How to measure informal settlement expansion from satellite imagery

Informal settlements rarely wait for a permit. A cluster of rooftops shows up on a slope that was zoned as a buffer, a strip along a drainage channel fills in over two dry seasons, and by the time a field team drives out to confirm it, the settlement already has a water line. Planning offices that still rely on annual flyovers or five-year census updates find out about this kind of expansion long after it's finished growing. Satellite-based measurement closes that gap, but only if you set it up to answer the actual question: how much built-up area appeared outside the line this year, and where.

Start with the boundary, not the image

The growth boundary is the fixed reference. Before you load a single scene, you need the boundary as a vector layer, the same one the ministry or municipality uses for permitting. Every subsequent measurement gets checked against it. Skip this step and you end up with a change map that shows expansion everywhere, inside the plan and outside it, with no way to separate the two without going back and doing the overlay by hand.

What medium-resolution imagery actually shows

For settlement tracking, imagery in the 3 to 10 meter range is the right tool for the job. At that resolution you can reliably pick up a new cluster of rooftops, a road grid extending into open land, or a densifying edge along an existing settlement. You will not resolve individual structures or count households from it. Multispectral bands matter more here than raw resolution: built-up surfaces, bare soil cleared for construction, and vegetation each have a distinct spectral signature, which is what lets a classifier separate "this pixel turned into rooftop and road" from "this pixel is just a dry season" without someone eyeballing two photos side by side.

Annual imagery is the right cadence for this use case. Settlement expansion in most cities moves over months and seasons, not weeks, and a once-a-year pass gives you a clean built-up footprint for this year against last year's, timed to whatever planning calendar your office runs on.

Turning two footprints into a number

Once you have this year's built-up layer and last year's, the measurement itself is a change detection problem: classify built-up area in both years, difference the two layers, then clip that difference against the growth boundary. The area that falls outside the line is the number that matters: hectares of expansion the plan didn't call for. That hectare figure is the thing you hand to a planning director or a bank's monitoring unit, because it's the same unit permit ledgers and zoning reports already use.

This is also where manual methods tend to break down. The comparison itself isn't conceptually hard. Doing it by hand means pulling old aerial photos, cross-referencing a stack of permit records, and redrawing the boundary overlay every single year. It works for one neighborhood. It does not hold up across a whole metro region, and it rarely gets repeated consistently enough to catch a slow creep before it's a few hundred hectares.

Informal settlement growth tracking as an annual habit

The value of this kind of monitoring compounds. A single year's expansion number tells you where to send a field team. Three or four years of the same measurement, run the same way each time, tells you which boundary segments are actually holding and which ones get breached every cycle regardless of enforcement. That's the pattern a development bank wants before it signs off on the next tranche of a planning-linked loan, and it's the pattern a ministry wants before it revises a growth boundary instead of just redrawing it reactively.

If your office is still stitching together old aerial photos and permit paperwork to answer this question, Urban Growth Monitor runs that comparison every year and hands back the built-up change map and the hectare figure directly. Get in touch to see what this year's boundary breach looks like for your city.