Weather & Climate Forecasting
Many models.
One forecast.
Inaccurate forecasting leads to disaster. No single weather model wins everywhere. They win on different variables, regions and lead times — so we run many and combine them, and we tell you how much they disagree.
Free tier: 1,000 calls a day, ensemble spread included, no sales call.
What we forecast
Four things that cost people money and lives
Severe
Severe weather
Tropical cyclone tracks with full member spread, plus the environments that produce tornadoes, damaging wind and hail.
Read the spec →
Nowcast
Precipitation nowcasting
Minute-by-minute rain and flash-flood risk at neighbourhood scale, from radar, out to two hours.
Read the spec →
Climate
Long-range climate
Seasonal drought outlooks, ENSO state and forecast, and decadal temperature trends — issued as briefings.
Read the spec →
Air quality
Air quality
Wildfire smoke dispersion, PM2.5 and ground-level ozone, and pollen by species.
Read the spec →
The difference
A number is not a forecast. A distribution is.
Most weather APIs return a single value and leave you to guess how much to trust it. An ensemble knows how confident it is, and that confidence is the part you can actually act on.
Every response carries the spread across members and the weight of each contributing model. A utility does not want 28°C — it wants to know the models disagree by four degrees, and hedge accordingly.
- p10 / p50 / p90 on every value Never a bare mean, on any endpoint.
- Named model provenance Which models contributed, and at what weight — so a bad forecast is auditable.
- Per-member access The full ensemble, not a summary, on Professional and above.
{
"init_time": "2026-09-23T00:00:00Z",
"valid_time": "2026-09-26T12:00:00Z",
"location": { "lat": 29.76, "lon": -95.37 },
"members": 42,
"models": [
{ "name": "graphcast", "weight": 0.31 },
{ "name": "aurora", "weight": 0.29 },
{ "name": "aifs", "weight": 0.22 },
{ "name": "gencast", "weight": 0.18 }
],
"spread": {
"temperature_2m": { "p10": 24.1, "p50": 28.3, "p90": 32.4 },
"wind_gust_10m": { "p10": 7.2, "p50": 11.8, "p90": 19.6 }
},
"units": "metric",
"resolution_deg": 0.25
}Published scoring
We publish our losses as well as our wins
Every forecast is verified against ERA5 reanalysis and scored against the ECMWF IFS baseline. Where we lose — currently precipitation beyond day seven — that goes on the page too. A scoreboard that only shows wins is marketing, and nobody believes it.
See our forecast accuracy →- Ground truth
- ERA5
- Baseline
- ECMWF IFS
- Grid
- 0.25°
- Metric
- RMSE / CRPS
Briefings
What we are watching
Who uses this
Built for decisions where being wrong is expensive
Energy & utilities
Load forecasting, renewable generation, and storm-hardening crew positioning. Wind and solar output tracks the forecast directly, so accuracy converts to money at a known rate.
Insurance & reinsurance
Catastrophe modelling, pre-event claims reserving, and parametric trigger pricing for hurricane and flood policies.
Agriculture
Frost and hail warnings, irrigation scheduling, seasonal drought outlooks, and planting-window guidance.
Logistics & aviation
Route disruption forecasting, port and airport closure risk, and de-icing demand.
Emergency management
Evacuation lead time, neighbourhood-scale flash-flood nowcasting, and wildfire smoke dispersion for public health advisories.
Developers
A self-serve API for products that need forecasts without building a data pipeline. Free tier, no sales call.
Start with the free tier. No sales call.
A thousand calls a day, ensemble spread included, so you can check our numbers against your own before you talk to anyone.