Oasis Weather
Calibrated forecast API · Free tier available

Forecasts that ship with a confidence interval — not just a number.

Daily and seasonal weather forecasts anchored to NOAA Weather Forecast Office guidance. Every value comes back with a high, a low, and an explicit standard deviation, so the caller can size decisions against the uncertainty instead of guessing at it. Every value also names where it came from.

1,000 free calls / month WFO-anchored Provenance per day
POST /v1/forecast/point
200 OK · 78 ms
{
  "succeeded": true,
  "station_id": "NYC",
  "day_points": [
    {
      "date":     "2026-06-21",
      "high_f":   { "mu": 82.1, "sigma": 3.4, "p10": 77.7, "p90": 86.5 },
      "low_f":    { "mu": 66.4, "sigma": 2.8, "p10": 62.8, "p90": 70.0 },
      "provenance": {
        "source":     "nws-wfo-gridpoint",
        "wfo":        "OKX",      // New York / Upton
        "issued_at":  "2026-06-20T18:14:00Z",
        "model_cycle": "12Z"
      }
    }
  ],
  "method": "wfo+climatology-residual-ci",
  "calibration": { "coverage_p80": 0.81, "coverage_p90": 0.89 }
}

What you get

The interval is the product.

A single forecast number is the wrong input to a decision. The Oasis Weather API ships the distribution instead, anchored to the same Weather Forecast Office guidance the NWS issues, with calibration receipts and per-day provenance you can audit before you commit.

Calibrated confidence intervals

Every value ships with μ, σ, and a p05–p95 confidence interval fit so its stated coverage actually holds on out-of-sample data. The receipts endpoint lets you verify coverage on your own historical window before wiring the API into anything load-bearing.

Provenance on every value

Every day in every response names its source, the issuing Weather Forecast Office, and the model cycle that produced it. If the engine falls back to climatology, the response says so. No silent substitutions, no opaque blends.

WFO-anchored, not vendor-blended

Forecasts ride on the actual gridpoint guidance from the WFO covering your station — the same authoritative source emergency managers and broadcasters use. Climatology and a 30-year observational baseline fill the gaps between forecast lead times.

Endpoints

Five surfaces, one shape.

All endpoints return the same confidence-interval envelope and the same provenance fields. The credit cost varies by how much work the request does. Coverage is the contiguous US in v0; Alaska, Hawaii, and US territories are on the roadmap.

POST /v1/climatology Climatology with calibrated bands Daily-temperature climatology over a date range, anchored to a 30-year station-observation baseline with a residual-sigma inflation fit on independent back-test years. 1 credit
POST /v1/forecast/point WFO-guided point forecast Daily high and low for a station or (lat, lon), riding the same Weather Forecast Office gridpoint guidance the NWS issues. Provenance names the WFO, model cycle, and issue time on every day. 1 credit
POST /v1/history/observations Observed history with rolling statistics Daily observations for any window, returned alongside configurable rolling-window statistics (mean, σ, min, max, skew, kurtosis) so the same call answers "what was it?" and "what's the distribution shape look like?" 5 credits
POST /v1/window/compare · Pro+ Window comparison + non-stationarity Statistical comparison across two date windows for the same station: distribution-equality tests, mean-shift detection, Granger causality between paired stations, and probabilistic mixture-model fits that surface drift, regime change, and ENSO/warming components as named parameters. 10 credits
POST /v1/seasonal Seasonal outlook Monthly mean with a tight confidence interval over a 1–12 month horizon. Method tag rides on the response; warming-trend and ENSO legs swap in without changing the call shape. 2 credits

Pricing

Start free. Scale by call volume.

Credits debit per call, weighted by endpoint cost. Plans renew monthly; unused credits do not roll over.

Free

$0 /mo

  • 1,000 credits / mo
  • 1 API key
  • All 5 endpoints
  • Community support
Start free

Basic

$29 /mo

  • 10,000 credits / mo
  • 1 API key
  • All 5 endpoints
  • Email support
Subscribe

Team

$499 /mo

  • 300,000 credits / mo
  • 25 API keys
  • All 5 endpoints
  • Slack/email channel
Subscribe

Enterprise

Custom

  • Custom credits
  • 100+ API keys
  • SLA + dedicated regions
  • Direct engineering contact
Contact sales

Why this exists

Calibration is the engineering problem.

Most weather APIs sell coverage.

They give you a number — a high, a low, a probability of precipitation — and leave you to figure out how much to trust it. That is fine for a weather widget; it is not fine for a trade, a control loop, or a multi-day plan.

The interval is what tells you how seriously to take the number. Oasis Weather is built around shipping intervals that hold their stated coverage on real out-of-sample data, with the per-day provenance and the receipts to prove it.

Built by an engineering team that ships calibrated systems.

Oasis Weather is the productized version of the forecast-calibration and verification work we already do as bespoke engagements — energy-system modeling, microclimate work for siting, and feasibility studies where the uncertainty itself is the deliverable.

If the API does not fit your shape and you need calibrated weather wired into a larger decision system, that work is a separate conversation. Either way, the calibration discipline is the same.

Who this is for

When the interval is load-bearing.

The angle — calibrated intervals, WFO-anchored guidance, full provenance — opens specific kinds of work that a point-forecast API can't support cleanly.

Trading & risk

Power, gas, agricultural, and weather-derivative desks that need a real distribution to size positions against. The receipts endpoint is the audit trail that satisfies a risk committee.

Operations & control

Grid operators, HVAC fleets, agricultural irrigation, supply-chain ETA models. Anywhere a control loop needs to weigh a worst-case against a best-case before it commits.

Scientific & research

Atmospheric researchers, climate-impact modelers, and backtest harnesses that need provenance per value to keep results reproducible. Mixture-model parameters expose drift, regime change, and seasonal heterogeneity in a way you can cite.

Consumer & device UX

Product teams building forecast UIs that want to show the range, not just the headline number — "85°F with a 7° spread, low confidence" beats "85°F" every time. Same call shape, your app decides how much of the distribution to expose.

Permitting & siting studies

Engineering and planning teams that need a defensible probabilistic baseline — for HVAC sizing, microclimate impact, renewable-resource assessment, or stormwater design — with cited provenance per data point.

Weather-risk pricing

Parametric coverage, energy hedges, weather derivatives, and event-cancellation triggers — where the strike, the payout, and the price all depend on the calibrated tail, not just the central forecast.

Ready to evaluate the API?

The free tier is enough to evaluate end-to-end against a real workload. Send a short note with the use case and the rough monthly call volume; we will mint a key while the self-serve signup is being finished.