Weatherbit

Weatherbit (weatherbit.io) is a US-based weather data provider offering a unified REST API spanning current conditions, forecasts (16-day daily, 240-hour hourly, 60-minute minutely), 30 years of point historical data, severe-weather alerts, lightning, air quality, agricultural weather (soil temperature/moisture, evapotranspiration), and energy/degree-day datasets with solar irradiance. Its positioning is "accuracy through machine learning": rather than reselling a single raw model, Weatherbit blends regional (e.g. NOAA HRRR) and global (GFS) models and applies statistical/ML bias-correction, ingesting data from 50,000+ weather stations plus satellites, radiosondes, and lightning networks. This makes it a strong fit for operational dashboards, IoT, energy/utility demand modeling, and agritech, where frequent (hourly-refreshed) forecasts and long historical backfills matter more than a consumer-grade widget.

The developer experience is a recognized strength: third-party roundups (Visual Crossing, Tomorrow.io) repeatedly single out Weatherbit's documentation quality and fast response times, and the always-free tier (50 calls/day, 16-day forecast, non-commercial) lowers the barrier to evaluation. The flip side is operational nuance that catches teams off guard. Historical requests are metered unusually, anything over 10 days counts as multiple requests against the daily quota (30 days = 3 calls), so heavy backfill jobs burn quota fast. Archived forecast *runs* aren't available from the live endpoints at all; they require a separate custom-report system. Historical records can also be silently revised as better source data arrives, so research-grade pipelines must check revision_status/revision_version fields. The official SDK story is thin, a single MIT-licensed Python wrapper (28 stars) on GitHub, with everything else driven off the published Swagger/OpenAPI spec.

Reliability is a selling point but tier-gated: Weatherbit claims geographically redundant data centers (North America + Europe) with "nearly 100% uptime" since 2017, but the contractual 99.9% SLA and 200ms average forecast response guarantee apply only to Advanced/Enterprise plans, Basic/Developer accounts get a 99.5% target with no SLA. Public third-party review presence is genuinely sparse (no meaningful G2/Capterra/Trustpilot footprint, unclaimed Datarade profile), which is the main caveat for buyers who want independent validation: you are largely relying on the vendor's own accuracy comparison page rather than neutral benchmarks.