Somalia is not an unknown country. It is an unmeasured one. Decisions about cropland, water, coastline, settlement and infrastructure are routinely made on estimates, sampling and last year’s figures — while the same ground has been imaged from orbit every few days for the better part of two decades. Almost nothing has been converting that into answers here.
This is a practical account of what is genuinely answerable from satellite data in a Somali context, what is not, and what separates a useful Earth observation project from an expensive one.
The distinction that decides everything
There is a difference between buying imagery and buying an answer.
Imagery is a file. Someone still has to correct it, compare it against a baseline, decide what the difference means, and turn that into a number a department can act on. That work is most of the cost and all of the value, and it is the part that quietly does not happen when a programme procures “satellite data” without asking who will process it.
The useful conversation therefore does not start with an area. It starts with a question — a specific one, with a decision attached. “Imagery of the Shabelle” is a purchase. “How much cropland was under water in week 32, by district, and how does that compare with the same week in the last five years” is a question, and it determines the sensor, the revisit rate, the method and the price, in that order.
What is genuinely answerable
Cropped area and crop condition
How much land went under cultivation at the start of a season is measurable by district, and comparable year on year against the same imagery source. Through the season, vegetation indices track crop vigour — a stressed field diverges from a healthy one in the data well before the divergence is visible from a road. That lead time is what turns reporting into early warning.
Flood extent
This is the strongest case in the country. Floods arrive with weather, and weather means cloud, which defeats optical imagery precisely when you need it. Radar sees through cloud. Inundation mapping from radar can establish which settlements, roads and cropland are actually affected within days of the water arriving, rather than weeks later from field reports that are necessarily partial.
Settlement and urban growth
Somalia’s cities have expanded enormously in two decades without a cadastre keeping pace. Built-up extent, building footprints, and change detection across a multi-year archive answer questions a planning department cannot answer from its own files — including when a particular area was developed, which matters in disputes.
Vessels
Transponder data shows the fleet that consents to be seen. Imagery and radar detect hulls regardless of what is transmitting. Neither is sufficient alone: imagery tells you something is there, transponders tell you who. Cross-referenced, they identify the gap — vessels present and not broadcasting — which is generally the entire point of the exercise.
Slow environmental change
Shoreline retreat, mangrove loss, channel migration, rangeland degradation. None of it is visible on any single day and all of it is obvious across a decade. The value here is the archive: comparable imagery going back years means a trend can be established rather than asserted.
What is not answerable, or not yet
Being straight about this is more useful than a longer capability list.
Satellites do not identify individuals or read documents; that is a film convention. They do not see through roofs. Optical imagery does not see at night or through heavy cloud — radar does, at lower resolution and higher cost. Revisit rate is a hard constraint: an event between two passes is not observed, and “daily” imagery of a specific point is considerably more expensive than most first conversations assume. And a detection is a detection, not an identification: knowing a vessel is present is not the same as knowing whose it is.
Where a question needs ground truth — soil samples, household surveys, inspection — satellite data narrows where to look, it does not replace going there.
Why the cost picture has changed
Two things shifted. Public programmes now provide large volumes of free, regularly refreshed imagery at resolutions adequate for many of the questions above — which means a serious agricultural or flood monitoring programme no longer starts with a large imagery bill. And commercial high-resolution tasking, while still expensive, has become affordable for targeted work where the decision justifies it.
The practical consequence is that cost has moved from acquisition to analysis. That is a better place for it to sit, because analysis capability can be built locally and kept, whereas an imagery licence expires.
The argument for doing it in country
Bringing imagery down and processing it requires a ground segment, compute, storage and people. All of those can sit anywhere. Where they sit determines who holds the resulting data, who can be asked questions about it, and whether the capability survives the end of a programme.
For land records, maritime enforcement and anything feeding a national statistic, the case for keeping the data and the analysis inside the country is not sentimental — it is the same residency argument that applies to any other sensitive dataset, with the added point that a capability built here stays here.
Where to start
Pick one question with a decision attached and a deadline. Establish what the answer would change. Then ask what can be measured, at what accuracy, how often, and what it costs — in that order. Programmes that begin with an area and a budget tend to produce a great deal of imagery and very few decisions.
