Terminal Weather Services SystemDegraded
UTC
A wet airport runway at first light, with approach lights receding into low cloud.

Private technical preview / aerodrome data observability

One observation.
Three paths.
Every divergence visible.

Detect missing, corrected or divergent METAR representations and source-path lag before discrepancies disappear into downstream processing.

Observability traceLEMD / METAR
Nominal observation151230Z
  1. 01AWCreceiptTimeexact TAC
  2. 02MET NorwayLast-Modifiedexact TAC
  3. 03NWS mirrorfile clockexact TAC
Normalized station eventCONTENT + TIME + PROVENANCE{ observation.changed }
Retained network snapshotCollected 11m ago · 12 Sep 15:23:24Z

01 / THE PROBLEM

A METAR can be the same observation and still arrive through different paths with different timing, wrappers, remarks or revision state.

Live source comparison / public network

Inspect a current source comparison

Aerodromes12
Aligned11
Content difference0
Source lag1
Snapshot11m ago
Adolfo Suárez Madrid–Barajas Airport / Spain Aligned
01

Aviation Weather Center

Aligned
Observation time
12 Sep 15:00Z
AWC receiptTime (report-specific)
12 Sep 15:04:25Z
METAR LEMD 121500Z 21006KT 170V270 CAVOK 32/04 Q1020 NOSIG
SHA-256 TAC fingerprintd04876625abe102580cd9d8e4ceb54c13f4b9625516db713a309adb0a5718f42
Open source ↗
02

MET Norway

Aligned
Observation time
12 Sep 15:00Z
HTTP Last-Modified (response-wide)
12 Sep 15:23:12Z
LEMD 121500Z 21006KT 170V270 CAVOK 32/04 Q1020 NOSIG=
SHA-256 TAC fingerprint8f5ff142982b5b0dae83802817f3cb3bf8668151bb288fd7b1e662dff999767c
Open source ↗
03

NWS station mirror

Aligned
Observation time
12 Sep 15:00Z
HTTP Last-Modified (station file)
12 Sep 15:04:30Z
LEMD 121500Z 21006KT 170V270 CAVOK 32/04 Q1020 NOSIG
SHA-256 TAC fingerprint6040add3c65cece7e79d8e3987158974cb815186ae8eb00106a087b840672a7b
Open source ↗

Observation time comes from the report and identifies the weather observation.

Source time is labelled according to what that endpoint actually supplies.

Terminal Weather fetch records when this comparison snapshot was collected.

All monitored aerodromes

European network board

AerodromeAWCMET NorwayNWS mirrorResultOpen
EFHKHelsinki Airport
12/15:20Z12/14:50Z12/15:20ZSource lag
EPWAWarsaw Chopin Airport
12/15:00Z12/15:00Z12/15:00ZAligned
LFPBParis–Le Bourget Airport
12/15:00Z12/15:00Z12/15:00ZAligned
EHAMAmsterdam Airport Schiphol
12/14:55Z12/14:55Z12/14:55ZAligned
EGLCLondon City Airport
12/14:50Z12/14:50Z12/14:50ZAligned
LEMDAdolfo Suárez Madrid–Barajas Airport
12/15:00Z12/15:00Z12/15:00ZAligned
LTACAnkara Esenboğa Airport
12/14:50Z12/14:50Z12/14:50ZAligned
LIMCMilan Malpensa Airport
12/14:50Z12/14:50Z12/14:50ZAligned
ESSAStockholm Arlanda Airport
12/14:50Z12/14:50Z12/14:50ZAligned
EDDFFrankfurt Airport
12/14:50Z12/14:50Z12/14:50ZAligned
LOWWVienna International Airport
12/14:50Z12/14:50Z12/14:50ZAligned
EIDWDublin Airport
12/15:00Z12/15:00Z12/15:00ZAligned
Product preview

From evidence to service

One evidence model.
Two operational surfaces.

Evidence in

Preserved source record

LEMD 151230Z · TAC · timestamps · SHA-256
Terminal Weather event model

Align · validate · classify

  • Station state
  • Revision chain
  • Path timing

Case files

Where observability changes the answer

An unbranded passenger aircraft taxiing through heavy rain on a reflective apron. Current snapshot / dynamic

CASE 01

LIVE / SOURCE LAG

EFHK: latest observation key differs by path

At collection time, at least one endpoint still exposed an older nominal observation than the newest available path.

AWC 12/15:20Z / MET Norway 12/14:50Z / NWS mirror 12/15:20Z

Useful forFeed migration checks, parser verification and source-path incident triage.

CASE 02HISTORICAL ARCHIVE CASE · 15 JUL 2025 / 14:20Z

One observation key. Three successive states.

An archived EGCC bulletin sequence contains a NIL report, an automatic report, and a corrected conventional report for the same nominal observation. A latest-only table loses that chain.

  1. NILMETAR EGCC 151420Z NIL=
  2. AUTOMETAR EGCC 151420Z AUTO 21011KT …
  3. CORMETAR COR EGCC 151420Z 21011KT …

Useful forRevision tracking, immutable archives and downstream replay.

CASE 03RECENT AWC SAMPLE

Timing has a shape, not just an average.

1 minMinimum
2 minMedian
6 min90th pct.
6 minMaximum

The prototype keeps observation and source-receipt timestamps separate, making station-specific timing patterns visible without relabelling them as end-to-end latency.

Station
EFHK
Median interval
2 min
Sample
12 reports

Useful forStation baselines, threshold design and investigating a change in delivery behaviour.

The method

From endpoint response to evidence

  1. 01

    Collect

    Fetch each public path centrally and cache the result.

  2. 02

    Preserve

    Keep the exact TAC representation extracted from each endpoint response.

  3. 03

    Fingerprint

    Hash every extracted TAC representation so identical bytes remain identifiable.

  4. 04

    Align

    Group messages by station and nominal observation time.

  5. 05

    Compare

    Separate wrapper changes, content differences and source lag.

An aerodrome weather sensor beside a runway during a passing rain shower.
OBSERVATIONDISTRIBUTIONDELIVERY

Practical applications

Built for the gaps between systems

The weather is only half the operational question. The other half is whether the expected report arrived intact, on time for that path, and in the form downstream software expects.

Built for

  • Aviation weather data providers
  • Airline and airport technology teams
  • Flight-planning and dispatch platforms
  • Data engineering and integration teams
  • Researchers and feed-migration teams

Not a consumer weather application, pilot briefing service or official meteorological source.

  1. 01

    Feed migration verification

    Run old and new feeds side by side before changing a downstream system. Surface missing markers, stripped remarks and timestamp differences.

  2. 02

    Distribution incident triage

    See whether a station is absent everywhere or only late on one route, reducing guesswork when a feed looks stale.

  3. 03

    Correction and revision tracking

    Retain COR, AUTO, SPECI and NIL states instead of silently replacing the previous message.

  4. 04

    Parser regression testing

    Replay exact source representations against a new decoder and detect fields lost during normalisation.

  5. 05

    Station reporting baselines

    Measure completeness and source-receipt patterns by aerodrome, schedule and distribution path.

Preview coverage

Twelve aerodromes.
One comparable surface.

The initial set spans northern, central, western and southern Europe, combining major hubs and distinct reporting environments. Stockholm Arlanda is the initial Swedish reference station.

  • ESSAStockholm ArlandaSweden
  • EFHKHelsinkiFinland
  • EPWAWarsaw ChopinPoland
  • EHAMAmsterdam SchipholNetherlands
  • LFPBParis–Le BourgetFrance
  • EGLCLondon CityUnited Kingdom
  • LEMDMadrid–BarajasSpain
  • LTACAnkara EsenboğaTürkiye
  • LIMCMilan MalpensaItaly
  • EDDFFrankfurtGermany
  • LOWWViennaAustria
  • EIDWDublinIreland

Source clocks

Know what each clock means.

01

Aviation Weather Center Data API

JSON METAR records with nominal observation time and AWC receiptTime. Polled centrally no more than once per cache interval.

API documentation ↗
02

MET Norway Tafmetar

Global text product for recent METARs. Its response-wide HTTP Last-Modified value and the Terminal Weather fetch-completion time are retained; it does not supply a per-report receipt timestamp.

Product documentation ↗
03

NWS Telecommunication Gateway

Latest station-coded report and HTTP file modification time. NWS describes the international observations as received from worldwide sources.

Gateway documentation ↗

Private technical preview

Evaluate the evidence model with your workflow.

We are inviting a small number of technical pilot conversations around feed verification, station coverage and downstream integration. No production or safety-critical use is offered.

Request pilot access