Chart No. AWX-001 · Pre-release
A sky you can brief.
ArcusWX reads what a forecaster reads — GFS and Open-Meteo grids, METARs, TAFs, terrain — resolves it into one coherent three-dimensional atmosphere, and holds the simulator to it. Winds and temperature aloft, cloud bases and tops, thermal, convective and mechanical turbulence, visibility, QNH.
Two messages, total. First public flight test, and release. No newsletter, no partner mail.
- Status
- In development
- Platform
- Windows x64
- Targets
- MSFS 2020 · 2024 · X-Plane 12
- Sources
- GFS · METAR · TAF · ERA5
Constructed case, drawn to convention. Not a live observation.
Weather is a volume, not a value.
Most sim weather is a surface observation with a sky drawn on top. The column above the field is where the flying actually happens: the shear at three thousand feet, the freezing level you are about to climb through, the inversion holding the haze down, the tops you are trying to get above.
ArcusWX resolves the whole column first and derives the picture from it, never the other way round. One continuous three-dimensional field, sampled at every level, instead of four sliders and a cloud preset.
Any hour you can name. Reanalysis replay puts the atmosphere back the way it was — the day you flew it, or the day someone else did. Sandbox mode lets you build a column by hand and fly that instead.
- Sky cover
- Fills the circle. Comes from the modelled layer structure, not from a single METAR octa.
- Temperature
- Upper left. Read at every level, not just the field.
- Dewpoint
- Lower left. The spread drives base height. It is not a global slider.
- Pressure
- Coded to tenths, upper right. The QNH the sim actually flies.
- Tendency
- Lower right. A 4 hPa rise in three hours is a gust front arriving.
- Wind barb
- Five, ten and fifty knots. The same convention runs from the surface to the tropopause.
One sounding per position, not one number per airport.
The plate above is a plan view. This is the same air seen edgewise: the profile at the station immediately ahead of the boundary, plotted on a skew-T. Everything annotated here is read off the trace, not asserted next to it.
- CAP
- 1.5 °C inversion between 850 and 800 hPa. The layer that decides whether anything happens.
- FREEZING LEVEL
- 0 °C at 557 hPa, ≈16,300 ft. Derived from the profile, not set by a slider.
- MOIST LAYER
- 7 °C spread at the surface, closing to 2.5 °C at 850. Cloud base follows from it.
- TROPOPAUSE
- The top of the column ArcusWX builds. Every level between here and the runway is modelled, not interpolated from two.
- 01
Ingest
GFS 0.25° and Open-Meteo grids, METARs and TAFs, ERA5 reanalysis, and an airport and runway database with elevation, surface and orientation.
- 02
Model
One continuous three-dimensional field. Not a stack of presets, and not four layers stretched to fit an observation.
- 03
Inject
Written to the running simulator, eased between states, at a rate you set. You decide how fast the sky is allowed to change.
- 04
Explain
Every value the engine hands the simulator is traceable to the observation or grid point it came from. If the freezing level looks wrong, you can see why.
Where the numbers come from.
| Source | Resolution | Cadence | Use |
|---|---|---|---|
| GFS | 0.25° global | 6 h cycle, 1 h steps | Wind, temperature and humidity through the column |
| Open-Meteo | 0.1°–0.25° blended | Hourly | Model blending, and fallback where GFS is thin |
| METAR | Point | ~1 h, sooner on change | Surface truth: wind, visibility, ceiling, QNH, dewpoint |
| TAF | Point | 6 h cycle | Forward-looking correction at the destination |
| SIGMET / AIRMET | Polygon | On issue | Convective, turbulence and icing volumes, as geometry |
| ERA5 reanalysis | 0.25° global | Hourly, 1940 → present | Historical replay |
| Airport and runway database | Point and polyline | Per AIRAC cycle | Elevation, surface and orientation for mechanical turbulence |
Sources may change before release. This table will change with them.
There are no screenshots on this page.
There is nothing yet we would want picked apart on a forum. When there is, this list sees it first — the flight-test notes, the model-versus-observation comparisons, and the parts that are still wrong.
Two messages, total. First public flight test, and release. No newsletter, no partner mail.