🔭 astrophysics

Real-time detection of solar flares from ground-based VLF data

This paper presents a real-time solar flare detection method using ground-based VLF phase data and propagation models to estimate X-ray flux and D-region electron density, offering a resilient, low-latency alternative to satellite-based monitoring that detects most M and X-class flares within a quarter of their rise time.

Original authors: Pauline Teysseyre, Carine Briand, Morris Cohen

Published 2026-05-13

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Pauline Teysseyre, Carine Briand, Morris Cohen

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the Earth is wrapped in a giant, invisible blanket of air called the ionosphere. The very bottom layer of this blanket, sitting just above our heads (about 60 to 90 kilometers up), is called the D-region. This layer is like a sponge that soaks up radio waves used for long-distance communication. Usually, this sponge is dry and steady. But when the Sun sneezes a burst of energy (a solar flare), that layer gets soaked instantly, becoming a heavy, wet sponge that swallows radio signals.

The problem is that this "sponge" is too high for satellites to hover over easily, and too low for balloons to stay in. So, how do we know when it gets wet?

The Ear to the Ground

The authors of this paper built a system that listens to the Earth's "heartbeat" using VLF (Very Low Frequency) radio waves. Think of these waves like a giant, invisible guitar string stretched between a radio tower and the ground. The Sun's flares change the tension of that string.

The team created a new tool, a Python package called vlf4ions, that acts like a super-sensitive ear. It listens to radio signals from four different transmitters around the world (in the UK, USA, Iceland, and Italy) as they bounce off the Earth's ionosphere.

How It Detects a Flare (The "Trend" Analogy)

Normally, the radio signal travels in a smooth, predictable rhythm. When a solar flare hits, the rhythm suddenly changes—it speeds up or slows down.

The paper describes an "incremental algorithm" that acts like a detective looking for a change in the story.

The Detective: It watches the radio signal minute by minute.
The Clue: It doesn't wait for the whole story to change; it looks for the first sign that the plot has shifted.
The Result: It can spot a solar flare 25% of the way through its rise time. Imagine a wave building up in the ocean; this system spots the wave while it's still just a small ripple, long before it crashes.

The "Weather Forecast" for Space

Once the system spots a flare, it tries to guess how big the "storm" is.

The Analogy: Imagine you are trying to guess how hard it is raining by looking at how wet a few different umbrellas are. If one umbrella is slightly wet, it's a drizzle. If three umbrellas are soaking wet, it's a downpour.
The Method: The system looks at how much the radio signals from multiple towers changed. It uses a bit of math (probability) to estimate the Sun's X-ray energy. It's not perfect, but it's good enough to say, "Hey, a storm is coming," and give a rough idea of its strength.

Why This Matters (The "Backup Plan")

Currently, most space weather alerts come from satellites (like NOAA's GOES satellites). These are like high-tech weather stations in space. They are great, but they have two weaknesses:

They are slow: It takes them 4 to 6 minutes to send a signal back to Earth.
They can break: If a satellite glitches or loses power, we are blind.

The vlf4ions system is the ground-based backup.

It's faster: Because the data is collected right here on Earth, the alert can be sent in less than a minute.
It's tough: Ground equipment is much harder to break than a satellite. If the satellite fails, this system keeps working.
It's cheap: You can build a receiver for a fraction of the cost of a satellite.

The Bottom Line

The paper presents a new, real-time way to detect solar flares using only ground-based radio receivers. It successfully detected 82.7% of major solar flares very early in their development. It also provides a rough estimate of the flare's strength and how much it will disrupt radio communications.

Think of it as a seismograph for the Sun. Just as seismographs detect earthquakes by feeling the ground shake, this system detects solar flares by feeling the "shake" in the Earth's radio waves. It's a simple, resilient, and fast way to keep our radio communications safe from solar storms.

Technical Summary: Real-time detection of solar flares from ground-based VLF data

Problem Statement
The D-region of the ionosphere (60–90 km altitude) plays a critical role in High Frequency (HF) wave absorption, affecting civil, broadcasting, and military communications. However, continuous monitoring of this region is impossible via satellite or balloon due to its low altitude. Current flare detection systems rely primarily on Geostationary Operational Environmental Satellite (GOES) data, which measures solar X-ray flux directly. While reliable, GOES alerts suffer from a latency of 4 to 6 minutes and can saturate at high flux levels (X17). Furthermore, satellite-based systems are vulnerable to space weather disturbances. There is a need for a resilient, ground-based alternative that offers real-time detection with lower latency and independent verification capabilities.

Methodology
The authors present a method and a Python package, vlf4ions, designed to detect and characterize solar flares using ground-based Very Low Frequency (VLF, 15–45 kHz) data. The system monitors VLF waves propagating in the Earth-ionosphere waveguide, where solar flares increase D-region electron density, altering wave phase and amplitude.

The workflow consists of four main stages:

Data Preprocessing: Raw data from an AWESOME receiver (Nanc¸ay, France) undergoes amplitude calibration and phase correction. Phase correction involves unwrapping and detrending to remove artificial slopes caused by transmitter frequency drifts, utilizing a frequency shift parameter ( $\delta f$ ) derived without reliance on historical data. A moving median filter and a flagging system (based on signal amplitude and solar zenith angle) remove noise and nighttime periods.
Flare Detection: An incremental algorithm detects breakpoints (trend changes) in the VLF phase data. The algorithm minimizes the likelihood criterion ( $L$ ) to determine if a new data point follows the existing trend or introduces a break. A breakpoint is confirmed if the likelihood improvement exceeds a user-defined threshold ( $\delta$ ).
Classification and Estimation: Detected breakpoints are classified as quiet periods, flare onset, maximum, or decay based on phase slope and variation ( $\Delta\Phi$ ). To estimate the solar X-ray flux, the system employs a probabilistic approach using Bayesian statistics. It combines Probability Density Functions (PDFs) derived from the phase variations of multiple transmitters (e.g., NRK, NAA, NSY, GQD) to calculate the posterior probability of the X-ray flux.
Electron Density Profiling: The system estimates D-region electron density profiles by comparing measured VLF amplitude and phase against modeled values from the Longwave Mode Propagator (LMP), utilizing the Wait profile parameters ( $h'$ and $\beta$ ).

Key Results
The method was validated using one year of data (November 2023–October 2024) covering 407 M and X-class flares.

Detection Performance: The system detected 82.7% of M and X-class flares. Crucially, it identified these flares within one-fourth of their rise time on average. For isolated flares above M3, the median detection time was 0.22 times the rise time.
Latency: The system detects the start of a flare approximately 7.1 minutes after the X-ray flux increase begins, but only 2.6 minutes after the VLF phase begins to rise. This is faster than the 4–6 minute latency typical of GOES alerts.
Flux Estimation: Combining data from multiple transmitters reduces the confidence interval of the estimated X-ray flux. While the current PDFs (based on limited training data) have wider confidence intervals for extreme events (e.g., [M1.7, X3.6] for NRK), the method successfully estimates flux within a factor of two for the validation set.
Implementation: The vlf4ions Python package processes data for a single transmitter in less than one second, with flux and density estimates taking a maximum of two seconds, confirming its suitability for real-time operation.

Significance and Claims
The paper claims that this approach provides a resilient, ground-based backup for space weather monitoring that is independent of satellite infrastructure. By relying solely on ground-based data, the system avoids the vulnerabilities of space-based assets (e.g., electronic glitches, loss of communication) and offers reduced data latency.

The authors emphasize that the system does not require prior knowledge of quiet-day profiles or external GOES data to function, making it robust during active solar periods. While the current implementation focuses on phase data, the authors note that the system detects changes in D-region ionization generally; thus, it can serve as a general alert system for ionospheric disturbances. The paper positions vlf4ions as a step toward a more resilient detection network, capable of providing alerts comparable to or faster than satellite-based systems, while also offering estimates of D-region electron density and HF absorption impacts. Future improvements suggested by the authors include the use of synthetic data from chemistry models to refine PDFs for C-class and extreme X-class flares, and the integration of amplitude data to further reduce estimation errors.

The Ear to the Ground

How It Detects a Flare (The "Trend" Analogy)

The "Weather Forecast" for Space

Why This Matters (The "Backup Plan")

The Bottom Line

More like this