Skip to main content

National Water Model Data Access

Within projects at CIROH (learn more about the consortium here), we encounter a wide range of data resources and data access inquiries. Frequently asked questions include what type of hydrological data is available to use, where it is stored, and how to access it. The purpose of this page is to illustrate what types of hydrological data are available, where it is stored, and to provide a few examples of how to access it.

What Type of Data is Available?​

The official national water model (NWM) meteorological inputs and hydrology and routing outputs are available for use. The inputs and outputs are accessible through HTTP and FTP. These resources are provided by the National Center for Environmental Prediction (NCEP). All outputs from the NWM are stored in NetCDF format in one of these three file types:

  1. 1 km gridded NetCDF (land surface variables and forcing data)
  2. 100 m/250 m gridded NetCDF (ponded water depth and depth to soil saturation)
  3. Point-type NetCDF (stream routing and reservoir variables)

(see https://water.noaa.gov/about/nwm for more details about outputs spatial coverage)

In addition, output from the NWM is visualized using an experimental interactive map and image view, provided by the Office of Water Prediction (OWP). To explore those tools, go here and here.

Input and Output Data of the National Water Model​

Below are listed the major resources that grant access to the input data used and the output data produced by the operational national water model.

Official NOMADS Resource​

Through NOAA Operational Model Archive and Distribution System (NOMADS) and National Centers for Environmental Protection (NCEP) FTP server, the full set of NWM output data and a subset of forcing files are both through HTTP and FTP. They are at the following locations:

As of June 4, 2026, these resources include the following directories

para_post-processed/    31-Mar-2025 19:36   -
post-processed/         02-Nov-2020 14:31   -
prod/                   04-Jun-2026 00:16   -
v3.0/                   04-Jun-2026 00:18   -

The para\_post-processed directory lacks specific documentation, although the "para" designation suggests it is a "parallel" execution, indicating a candidate production run under testing for operational use. When the National Water Center (NWC) is experimenting with new or proposed products, these will often be placed in the para\_post-processed folder for examination before official adoption.

  • NOMADS post-processed
    • RFC: Outputs filtered down to RFC locations.
    • WMS: Contains re-indexed/reformatted outputs in per-forecast netCDFs suitable for rapid querying and responsive for graph visualizations on the water.noaa.gov/map site.
    • IMAGES: .png-formatted renderings of NWM output for various domains and variables.
    • logs: Logs.

The NWM is a rolling, real-time forecast that produces terabytes of data every day. Because of this, only data from the last 48 hours are held in the NOMADS service. Data from any other time period must be accessed through partner services, as described below.

NODD – NOAA Open Data Dissemination Program​

"The NOAA Open Data Dissemination (NODD) Program provides public access to NOAA's open data on commercial cloud platforms through public-private partnerships. These partnerships remove obstacles to public use of NOAA data, help avoid costs and risks associated with federal data access services, and leverage operational public-private partnerships with the cloud computing and information services industries." (For more information, visit NODD https://www.noaa.gov/information-technology/open-data-dissemination)

The NODD datasets, as made available through several public cloud vendors, are an incredible resource for accessing NWM data for research and evaluative purposes. The following NODD datasets have proven to be of particular value for hydrological modeling and experimentation with the NWM.

Amazon Web Services​

Amazon Web Services (AWS) hosts two repositories as part of their sustainability data initiative. Information about their sustainability data initiative can be found at this link. All data is stored in S3 buckets.

NOAA National Water Model Short-Range Forecast Dataset​

The first repository is the NOAA National Water Model Short-Range Forecast and contains the operational data. It was first launched in August 2016 and hosts the 4-week rolling collection of all model output and the corresponding forcing data. The forcing data for this model comes from the High Resolution Rapid Refresh (HRRR) and the Rapid Refresh (RAP) models. The Short-Range Forecast configuration cycles hourly and produces hourly determinist forecasts of streamflow and hydrologic states out to 18 hours. This repository is updated daily. (See this site for more information and relevant AWS CLI paths.)

All data is in NetCDF format and follows these general configurations:

  • 1 km gridded NetCDF (land surface variables and forcing)
  • 250 m gridded NetCDF (ponded water depth and depth to soil saturation)
  • Point-type NetCDF (stream routing and reservoir variables)
  • Point-type reservoir NetCDF (water surface elevation, inflow, outflow)

To browse the bucket, go here. More detailed documentation about data file naming conventions and configuration can be found here.

NOAA National Water Model Retrospective Datasets​

The second AWS repository is called the NOAA National Water Model CONUS Retrospective Dataset. AWS hosts four different versions of the retrospective dataset corresponding to different versions of the National Water Model. Below is a description of each version that includes what geographical areas are represented, which years are represented in each retrospective simulation, what outputs and forcing data inputs are provided, and what file formats are the data stored in. Each version also has links to the corresponding buckets holding the data.

Version 3.0

Version 3.0 is a 44-year retrospective simiulation using v3.0 of the NWM spanning from February 1979-January 2023. It provides output for all of the continential United States (CONUS) as well as coverage fro Alaska, Hawaii, Puerto Rico, and US Virgin Island domains. This makes v3.0 unique because all other versions have output for CONUS

Forcing data for the model is avaliable. For CONUS domains, the simulation uses v1.1 forcing data from the Office of Water Prediction Analysis of Record for Calibration (AORC) dataset. AORC Alaska forcing data is used in Alaska, North American Regional Reanalysis (NARR) data along with precipitation from the Alaska Pacific River Forecast Center (APRFC) was used to drive the Hawaii retrospective simulation. Similarly, the Puerto Rico & US Virgin Island retrospective simulation was driven by NARR data along with precipitation from the Southeast River Forecast Center.

NWM Retrospective data is avaliable in 2 formats: NetCDF and Zarr. The NetCDF files for v3.0 contain the full set data (forcing and output) while the Zarr files contain all the output data and precipitation forcing fields.

  • Browse either the NetCDF or Zarr files here.

Version 2.1

Version 2.1 is a 42-year retrospective simulation using v2.1 of the NWM spanning from February 1979-December 2020. Its output domain is CONUS. Forcing data is avaliable for this similation. The model uses forcing data from AORC. For 1979-2006, NWM v2.1 uses forcing data from AORC v1.0, and for 2007-2020, NWM v2.1 uses forcing data from AORC v1.1.

All model output and forcing input fields are avaliable in NetCDF format. In Zarr formats, many of the output fields and the forcing precipitation fields are avaliable.

Click to see an example of pulling data from the channel output zarr 2.1 archive and writing the results to csv. 
'''
#install these libraries if they aren't already installed
!pip install zarr
!pip install xarray
!pip install s3fs
!pip install numpy
'''
# Import needed libraries

import xarray as xr
import numpy as np
import s3fs
from datetime import datetime, timedelta

# open the zarr store
url = "s3://noaa-nwm-retrospective-2-1-zarr-pds/chrtout.zarr"
fs = s3fs.S3FileSystem(anon=True)
store = xr.open_zarr(s3fs.S3Map(url, s3=fs))

# Function to get the time series for a specified reach id and and time range
# then write it out to a csv file.
def GetAndWriteTimeSeriesAtReach(reach_id, start_time_index, end_time_index):
    flows = streamflow_array.where(feature_id_array==reach_id, drop=True)
    df_flows = flows[start_time_index:end_time_index].to_dataframe()
    df_flows.to_csv(f'flows_{reach_id}.csv')

# get an xarray array of the various values
time_array = store['time']
feature_id_array = store['feature_id']
streamflow_array = store['streamflow']

# Define the feature IDs to check for
feature_ids = [5781221, 5781223, 5781703]

# Specify the start and end times of interest
start_time = datetime(2015, 5, 23, 0, 0, 0)
end_time = datetime(2015, 6, 24, 0, 0, 0)

# Get the indices for the needed dates
zero_start_time = start_date = datetime(1979, 2, 1, 0, 0, 0)
start_time_index = int((start_time - zero_start_time).total_seconds() / 3600)
end_time_index = int((end_time - zero_start_time).total_seconds() / 3600)

for reach_id in feature_ids:
    GetAndWriteTimeSeriesAtReach(reach_id, start_time_index, end_time_index)

'''
Simple Script for Retrieving Retrospective NWM Data from AWS Store
Dan Ames, 11/17/2023
dan.ames@byu.edu
'''


Version 2.0

Version 2.0 is a 26-year retrospective simulation using v2.0 of the NWM spanning from January 1993-December 2018. Its output domain is CONUS. No forcing data for this version is avaliable through AWS.

All model output fields are avaliable in NetCDF format. Model channel output including streamflow and related fields are available in Zarr format.

Version 1.2

Version 1.2 is a 25-year retrospective simulation using v1.2 of the NWM, spanning January 1993-December 2027. Its output domain is CONUS. No forcing data is avaliable through AWS for this version.

All model output fields are avaliable in NetCDF format. No Zarr format files are avaliable for this version.

For more general information about the retrospective data sets see this site, and for more detailed documentation, go here.

The AWS retrospective resource is the primary publicly available source for the version 1.0 of the Analysis of Record for Calibration dataset (AORC), which is a 40-year best-available estimate of most common meteorological parameters required for hydrological modeling. Version 1.1 of the dataset accompanied the release of the NWM model version 3.0 retrospective and aimed to resolve issues from version 1.0.

Additional Jupyter notebook instructions for processing NWM Zarr and NetCDF output formats are available here.

Google – Operational NWM Data​

  • Operational data can be browsed here.
  • A copy of the NWM v1.2 retrospective data can be browsed here.

NWM BigQuery API​

Some of the datasets from the NWM operational and retrospective simulations are available on BigQuery for ultra-high-bandwidth access. For more information, visit the NWM BigQuery API page.

Microsoft Azure Blob Storage​

Microsoft hosts the NWM input and output datasets in Azure Blob Storage. The dataset is not yet avalible through Microsoft Planetary Computer API. All data files are in NetCDF4 format. Documentation details and file name syntax can be found here.

There are 5 main categories of data stored in Azure Blob Storage, listed below.

NWM Analysis and Assimilation

This section has 5 subcategories of data: Analysis and Assimilation, Analysis and Assimilation Long, Analysis and Assimilation Forcing, Analysis and Assimilation (No DA), Analysis and Assimilation Long (No DA). For the Analysis and Assimilation, Analysis and Assimilation Forcing, and Analysis and Assimilation (No DA), data is available for the Continential United States, Hawaii, and Puerto Rico. For the two "Long" data sets, only data avaliable is for the Continetial United States. The "No DA" stands for "No Data Assimilation."

NWM Short Range

Both the short range forecast and the short range forcing data are included in this section. Data is avaliable for the Continetial United States, Hawaii, and Puerto Rico for both data sets.

NWM Medium Range

Medium range forecast data and medium range forcing data are only avaliable for the Continential United States.

NWM Long Range

Long range forecasts are avaliable for the Continental United States only. There is no associated forcing data avaliable.

USGS Observations

This data set has USGS Timeslices, which are 15-minute observation data for the Continental United States.

Tom Augspurger of Microsoft has a series of notebooks providing examples of how to use this data from his workshop at the first CIROH Developers Conference. View them here.

CIROH Resources​

More detailed information and example usage will be available soon.

  • Kerchunk Retro (points to AWS 2.1 NetCDF Retro)
  • Kerchunk Operational (points to Google assets – a simple text change can point to AWS short range, if desired)

Other resources​

ESRI Living Atlas​

ESRI Living Atlas provides a map-enabled version of the NWM output, which can be accessed here.

Description of WRF-Hydro code:​

A detailed description of various aspects of the WRF-Hydro code, which produces the current NWM, can be found here.