National Blend of Models
NBM Overview
The National Blend of Models (NBM) is a nationally consistent and skillful suite of calibrated forecast guidance based on a blend of both NWS and non-NWS numerical weather prediction model data and post-processed model guidance. The goal of the NBM is to create a highly accurate, skillful and consistent starting point for the gridded forecast. This new way to produce NDFD grids will be helpful providing forecasters with a suite of information to use for their forecasts. The NBM is considered an important part of the efforts to evolve NWS capabilities to achieve a Weather-Ready Nation.
Cycles and Domains
The National Blend of Models (NBM) runs several times daily, with its cycle frequency varying depending on the specific product suite and domain. This lack of uniformity is a result of processing a wide array of model outputs, each with unique arrival times and product output. Altogether the NBM runs hourly and covers the following domains: CONUS, Alaska, Hawaii, Puerto Rico, Guam, and Oceanic.
Data Availability
The availability of the NBM varies by product, region, and cycle.
| Core NBM | |
|---|---|
| Cycles | Timing (UTC) |
| 00 | 00:55 - 01:00 |
| 07 | 08:05 - 08:10 |
| 12 | 12:55 - 13:00 |
| 18 | 18:40 - 18:50 |
| 19 | 20:00 - 20:10 |
| 01, 02, 03, 04, 05, 06, 08, 09, 10, 11, 13, 14, 15, 16, 17, 20, 21, 22, 23 | HH:30 - HH:45 |
| Quantile Mapping and Dressing (QMD) | |
|---|---|
| Cycles | Timing (UTC) |
| Alaska/CONUS/Oceanic | |
| 00 | 06:50 - 07:20 |
| 06 | 12:50 - 13:40 |
| 12 | 18:50 - 19:20 |
| 18 | 00:50 - 01:30 |
| Puerto Rico | |
| 06 | 12:40 - 12:45 |
| 18 | 00:40 - 00:45 |
| Winter NBM | |
|---|---|
| Cycles | Timing (UTC) |
| 01 | 02:05 - 02:15 |
| 07 | 08:00 - 08:05 |
| 13 | 14:05 - 14:15 |
| 19 | 20:00 - 20:10 |
| 00, 02, 03, 04, 05, 06, 08, 09, 10, 11, 12, 13, 14, 15, 16, 17, 18, 20, 21, 22, 23 | HH:30 - HH:35 |
| Oceanic NBM | |
|---|---|
| Cycles | Timing (UTC) |
| 00 | 01:00 - 01:10 |
| 07 | 08:10 - 08:20 |
| 12 | 13:00 - 13:10 |
| 19 | 20:10 - 20:20 |
| Global NBM | |
|---|---|
| Cycles | Timing (UTC) |
| 00 | 11:30 - 11:35 |
| 12 | 23:30 - 23:35 |
| Text NBM | |
|---|---|
| Cycles | Timing (UTC) |
| 01 | 00:55 - 01:00 |
| 07 | 08:05 - 08:10 |
| 13 | 12:55 - 13:00 |
| 19 | 18:40 - 18:50 |
| 00, 02, 03, 04, 05, 06, 08, 09, 10, 11, 12, 14, 15, 16, 17, 18, 20, 21, 22, 23 | HH:30 - HH:40 |
Data
A live data feed (past couple of days of data) of Core and QMD NBM is available on NOAA’s Operational Model and Distribution System (NOMADS) A full archive going back to about May 2020 of operational NBM data is available on our Amazon Web Services (AWS) big data project; now part of the NOAA Open Data Dissemination (NODD) database. An additional location of archive data is at the MDL Big Data Archive Viewer.
Inputs and Interpolation
The NBM integrates forecasts from over one hundred individual models, many of which are not the same resolution as the NBM grids. This section provides a general overview of the models incorporated into the NBM and their characteristics. The models listed here are not universally applied to all elements. This is due to the varying data availability within each model.
Weighting
The NBM employs a wide variety of weighting techniques when blending multiple model solutions together depending on the element in question. These weights can be derived in a subjective manner, where developers manually assign weights based on subject matter expert input, or in an objective, statistical manner. Weighting schemes applied include decaying average mean absolute error weighting, expert weighting, and dynamic resolution weighting.
Probabilities and Threshold Exceedances
The data points listed for each region are structured around percentiles (0-100), which denote the range of possible outcomes, and probabilities of exceedance (the chance of a variable surpassing or falling below a specific threshold). These products vary by region, and are detailed below. Note that thresholds in blue have been added to the NBM as of version 5.0, while those in red have been removed.
Quantile Mapping & Dressing (QMD)
QMD is a bias correction technique that leverages the entire distribution of events in the form of Cumulative Distribution Functions (CDFs) for a given variable over some N-day training period for the model forecast system and for the analysis (truth). Given today’s forecast value at that grid point, one can determine the associated quantile from the forecast CDF and then replace the forecast with the analyzed value associated with that same quantile. Quantile mapping adjusts for bias conditioned on the forecast precipitation amount, and it does so in a way that avoids the collapse of spread common with regression approaches when there is little relationship between the forecast and observed. It should be noted here that the “D” in QMD represents the dressing of the quantile-mapped value. At this time, no dressing is performed in the QMD system.
Aviation Products
A guide overviewing the configuration and technical details of NBM Aviation products, including wind speed and gust, can be viewed here: .
Precipitation
The NBM produces calibrated precipitation guidance that is generated using the QMD correction technique for the Alaska, CONUS, Hawaii, Oceanic, and Puerto Rico National Digital Forecast Database (NDFD) domains and initialized daily at 0000, 0600, 1200, and 1800 UTC. Probabilistic guidance in the form of percentiles of QPF (1 through 99); probability exceedance thresholds; and a single deterministic value (the mean of QMD values per grid point) is generated for precipitation duration periods of 6-, 12-, 24-, 48-, and 72-hours, out to Day 11. The probability exceedance thresholds change per the precipitation duration period.
Winter Weather Products
The NBM does not ingest a model input's native winter product to produce a forecast. Instead, the NBM computes its own winter accumulations using QPF, precip type, and thermodynamic information from the individual inputs.
Marine Products
In order to further enhance the probabilistic guidance for wave height, v5.0 employs the quantile mapping and dressing (QMD) approach. This will leverage all the ensemble members and deterministic forecasts that are available from each wave model family.
Fire Weather Elements
A guide to the configuration and technical details of NBM fire weather products, including the v5.0 introduction of joint fire weather probabilities, can be found in the document .
Tropical Products
Active tropical cyclones are incorporated in the NBM as separate wind speed, gust, and direction products utilizing the NBM windfield as a background and a combination of the gridded windspeed probabilities from the Tropical Cyclone forecast advisory Message (WTCM) and the Hurricane Analysis and Forecasting System (HAFS) models to represent the storm.