Optical vs SAR: Choosing the Right Satellite for Port Surveillance

Most analysts come to satellite imagery through optical: Sentinel-2, Landsat, Planet. The images are intuitive — the eye reads them at once. Ships are ships, runways are runways, a fresh dredge is the colour of wet sand. Then a major news event happens at night, or under a low pressure system, and the discipline runs into its first wall: cloud cover.

This is where Synthetic Aperture Radar — SAR — earns its keep. Sentinel-1 transmits its own signal and reads the reflection. Cloud, smoke, night, rain — none of it matters. The cost is that the image looks alien at first, and learning to read it takes time. The benefit is that the satellite simply works when you need it.

When optical is the right tool

Sentinel-2 is the default for any task that involves:

• Confirming the presence of a specific vessel type in a known berth on a known sunny morning.
• Reading dredging, oil spills, or hull markings.
• Measuring vegetation, surface temperature, agricultural change over time.
• Producing imagery that a non-specialist editor or reader can interpret without a caption longer than the photo.

The 10-metre resolution of the visible bands is enough to count container stacks and identify large vessel classes. The five-day revisit cycle is enough for most slow-moving stories.

When SAR is the only tool

Sentinel-1, by contrast, is what you reach for when optical lets you down. The classic scenarios:

• A vessel of interest is in a port subject to persistent cloud cover (North Korean ports in winter, parts of the Russian Far East, the West African coast during the harmattan dust season).
• The event of interest happened at night and you need confirmation by morning.
• You are tracking activity that hides in clouds by design — informal STS transfers in fog corridors, dredging during monsoons, port expansion under cover of weather.
• You need to detect change in water levels, infrastructure subsidence, or dam integrity, where the InSAR coherence change is the story itself.

A useful property of SAR for maritime work specifically: metal vessels reflect strongly against the dark water background. Even a small fishing boat shows up as a bright dot. With a backscatter threshold filter you can count ships in a harbour at night, in a storm, behind a typhoon. Try doing that with Sentinel-2.

The compare-tool reflex

A single image is rarely the story. The story is the difference between two images: before and after the strike, before and after the spill, before and after the new berth was dug.

Inside the Sentinel satellite view this is a one-click operation: pick the AOI, choose two acquisitions (which can be of different sensors — a Sentinel-2 baseline and a Sentinel-1 post-event SAR scene, for example), align them, and overlay them. The tool handles the projection and the geo-registration. The analyst keeps the cognitive budget for what matters: what changed, and what it means.

A worked example

Late last year we were asked to verify a tip about increased activity at a port on the eastern Black Sea — specifically, whether a series of vessels with mixed flags had begun calling at a terminal previously thought to be dormant. The local weather was, predictably, terrible. Optical was useless for two weeks.

We acquired four Sentinel-1 scenes across the period. Even with the basic VV polarisation, the vessel signatures were obvious — bright pixels against the radar-dark water. We counted, dated and cross-referenced each one against the terrestrial AIS feed for the same time window. Three vessels had been broadcasting normally. Two had not. Those two we tracked back via S-AIS, found a clean AIS-gap match, and from there the story almost told itself.

A pure optical workflow would have given us nothing for the duration of the cloud cover. A pure SAR workflow would have given us silhouettes without context. Combining the two — and stacking AIS, flag and ownership data on top — is the kind of work that turns a tip into a publishable piece in a single shift.

A final note: resolution honesty

Free Copernicus data is good. It is not all-seeing. Sentinel-2 is 10 metres; Sentinel-1 IW mode is around 5 by 20 metres. You will not read aircraft tail numbers and you will not identify individual people. The right time to bring in commercial high-resolution imagery — sub-metre Maxar, Airbus Pléiades Neo — is when the Copernicus image has told you exactly where and when to point a paid sensor. That sequencing keeps budgets sane.