GribStream Blog
HRRR sub-hourly (15-minute) surface fields now available
GribStream added NOAA HRRR sub-hourly surface output (wrfsubhf) with 15-minute valid times out to 18 hours for intraday workflows.
On February 4, 2026, we added NOAA HRRR sub-hourly surface fields to GribStream.
This is both a new dataset and a new capability: GribStream now supports datasets whose lead times are not whole hours (e.g. 15-minute valid times).
About HRRR sub-hourly
HRRR sub-hourly (the wrfsubhf product family) is a surface-focused HRRR output stream for CONUS with 15-minute valid times.
- GribStream code:
hrrrsubh(model page) - Coverage: CONUS
- Resolution: 3 km HRRR Lambert grid
- Update cadence: hourly cycles
- Lead time: 0-18 hours, with valid times at
:00,:15,:30,:45
If you need the full HRRR catalog (hourly fields, longer horizons, and broader parameter coverage), use HRRR.
Why this matters
Most weather APIs stop at hourly resolution. But many real-world decisions do not.
15-minute HRRR fields are useful for:
- Power markets / ISO operations: intrahour bidding, dispatch, and curtailment workflows that care about ramps.
- Wind and solar: short-horizon ramp detection, persistence checks, and blending with site telemetry.
- Aviation: fast-changing surface winds and near-term boundary-layer evolution.
- Severe weather monitoring: sharper timing for instability and boundary interactions in the next few hours.
How to query it
Use the standard GribStream endpoints:
/api/v2/hrrrsubh/runsfor a single model run (best for clean, continuous forecasts)/api/v2/hrrrsubh/timeseries(alias:/history) for the "best available" forecast at every valid time
For request/response examples and payload structure, see Quick-start. To browse available variables/levels and copy selectors, use the inventory on the HRRR sub-hourly model page.
A note on "history" vs "forecast"
HRRR sub-hourly is still an hourly cycling system. That means:
- The analysis points are at
horizon=0(hourly). - The :15/:30/:45 points are short-lead forecasts.
If you stitch "best available" values across runs (the timeseries walker), you can see hour-boundary jumps when the chosen cycle changes. For "smooth" intrahour curves, either query a fixed run (/runs) or interpolate between consecutive horizon=0 analysis points in downstream tooling.