GribStream

GribStream Blog

NBM v5.0 goes live April 15, 2026: what to check before cutover

|
nbm nbmpar nbmparqmd

NWS has set April 15, 2026 for NBM v5.0, with longer hourly guidance, upgraded QPF and snow products, and a clear cutover date for validation.

On March 12, 2026, NOAA/NWS issued SCN 26-24, setting April 15, 2026 as the operational implementation date for NBM v5.0.

If you use NBM in production, April 15 is the date to mark in validation, alerting, and backtest workflows.

New products and removals

The SCN is specific about what NBM v5.0 adds.

New products include:

  • a 24-hour probability matched mean QPF product for CONUS and Alaska
  • a 24-hour probability matched mean snow and ice product for CONUS and Alaska
  • joint fire weather probabilities for combinations of wind and RH thresholds
  • quantile-mapped 10 m wind speed and gust guidance for Alaska, Hawaii, Puerto Rico, Guam, and Oceanic domains
  • new deterministic and probabilistic precipitable water guidance
  • 24 h / 48 h / 72 h calibrated snow exceedance guidance for CONUS
  • deterministic snow depth and related exceedance probabilities for CONUS and Alaska
  • apparent temperature guidance
  • CAPE weighted percentiles for CONUS

The main removal is also explicit: Haines Index is gone in v5.0.

Method changes that can move real results

The most useful part of the SCN is not the product list. It is the list of underlying methodology changes:

  • NBM v5.0 pushes for greater consistency between deterministic and probabilistic outputs instead of computing them more independently
  • QPF and PoP move away from equal weighting of all inputs and toward heavier weighting of higher-resolution guidance, with WPC MMEBC QPF added as a CONUS input
  • the winter suite changes model weighting, accumulation handling, downscaled temperature treatment, and snow/ice to rain conversion logic
  • instantaneous temperature, dew point, apparent temperature, RH, and 12-hour max/min RH move from decaying-average logic to quantile mapping
  • deterministic wind speed and gust switch to a new percentile-picking method instead of using the mean QM value
  • significant wave height bias correction also moves to quantile mapping, and the number of inputs increases from 13 to about 120 by using ensemble members instead of just ensemble means

Those are the changes most likely to show up in threshold behavior, percentiles, and time series comparisons.

Longer hourly guidance and different inputs

NOAA also calls out two operationally important changes:

  • hourly guidance extends from 36 hours to 48 hours for most parameters
  • several inputs get replaced or upgraded, including higher-resolution ECMWF, ECMWF Ensemble, GDPS, and REPS guidance

The SCN also says NBM v5.0 removes SREF input usage and adds both ECAIFS and AIGFS as inputs for temperature, wind speed, and QPF products, with higher-resolution GEFS surface guidance also feeding the blend.

What to check

  • hourly workflows that currently stop at 36 hours
  • precipitation and snow thresholds
  • wind logic that depends on deterministic versus percentile behavior
  • significant wave height workflows
  • any analysis that assumes the older v4.3 field mix or older upstream input stack

The SCN also warns that runtime and dissemination timing will shift for some quantile-mapped and core products. Some products may arrive later, some earlier. If you key jobs off NBM file arrival times, that is worth checking before cutover.

For text-product users, the notice also calls out a few concrete changes:

  • wet bulb globe temperature (WBG) is added to NBH/NBS/NBE/NBX
  • significant wave height (SWH) becomes available for all cycles in NBX
  • Haines Index (HID) disappears from NBM text products

If you want to compare before and after without waiting for the live cutover, NOAA's public parallel line is already available. See NBM v5 parallel datasets now available.

Related reading

Sources

Related models

nbm nbmpar nbmparqmd
Try GribStream