Ascending vs Descending Orbits in SAR: Why It Matters More Than You Think

Early in my SAR career, I detected what appeared to be a massive landslide on a mountainside in central Japan. The backscatter had changed dramatically between two Sentinel-1 acquisitions. I was about to report it when a senior colleague pointed out that I'd compared an ascending pass with a descending pass. The "landslide" was just the mountain's radar shadow switching sides.

That mistake taught me a lesson I never forgot: in SAR analysis, orbit direction isn't a minor detail — it's fundamental.

How SAR Geometry Changes Between Passes

SAR satellites in sun-synchronous orbits make two types of passes over any given latitude:

Ascending pass: The satellite travels roughly south-to-north (ascending in latitude). For Sentinel-1, this typically occurs in the evening local time. The radar looks to the right of the flight direction — in this case, roughly eastward.

Descending pass: The satellite travels north-to-south. This typically occurs in the morning. The radar looks to the right — roughly westward.

The critical difference: the look direction flips by approximately 180°. A slope facing east is illuminated head-on during ascending passes but faces away from the radar during descending passes. The resulting backscatter difference can be enormous — 10 dB or more in mountainous terrain.

What Changes Between Orbits

Layover and Foreshortening

Radar signals travel at an angle to the ground. Tall features like mountains and buildings are "seen" from the side, not from above. This creates geometric distortions:

Foreshortening: Slopes facing the radar appear compressed. A 30° slope illuminated from the front gets squeezed into fewer pixels than it actually occupies on the ground.

Layover: Extreme foreshortening where the top of a feature is mapped closer to the radar than its base. The mountain top appears to "lay over" onto the valley in front of it. This effect reverses completely between ascending and descending orbits.

Radar shadow: Slopes facing away from the radar at steep angles receive no illumination and return no signal — they appear black. The shadow falls on opposite sides of terrain features in ascending vs. descending data.

Building Appearance

Urban buildings are affected similarly. A row of buildings along a north-south street will show:

• Ascending: Bright returns from east-facing walls, shadow on the west side
• Descending: Bright returns from west-facing walls, shadow on the east side

The overall brightness of the urban area may be similar, but the spatial pattern of bright and dark pixels is completely different. This makes pixel-level comparison between orbits unreliable.

Incidence Angle Differences

Even flat terrain is affected. The local incidence angle (the angle between the radar beam and the surface normal) differs between ascending and descending passes because the satellite approaches from different azimuths. This produces systematic backscatter differences for any non-flat surface.

The Practical Rules

Rule 1: Same Orbit for Change Detection

This is the most important rule in SAR time-series analysis. When comparing two dates to detect change:

• Always use the same orbit direction (both ascending or both descending)
• Ideally use the same relative orbit number (Sentinel-1 has 175 relative orbits in each direction)
• Same orbit means same viewing geometry, so backscatter differences represent actual surface changes

Rule 2: Same Orbit for InSAR

Interferometry requires near-identical viewing geometry. Interferograms are computed from same-orbit-direction pairs with small perpendicular baselines. Cross-orbit interferograms are physically meaningless.

Rule 3: Both Orbits for Deformation Decomposition

InSAR measures displacement in the satellite's line-of-sight direction, which differs between ascending and descending. By combining deformation measurements from both orbit directions, you can decompose the signal into vertical and east-west components — providing quasi-3D deformation information.

Rule 4: Both Orbits for Improved Temporal Sampling

If you need more frequent observations and can account for geometric differences (through normalization or by analyzing trends rather than absolute values), combining both orbit directions effectively doubles your temporal sampling. Sentinel-1 ascending and descending passes over the same area are typically offset by 3-6 days.

Sentinel-1 Orbit Configuration

Sentinel-1 repeats its exact orbital track every 12 days (one satellite) or 6 days (two satellites). Each track is identified by a relative orbit number (1-175).

For any given area, the data is typically available from:

• 1-2 ascending relative orbits
• 1-2 descending relative orbits

The overlap between adjacent tracks means some areas are covered by multiple orbits in the same direction, with acquisitions offset by a few days. This is particularly true at higher latitudes where orbital tracks converge.

To check which orbits cover your area of interest:

1. Search for available data over your area
2. Note the orbit direction and relative orbit number of each acquisition
3. Group by orbit direction and relative orbit number
4. Work within consistent groups for time-series analysis

Choosing the Better Orbit

When both ascending and descending data are available, which should you use? Consider:

Terrain orientation: If your study area has predominantly east-facing slopes, descending (westward-looking) data will have less shadow but more foreshortening. Ascending (eastward-looking) data will illuminate these slopes head-on. Choose based on which distortion is more tolerable.

Shadow avoidance: Radar shadow is pure data loss — no information is recovered. Choose the orbit that minimizes shadow over your area of interest.

Temporal sampling: If one orbit direction provides more frequent coverage (due to overlapping tracks), it may be preferred despite suboptimal geometry.

Existing time series: If a long time series already exists in one orbit direction, continuing with the same direction maintains consistency.

A Common Mistake and Its Cost

The orbit-mixing mistake is more common than you'd think, partly because data catalogs sort by date, not by orbit. An analyst searching for "all Sentinel-1 data over Area X in January 2025" gets a mix of ascending and descending acquisitions. Computing NDVI-like backscatter changes between consecutive dates (which alternate between orbit directions) produces spurious signals that correlate with terrain more than with actual surface change.

I've reviewed reports where "deforestation" was detected on steep slopes — and the detection was entirely an artifact of comparing different orbit directions. The forest was fine. The radar geometry was different.

The fix is simple: filter your data by orbit direction before any analysis. It adds one step to your workflow and prevents an entire category of errors.

When Geometry Differences Are the Signal

There's one scenario where comparing ascending and descending data is useful: characterizing 3D structure. Buildings, bridges, and tall vegetation respond differently to eastward vs. westward illumination. These differences carry information about structure orientation, height, and geometry.

Researchers have used ascending-descending comparisons to:

• Estimate building heights in urban areas
• Map forest canopy structure
• Characterize slope-dependent surface roughness

In these applications, the geometric difference between orbits isn't noise — it's the signal. But you need to know you're measuring geometry, not surface change.

The orbit direction is metadata that's always available but rarely gets the attention it deserves. Treating it as a first-class parameter in your analysis workflow prevents the most common SAR analysis mistake and opens the door to more sophisticated multi-geometry analyses.