Archives and dissemination
A national imagery archive is a strategic asset, not a hard drive. Getting the storage arithmetic, access law and dissemination standards right before first light determines whether the data remains sovereign and usable long after the constellation is retired.
The arithmetic arrives before the first image does
A modest constellation of four optical satellites, each carrying a pushbroom imager producing 1-metre-resolution imagery at a 30-kilometre swath, can generate somewhere between 2 and 6 terabytes of raw scene data per satellite per day depending on tasking density and onboard compression. Four satellites at conservative duty cycles produce roughly 4 to 8 TB of downlinked data daily. Over a five-year operational life, that is 7 to 15 petabytes of raw imagery before derived products, metadata, calibration records or change-detection stacks are counted. Synthetic aperture radar constellations, which operate day and night regardless of cloud, can exceed those figures because operators often task them more aggressively and the complex (I/Q) data format is inherently bulkier than optical rasters.
The point is not that the numbers are large. It is that they are knowable before procurement begins. Storage architecture, ingest bandwidth from ground stations, and tiered retention policy must all be sized against a credible data-volume model, not a round-number estimate. A government that signs a constellation contract without a parallel storage-and-retention specification will find itself making expensive ad-hoc decisions under operational pressure, usually in the wrong direction.
Tiering is not a cost-saving trick; it is a legal instrument
Modern archive design uses at least three tiers. Hot storage, typically NVMe or high-throughput SAN, holds recent imagery and products under active exploitation. Warm storage, usually object-store systems on spinning disk or high-density flash, holds the past one to three years of scenes that analysts retrieve occasionally. Cold storage, tape or deep-archive object tiers, holds the historical record that may not be touched for years but must be recoverable within hours when a boundary dispute, a disaster-response inquiry or a parliamentary audit demands it.
The tier boundaries are not purely technical. They encode retention law. Many national archives legislation frameworks require government records, including geospatial records, to be preserved for periods of 20 to 30 years or indefinitely for certain categories. Imagery of a disputed border, a coastline affected by sea-level rise, or a disaster site is evidence. Deleting it on a cost-driven schedule without legal authority is a governance failure, not an engineering decision. The archive specification should name the applicable national records law and map each data category to a retention class before a single byte is ingested.
Access control across agencies is harder than access control within one
A national programme typically serves multiple ministries: defence, agriculture, environment, disaster management, urban planning. Each has different classification requirements, different legal authority to share data externally, and different technical capacity to consume it. A single flat access model fails all of them. Defence imagery of a border region may carry handling caveats that preclude sharing with a municipal planning department; agricultural multispectral data from the same satellite pass may be entirely unclassified and should flow freely to extension services.
The standard approach is attribute-based access control (ABAC) layered over a metadata catalogue that tags every scene with classification level, geographic extent, sensor type, acquisition date and permitted-use category. Agencies query the catalogue and receive only the scenes their credentials authorise. The catalogue itself becomes a sensitive asset: knowing what a government has imaged, and when, is intelligence even if the images are withheld. Catalogue access therefore needs its own tiered permission model.
Dissemination standards matter here. OGC Web Map Service (WMS) and Web Coverage Service (WCS) interfaces allow agency GIS platforms to consume imagery without copying it, which simplifies access auditing. STAC (SpatioTemporal Asset Catalog) has become the de facto standard for machine-readable imagery metadata and is supported by most modern exploitation tools. Mandating STAC-compliance at ingest, rather than retrofitting it later, saves significant integration effort.
The honest limits: what archives cannot fix
An archive preserves what the constellation collected. It cannot retroactively fix collection gaps. If cloud cover prevented optical imaging of a flood-affected province during the critical 72-hour window, no amount of archival sophistication recovers that scene. This is a sensor and tasking problem, not a storage problem, but it is worth stating plainly because governments sometimes treat archive investment as a substitute for adequate constellation capacity.
Long-term readability is a genuine risk. Proprietary imagery formats, vendor-specific compression codecs and bespoke metadata schemas can render archived data unreadable within a decade if the originating vendor ceases to support the format or the software licence lapses. The mitigation is to mandate open or standardised formats at ingest: GeoTIFF with cloud-optimised tiling (COG) for optical rasters, HDF5 or NetCDF for SAR complex data, STAC-JSON for metadata. This adds modest processing overhead at ingest but avoids format-obsolescence risk over a 30-year retention horizon.
Disaster recovery deserves explicit architecture, not an assumption. A single in-country data centre that suffers a power fault, a flood or a physical security incident can destroy years of irreplaceable imagery. Geographic replication to a second site, whether in-country or under a bilateral agreement with a trusted partner state, is the minimum credible posture. Replication lag and recovery-time objectives should be specified contractually, not left to the operations team to negotiate after an incident.
Why the archive outlives the satellites, and why that matters
A satellite has a design life. An archive, if governed correctly, does not. The Landsat programme, operated by USGS and NASA, made its full historical archive freely available in 2008; researchers immediately used decades of imagery to reconstruct land-cover change, glacial retreat and urban expansion that had never been systematically analysed. The scientific and policy value released by that decision was substantially larger than the value extracted during the original operational period. The archive had outlived the satellites that built it and was still generating returns.
A national government establishing a sovereign constellation today is making a decision whose archival consequences will extend well past the current administration, the current satellite bus and the current ground-segment vendor. Sovereignty over the archive means the data remains available for future analysis methods, future legal proceedings and future policy questions that cannot be anticipated now. Source-access terms for archive software, hardware audit rights over storage infrastructure, and clear contractual provisions for data migration at end-of-vendor-relationship are therefore not procurement details. They are the conditions under which the archive remains genuinely sovereign rather than nominally so.
Engineering parameters
| Typical raw ingest rate (4-sat optical constellation) | 4 to 8 TB per day at moderate tasking; up to 20 TB/day for aggressive SAR constellations |
| 5-year raw archive volume (4-sat optical) | 7 to 15 PB before derived products; plan for 2 to 3× with products and metadata |
| Hot-tier latency (scene retrieval) | Seconds to low minutes; NVMe or high-throughput SAN typical |
| Cold-tier recovery time (tape or deep-archive object) | Hours to 24 hours; must be specified contractually against legal-hold requirements |
| Metadata catalogue standard | STAC 1.0 (SpatioTemporal Asset Catalog); OGC WMS/WCS for dissemination interfaces |
| Recommended raster format for long-term retention | Cloud-Optimised GeoTIFF (COG) for optical; HDF5 or NetCDF for SAR complex data |
| Minimum replication posture | Two geographically separated sites; recovery-time objective typically 4 to 24 hours |
| Access control model | Attribute-based access control (ABAC) over STAC catalogue; per-scene classification tagging |
| Retention period (legal minimum, varies by jurisdiction) | 20 to 30 years for government records; indefinite for designated evidence categories |
One contract, one accountable engineer
Commissioned as one programme, not a stack of contracts: spacecraft, launch, ground segment, mission control, training and handover are priced together. Source-access terms and audit rights are agreed in writing before signature. Review archive specification with our engineers.