Orbit, Attitude & Imaging Geometry

Where the satellite sits determines everything it can -and cannot see.

There’s a question I used to skip over. I’d look at a satellite image, note the date and time, maybe check the sensor spec-and then get straight to the analysis.

I didn’t ask: why is this satellite here, looking at this place, at this angle, at this moment?

It took a few embarrassing conversations with colleagues to realise that this question matters enormously. The geometry of how a satellite observes the Earth isn’t background information. It’s baked into every pixel of the data. Ignore it and you will misread what you’re looking at.

The orbit -not a choice, a trade-off

Most Earth observation satellites operate in Low Earth Orbit, or LEO roughly 300 to 1000 km altitude. Higher orbits give you wider coverage but lower resolution. GEO satellites at 36,000 km can see an entire hemisphere simultaneously but can’t resolve individual buildings. The Earth observation satellites that give us our most useful land imagery Landsat, Sentinel, WorldView, Planet are all in LEO.

Within LEO, the choice that defines an Earth observation satellite more than any other is sun-synchronous orbit, or SSO. It sounds exotic but the principle is simple: engineer the orbital parameters so the satellite crosses any given latitude at approximately the same local solar time on every pass.

Why does this matter? Because sun angle determines shadows, and shadows determine whether your images are comparable across dates. If a satellite imaged a field at 7am on one pass and 3pm on the next, the difference in shadow length, illumination angle, and atmospheric path length would swamp any real changes in the vegetation. Science would become impossible. SSO fixes the illumination geometry making change detection, time-series analysis, and any kind of calibrated comparison between dates actually meaningful.

Most optical Earth observation satellites are SSO, passing over equatorial regions around 10am to 11am local time. It’s become the de facto standard for a reason.

Revisit time-the hidden variable

A satellite in LEO moves fast one full orbit every 90 minutes or so. But the Earth is also rotating underneath it. The ground track shifts westward with each pass. The satellite doesn’t see the same ground strip every orbit.

Revisit time is how long it takes for the satellite to return to the same location. For a single satellite at 600 km altitude, this is typically several days. For Landsat 8, it’s 16 days. That’s a long time if you’re trying to monitor flood extent, crop growth, or a wildfire.

The solutions are constellations ,multiple satellites in the same orbital plane, spaced so that the combined revisit rate is much shorter. Planet’s Dove constellation achieves daily global coverage. Satellite operators also use off-nadir pointing tilting the sensor to look at targets not directly below the spacecraft, which can compress revisit to hours for specific high-priority targets. But there’s a cost to pointing off-nadir: geometry changes, and so does resolution.

Nadir, off-nadir, and why the angle matters

Nadir is straight down. The vector from the satellite directly to the centre of the Earth. Imaging at nadir gives you the most symmetric view, the least distortion, the most accurate measurement of ground area.

Off-nadir means the sensor is pointed to the side the satellite tilts its attitude so the camera looks at a target that isn’t directly beneath it. This extends coverage, enables stereo imaging (capturing the same point from two different angles to derive elevation), and increases revisit frequency.

But look angle has consequences. At nadir, a 30 cm GSD sensor gives you 30 cm pixels. At 30 degrees off-nadir, the same sensor sees the ground at an oblique angle -the effective GSD degrades, tall objects appear to lean away from the camera, and terrain relief creates displacements that require careful correction to undo.

Every satellite image is a view from somewhere specific, at a specific angle, at a specific moment. The image doesn’t show you the Earth. It shows you the Earth from there, then.

Attitude -the three axes that define pointing

Knowing where the satellite is (its position in orbit) is only half the problem. You also need to know which way it’s pointing.

A satellite’s orientation-its attitude is described by three rotations: roll, pitch, and yaw. Roll is rotation around the along-track axis (banking left-right like an aeroplane). Pitch is rotation around the cross-track axis (nosing up or down). Yaw is rotation around the nadir axis (spinning like a compass).

Small errors in any of these even fractions of a degree translate to significant displacements in where each pixel actually hits the ground. At 600 km altitude, a 0.1 degree pointing error produces roughly a 1 km position error on the ground. For a sensor with 30 cm resolution, 1 km is thousands of pixels.

This is why every serious Earth observation satellite carries precise attitude determination systems -star trackers, gyroscopes, GPS receivers and why the geometry calibration of these instruments is an ongoing engineering challenge.

The imaging geometry equation

Put this all together and you can see why understanding the geometry chain matters so deeply for anyone working with satellite data.

The path from a real-world feature to a pixel in your image runs through: orbital position → attitude → look angle → atmospheric refraction → terrain elevation → sensor geometry → pixel coordinates. Every step introduces a potential source of error. Every step needs to be modelled and corrected before the image can be used for measurement.

We’ll walk through how that correction pipeline actually works in Articles 07 and 08. For now, it’s enough to hold the shape of the problem: a satellite image is not a map. It is a projection of the Earth’s surface through all of that geometry onto a sensor. Turning it back into a map orthorectification is one of the most technically demanding steps in the entire data pipeline.

Orbit is not a detail. It is architecture. The decisions made about altitude, orbital plane, inclination, and pointing capability define what data the satellite can produce and for whom.

Understanding orbit means understanding the limits of your data and the opportunities hidden within those limits. It’s one of those things where the more you know, the more you see in every image you open.

Coming up in Article 05 ↓ We’re finally at the moment of capture. What actually happens in the milliseconds when a satellite sensor integrates light? What is a Digital Number and where does it come from? And what does raw, uncorrected satellite data actually look like?