Model runs — Aus Alpine Weather

Refresh

Syncing fetches the chosen products deliberately and atomically publishes a local last-known-good artifact. Tick the products, run the command, then reload local dev — page navigation itself never contacts the provider. Run npm run build separately only when you intend to deploy the synced state.

npm run models:sync -- --all

Selected deterministic models share one batched all-resort request, then publish as independent products. A failed or incomplete sync retains the previous product.

Model products, their latest published run, and the run in the local snapshot
Refresh	Product	Provider's latest	Run in snapshot	Retrieved	Status — snapshot, then checked
	ECMWF IFS 0.25° Deterministic · ecmwf_ifs025	not published	Tue 29 2200hrs AEST · 1200Z 13 h ago	10 min ago	Unknown
Provider metadata was not checked, so the synced run has not yet been compared with the latest one. Used by /forecast, /d0-3, /resort/*, /operational-1500m
	ECMWF AIFS 0.25° Deterministic · ecmwf_aifs025_single	not published	Wed 30 0400hrs AEST · 1800Z 7 h ago	10 min ago	Unknown
Provider metadata was not checked, so the synced run has not yet been compared with the latest one. Used by /forecast, /d0-3, /resort/*
	GFS 0.13° Deterministic · ncep_gfs013	not published	Wed 30 0400hrs AEST · 1800Z 7 h ago	10 min ago	Unknown
Provider metadata was not checked, so the synced run has not yet been compared with the latest one. Used by /forecast, /d0-3, /resort/*
	ICON global Deterministic · dwd_icon	not published	Wed 30 0400hrs AEST · 1800Z 7 h ago	10 min ago	Unknown
Provider metadata was not checked, so the synced run has not yet been compared with the latest one. Used by /forecast, /d0-3, /resort/*
	UKMO Global 10 km Deterministic · ukmo_global_deterministic_10km	not published	Tue 29 2200hrs AEST · 1200Z 13 h ago	10 min ago	Unknown
Provider metadata was not checked, so the synced run has not yet been compared with the latest one. Used by /forecast, /d0-3, /resort/* A comparison track only — it is not one of the four models the primary precipitation split is built from.
	JMA GSM Deterministic · jma_gsm	not published	Tue 29 2200hrs AEST · 1200Z 13 h ago	10 min ago	Unknown
Provider metadata was not checked, so the synced run has not yet been compared with the latest one. Used by /forecast, /d0-3, /resort/* A comparison track only — it is not one of the four models the primary precipitation split is built from.
	ARPEGE World 0.25° Deterministic · meteofrance_arpege_world025	not published	Wed 30 0400hrs AEST · 1800Z 7 h ago	10 min ago	Unknown
Provider metadata was not checked, so the synced run has not yet been compared with the latest one. Used by /forecast, /d0-3, /resort/* A comparison track only — it is not one of the four models the primary precipitation split is built from.
	GDPS 15 km Deterministic · cmc_gem_gdps_15km	not published	Tue 29 2200hrs AEST · 1200Z 13 h ago	10 min ago	Unknown
Provider metadata was not checked, so the synced run has not yet been compared with the latest one. Used by /forecast, /d0-3, /resort/* A comparison track only — it is not one of the four models the primary precipitation split is built from.
	ECMWF IFS ~9 km surface Surface reference · ecmwf_ifs	not published	Wed 30 0400hrs AEST · 1800Z 7 h ago	10 min ago	Unknown
Provider metadata was not checked, so the synced run has not yet been compared with the latest one. Used by /forecast Surface fields only — no pressure profile, so it never supplies a snow level.
	ECMWF IFS ENS Ensemble · ecmwf_ifs025_ensemble	not published	Tue 29 2200hrs AEST · 1200Z 13 h ago	10 min ago	Unknown
Provider metadata was not checked, so the synced run has not yet been compared with the latest one. Used by /d7-plus
	ECMWF AIFS ENS Ensemble · ecmwf_aifs025_ensemble	not published	Tue 29 2200hrs AEST · 1200Z 13 h ago	10 min ago	Unknown
Provider metadata was not checked, so the synced run has not yet been compared with the latest one. Used by /d7-plus
	GEFS 0.25° Ensemble · ncep_gefs025	not published	Wed 30 0400hrs AEST · 1800Z 7 h ago	10 min ago	Unknown
Provider metadata was not checked, so the synced run has not yet been compared with the latest one. Used by /d7-plus
	GEPS 0.5° Ensemble · cmc_gem_geps	not published	Tue 29 2200hrs AEST · 1200Z 13 h ago	10 min ago	Unknown
Provider metadata was not checked, so the synced run has not yet been compared with the latest one. Used by /d7-plus, /forecast/models
	MOGREPS-G Ensemble · ukmo_global_ensemble_20km	not published	Tue 29 2200hrs AEST · 1200Z 13 h ago	10 min ago	Unknown
Provider metadata was not checked, so the synced run has not yet been compared with the latest one. Used by /d7-plus
	ICON-EPS Ensemble · dwd_icon_eps	not published	Tue 29 2200hrs AEST · 1200Z 13 h ago	10 min ago	Unknown
Provider metadata was not checked, so the synced run has not yet been compared with the latest one. Used by /d7-plus
	ACCESS-G Deterministic · bom_access_global	not published	not synced	—	Unknown
Provider metadata was not checked, so the synced run has not yet been compared with the latest one. Used by /d0-3/compare, /d0-3/access-g Open-Meteo's ACCESS-G domain has not initialised since 2025-06-26. It is kept visible because its absence is the point; it is not on any panel.

Direct-source products

Not everything here comes through Open-Meteo. The synoptic comparison maps, the official ECMWF single-cycle reference and the three NOAA AI products are each drawn by a standalone producer that reads GRIB directly, on its own cadence.

Synoptic comparison maps Used by /d7-plus/synoptic and /d7-plus/anomaly

Current Tue 29 2200hrs AEST · 1200Z run · retrieved 14 min ago

This refreshes on a schedule; the verdict beside it is the authority on what you are seeing. To refresh immediately: npm run synoptic:refresh

ECMWF IFS single cycle Used by /forecast, /forecast/models and resort pages

Current Tue 29 2200hrs AEST · 1200Z run · retrieved 9 min ago

A single exact 00Z/12Z cycle read directly from ECMWF's open data, refreshed once daily. This deploy never fetches it on demand — to regenerate the current cycle locally: npm run ecmwf-oper:refresh Or to pull what the last scheduled run already published, without waiting on ECMWF: npm run ecmwf-oper:pull

NOAA AI-GFS Used by /forecast, /forecast/models and resort pages

Current Wed 30 0400hrs AEST · 1800Z run · retrieved 5 min ago

NOAA's AI deterministic run, read from NOMADS as one exact cycle: precipitation to +384 h and a wet-bulb profile to +240 h, so it carries a derived Snow P but no provider SWE — the centre publishes no snowfall field for it. Refreshed once daily alongside the synoptic sweep. To regenerate the current cycle locally: npm run aigfs-point:refresh Or to pull what the last scheduled run already published: npm run aigfs-point:pull

NOAA AIGEFS Used by the Model Explorer run comparison

Current Tue 29 1600hrs AEST · 0600Z run · retrieved 11 h ago · 31 members

NOAA's AI ensemble: 31 members captured individually, because percentiles taken from the published mean and spread would assume a normal distribution that precipitation does not have. It has its own workflow and its own daily schedule — 31 members through a host gated to two connections does not fit inside the synoptic job. NOMADS keeps AIGEFS for two days only, so a cycle missed here is gone for good; the Model Explorer catalogue is the only place a captured one survives past that window. To capture the current cycle locally (slow — expect ~25 minutes): npm run aigefs:refresh Or to pull the cycle the scheduled capture already published: npm run aigefs:pull

NOAA HGEFS mean Used by the snow-level charts on /forecast and resort pages

Current Wed 30 0400hrs AEST · 1800Z run · retrieved 1 min ago

NOAA's hybrid AI-and-physics ensemble, and the one product here that is a mean rather than a run. It publishes ensemble statistics only — no individual members — so there is no percentile range beneath it and no member series to keep; only the published mean is retrieved, to +240 h, its own hard cap. Its bars are unclassified and it carries no snow level: a wet-bulb profile averaged across members that disagree is a state none of them forecast. Refreshed once daily with the synoptic sweep. To regenerate the current cycle locally: npm run hgefs:refresh Or to pull what the last scheduled run already published: npm run hgefs:pull

What these columns do and do not prove

After you press the check button, Latest run published is read from the provider's own meta.json for that storage domain — often not the same name as the model (gfs_global is served from ncep_gfs013, icon_global from dwd_icon). An unreachable endpoint is reported as unknown, never as unchanged.
Run in snapshot is read from the same last-known-good artifact as the forecast pages. Where resorts disagree, the oldest is shown, because that is the oldest data a reader can still be handed.
The snapshot column is local and makes no request. The amber Live badge appears only after you choose Check providers now. A row can then read Current and Live: 6 h behind at once: the snapshot was current when last compared, and the provider has published another run since.
The deployed site is rebuilt twice a day and providers publish every six hours, so it is normal for it to sit a cycle behind between rebuilds. Amber here means the site has not been synced and redeployed since the last run landed — not that anything has failed.
Staleness is judged from when the provider published a run, not when the run initialised. ECMWF's long ensemble runs come off 00Z and 12Z, so the freshest available ENS is routinely more than six hours old with nothing wrong.
A product marked Not running cannot be fixed by refreshing. Open-Meteo's ACCESS-G domain has not initialised since June 2025; it stays listed because its absence is the point.