Earth Observation and maritime intelligence in Africa

Africa has roughly 30,000 km of coastline, and for many coastal states the exclusive economic zone is comparable to or larger than the land territory. These are areas that have to be managed: fisheries, shipping, conservation, and increasingly offshore energy and infrastructure.

Monitoring them by vessel and aircraft is expensive, and coverage scales with the size of the fleet. This is not a question of institutional will. A patrol vessel observes a few square kilometres at a time and cannot be everywhere.

What changed

Two things changed the arithmetic.

The first is open data. The European Union’s Copernicus programme makes Sentinel data freely available under an open licence. Radar and multispectral imagery covering African waters, acquired on a regular schedule, can be obtained at no cost by anyone. The marginal cost of one more observation is effectively zero.

The second is that processing has moved to where the data already is. Downloading a time series of radar scenes over a national EEZ is a serious bandwidth problem. Running the analysis next to the archive and retrieving only results is not.

Together these remove the two barriers that historically made satellite-based monitoring the preserve of well-resourced agencies.

What is actually scarce

Not data. Analysis capacity.

Turning a radar scene into a decision requires calibration, geometric correction, detection, association with other data sources, and enough context to know what a detection means. Each of those steps is a specialist task. The gap between freely available imagery and a usable operational picture is a technical one, and it is where the work is.

This reframes the opportunity. The question is not how African institutions get access to satellite data. They already have it. The question is who builds the downstream layer that makes it usable, and whether that layer is built by people who understand the local operating context.

Why context is a technical input

A detection model tuned on North Sea shipping will underperform on the Kenyan coast, and not because of anything cultural. The vessel mix is different, with a large proportion of small wooden and fibreglass craft that return radar energy weakly. The sea states are different. The seasonal cloud regime is different. The regulatory thresholds that determine which vessels report their positions are different.

Every one of those is a parameter that affects model performance. Building around local conditions is not a preference, it is what makes the output correct.

What good looks like

A maritime intelligence capability worth having is not one that produces verdicts. It is one that:

• observes independently of what vessels choose to report
• works in the conditions that actually occur, including cloud and darkness
• states what it saw, when, and with what confidence
• makes it obvious which observations deserve a human’s limited attention first
• connects to the systems the responsible institutions already use

The last point is the one most often skipped. A monitoring product that requires an agency to abandon its existing workflow will not be adopted, however good the analysis.

The honest position

Earth Observation does not solve maritime management. It observes, at a scale and frequency no patrol fleet can match, and it observes independently of cooperation.

What it makes possible is better allocation of attention. Given that patrol capacity will always be limited, directing it toward the observations that most warrant a look is a meaningful improvement, and a realistic one to claim.