Can satellites actually count people, or just buildings?

Satellites count proxies — rooftops, building footprints, lit pixels at night — not people directly. Machine-learning models then apply occupancy rates (persons-per-dwelling) calibrated against sample surveys or prior census data to produce gridded population estimates. The accuracy of that conversion step is the main source of uncertainty, and it varies significantly between urban cores, peri-urban sprawl, and rural dispersed settlements.

How does this differ from what Facebook/Meta's HRSL or WorldPop already offer for free?

Free global products like Meta's High Resolution Settlement Layer and WorldPop deliver 30–100 m gridded estimates that are adequate for regional planning. A sovereign programme adds three things those products cannot: sub-10 m resolution tailored to national cadastral boundaries, guaranteed update cadence tied to policy cycles rather than donor funding, and full control over the raw imagery and model weights — so the government is not reliant on a US or EU entity to re-run the model after a disaster or census cycle.

What orbit and sensor type should a nation choose for this application?

A LEO microsatellite constellation in a sun-synchronous orbit at 450–550 km altitude is the right default. Optical payloads at 0.5–1 m resolution handle dwelling-count tasks; a SAR payload on at least a subset of satellites ensures all-weather, day-night coverage. A 12–20 satellite constellation provides daily revisit over all major urban areas, sufficient for quarterly population-change products.

How quickly can a sovereign population map be produced after a sudden displacement event?

With a national constellation providing 24-hour revisit, a change-detection product flagging new settlement structures or abandoned dwellings can be produced within 48–72 hours of an event. That compares favourably with field-survey estimates from UNHCR or ICRC, which typically take 2–4 weeks to reach comparable spatial coverage. Speed at scale is the core operational advantage of the satellite approach.

What is the realistic accuracy of a satellite-derived population map compared to a traditional census?

Peer-reviewed studies report mean absolute percentage errors of 15–25% at the 1 km² grid-cell level when satellite methods are compared against census ground truth in low-income urban settings. Accuracy improves to under 10% when the model is trained on local survey data. Traditional censuses carry their own errors — undercounting of informal settlements routinely exceeds 20% — so satellite methods are often more accurate in practice, not just cheaper.

Does a government need to operate the satellites itself, or can it commission imagery from a commercial operator?

Commissioning imagery is faster to stand up but creates three durable risks: foreign-operator access controls, cessation of service during geopolitical friction, and no rights to the underlying imagery archive. Owning the constellation gives the government perpetual access to its own archive, the ability to retask within hours of a domestic decision, and the option to sell or share derived products with neighbours — a revenue and soft-power asset a service contract cannot replicate.

Which international standards govern how population map data should be structured and shared?

ISO 19115-1 (metadata), ISO 19157 (data quality), and OGC Web Feature Service / WCS standards govern data interoperability. UN Statistics Division Series F No. 103 provides the census-specific geospatial framework. Compliance with these standards ensures national population datasets can be ingested by UN agencies, development banks, and bilateral partners without format-conversion barriers.

How does nighttime light data complement daytime optical imagery for population mapping?

VIIRS and DMSP-OLS nighttime light composites are strong proxies for economic activity and electrification density, which correlate with population at regional scales. They add an independent signal when daytime imagery is cloud-affected, and they reveal peri-urban growth corridors that building footprints alone may undercount. However, nighttime lights saturate in dense city centres and miss unelectrified informal settlements, so they are best used as a covariate layer rather than the primary input.

9.7.1Live

High-Resolution Population Mapping

Census worker with tablet interviewing residents in a crowded informal urban settlement — Houses nestled beside mid-rise buildings reveal the urban density data that satellite population mapping transforms into smarter city planning.

Satellite Population Mapping: Space to City Plan

1
Constellation
Sub-metre optical imagery captures rooftop density
2
Downlink & Archive
Raw tiles ingested into cloud data lake
3
AI Disaggregation
ML models estimate population per 100 m grid
4
Urban Planners
Interactive maps guide infrastructure and services

Satellite imagery feeds AI models that disaggregate census counts into fine-grained population grids for city planners.

Page in the build queue

This application has its place in the atlas. The full page — short and long descriptions, what-matters bullets, sovereignty rationale, and the eight-field architecture block — is being written next.

Maturity reading: Operational systems exist with paying customers. Phase 1 built §4.1 (Maritime Intelligence) end-to-end as the depth-and-format reference. Section 4 is the next section to graduate from scaffold to full content; the rest follow on the roadmap.

Related applications

All in population intelligence

9.7.2 · Next

Daytime/Nighttime Population Shifts

Bring this into your sovereign programme.

All solutions Speak with engineering