Aeris Forecast

Every aviation weather tool I used showed the same picture: a beautiful isobar map that answers "what is the weather" but never "can I fly this corridor now, and where exactly would it get dicey."

Aeris is my attempt at the second question — a personal pipeline that turns public gridded forecasts into a decision surface for pilots and field crews.

Pipeline

Four moving parts, deliberately boring:

1. Ingest — pull model runs (wind, temp, precip, ceiling) on schedule.
2. Reshape — regrid to multi-resolution storage. Wind shear wants fine cells; pressure gradients don't. Same idea as the thesis quadtree, applied to weather.
3. Query — a PostGIS layer that answers "along this polyline, when does any threshold break?" in one round trip.
4. Show — a MapLibre client that paints risk, not raw variables.

SELECT segment, time,
       max(wind_gust)  AS gust,
       ceil(min(ceiling)) AS floor
FROM aeris.enroute
WHERE flight_id = $1
GROUP BY segment, time
HAVING max(wind_gust) > 22 OR min(ceiling) < 1000;

What it taught me

• Public weather data is a solved problem; legibility is not.
• P95 tile latency is a UI feature. People forgive slow models, not slow maps.
• Multi-resolution storage makes "is this flight legal and wise right now" cost one query instead of a dashboard scan.

Status

Runs quietly on a small box, mostly for me and a few pilot friends. If the geospatial or simulation side of this interests you — or you have a corridor worth testing — say the word.