar flare data
- Space-Track orbital history data
- Statistical analysis across more than 229,000 satellite-event observations
The goal was simple:
Quantify how solar activity measurably alters Starlink orbital decay.
The results turned out to be far more interesting than expected.
Why This Matters
Modern satellite constellations operate on precision.
Systems like Starlink rely on:
- autonomous station-keeping
- predictive drag modeling
- collision avoidance systems
- coordinated orbital traffic management
All of those systems assume satellite trajectories remain reasonably predictable.
Solar storms break those assumptions.
As Earth’s atmosphere expands outward during geomagnetic events:
- drag increases
- satellites sink faster
- orbital models drift
- conjunction calculations become noisier
- fuel consumption rises
At small scale, this is manageable.
At constellation scale, with thousands of active satellites sharing crowded orbital shells, it becomes a serious engineering problem.
And as Solar Cycle 25 intensifies, these events are becoming more frequent.
Building the Dataset
To isolate the impact of solar activity, I collected historical data spanning January 2022 through December 2025.
Solar Flare Dataset
Using NASA’s DONKI API, I identified:
- 2,375 high-energy flare events
- 1,902 M-class flares
- 92 X-class flares
These are the events most capable of significantly heating Earth’s upper atmosphere.
Satellite Dataset
Using historical TLE data from Space-Track.org, I tracked:
| Group | Count | Altitude |
|---|---|---|
| Starlink Shell 1 | 50 | ~550 km |
| Starlink Shell 2 | 50 | ~570 km |
| Control Debris Objects | 12 | 500–600 km |
The debris objects served as a control group.
This was critical.
If only Starlink satellites showed altitude changes, the signal could simply be caused by routine station-keeping maneuvers. But if dead debris objects showed the same pattern, then the effect must originate from atmospheric drag itself.
Solar Activity Timeline
The timeline above shows the density and intensity of major flare activity during the observation period. Solar Cycle 25 produced frequent bursts of M-class and X-class events, especially during 2024.
The Analysis Pipeline
The workflow looked like this:
fetch_flares.py -> NASA DONKI flare events
fetch_tles.py -> Space-Track TLE history
compute_decay.py -> Daily orbital decay rates
align_events.py -> Flare-relative event windows
analyze_decay.py -> Statistical analysis
visualize_decay.py -> Charts and distributions
For each flare event:
- Satellite altitude changes were measured
- Daily decay rates were computed
- Flare windows were compared against pre-flare baseline behavior
The final dataset contained:
229,646 satellite-event decay measurements
Does Orbital Decay Actually Increase?
Yes, very clearly.
Across the full dataset:
| State | Mean Decay |
|---|---|
| Baseline | 8.5 m/day |
| Flare Window | 13.0 m/day |
That represents:
a 1.537x acceleration in orbital decay during flare windows.
The statistics were extremely strong:
- Paired t-test: p < 0.001
- Wilcoxon signed-rank: p < 0.001
The signal was not subtle.
Orbital Decay Distribution
The decay distribution shows a strong right-skewed pattern. Most events are moderate, but extreme geomagnetic storms dramatically increase average decay rates across the constellation.
Why Solar Flares Increase Drag
Solar flares release massive bursts of:
- X-rays
- ultraviolet radiation
- energetic particles
That energy heats Earth’s thermosphere.
When the thermosphere heats up:
- it expands outward
- atmospheric density rises at orbital altitudes
- satellites encounter more resistance
Even at 550 km altitude, the atmosphere is not truly empty.
During strong geomagnetic storms, the density increase becomes large enough to measurably alter orbital decay rates across entire constellations.
Does Bigger Solar Activity Cause Bigger Decay?
Again, yes.
Using non-parametric statistical testing, the data showed a clear scaling effect.
| Flare Class | Median Decay Ratio |
|---|---|
| M1–M5 | 1.711 |
| M5–M9 | 1.733 |
| X1–X5 | 1.914 |
| X5+ | 1.970 |
The stronger the flare:
- the hotter the thermosphere
- the larger the atmospheric expansion
- the stronger the drag increase
X-class events consistently produced the strongest orbital effects.
Decay vs Flare Class
The relationship between flare intensity and orbital decay becomes increasingly visible as flare class increases.
The Surprising Part: The Flare Hit Harder Than the CME
Initially, I assumed the CME phase would dominate the drag increase.
But the data suggested otherwise.
| Window | Mean Decay |
|---|---|
| Flare Window | 12.6 m/day |
| CME Window | 11.3 m/day |
The immediate flare radiation produced the sharpest atmospheric response.
This makes sense physically:
- flare radiation reaches Earth at light speed
- thermospheric heating begins almost immediately
- CME plasma clouds arrive later
The atmosphere reacts to the radiation burst first.
The Control Group Validation
This was one of the most important sanity checks in the project.
The dead debris objects showed the same acceleration pattern as Starlink satellites.
| State | Mean Decay |
|---|---|
| Baseline | 1954.2 m/day |
| Flare Window | 2129.9 m/day |
Statistical significance:
- p < 0.001
This confirmed something important:
The observed signal was real , not just unique to Starlink.
Without the control group, that conclusion would have been much weaker.
The Thruster Signature
One of the most interesting discoveries appeared during the May 2024 storm itself.
When plotting orbital altitude over time, a strange pattern appeared:
massive upward spikes during peak drag periods.
At first glance, it looked incorrect.
But then the explanation became obvious.
Atmospheric drag can only pull satellites downward.
Those positive spikes were:
the signature of SpaceX firing autonomous krypton Hall-effect thrusters across the constellation to counteract atmospheric drag.
The constellation was actively fighting the atmosphere in real time.
Autonomous Maneuver Activity During the G5 Storm
The upward spikes visible during the storm window likely represent coordinated station-keeping maneuvers executed across the constellation.
The May 2024 Gannon G5 Storm
Statistical averages are useful.
But extreme events reveal what operators are actually afraid of.
When the Gannon G5 storm hit, it created what can best be described as a constellation-wide sinking wave across Low Earth Orbit.
Analysis of 100 Starlink satellites revealed:
- 60% lost more than 100 meters of altitude in a single day
- the maximum recorded drop was 642 meters in 24 hours
That is enormous for operational satellites.
The Orbital Traffic Jam
The problem was not simply altitude loss.
The real problem was unpredictability.
When thousands of satellites begin sinking simultaneously:
- orbital prediction models drift
- conjunction calculations become unstable
- collision risk estimation becomes noisier
- Conjunction Data Messages (CDMs) surge
Even though Starlink remained well above the ISS altitude (~410 km), the sudden instability in orbital trajectories stressed space traffic management systems worldwide.
This may have been:
the largest coordinated autonomous maneuver event in history.
Recovery Time
Most satellites returned to near-baseline decay rates within approximately 3–4 days after flare activity subsided.
The Cost of Survival
How expensive was recovery?
Using the Tsiolkovsky rocket equation:
Assumptions:
- Starlink V1.5 mass: ~295 kg
- Krypton Hall-effect thrusters
- ~400 meter altitude recovery
Required:
Delta-V ≈ 0.22 m/s
Fuel usage:
- ~4.4 grams krypton per satellite
- ~22 kg total across 5,000 satellites
Financially, that is manageable.
Operationally, it is much more important.
Satellites carry finite fuel reserves designed to last roughly 5–7 years.
Every major geomagnetic storm permanently consumes part of that lifetime.
The real cost is not the gas.
The real cost is:
orbital lifespan.
This Was Not the First Warning
In February 2022, shortly after launch, a geomagnetic storm struck a newly deployed group of Starlink satellites at only ~210 km altitude.
Atmospheric density increased sharply.
The satellites could not overcome the drag increase.
Up to 40 satellites were lost and re-entered within days.
That incident demonstrated how dangerous solar activity can become for low-altitude orbital operations.
Limitations
No orbital analysis using public TLE data is perfect.
Several limitations remain:
- TLE precision introduces noise (~1 km uncertainty)
- subtle low-thrust maneuvers may evade detection
- solar wind speed and F10.7 index were not independently modeled
- overlapping flare events during solar maximum complicate attribution
Still, the overall signal remained statistically strong across the dataset.
Engineering Takeaways
The most important conclusion is not:
“solar flares increase drag.”
That has been understood for decades.
The more important realization is this:
Mega-constellations are becoming space-weather-sensitive infrastructure.
As humanity deploys tens of thousands of satellites:
- atmospheric modeling becomes increasingly important
- autonomous maneuvering becomes mandatory
- fuel budgeting becomes linked to solar activity
- orbital traffic management becomes more complex
Space weather is no longer just an astrophysics topic.
It is becoming an operational engineering problem.
Project Repository
The complete analysis pipeline, datasets, and visualization scripts are available on GitHub:
How the Gannon G5 Solar Storm Almost Dragged Down Starlink (And What 229,000 Data Points Reveal About It)
Executive Summary
In May 2024, Earth was hit by the strongest geomagnetic storm of Solar Cycle 25: the Gannon G5 storm. Triggered by multiple X-class solar flares from active region AR3664, the storm dramatically expanded Earth’s upper atmosphere, increasing drag across thousands of satellites in Low Earth Orbit (LEO).
SpaceX’s Starlink constellation, the largest satellite network ever deployed , suddenly faced a constellation-wide orbital decay event. Satellites began sinking unpredictably, orbital models became unreliable, and autonomous station keeping maneuvers had to be executed at unprecedented scale.
We know solar storms increase atmospheric drag.
But how much?
And what does that actually mean operationally for mega-constellations?
To answer that, I built a Python analysis pipeline combining:
- NASA DONKI solar flare data
- Space-Track orbital history data
- Statistical analysis across 229,646 satellite-event pairs
- Orbital decay…
Final Thoughts
The Gannon G5 storm offered a rare real-world glimpse into how fragile large orbital systems can become during extreme solar activity.
For a brief moment:
- thousands of satellites began sinking together
- orbital models became unstable
- autonomous systems had to react at constellation scale
And all of it was triggered by activity occurring 150 million kilometers away on the surface of the Sun.
As orbital infrastructure continues expanding, events like this will become increasingly important to study.
Because in the era of mega-constellations, the atmosphere itself is becoming an active participant in orbital engineering.
Top comments (0)