Governments and humanitarian coordinators cannot plan water, food, health or shelter delivery without knowing how many people they are serving. Ground-based census methods in large, rapidly expanding camps are slow, politically contested and frequently manipulated — populations are over-reported to attract aid or under-reported to avoid scrutiny. Satellite imagery cuts through that fog: structure counts derived from sub-metre optical or SAR data, combined with validated occupancy coefficients, produce population estimates accurate to within 10–15 % within days of a tasking request.
The satellite stack combines very-high-resolution optical imagery for shelter delineation and SAR for cloud-penetrating revisits during rainy-season crises when optical tasking fails. Machine-learning pipelines segment individual shelter footprints automatically, flag new construction since the last pass and generate change-detection alerts when population surges occur. A sovereign national system adds one critical advantage: the host government can task the constellation on its own schedule rather than queuing behind commercial customers or waiting for a foreign donor to authorise a re-tasking.
The operational outcome is a rolling, auditable population baseline that drives logistics allocation, donor reporting and protection planning. When a camp expands overnight after a new displacement event, the national emergency management authority receives an updated estimate within 24 hours rather than weeks. That speed is the difference between a well-supplied response and a mortality crisis.
Frequently asked
Why can't humanitarian agencies just use census data or registration records to count camp populations?
Registration data is almost always incomplete: newly arrived populations avoid registration due to fear, distance or benefit calculus, and deregistration of departures is even patchier. UNHCR's own assessments consistently show registration undercounts of 15–40% in active emergency settings. Satellite shelter counts provide an independent, unbiased cross-check that does not depend on anyone presenting themselves to an authority.
How does a satellite actually estimate population — does it count people directly?
In most operational deployments the satellite counts proxy structures: rooftops, tents, shelters or dwelling footprints identified by computer-vision models trained on labelled imagery. That count is then multiplied by a locally calibrated mean household size (typically 4.5–6.5 persons per shelter) derived from periodic ground surveys. Direct detection of moving persons in imagery is possible at 0.3 m GSD but is rarely used at scale due to privacy concerns and processing cost.
Is SAR or optical imagery better for this application?
Optical VHR is generally preferred because shelter materials (plastic sheeting, canvas, corrugated iron) are visually distinct and model training datasets are abundant. SAR is essential when cloud cover is persistent or when night-time monitoring is required, but backscatter signatures from informal shelters are ambiguous and occupancy inference is harder. Best practice is a fused pipeline: SAR for change detection and gap-filling, optical for detailed structure classification.
What level of accuracy is good enough to trigger aid allocation decisions?
The humanitarian coordination system (cluster approach, OCHA) typically uses population figures as order-of-magnitude planning inputs rather than precise counts; errors within ±15% are generally acceptable for supply-chain planning. However, for per-capita funding formulae (e.g. World Food Programme ration calculations) tighter accuracy is demanded and ground validation becomes mandatory alongside the satellite estimate.
Why should a sovereign nation invest in its own imaging satellite rather than buying data from Planet or Maxar?
Foreign commercial providers can restrict data exports, reprice contracts, or simply not task your region during competing crises. A nation hosting a large refugee population has a direct sovereign interest in continuous, unmediated monitoring: for security, public-health planning, and international burden-sharing negotiations. Owning a microsatellite with a VHR camera also builds domestic remote-sensing capacity applicable across agriculture, disaster response, and border security — the refugee use case is often the catalytic justification for a broader national EO programme.
How frequently does imagery need to be refreshed to be operationally useful?
For stable, protracted settlements a monthly update is generally sufficient for supply planning. During acute influx phases (first 30–90 days of an emergency) daily or even sub-daily revisit is desirable to track population growth rates. A small national constellation of 4–6 microsatellites in LEO can achieve 12–24 h revisit over a fixed region of interest, which covers both modes.
Does this application require ground infrastructure the host nation must operate?
Yes: a ground receiving station or cloud-downlink arrangement, data processing pipelines running computer-vision inference, and secure data-sharing interfaces with UNHCR, IOM and national authorities are all needed. The processing burden is significant — GPU-accelerated inference is now standard — but open-source toolchains (ESA SNAP, NASA Earthdata, QGIS) reduce cost for nations with limited budgets.
Are there legal restrictions on collecting imagery over refugee camps?
Commercial imagery licensing (governed by national remote-sensing laws and ITU-R frequency assignments) generally permits collection, but several host governments restrict foreign satellite tasking over sensitive domestic sites. UNHCR's Data Protection Guidelines (2015, updated 2021) govern how population data derived from imagery may be stored and shared. Nations with a domestic imaging programme must enact their own national space law or remote-sensing regulation — often modelled on UN-OOSA principles — to clarify permissible use.