How this atlas is built

01How to read a record

Every record declares whether it is confirmed, and so do its sensitive fields, one by one. A developer can be confirmed while its commissioning date is not, and the record says so field by field. The four states are Confirmado (confirmed, solid mark), Inferido (inferred, near-solid with a dashed ring), Parcialmente verificado (partially verified, faint fill and dashed ring) and No verificado (not verified, hollow and dashed):

Some claims are not made by any single document. The name of a cable may come from joining the concession for its landing with the project of another cable that describes that same landing. The atlas marks these claims as inferred and writes the chain into the record itself: each document, the literal sentence it contributes, what joins them and which finding would break the link. Validation looks for each sentence in its archived document in every edition. Whether the link is sound is something no machine can check, and a person signs it.

Every claim cites the source that supports it, classified by what that source can support: primaria (official documents), prensa (press), corporativa (corporate) and hueco, the missing source, declared:

Only a primary source supports a confirmed fact. A corporate announcement or a news story is recorded with its origin, because it has value, but it does not rise to fact. Whenever a source is cited it is archived: web addresses stop working over the years and the archived copy remains.

Intelligence analysts grade what they read with the NATO two-axis code: a letter for the reliability of the source, from A to F, and a number for the credibility of the information, from 1 to 6. The atlas does not write that code into its data. It derives it from what the data already declares: each source type carries its letter and each state its number, in the two tables above, and the code of a field is the letter of the source it cites and the number of its state. A field confirmed by a primary source is an A2, and one inferred from two primary sources is an A3: its documents are of the best class, but none of them makes the whole claim. The record shows it on the mark of each field and in the list of sources.

It has two limits. No field carries a 1, which means confirmed by independent sources, because the atlas does not check that two sources are independent. An inferred field does not carry it either: its documents do not say the same thing, each one contributes a part. And the letter grades the class of document, not the track record of whoever issues it: an official register that publishes what it is told carries the same A as the decision that grants an authorisation, and where that difference matters, the record says so in its text.

Each layer also declares what it promises about its own completeness. In a closed census layer the atlas asserts that the list is complete: submarine cable landings with an administrative act, spent nuclear fuel stores, planned electricity interconnectors. A missing one is a failure of the atlas. In an open census layer the atlas records what an administrative act names and locates, and does not claim it is everything: data centres, for which no public register exists, or desalination plants. There, a gap measures what remains to be found. The dashboard states which is which.

02Gaps are not a defect

A gap is a missing source, declared, with its own place in the record. It is the central mechanism of the atlas. Filling a gap with something plausible is the worst failure this project can commit, because an invented figure that looks reasonable is the one nobody catches.

Three of the gaps published right now, as they read in their records:

• The nuclear power plants. The closure calendar agreed in 2019 has no public document, because it is a private protocol. In five of the seven reactors the date is empty and declared, instead of being filled with the one everybody quotes.
• The sixteen mineral domains. None has primary-source cartography: they are freehand sketches, and the whole layer is a declared gap. They are drawn translucent with dashed borders so they cannot be read as measured boundaries.
• Tropic Seamount. The atlas records where it is and what it is called, citing the hydrographic authority that names it. It does not publish the tellurium and cobalt grades that circulate: they come from scientific campaigns this atlas has not archived.

Reporting a gap, a figure taken on faith without a primary source, is worth as much as contributing a new record.

Every year the atlas measures which gaps were opened and which were closed, in which edition, and how long the others have been open; and, on the same page, what each watcher brought, what was ruled out and what arrived twice: the collection, measured (in Spanish).

03The photo and the film

Each record is a photo: its state as of the date of its bulletin. Since contract 1.35, some layers also carry their film, the full time series from the same primary source, drawn on the record page itself. The first one is reservoir storage: every reservoir with its weekly bulletins since 1988.

Three rules govern it, all three enforced on every change. There is only film where there is a negative: a series requires its complete history in one archived primary source, nothing is reconstructed from sources that no longer exist, and different sources are never spliced together. Gaps stay gaps: a week without a bulletin is a week without a point, and the line breaks instead of bridging it, because interpolation is invention. And the photo and the film may not contradict each other: the last point of the series must match what the record states, or validation blocks the whole release.

Water also has a page of its own (in Spanish): the week's bulletin, the thirty-eight-year curve and the breakdown by river basin district. There reservoirs are added up and divided to get a fill level. Deriving is allowed; writing it down is not. None of it enters the data: it is computed when the page is built, and the page says so. Gas has one too, monthly and under the same rule, and the library gathers both along with every earlier bulletin.

The chart draws no percentages and no trends: it draws the bulletins. What you see moving is the source.

04What is published today

The viewer does not read the live branch: it serves a tagged release. What you see below is what release 2026.11.182 serves, read from its own manifest. Every layer can be downloaded as published, as GeoJSON under CC BY 4.0.

All of it is machine-readable. The catalogue lives at /datos/manifest.json and each layer at /datos/capas/<id>.geojson, served static, without a key and open to any origin. Which routes are promised, and which are internal organs of the viewer that will change without notice, is stated in the data contract. The promised routes are never served pre-compressed. It is a written guarantee, and it is what lets you read them with plain curl, from a script without libraries, or from a desktop GIS.

An edition can be pinned, not just cited. Every release of the public archive carries its data bundle attached: the layers, the series, the datasets and the manifest of that edition, exactly as published. All of them have one, not just the latest, so an analysis made a year ago can be rebuilt with the data of the time, not today's. The release list gathers them.

Greyed branches are declared and empty. A layer that is going to exist appears before it has a single record, so that silence is never mistaken for absence.

05How to correct a record

1. Open an issue in the repository naming the record by its identifier, for instance minerales-proyectos:aguablanca, and the source that supports the correction. A correction without a source cannot be applied. It is the only hard rule.
2. There is no pull request route for the data. The layers are served already built, so what is needed from you is the source, not the file.
3. Nothing is deleted. A record that stops being valid changes state, from current to historical or withdrawn, and the change is recorded, with its evidence, in the changelog of the release that touches it.

Corrections are proposed in the atlas's public archive: the archived sources, the data contract and the changelog, open to anyone, no account needed. The viewer and the pipeline are developed in a private workshop. What makes a datum verifiable is the citation, and the citation lives whole in the archive.

06What this atlas cannot guarantee

These are the real limits of everything above, which is why they are here and not in the small print.

• A citation can be dead and still answer that it is alive. A weekly check verifies that sources are still standing, but EUR-Lex and the IGME return “200 OK” while serving an error page. When the URL promises a PDF the deception is caught; when it does not, there is no cheap way to tell, and the check reports how many citations remain genuinely unverified. This is the reason the documentary archive exists.
• An official source can contradict itself. Two mining rights in the Mining Cadastre hold two different legal statuses today depending on which export you download, same day, same registry. When that happens the atlas publishes the disagreement and downgrades the field to partially verified. It does not pick a winner.
• An official download can arrive silently mutilated. The CSV export of that same cadastre truncates vertex lists at 424 characters, and a polygon missing a corner still closes and looks correct. All geometry comes from the shapefile for that reason.
• A record has the precision it declares, not the precision it appears to have. A point drawn on good cartography asserts exactness even when the record says otherwise, and on a map what you see wins. That is why geographic precision is a mandatory field with its own source, and “municipal centroid” is a legitimate result, stated as such.

07How to cite this atlas

The data is CC BY 4.0: free to reuse, commercially included, crediting the source. To cite the whole atlas:

Vázquez Paumard, A. D. — Atlas Estratégico de España (Strategic Atlas of Spain), release 2026.11.182 · 10.5281/zenodo.21918595 · atlas.eltercioviejo.com · data CC BY 4.0

That DOI is the concept one: it always resolves to the latest deposited edition, and each release also has its own. It is worth knowing what it answers for, because a DOI promises permanence. The Zenodo deposit archives the citation apparatus, that is, the archived sources, provenance, contract and changelog. It does not archive the data layers, which are served from here and from the explorer.

To cite a single record, every record page carries a ready-made citation in the «Citar este registro» disclosure at the foot of the dossier; in the viewer, with a copy button. It also has its own permanent link, which the corner button copies. Its identifier is stable forever: layer:record, for instance nuclear:almaraz-i. Every downloaded layer carries its release and version inside.

Every fact in the atlas cites, in turn, the source that supports it, with the attribution formula each emitter demands.

08Embedding the map

The map can be embedded anywhere with an iframe. Every record's file gives the code for that record, next to its citation (in Spanish), and every layer's page gives the code for the whole layer. Copy it and paste it:

<iframe src="https://atlas.eltercioviejo.com/incrustar.html?capas=nuclear&sel=nuclear:almaraz-i&ed=2026.11.182" width="100%" height="420" style="border:1px solid #DAD7CC" title="Almaraz I — Atlas Estratégico de España" loading="lazy"></iframe>

The address can also be written by hand, and it is one of those the data contract promises: it does not change shape without a major version. capas is the layers drawn, comma-separated; v, the framing, as zoom, latitude and longitude; sel, one record, which is marked, framed and named with a link to its file; ed, the edition the code was copied with; and base, the background map.

Along the bottom of the map runs a band that does not fold away, with the attribution of every layer and of the base map, which each emitter's licence requires, and the data edition. The map always shows the live edition: if it is no longer the one in the code, the band says which edition it was embedded with and links to what has changed since. Every record is clickable and opens the full atlas in a new tab, with its file and sources (if the iframe carries the sandbox attribute, it needs allow-popups). The badge links back to the same view in the atlas. You can also download the current view as an image from the viewer itself, with the button under the zoom controls: a PNG with the attribution inside the image.

09On paper, or as PDF

The record sheets, the water and gas reports, the datasets, the dashboard, the library and the project dossier are made to be printed. There is no button: print with Ctrl+P (or ⌘P on a Mac) and choose «Save as PDF». Each of those pages says so at the foot.

It is not the dump a browser gives by default. The sheet inverts the ink, because the site is dark and browsers do not print backgrounds; navigation and buttons are dropped, because on paper they are nothing; and what exists only there is added: the address of the page and the release its figures came from. A printed sheet has no address bar, and paper that does not say where it came from is what this atlas does not publish.

That is why the PDF you get here stays citable: it says which edition it was and where to look again. A datum that changes tomorrow does not turn the paper into a lie. It leaves it dated.

10Licences and cadence

The data is CC BY 4.0: free reuse, commercial included, crediting the source. That choice forbids incorporating datasets with viral licences, and forces rebuilding from primary sources whatever cannot be used. The code is neither published nor licensed: it is personal work and all rights are reserved. What the public repository serves is the documentary archive, that is, the contract, the changelog, the provenance and the archived sources. It is what you need to verify a datum, not to rebuild the viewer. The data does not depend on that decision: it remains CC BY 4.0, and that is the half you can audit.

Every release has its entry in the changelog: what changed, why, with what evidence, and what remains unknown, which is a mandatory section. A release that does not declare what it ignores is claiming to know everything. Every release is also announced via RSS, for whoever prefers the changes to come to them.

The atlas is the work of one person: it is curated and maintained by Arturo David Vázquez Paumard. The signature does not turn the record into opinion. Every claim still hangs from its source, not from whoever writes it down.

11What this atlas does not collect

A collection policy is written down so that it can be checked, which is why it is here. The atlas collects assets and administrative acts, not people or operations.

• Nothing that is not public. Every record comes from a document anyone can consult: an official gazette, an administrative register, an inventory published by its own body. There are no restricted sources, leaks or privileged access, and none are accepted.
• No private individuals. The holder of a right or the developer of a project is recorded when it is a company or a public body. When an act names a private person, as in expropriation annexes or in applications for private installations, that name does not enter. The document is archived and the asset is recorded. From the Official Gazette of the Commercial Registry the atlas reads companies and their structural changes, never directors or officers, nor a sole shareholder who is a private person: it says so without naming them. From public procurement it reads the contracting body, the subject and the type of each contract, never the body's contact people or the winning bidders, who are sometimes private persons. From environmental impact assessment it reads the title, type, location and status of each project, never the developer, whose entry sometimes names its representative.
• No operational security detail. The atlas locates an infrastructure and states which act authorises it. It does not collect protection plans, security arrangements, access points, shifts or anything describing how it is protected. Everything it publishes is in the public domain, and that detail still stays out by decision.
• Aerial and satellite imagery, bounded. No imagery of defence, space tracking, nuclear or radioactive waste facilities; of the rest, only a window of up to one kilometre around a record that an act locates exactly, with the image left unannotated.
• No estimates of its own. A figure the atlas cannot cite is not published as data, even when it could be computed with good judgement. What derives from the data itself, such as a fill percentage, is computed when the page is built and labelled as derived.

If a reader finds anything in the atlas that contradicts this list, it is corrected like any other record: with an issue in the public archive.

12How to read an image comparison

A decision states that a data centre is authorised, and does not state whether anything has been built. To get closer to that question, some record pages carry an image comparison (cotejo de imagen): two dated official images of the same window, side by side, and what a person saw when comparing them. Both belong to the same family: two orthophotos from the National Aerial Orthophotography Plan (PNOA), by the National Geographic Institute, or two scenes from the Sentinel-2 satellite, by the Copernicus programme. The window is a square of up to one kilometre a side, centred on the coordinate that an act publishes for that record.

The reading says one of four things. That in the later image a shape appears (Aparece) that was not there before, or that one disappears (Desaparece). In both cases it names the shape: a roof, an area of levelled ground, scattered objects, a linear trace or a regular patch. That there is no change at this scale (Sin cambio a esta escala). Or that the pair is not conclusive (No concluyente), for instance because there is cloud in the window. A short text says where the shape lies within the window, with no figures and no months.

Every reading is tied to its scale, which is the coarsest pixel of the two images: one metre or less in an orthophoto and ten metres in Sentinel-2. A shape is named only if that scale can tell it apart, which is why no roof is named at ten metres. Below a threshold of five pixels, fifty metres in Sentinel-2 and five or less in an orthophoto, what is not seen is not denied. A no-change comparison says that no new shape larger than that threshold can be seen, never that there is nothing. The record page gives the threshold in metres next to the signature of each reading. Shapes that the season can mimic, such as a trace or bare ground, are only compared between dates at the same time of year.

It is always a pair. A single image does not show what has changed, and an empty window does not prove that something does not exist. Both images, the documents that date them and the record of how they were requested are archived with their hashes, so anyone can check that nobody has touched them. The service may serve a different image at the same address today, and the one in the comparison is the archived copy.

The sources of a comparison carry no letter of the two-axis code, and the reading is labelled as a signed reading. The letter grades what a document can support, and a comparison supports nothing in the record: it does not change its phase, closes no gap, does not enter its chronology and counts towards no indicator. A person signs it because no automatic measure separates a real change from an apparent one. In September 2026, the mean difference between two scenes was of the same order where there was a new shape and where there was none.

A comparison never says what is seen, whose it is, whether it is the installation in the file, whether it is finished or working, or how big it is. A mistaken reading is not rewritten: it is withdrawn with its reason and a new one is published.

13Who is behind an asset

Every record transcribes whom the act names: a developer, a holder, an operator, in the exact words of the document. Who that company or body is lives elsewhere, in the register of entities (in Spanish), with one page per legal person. It carries its Spanish tax number or its Legal Entity Identifier, its country, the exact texts with which the acts write it, and its relations with other entities, each with its source and its status. The text of the record never changes. The link between the two is computed when the site is built by matching exact texts, never by resemblance. The record page shows it in the «Quién está detrás» block.

Who owns the asset, and what share, is sometimes stated by another document: the landing licence at the other end of a cable, a commercial registry entry, consolidated accounts. That fact is written in the owning entity, as a relation to the record, and not in the record. It carries a percentage only if the document gives one, states what the percentage is of in the words of the document, and how much the document distributes in total. The record page shows it in the «Propiedad documentada» block: each owner, its share, the country of its ultimate parent, and the sum against what the document distributes. What is missing is computed when the page is built; it is not written down.

The holder of the file and the owner of the asset are two distinct facts, and the atlas does not derive one from the other. They may coincide or not. At the landing of the Marea cable in Sopela, the holder of the landing concession owns a quarter of the segment in Spanish waters and of the landing station, according to the notice of the United States communications authority that distributes the system. The record keeps saying what its act says, and the entity, what its own document says.

A relation is confirmed only if a primary source supports it. When a company declares that it does not report its parent, that is published as a gap: not reporting one is not the same as not having one. And no private individual enters the register: only companies and public bodies.