What has changed

Rasmus Pedersen, Northing · corrections log of The State of European Renewable-Siting Data, 2026

Every correction to the index since publication, newest revision first. Each entry names the country, what changed, and why. A grade never moves on a claim; it moves on a re-probe of the live service, and every change is an error of mine or a rule written down before the re-derivation, never a source that changed since. Verified corrections from readers ship in the next revision, credited.

v1.2, 2026-09-21. This revision re-derives the two grid axes, grid geometry and connection capacity, for all 31 countries. The July grades on those axes were a desk audit: I read what each operator says it publishes and graded it. For v1.2 an evidence sheet was rebuilt per country from live probes of the actual sources, every request archived and reproduced from a second network wherever a wall was involved. Each sheet was then masked so that no July grade was visible, and a reviewer with no memory of the country derived the grade clause by clause from the evidence alone, re-run from scratch whenever the evidence or the rules changed. Two things follow. A grade now measures what is actually served and how: numeric or banded, per substation or per region, open or walled, nationally complete or partial. And a cell can come back unresolved when a deciding clause could not be settled; it prints a floor and a ceiling instead of a letter and counts as not clean. July had no such value, which is one reason the July matrix looked more finished than it was. Capacity: 13 cells up, 7 down, 10 unchanged, 1 unresolved. Geometry: 2 up, 4 down, 16 unchanged, 9 unresolved. Estonia, Malta and the Netherlands leave the Complete band.

Five rules were written down before the re-derivation and applied to every country. Connection capacity A means one nationally complete numeric tier, weighted by customers across that tier's operators with 80 percent as the line; A+ means an explicit firm or N-1 basis stated by the publisher or by the method it cites, and the July A+ for "both tiers at node level" is retired. Currency is printed as a mark (^s), not graded. A publisher's own disclaimers (indicative, non-binding, confirm with us) are printed caveats, not downgrades; the governing method or regulation controls. Three channel classes count with a caveat attached: a site walled to ordinary browsers on every path tested, an API read by replaying an application's own data requests, and figures published for a capacity tender. A vendor-hosted service whose own metadata does not name the publisher carries ^u. Each entry below states the country, what changed, and why.

Connection capacity

• Austria, capacity A → A+ ^uv. The calculation ordinance the operators cite states an N-1 basis; the operators' non-binding wording is a printed caveat. The national picture is a verified union of the operators' own channels, undocumented as an interface. Band unchanged (Fragmented).
• Belgium, capacity A → B. Only the Flemish distributor, a little over half of the country's customers, publishes per-installation numeric figures; the TSO's site challenge-walls a real browser on every path tested. Band unchanged (Fragmented, now 1 of 6).
• Bulgaria, capacity A → A+ ^u. The TSO's methodology page names the grid-code security criterion behind its figures: an explicit basis. Band unchanged.
• Switzerland, capacity C → B. The TSO publishes a per-substation capacity table for the transmission tier, complete and current, as a PDF whose unit is stated on the companion map rather than in the table. "No workable channel" was wrong; document-bound is the limit. Band Partial → Fragmented, with the geometry change below.
• Cyprus, capacity B^d → A ^su. The per-substation hosting-capacity service, published under the TSO's and the distributor's joint responsibility, is anonymous and numeric, so the desk mark comes off. No cadence is declared; its latest edition is eleven months old, 1.7 times the interval between the two releases I observed, hence the stale mark. The publisher calls the figures indicative, hence A rather than A+. Band Broad, now 5 of 6.
• Czechia, capacity A → B. Numeric per-substation figures exist for one distributor, about a seventh of customers; the largest publishes bands, and the third switched its numeric output off. Band Broad → Partial, with the geometry change below.
• Germany, capacity C^d → B. One TSO publishes substation records in megawatt bands, one a colour picture, one is bot-walled. No distributor figure was found, though four of the largest 110 kV distributors sit behind bot walls and could not be content-checked. Structured but not numeric is B, not C. Band unchanged.
• Denmark, capacity A+ → A. The TSO dataset carries the 132/150 kV tier only, so July's "both tiers at node level" was wrong, and that A+ is retired in any case. The distribution tier is numeric and current; no basis is stated. Band unchanged.
• Estonia, capacity A → B ^u. The distributor's map carries bands. There is no megawatt field behind them; I checked the service's full field list. Band Complete → Broad.
• Spain, capacity A → A+. The TSO's own file names its security criterion per node, an explicit statement inside the dataset. Band unchanged.
• Finland, capacity A → A ^u. The TSO's figures are indicative, with no methodology statement and no validity date; both are printed, neither lowers the grade.
• France, capacity A → A ^u. Graded on the transmission tier, where the figure is indicative and the cited method was not read for a firm statement. The distribution figure carries a flexible input in its schema.
• Hungary, capacity B → A. The TSO's substation list for the 2026 capacity tender carries an explicit MVA column on its medium-voltage cells for five of six distributors, 87 percent of customers; one distributor lists no medium-voltage cell. The figures are published for a tender round and say so in the caveat; the basis is deferred to the network studies, so A rather than A+. Band Fragmented → Partial, with the geometry change below.
• Ireland, capacity A → B. The distributor's per-substation workbook is anonymous and numeric, but the site terms restrict downloads to personal or non-commercial use. An express restriction on commercial use fails the open test. Band Broad, now 4 of 6.
• Iceland, capacity C^d → unresolved (ceiling B). The TSO's map API blocked every client I tried from two data-centre networks, a residential network was not tried, and the negative search behind the July C is not complete. Neither letter can be printed honestly. Band unchanged.
• Italy, capacity B^d → A. The TSO's download centre serves its full connection-request queue by stage, aggregated by municipality, updated monthly, retrieved from two networks by replaying the page's own data requests (printed as a caveat). No basis statement, so A. Band Fragmented → Partial, with the geometry change below.
• Lithuania, capacity A+ → A+ ^u. The TSO publishes with- and without-N-1 columns; the with-N-1 column is graded. The channel is undocumented, hence the mark.
• Luxembourg, capacity B^d → A+ ^u. The distributor's map field is labelled available firm capacity, numeric per substation, 93 percent of customers. The explicit firm statement gives A+. Band Partial → Broad.
• Latvia, capacity A → B ^u. The distributor's per-substation megawatt feed is anonymous and current, but the publisher's terms prohibit copying and commercial use. Band Broad, now 4 of 6.
• Malta, capacity A → B. The numeric feed covers the low-voltage tier only, which is not a criterion tier. Band Complete → Broad, with the geometry change below.
• Netherlands, capacity A → A ^v. The national capacity map is association-run with the operators' own values, which the union mark records. The definition it publishes reads as a basis statement without the words firm or N-1.
• Norway, capacity B → A. The TSO's capacity queue, per station and per case with an hourly refresh, is readable from its dashboard's own data requests (printed as a caveat). The graded figure is the ordinary-terms reservation column; the headline column adds special-terms reservations and is printed as a blend. Band Broad, now 5 of 6.
• Portugal, capacity A → A+. The regulator's access rules state a firm-access principle that governs the distributor's figure; the distributor's "approximation, consult us" is a caveat. Band unchanged.
• Romania, capacity A → A ^u. The channel is undocumented, hence the mark; the letter holds.
• Sweden, capacity A → A ^su. The TSO's N-1 statement describes a banded field, not the graded county totals, so A. The edition is four months old against a declared monthly cadence, hence the stale mark. Band Broad, now 4 of 6, with the geometry change below.
• Slovenia, capacity A → A+ ^u. The TSO states that its figures rest on N and N-1 criteria: an explicit basis. Band unchanged.
• Slovakia, capacity B → A ^uv. The three regional distributors, with mutually exclusive territories, each publish per-substation megawatt figures, two as files and one through its public report's own data requests (printed as a caveat). Together they cover every customer. Two call their figures non-binding, so A. Band Partial → Broad, with the geometry change below.
• Greece, Croatia, Poland, United Kingdom, capacity retained. A for Greece and the United Kingdom, B document-bound for Croatia and Poland.

Grid geometry

• Switzerland, geometry A → B. The national line layer is incomplete by the publisher's own statement: not every operator has delivered its data yet. Partial coverage.
• Malta, geometry A → B. The geometry services carry a non-commercial, no-derivatives licence, which bars commercial screening: access friction.
• Sweden, geometry A → B. One tier only.
• United Kingdom, geometry A → B. The distributors' line packages sit behind registration walls, and the one open layer sampled is a medium-voltage pocket, not the major-distribution tier. Access friction. Band Broad → Partial.
• Greece, geometry C → B ^u. The TSO's own map page carries its lines as a vector payload, one tier, undocumented as an interface; "no machine-readable geometry at any tier" was wrong.
• Hungary, geometry C → B. Substation points without lines.
• Romania, geometry B^d → B. Live-probed; the desk mark comes off.
• Six retained geometry B cells gain ^u: Bulgaria, Denmark, Spain, Ireland, Iceland, Lithuania. The graded channel is undocumented as an interface in each; no letter changes.
• Nine geometry cells unresolved: Czechia, Germany, Finland, Croatia, Italy, Luxembourg, Netherlands, Poland, Slovakia. In each the reviewer could not close one deciding clause, most often whether the search across distribution operators was complete or whether a national topographic line layer carries the tier and coverage it appears to. Croatia, the Netherlands and Slovakia print a floor of B, which the evidence supports, and a ceiling of A. Luxembourg's ceiling is B. The others print a ceiling of A with no floor the evidence can carry yet. Each lock names the one probe that would close it; those run next.

Marks and legend. Two superscripts join the legend: ^s for a stale edition and ^v for a verified union. ^u now also covers a vendor-hosted service whose own metadata does not name the publisher. The desk mark ^d leaves every grid cell; all of them are live-probed now. The axis table in finding 3 gains an unresolved column.

v1.1, 2026-09-07. After publication I re-probed every country's permit axis with a stricter method: pressing the export controls inside search tools, calling the endpoints behind result grids, opening the national registers' case entities and file catalogues, and reading each record schema before naming its unit. The July research had stopped at the search screen. Every grade change below is an error I made in July, not a source that changed since; the channel existed at the time. No grade moved down. Each entry states the country, what changed, and why.

• Denmark, permits B → A. The national building register publishes every building-permit case, with application, permit, start and occupancy dates, as a national file behind the same self-service key the parcel data already uses. July had probed two portals and stopped. Band unchanged (Broad, now 5 of 6).
• Hungary, permits B → A. Both public search tools are fed by endpoints that page through the whole national corpus without login, one with an export control; the documented 90-day deletion in the environmental register is not what the register does. Band unchanged.
• Bulgaria, permits B → A. The national environmental-assessment register returns its entire case corpus in one request once the page size is set; July had read the default ten-row page. Band Partial → Broad.
• Croatia, permits C → B. A national aggregate series of issued building permits exists at the statistics office, a permit-register module sits behind a captcha rather than an identity wall, and the economy ministry publishes a national register of renewable-energy projects, though without decision dates. "No workable public channel" was wrong.
• Luxembourg, permits B^d → B. Every channel behind the cell has now been live-probed; the desk-only mark comes off. The letter is unchanged.
• United Kingdom, permits B×3 C×1 → A. The government's renewable-energy planning database covers all four nations in one open file. It holds every renewable and storage planning application at or above 150 kW since 2021 (1 MW before), case-level, with a documented status field that runs from application through refusal to decommissioning, and day-precision dates. July had graded planning registers nation by nation and never opened it. The threshold is part of the grade: projects below it are not in the register, and for onshore wind in Scotland it holds a small fraction of the sites the national statistics count. Band Partial → Broad.
• Cyprus, permits B^d → A. The energy regulator, by law the sole licensing authority for generation, storage, supply and interconnection, publishes its licence, exemption and notification registries for every project above 20 kW as downloadable files, case-level, with dates and district; granted, pending and revoked sit in separate registries. Refusals are not recorded in any of them, which the grade discloses. July had read the building-permit statistics and the environmental pages and stopped. Band Partial → Broad.
• Germany, permits B×16 → A. The federal market master data register publishes, in its open bulk export, the permit that authorised every registered wind, solar, biomass, hydro and storage unit, with issuing authority, decision date and case-file number, across all 16 Länder and offshore. The population is permits that produced a registered unit: refusals and permits that never led to a unit are outside it by law, and a share of operating turbines have no permit record linked. July had graded building and immission-control permits Land by Land. Band unchanged (Fragmented, now 2 of 6).
• Austria, permits B×9 → B. The federal environment agency's public map of approved environmental-impact-assessment projects is one national channel covering all nine Länder, fully enumerable, with year-level decision dates. Its population is set by EIA thresholds, which by statute exclude solar entirely, so the letter stays B; the cell is now one national grade instead of nine regional ones. Band unchanged.
• Romania, permits B, reviewed. The energy regulator's licence and authorisation register turned out to be enumerable in full through an open endpoint and to carry refusals, but its records have no stable case identifier, locate the holder's registered office rather than the site, and cover only installations above 1 MW. B stands.
• Slovakia, permits A, reviewed. The site terms behind the national EIA register restrict its web content to personal use. The statute that obliges the register to publish also grants the right to redistribute what it publishes, and the terms do not name the register's data endpoint. A stands.
• Seven cells, two new marks. Two superscripts join the legend. ^e marks a cell where a case-level register exists behind a national identity login that a cross-border reader cannot use; it sits on Croatia. ^u marks a grade that rests on a channel with no published documentation or export control, reachable today because I read the search tool's own code, and gone the day the portal changes: Greece, Hungary, Portugal and Slovenia at A, Austria and Switzerland at B. No letter changes. Cause: Method addition, after the same re-probe showed how many working channels are undocumented.
• France, wording. Finding 1 called the 138,000-record environmental register a corpus of "permit cases". It is a register of classified installations, a quarter of them not regulated under that regime. The count is unchanged and reproduces. France's permit grade rests on the other half, a national permit-level dataset of building permits, now named.
• Norway and Sweden, wording. Finding 4 said neither aggregates building-permit decisions nationally "at all". Norway's national case index returns building-permit decisions municipality by municipality; Sweden's statistics office holds every granted permit and publishes totals, and a national gazette duty for permits ran until December 2025. Neither publishes an open case-level register, which is what the sentence now says.

← Back to the index