What You Can See With a 10 MHz Radar

Exploring Ionospheric Backscatter Using SuperDARN Open Data

Most people think of radar as something that tracks airplanes or measures your speed on the highway. But point a 10 MHz radar at the sky and something remarkable happens: the signal bounces off the ionosphere, travels hundreds or thousands of kilometers, and returns carrying information about plasma flowing in the upper atmosphere.

The Super Dual Auroral Radar Network (SuperDARN) is a collection of over 35 HF radars spread across the globe, mostly at high latitudes. Each radar transmits in the 8-20 MHz range and listens for echoes bouncing off ionospheric irregularities and the ground. The network has been operating since the 1980s, and all the data is freely available.

I wanted to see what I could extract from the raw data myself. Not the preprocessed products, but the actual raw autocorrelation functions recorded by the radar. Here is what I found.

The Raw Data: Autocorrelation Functions

SuperDARN radars use a 7-pulse transmit sequence. The received signal is correlated against itself at various time delays (lags) to produce autocorrelation functions (ACFs). Each range gate gets a set of 22 complex-valued lag products. There is no audio here, no waveform to play. Just 22 complex numbers per range gate, per beam, per integration period.

From these 22 complex numbers you can extract three fundamental parameters: signal power (from the lag-zero magnitude), Doppler velocity (from the phase rotation across lags), and spectral width (from how quickly the ACF decorrelates). The range-Doppler map below shows the power spectrum at each range gate, computed by Fourier transforming the ACF lags. The bright features at specific range-Doppler coordinates are real ionospheric backscatter: plasma irregularities reflecting the radar signal back at a Doppler-shifted frequency.

Range-Doppler map, power profile, and Doppler velocity from Saskatoon RAWACF data
Three views of the same raw ACF data from SuperDARN Saskatoon (beam 0, averaged over 120 scans). Left: Range-Doppler map showing power density as a function of range and Doppler frequency. The bright features are ionospheric backscatter at specific range-Doppler coordinates. Middle: Power profile showing signal strength across Doppler frequencies. Right: Lag-1 Doppler velocity as a function of range and time.

Range-Time-Intensity: Watching the Ionosphere Breathe

The workhorse visualization in ionospheric radar is the range-time-intensity (RTI) plot. By stacking consecutive radar scans vertically, you get a 24-hour movie of what the ionosphere is doing. Time runs left to right, range (distance from the radar) runs bottom to top, and color represents signal strength.

Full-day RTI plot showing backscatter power
24 hours of backscatter power from the Christmas Valley West radar, beam 7 (boresight direction). 1,440 scans are stacked to cover the full day. The ionosphere “breathes”: during the day, higher electron density from solar radiation refracts the signal at lower altitudes (shorter skip distance). At night, reduced ionization means the signal must travel higher before being refracted back, pushing ground scatter to longer ranges. The band near 300-600 km is ground scatter, where the signal bounces off the Earth’s surface after reflecting from the ionosphere.

Notice the structure. During the daytime hours (roughly 14:00-02:00 UTC, which is local daytime in Oregon), solar radiation increases electron density, causing the signal to refract back at lower altitudes. At night, ionization drops and the signal has to reach higher altitudes before refracting, so backscatter shifts to longer ranges.

Doppler Velocity: Plasma in Motion

The phase rotation of the ACF across lags gives the Doppler shift, which tells you how fast the ionospheric plasma is moving toward or away from the radar. This is not the raw signal power anymore. This is physics: the line-of-sight component of plasma convection velocity.

Full-day RTI showing Doppler velocity
Doppler velocity from the same data. Blue is plasma moving toward the radar, red is away. The velocity structure reveals the large-scale convection pattern driven by solar wind interaction with Earth’s magnetic field.

Spectral Width: Measuring Ionospheric Turbulence

The third parameter extractable from the ACF is spectral width, which measures how quickly the ACF decorrelates across lags. Narrow spectra mean laminar, organized flow. Wide spectra mean turbulence, particle precipitation, or complex echo structure.

Full-day RTI showing spectral width
Spectral width for the same 24-hour period. Regions of enhanced spectral width at higher latitudes often mark the auroral boundary, where particle precipitation creates turbulent ionospheric conditions.

Elevation Angles: The Hard Part

This is where it gets interesting. SuperDARN radars have a main antenna array and a secondary interferometer array, spaced roughly 100 meters apart. The phase difference between the signal received at both arrays depends on the elevation angle of the incoming echo. This is the same principle behind radio interferometry in astronomy, but applied to radar echoes.

The complication is phase unwrapping. The phase difference between arrays can only be measured modulo 2π, which means the raw measurement is ambiguous. The algorithm that resolves this ambiguity was published by Shepherd (2017) and involves tracking the phase difference across range gates and choosing the unwrapping path that minimizes phase jumps.

Implementing this correctly required tracking down the interferometer baseline distance for each of the 35+ radars (buried in hardware configuration files), handling the geometric phase calculation for each range gate, and then implementing the phase unwrapping logic. The result, when it works, is stunning:

Elevation angles and reflection heights derived from XCF interferometry
Top: Elevation angles extracted from the cross-channel correlation (XCF) data using the Shepherd (2017) phase unwrapping algorithm. Typical elevation angles are 10-40 degrees. Bottom: The derived reflection height, computed as range × sin(elevation). This shows the ionospheric reflection layer hovering around 200-400 km altitude, consistent with the F-region peak. The diurnal variation is clearly visible: the effective reflection height drops during the day (higher electron density refracts the signal sooner) and rises at night (lower density means the signal penetrates deeper before bending back).

To validate the implementation, I compared our Shepherd 2017 elevation angles against the official FITACF3 fitted values produced by the SuperDARN processing chain. The correlation was R = 0.95, confirming that the phase unwrapping and geometry are correct.

Geographic Fan Plots: Seeing the Radar Footprint

Each SuperDARN radar electronically steers through 16 to 24 beams, covering roughly 52 degrees of azimuth. By mapping the backscatter from all beams onto a geographic projection using the radar’s known antenna pattern and range gate spacing, you get a “fan plot”: a wedge-shaped view of what the radar sees on the ground. With two radars located at the same site pointing in different directions, you can combine their fans into a single view that covers a much wider area of the sky.

The animation below shows the Fort Hays East and Fort Hays West radars in Kansas, which share a site but point in opposite directions (45 degrees and -25 degrees from north). Together they cover nearly 100 degrees of azimuth. Each frame is one complete beam scan (about 1-2 minutes), and the animation plays through a full day of backscatter power.

Fan beam animation from the Fort Hays dual radar system in Kansas. Each frame shows backscatter power from all beams projected onto a geographic map using cartopy. Range rings are marked in km. The radar’s two fans point in different directions, covering the ionosphere over the central United States. Ground scatter appears as the bright band at shorter ranges; ionospheric backscatter extends further out.

Polar Convection Maps

The SuperDARN community produces gridded convection maps that combine velocity data from multiple radars into a single polar view of ionospheric plasma flow. These MAP files contain fitted velocity vectors on a magnetic latitude / magnetic local time grid, along with the cross-polar cap potential (a measure of how strongly solar wind drives the convection).

24-hour animation of ionospheric convection over the North Pole on November 1, 2025. Arrows show plasma flow direction and speed. The coastline overlay provides geographic context: you can see North America, Greenland, and Scandinavia. The two-cell convection pattern, driven by magnetic reconnection on the dayside and nightside, is clearly visible.

What Makes This Possible

All of the visualizations on this page were produced using open source tools and open data:

  • Data access: SuperDARN data is distributed via Globus, a high-performance data transfer platform. Anyone can request access to the SuperDARN data collections.
  • File reading: pyDARNio handles the DMap file format that SuperDARN uses for RAWACF, FITACF, and MAP files.
  • Processing: All signal processing (ACF to Doppler velocity, spectral width, elevation angles) was implemented from scratch in Python using only numpy.
  • Visualization: matplotlib and cartopy for geographic projections.

Acknowledgments

The SuperDARN radars are operated by an international collaboration of scientists and institutions. Data is made freely available through the efforts of the SuperDARN Data Analysis Working Group and stored on servers managed by Virginia Tech and others. This work would not be possible without their commitment to open data.

SuperDARN is a collection of radars funded by national scientific funding agencies of Australia, Canada, China, France, Italy, Japan, Norway, South Africa, Sweden, United Kingdom, and United States of America.

All plots produced from SuperDARN RAWACF and MAP data (Christmas Valley West, June 1 2026; Saskatoon, December 24 2026; Northern Hemisphere convection map, November 1 2025). Elevation angles computed using the Shepherd (2017) algorithm from XCF data.