The Data Behind the Charts: GCAT

Analysis of space programs lives or dies on the quality of its numbers. Launch counts, orbital populations, debris totals, national shares. These figures circulate widely, get quoted across press releases and policy papers, and are almost never traceable back to a definition. How many satellites are “in orbit” depends entirely on what counts as a satellite and what counts as orbit. Two defensible answers can differ by a factor of two.

Space Policies takes a single position on this problem: every chart published here is built from one source, that source is named, and the definitional choices behind each figure are written down before the query is written. The source is GCAT, the General Catalog of Artificial Space Objects, compiled by astrophysicist Jonathan McDowell and published at planet4589.org/space/gcat .

Why GCAT Is Not Just Another Satellite Database

Most space datasets are byproducts. Tracking catalogs exist to support collision avoidance and record orbital elements; industry databases exist to support market forecasts and are priced accordingly; agency listings cover their own programs. GCAT is something rarer: a catalog built to be a historical record, by one person, held to a standard of documented provenance rather than operational convenience. Eight properties make it irreplaceable for policy work.

It is one continuous record, from 1942 to today. The launch table opens with the V-2 and runs to launches from the past few days. There is no break in methodology at the end of the Cold War, no change of custodian, no segment where coverage silently thins because a contract lapsed. Long-run trend claims, the ones policy arguments actually rest on, can be made against a single consistent series.

It counts what other catalogs discard. Suborbital flights, missile tests, deep-space missions, launches that failed before reaching orbit, objects that never orbited at all: all of them are records with a category code, not omissions. Orbital launches are under ten percent of the launch table. That residue is precisely where questions about military test cadence, emerging launch programs, and failure rates live: questions that a catalog of successful orbital launches cannot answer at all.

It is object-level, not aggregate. GCAT does not publish “the number of Chinese satellites.” It publishes every object, with its owner, operator, mass, orbit, application, status and disposal, and lets the aggregate be derived, which means the aggregate can be defined, re-derived under a different definition, and checked. Nothing in a headline figure is taken on trust.

Uncertainty is a field, not a footnote. GCAT dates carry their own precision: a record may be known to the second, to the day, to the month, or to the year, and the catalog says which. Uncertain values are flagged as uncertain rather than rounded into false confidence. Very few datasets in any field encode the limits of their own knowledge this explicitly, and it is what allows a chart to stop at the boundary of what is actually known.

Intent is recorded separately from outcome. On orbital launches, GCAT encodes the planned target orbit, the planned disposal mode of each upper stage that reached orbit, and the intended count of launch-vehicle debris, distinct from what subsequently happened. This distinction is the raw material of debris-mitigation compliance analysis. Whether an operator intended to deorbit a stage and whether the stage came down are different questions, and policy cares about both.

It is a referential system, not a table. Organizations, launch sites, launch vehicles, vehicle families, objects and payloads share one vocabulary of codes and reference each other. An operator resolves to a parent organization, to a state, and to a class: non-profit, commercial, civil government, defense. A payload resolves to its launch, its site, its vehicle, its application, and its UN registration status. That referential integrity is what makes cross-dimensional questions tractable: launch-service model by customer type, payload application by owner class, orbital regime by launching state, all from one catalog and without reconciling three sources that disagree.

It is honest about its own boundary. GCAT covers tracked objects. Statistically modeled small debris, the sub-centimeter population that dominates every fragment count in the popular press, is explicitly out of scope. Masses that upstream does not know stay empty rather than being imputed. A dataset that tells you where it stops is worth more than one that does not.

It is open and citable. GCAT is released under CC-BY. Anyone can reproduce a figure published here, from the same catalog, and disagree with it on the record. Behind a paywall, that check is not available to the reader, and an argument the reader cannot check is not analysis, it is assertion.

What GCAT Does Not Cover, and What It Offers Instead

GCAT is a catalog of physical activity in space. It carries no figures in currency (budgets, revenues, investment flows, cost per kilogram), no law, no system performance, and no personnel data beyond the organizational level. Where an analysis here needs those, it draws on other sources and says so; what it never does is quietly extend GCAT beyond its scope, the most common way a well-sourced chart becomes a wrong one.

It would be a mistake, though, to conclude that the catalog says nothing about economics. It says a great deal, in physical units rather than monetary ones. On orbital launches it records who operated the vehicle and who paid; it classifies every organization as commercial, civil government, defense or non-profit; it classifies every payload by application. From that follow the government/commercial split of launch demand over time, shifts in the application mix, supplier concentration, and mass to orbit by state and by operator class. As indicators of direction these are often more reliable than the money figures they stand in for: a reported revenue depends on consolidation scope, exchange rates and accounting standards, a public budget mixes line items that are not comparable across countries, and both are partly shielded by commercial confidentiality or classification. A launch either happened or it did not, and its customer has been classified by one consistent rule since 1957.

The limit deserves the same bluntness. These proxies measure volume and composition, not value: a commercial rideshare and an institutional launch each count as one while moving different orders of magnitude of money, and the classification follows customer type rather than funding source: a “commercial” launch may rest on a public anchor contract. Prices, margins and subsidy content cannot be derived from here, and this site does not try.

From Catalog to Local Mirror

The published catalog is a set of tab-separated files. Space Policies works against a local relational mirror of them, rebuilt from upstream on demand, because a chart must be reproducible: the same query, run in six months, has to return the same numbers or explain why it does not.

The rebuild is deliberately defensive. Downloads are cached against upstream change stamps, so a refresh transfers only what actually moved. Every table’s column layout is checked against a recorded manifest, because upstream does occasionally restructure a catalog, and a silent column shift is the failure mode most likely to produce a plausible, wrong chart. Loading is atomic, the previous database is retained for rollback, and each build runs sanity checks against known ground truths before the result is used for anything.

On top of the raw catalog the mirror adds a decode layer: lookup tables that translate GCAT’s compact single-letter vocabularies (object types, status codes, payload applications, organization classes, orbital regimes, launch-service models, disposal modes) into readable labels transcribed from GCAT’s own documentation, never invented. Views join objects to their owners, launches to their operators, payloads to their applications. Dates are parsed into typed columns that preserve the original precision. None of this changes what the catalog says; it makes what the catalog says queryable without the documentation open in another window.

Why the Charts Are Interactive

Charts on this site are rendered with Apache ECharts , and the choice is editorial rather than technical. A policy chart usually carries more than one legitimate reading: absolute counts and shares tell different stories about the same series, and which one matters depends on the argument the reader is testing, not the one the author happened to make.

So the reader gets controls, not just an image. Series can be switched off to isolate a comparison. Where both readings are meaningful, a view selector moves between counts and percentages. A time range can be zoomed. The underlying data table can be opened and read directly, and the chart can be saved as an image. The intent is that a reader who disagrees with the framing can re-frame the chart in place rather than take the author’s crop of the data on trust.

One rendering engine drives every chart type: lines, stacked areas, bars, scatter and bubble plots, heatmaps, flow diagrams. Behavior, colors and controls stay identical from one article to the next, and a reader learns the interface once. Each chart carries all of its display text for both site languages in a single payload, which keeps the English and Italian versions of a figure structurally identical rather than separately maintained. Where a figure is better served as a fixed image, for dense print layouts or PDF reports, the same bundle is rendered as a static vector graphic instead, from the same data and the same generator.

Reading the Charts Honestly

A few conventions apply throughout, and are worth stating once.

The current year is partial. Any series that runs to the present ends on an incomplete year. Where a comparison would be distorted by that, the chart either stops at the last complete year or labels the partial one.

“Tracked” is not “all”. Debris figures count cataloged objects. The modeled small-debris population is larger by orders of magnitude and is not what these charts show.

Unknown is not zero. Where upstream does not record a value (the mass of most debris fragments, for instance), it stays unknown and is excluded from totals rather than counted as nothing.

Every chart is a snapshot. The catalog is republished continuously. Figures carry the date of the data behind them, and that date is what the chart is accountable to.