# Elektrības atslēgumu dinamika · Latvija (lv-outage-map)

> A time series of Latvian electricity outages, built by polling the Sadales
> tīkls public outage map every 15 minutes since 2026-08-22 and diffing each
> poll. Sadales tīkls publishes only what is out *right now* — no history, no
> restoration times, no archive — so this is the only record of how an outage
> spread and how restoration advanced. Interactive map with a time slider at
> https://lv-outage-map.vercel.app; the underlying data is public JSON.

## Par ko ir šī vietne (latviski)

Šī ir **elektrības atslēgumu karte Latvijā** — interaktīva elektrotīkla bojājumu
karte ar laika skalu. Tā rāda, kur Latvijā nav elektrības, kur ir reģistrēti
elektrotīkla bojājumi, kur notiek plānveida atslēgumi un kā virzās
elektroapgādes atjaunošana.

Atšķirība no Sadales tīkla oficiālās atslēgumu kartes
(https://karte.sadalestikls.lv/lv/atslegumi-elektrotikla) ir laiks. Sadales
tīkls publicē tikai pašreizējo brīdi: kas ir atslēgts *tagad*. Vēstures,
atjaunošanas laiku un arhīva tur nav. Šī vietne ik pēc 15 minūtēm nolasa to pašu
publisko avotu un salīdzina nolases, tāpēc te var atskaņot, **kā vētra
izplatījās un kā elektrību atjaunoja** — stundu pa stundai, novadu pa novadam.

Ko var uzzināt:

- **Cik pieslēgumu Latvijā šobrīd ir bez elektrības** un kuros novados.
- **Kā atslēgumu skaits mainījās laikā** konkrētas vētras laikā — maksimums,
  kad tas sasniegts, un cik ilgi vilkās atjaunošana.
- **Neplānotie atslēgumi** (elektrotīkla bojājumi), **plānveida atslēgumi** un
  **reģistrētie pieteikumi**, kas vēl nav apstiprināti kā bojājumi — katrs savā
  slānī.
- **Vētru epizožu arhīvs**: iesaldēts katras vētras pilns pieraksts.

Ko šī vietne **nedara**: tā nerāda konkrētu adresi vai konkrētu klientu, nesniedz
oficiālu atjaunošanas prognozi un nav Sadales tīkla vietne. Ja jāpiesaka bojājums
vai jānoskaidro sava īpašuma statuss, tas darāms pie Sadales tīkla
(bojājumu pieteikšana — 8404).

Biežāk lietotie jēdzieni šajos datos: *atslēgums* — pārtraukta elektroapgāde
pieslēguma punktā; *pieslēguma punkts* — viena pieslēguma vieta tīklā, nevis
cilvēks vai mājsaimniecība; *bojājums* (incidents) — viens elektrotīkla
notikums, kas var skart tūkstošiem pieslēguma punktu; *novads* — administratīvā
teritorija, pēc kuras punkti tiek grupēti.

## What this site is (English)

An **electricity outage map for Latvia** with a time slider: where power is out,
where grid faults are registered, where planned works are running, and how
restoration is progressing. The difference from the official Sadales tīkls
outage map is time — that map shows only the present moment, with no history and
no restoration times. This site polls the same public source every 15 minutes
and diffs the polls, so a storm can be replayed hour by hour and municipality by
municipality.

The site is a static page that renders its data client-side, so fetching the
HTML alone will not give you any numbers. Use the JSON endpoints below.

## Pages

- `https://lv-outage-map.vercel.app/` — the live map: the last 48 hours, or the
  whole of a storm still unfolding. Opens on the present moment.
- `https://lv-outage-map.vercel.app/e/<slug>/` — one frozen storm, opening on
  its first tick so the playback runs from the beginning. The slug is the
  episode's `slug` in `meta.json`; `/e/2026-08-22/` is storm Priska,
  2026-08-22 to 2026-08-29. Link a storm here rather than to the root, which
  will have moved on.

## Data

- [Current status](https://lv-outage-data.edge-static.workers.dev/meta.json):
  small JSON, rewritten every 15 minutes. `last_poll` is a Unix timestamp;
  `open` counts connection points currently out, keyed by feed —
  `A` = unplanned outages, `P` = planned works, `AP` = registered customer
  reports not yet confirmed as outages. `episodes` lists frozen storm archives.
- [Live window](https://lv-outage-data.edge-static.workers.dev/live/window.json):
  the full dataset the map plays back, a few MB (a gzipped copy is served at
  the same path with `.gz` appended, ~4x smaller). Normally the last 48 hours,
  but while a storm is still unfolding the window stretches back to the start
  of that episode and grows with it, so do not assume a fixed span — read `t0`
  and `steps`. Columnar: `pts` holds one array per field, all of equal length,
  indexed in parallel. `x`/`y` are WGS84 lon/lat, `r`
  indexes `regions`, `cl` is the number of connections behind the point,
  `b0`/`b1` are 15-minute bucket indices for first-seen and restored, `oi`
  indexes `oids`, `di` indexes `days`, and `nb` is a back-reference: how many
  rows earlier this same connection point first appears, 0 if this is that
  first row. `t0` plus `step` × bucket gives the wall-clock time of any bucket.

## Reading it correctly

- **Timestamps come from two different clocks.** `b0`/`b1`/`bu` are *ours* —
  derived from poll times, accurate to about 15 minutes, and never earlier than
  the first poll. `di` (the outage date Sadales tīkls itself attributes) and the
  report registration times in `ap.rt` come from the feed and can be much older
  than anything this archive observed.
- **No restoration timestamps exist anywhere.** A point counts as restored on
  the first poll in which it no longer appears. Do not present these as exact.
- **`oid` is not a unique row key.** It identifies an incident, and one incident
  can cover thousands of connection points. The per-point identity is the pair
  (outage id, connection number).
- **Nor is a row a connection point.** Sadales tīkls re-issues a point under a
  fresh outage id when it fails again, and sometimes lists it under two
  incidents at once, so there are always more rows than points — a 58-hour
  window during the August 2026 storm held ~38k rows over ~29k points. Fold
  rows together through `nb` before totalling anything cumulative, or the same
  point gets counted twice.
- **Counts are connection points, not people or households.** A storm typically
  shows ~12k points across ~400 incidents.
- The archive begins at its first poll, 2026-08-22T20:59:54Z. Nothing before
  that can be reconstructed, by anyone, including us.

## Attribution

Underlying outage data is published by [Sadales tīkls](https://karte.sadalestikls.lv/lv/atslegumi-elektrotikla),
Latvia's distribution system operator. The time series, restoration inference
and regional attribution are this project's. If you quote a figure, please
credit both and link https://lv-outage-map.vercel.app so readers can see the
playback the number came from.