Current evidence
Starting-state check pending
Waiting for a current observation and same-time model guidance.
Source status and forecast implications
Source completeness and the detailed forecast are still loading.
AVIATION FORECAST DESK
Build the atmospheric picture.
Current conditions, forecast evolution, and terminal guidance
Connecting to public weather data…
Model, valid time & display settings
Chart time does not change TAF validity. Shortcuts: Alt+1–4 changes workspace; arrow keys change chart time; / focuses ICAO.
Building the five-day transition and hazard overview.
MET watch Decision support
Waiting for a station.
AI forecast Experimental
Waiting for a station.
Hourly flight-category probabilities appear here for the six pilot airports.
TERMINAL FORECAST · 30 HOURS · UTC
AI TAF vs official TAF Experimental
Waiting for a station.
The AI TAF and its TAFVER record appear here for 28 airports: the six original pilot airports, KFAY and the 21-base pilot.
TERMINAL FORECAST · 30 HOURS · UTC
Terminal forecast worksheet.
Validity and review requirements
Enter a whole-hour UTC start as YYYY-MM-DD HH:00. TAF code and validity remain UTC regardless of the page display preference. Editing the start does not run the generator; check the calculated end, then use the button to generate a new draft. The evidence clock controls charts and profiles.
Copy and editor actions stay unavailable when the starting observation is missing, stale, or materially conflicts with same-time model guidance. The generated text remains visible for diagnosis.
Gathering raw guidance for this ICAO…
The 30-hour model-guidance draft will appear when source data loads.
Why these forecast groups were selected
Waiting for forecast decisions.
Forecast pilot
KFBG is the primary pilot. These tools measure review effort and monitor a saved forecast; forecast accuracy remains unproven.
No forecast selected for monitoring.
Start the timer when you begin reviewing.
What does SKY/CIG unresolved mean?
SKY is sky condition; CIG is ceiling. The generator has no usable cloud-layer or ceiling guidance for that hour. Nearby sounding coverage, the forecast horizon, or a missing cloud fraction may be responsible. This does not mean clear skies, and NSW describes significant weather rather than cloud coverage. Inspect element provenance and review the sky forecast before use.
Element provenance and unresolved fields
DAYS 1–5 · 12-HOUR PERIODS
Five-day forecast evolution.
Model-derived ranges. Review flagged fields and timing before accepting the draft.
Waiting for the detailed five-day guidance.
Five-day source provenance
Source cycles and retrieval status will appear with the forecast.
Generation method and limits
Extended outlook · days 6–12
EXTENDED OUTLOOK BOUNDARY · DAYS 6–12
Daily model envelopes
No days 6–12 outlook is available in this preview. Do not extrapolate the detailed day-5 blocks into week two. A future outlook belongs here only when ensemble scenarios, broad timing windows, probabilities, and confidence decay are available.
CURRENT CONDITIONS
Observed at the airfield.
No observation loaded.
Last 12 hours · station reports
MODEL INITIALIZATION
Initial-condition comparison.
Differences below are diagnostics. The product gate independently evaluates field completeness, current observations, moisture, and material conflicts; a highlighted difference is not a forecast accuracy ranking.
FIVE-DAY TRANSITIONS
Forecast transitions.
Regional and upstream context.
Click the map to inspect a grid point. Zoom in, then click and drag to move around.
Regional data is prepared in the background. First retrieval can take several minutes.
Waiting for forecast times
Surface
200 mb · upper atmosphere
Selected chart time
High-definition rendering · state boundaries · H and L mark closed, sampled height or pressure extrema. Regional fields use 3° sampling. Local GFS detail uses 0.25° fields; inspect the selected map detail and model horizon.
The vertical picture.
Inspect pressure-level values
FORECASTER-OWNED PRODUCT
Build the detailed days 1–5 draft.
Start with the sampled day 1–5 guidance, then add accepted timing, alternatives, uncertainty, and local impacts. The result remains a working draft.
Previous unclassified station draft retained
Select the intended editor product, then explicitly restore this text as a manual draft.
Draft formatting checks do not establish meteorological accuracy or military product compliance.
Verification, learning & source details
Learning from verified forecasts.
Inspect learning trials →Checking the learning collector…
Only prospectively validated updates change future guidance.
Latest learning decisions
Continuous verification is the AI TAF panel's TAFVER record against the official TAF. The Shadow forecast experiment and the published-TAF benchmark were retired on 14 Oct 2026; their archives are kept.
MEASURE THE OUTCOME
Prospective station verification.
Forecasts retrieved before their valid time are archived here. Skill appears only after later observations arrive.
This edition reports mean absolute error, signed bias, and root mean square error for matched temperature, wind, and uncensored visibility. The continuous-learning collector separately tests numeric ceiling corrections, seasonal and regime features, source-model weights, and IFR probabilities. Event onset and clearance verification remain development work; no gain is claimed before prospective validation.
Source availability and capability limits.
Load a station to inspect observations and model fields.
Integrated systems and manual-only official sources
| System | Workstation support | Coverage and use |
|---|---|---|
| METAR/SPECI | Integrated | Reports are age-stamped. No report is treated as unknown truth; automatic transfer blocks without a current observation. |
| Station forecasts | Integrated | Deterministic model guidance with per-field source and missing-data states. Model agreement is not a calibrated probability. |
| Vertical profiles | Integrated where available | Nearby native profiles or a labeled station-grid profile. Distance, model, valid-time, and horizon limits remain visible. |
| Regional maps | Integrated where domain is supported | Sampled regional fields, local 0.25° GFS detail, and polar stereographic charts. Available model horizons are explicit. |
| NWS point alerts | Integrated for U.S. locations | An empty alert list does not rule out aviation hazards. International stations need their national meteorological service. |
| Radar, satellite, lightning | CONUS official imagery on request | Official CONUS radar, GOES GeoColor and GLM imagery can be loaded with publication/retrieval age. International imagery remains a national-service review item. |
| PIREPs and aviation advisories | Integrated on request | Age-stamped AWC PIREPs, SIGMETs, G-AIRMETs, Alaska AIRMETs and CWAs are available on request. Coverage and unavailable feeds remain explicit. |
| Ensemble probabilities and verification | Not calibrated | No probabilistic forecast or verified ceiling, event-onset, or event-clearance skill claim is supported. |
| Issuance | Forecaster review required | Automatic transfer is a completeness gate for a working draft, not an issuance or operational approval. |
AWC observations, PIREPs and aviation advisories
Official NWS radar
NOAA satellite and GOES GLM lightning imagery
NWS warnings and forecasts
Official imagery has not been loaded.
Aviation hazard feeds have not been checked.
Official alerts have not been queried.
Local resource usage
Disposable caches are bounded; saved handoffs are preserved.
DD FORM 175-1 · FLIGHT WEATHER BRIEFING · DRAFT
Build a flight weather briefing.
Enter the route and times (UTC). The workstation fills the DD 175-1 from current observations, TAFs, GFS/HRRR guidance along the route and official advisories, for you to review, edit and sign. Blocks only the briefer can fill stay blank, and missing data is never treated as clear. Nothing you enter is saved.
REVIEW BEFORE DOWNLOAD