This data set contains snow pack properties, such as depth and snow water equivalent (SWE), from the NOAA National Weather Service's National Operational Hydrologic Remote Sensing Center (NOHRSC) SNOw Data Assimilation System (SNODAS). SNODAS is a modeling and data assimilation system developed by NOHRSC to provide the best possible estimates of snow cover and associated parameters to support hydrologic modeling and analysis.
Snow Data Assimilation System (SNODAS) Data Products at NSIDC, Version 1
This is the most recent version of these data.
Overview
|
Citing These Data
As a condition of using these data, you must cite the use of this data set using the following citation. For more information, see our Use and Copyright Web page.
National Operational Hydrologic Remote Sensing Center. 2004. Snow Data Assimilation System (SNODAS) Data Products at NSIDC, Version 1. [Indicate subset used]. Boulder, Colorado USA. NSIDC: National Snow and Ice Data Center. doi: https://doi.org/10.7265/N5TB14TC. [Date Accessed].Documentation
Overview
This data set contains output from the NOAA National Weather Service's National Operational Hydrologic Remote Sensing Center (NOHRSC) SNOw Data Assimilation System (SNODAS). SNODAS is a modeling and data assimilation system developed by NOHRSC to provide the best possible estimates of snow cover and associated parameters to support hydrologic modeling and analysis. The aim of SNODAS is to provide a physically consistent framework to integrate snow data from satellite, airborne platforms, and ground stations with model estimates of snow cover (Carroll et al. 2001). SNODAS includes procedures to ingest and downscale output from the numerical weather prediction (NWP) models and to simulate snowcover using a physically based, spatially-distributed energy- and mass-balance snow model. SNODAS also includes procedures to assimilate satellite-derived, airborne, and ground-based observations of snow covered area and snow water equivalent (SWE).
NOHRSC, located in Minneapolis, Minnesota, supplies snow information in a variety of products and formats to meet operational forecasting needs. Most of these products are available from the NOHRSC website. In partnership with NOHRSC, the NOAA program at NSIDC provides this subset of SNODAS output as a service for an extended community of users. We archive, document, and distribute selected SNODAS parameters. These output files are valuable for hydrologists, hydrologic modelers, climatologists, ecologists, and land surface modelers. These data are not suitable for snow fall events or totals for specific regions. For snow fall data, please see the state climatology reports for a particular state.
This documentation draws heavily on an assessment of SNODAS products by Barrett (2003) and on material provided by NOHRSC. Consult these sources for additional information. While we attempt to keep the documentation current, NOHRSC sometimes makes changes in processing or output format that are not caught by us before some time has passed. We depend on users to make us aware of inconsistencies in the product, and we correct them or document them as resources allow.
Detailed Data Description
Provided here are gridded data sets for the contiguous United States at a 1 km spatial resolution and a 24 hour temporal resolution. Data are stored in flat binary 16-bit signed integer big-endian format with header and metadata files. Both a masked version for the contiguous United States and an unmasked version that extends north into Canada are available.
The SNODAS product is model output and should not be confused with actual observations. These data are not suitable for snow fall events or totals for specific regions. For information on snowfall events or snowfall totals, please contact one of the climate centers listed below:
- American Association of State Climatologists website
- NOAA Regional Climate Centers website
- NOAA National Climatic Data Center website
Eight driving, state, and diagnostic parameters are archived by NSIDC. Through September 2018, driving parameters were ingested from the Rapid Update Cycle 2 (RUC2) NWP model and used to force the snow model. For data after September 2018 through the present, driving parameters are ingested from the Rapid Refresh (RAP) NWP model. State variables are defined here as parameters that the snow model keeps track of and that describe the state of the model snow pack. State variables are modeled snow pack characteristics that are also required to initialize the model. Diagnostic variables are model output but do not describe the internal state of the model. The parameters archived by NSIDC are listed in Table 1 and can be used to compute snow water balance.
Warning: In December 2018, NSIDC became aware that there were persistent, erroneous zeroes in SWE data primarily found around the perimeters of water bodies and along coastlines. In October 2019, the data provider, NOHRSC, was able to diagnose and fix the issue. Details of the issue and its consequences are described below.
The issue arose when NOHRSC adjusted their land/water mask in October 2014. Because of an error made at the time in the modification of the model state variables, the new mask was never properly initialized for many locations on the SNODAS grid. Exactly 126,950 cells on the grid, primarily found around the perimeters of water bodies and along coastlines, were affected. In the SNODAS archives at NSIDC, these cells will always have zero values for snow water equivalent and snow depth for the period from 09 October 2014 through 10 October 2019. Prior to 09 October 2014, these cells have missing values. After 10 October 2019, those cells are modeled and contain valid data (Table 2).
Date (YYYY-MM-DD) | SWE/depth values for 126,950 affected cells |
---|---|
2014-10-08 and earlier | no-data value (-9999) |
2014-10-09 through 2019-10-10 | 0.0 |
2019-10-11 and after | valid modeled SWE/depth |
SNODAS data files are supplied to NSIDC as flat binary 16-bit signed integer big-endian grids. A header file is also supplied to NSIDC as a text file, which includes metadata. The data files can be read by user-written routines such as Fortran and C programs; off-the-shelf image processing packages such as ENVI, IDL, MATLAB, and ERDAS IMAGINE; and by GIS and other mapping packages such as GMT, GRASS, and ARC/INFO. For instructions, such as importing files into ENVI, refer to Barrett (2003).
The header files contain information to georegister grids contained in the flat binary files. They also contain information about creation and modification of each file, data type of each file, georeferencing data, maximum and minimum values, calibration/scaling information, and a time stamp for each field. Two attributes of the header file that most users will want to pay attention to are the minimum/maximum x and y axis coordinates. These are the grid cell edges that define the extents of the grid. The Benchmark y-axis coordinate in SNODAS header files does change over time.
File Size
Table 3 shows the size of the files depending on level of compression or tarring and whether they are the masked or unmasked version.
File Type | Size Range |
---|---|
Masked tar files (.tar) | 1.2 MB - 35 MB |
Masked uncompressed data files (.dat) | 46.5 MB |
Masked uncompressed header files (.txt) | 4 KB |
Unmasked tar files (.tar) | 3.4 MB - 102 MB |
Unmasked uncompressed data files (.dat) | 67.1 MB |
Unmasked uncompressed header files (.txt) | 4 KB |
The masked and unmasked data files are organized on the FTP site in separate directories labeled masked
and unmasked
. Within these two directories are subdirectories labeled by a 4-digit year. Within the year directories, there are subdirectories for the months of the year of the form MM_mon
where MM
is the two-digit month number and mon
is the three-character month abbreviation. Each month directory contains the tarred archive file, usually one for each day of the month. See the File Naming Convention section of this document for information on the data file names. Figure 1 shows a sample of the directory structure.

This section describes the file names of the files on the FTP site. There are eight daily data files (one for each data parameter) and eight daily header files (one for each data file) that are compressed using gzip. These 16 gzipped files are packaged together into one daily tar file and placed on the FTP site.
Note that in July 2019, the filename extensions for the uncompressed header files (.hdr
) were changed to .txt
. The decision to rename the file extensions was made to reduce users confusing these files with standard .hdr
files read by software packages like ENVI or ArcGIS.
Follow the links below to see the file naming convention for that type of file:
Tarred Daily File Naming Convention (.tar)
The data are available through the FTP site as daily tar files with the following naming convention and as described in Table 4:
SNODAS_YYYYMMDD.tar
SNODAS_unmasked_YYYYMMDD.tar
Where:
Variable | Description |
---|---|
SNODAS |
Identifies this as SNODAS data |
unmasked |
Identifies this as an unmasked version of the SNODAS data |
YYYY |
4-digit year |
MM |
2-digit month |
DD |
2-digit day of month |
.tar |
Identifies that this file has been tarred |
Untarred Daily File Naming Convention (.gz)
Once you untar the daily files, you will find 16 gzipped files: Eight data files and eight header files. The gizipped files have the following naming format and are described in Table 5:
rr_mmmffppppSvvvvTttttaaaaTSyyyymmddhhIP00Z.xxx.gz
Where:
Masked Files
The masked data files represent snow cover in the contiguous United States, extending into Canada for certain drainage basins. The spatial coordinates of the area are listed below:
Southernmost Latitude: 24.9504° N
Northernmost Latitude: 52.8754° N
Westernmost Longitude: 124.7337° W
Easternmost Longitude: 66.9421° W
Figure 2 is an example of the SNODAS SWE field displayed as an image.

Unmasked Files
The unmasked data files represent snow cover in the contiguous United States, in addition to extending well into Canada as well as outlines the coast and contains parts of Mexico. The spatial coordinates of the area are listed below:
Southernmost Latitude: 24.0996° N
Northernmost Latitude: 53.9667° N
Westernmost Longitude: 130.5171° W
Easternmost Longitude: 62.2504° W
Note: This coverage is not consistent until 2013. SNODAS data coverage over eastern Canada is non-existent in 2010, extends up to 50° N in 2011 and 2012, and then goes up to the full 58.2329° N in 2013.

The grid for the masked data files is 6,935 columns by 3,351 rows; for the unmasked files, it is 8,192 columns by 4,096 rows. Grid values are 16-bit, signed integers (big-endian). The first value at (1,1) is the top-left corner of the array (NW corner in this context). The file is structured so that values are read across the rows. For example, the second value to be read would be the second column of the first row (2,1). Grid cells have a 30-arc second spacing (nominally 1 km on the ground). Model output and precipitation variables are point estimates for the center of each grid cell and not an areal estimate. However, for the purposes of hydrologic and snow cover forecasts, these point estimates are assumed to represent average conditions in each grid cell.
The x- and y- axis coordinates are listed in the header files (.txt) associated with each data file. The x-axis coordinate of the center of the upper left hand cell is given in the Benchmark x-axis coordinate, and the x-axis coordinate of the left edge of the upper left hand cell is given in the Minimum x-axis coordinate. The y-axis coordinate of the center of the upper left hand cell is given in the Benchmark y-axis coordinate, and the y-axis coordinate of the top of the upper left hand cell is given in Maximum y-axis coordinate. Note: The Benchmark y-axis coordinate in SNODAS header files changes over time.
The X-axis offset and the Y-axis offset in the header files are the distances between the origin and the center of the pixel that lays over the origin. The purpose of those parameters is to provide an easy way of confirming whether or not two grids are aligned, without regard to whether or not they occupy the same region.
Note: Please be aware of a small spatial data shift that occurs on 01 October 2013 in the SNODAS period of record. On this date, the data provider, NOHRSC, slightly changed the center coordinates of the grid to align with an integer value of latitude and longitude. The resulting shift is minute, on the order of 1/4 of a 1km grid cell. However, if this shift will cause significant impact to your study, see the tutorial How do I convert SNODAS binary files to GeoTIFF or NetCDF? for the spatial bounds for the data both before and after the 01 October 2013.
SNODAS fields are grids of point estimates of snow cover in latitude/longitude coordinates with the horizontal datum WGS 84. Estimates of SWE and snow depth, as well as other parameters, have no real areal extent. Therefore, projecting SNODAS output to a particular projection may not be necessary. Moreover, different users prefer different projections. For example, federal agencies are likely to use the Albers Equal Area projection, while researchers may prefer an alternative such as one of the projections used for the Equal Area Scalable Earth (EASE-Grid). Refer to All About EASE-Grid for more information. Given that SNODAS outputs are essentially point estimates, the decision to project the data and choice of projection can be left to individual users.
The masked files span 30 September 2003 to the present, and the unmasked files span 09 December 2009 to the present.
NSIDC archives fields representing the model state for 06:00 Universal Time (UTC). The time 06:00 UTC was chosen because this is closest to midnight for the United States. Snow data are for 01:00 local time for the East Coast and 22:00 for the West Coast. SWE, snow depth, and snow pack average temperature represent the state of the snow pack at 06:00 UTC. Snow melt runoff, sublimation and evaporation, and precipitation parameters that describe sources and sinks of snow pack water are integrated for the previous 24 hours, giving daily totals. Note that output for 06:00 UTC is a best estimate of snow pack characteristics. Because SNODAS only updates snow fields once a day, 18 out of 24 time steps in each day's model run do not use observations to update model estimates. Therefore, hourly data from SNODAS is model output only and does not represent the best possible estimate of the snow pack.
Near-Real-Time Data
NSIDC has scripts that run several times a day. If new SNODAS files are found, these scripts will automatically post them to our FTP server. If you need data sooner than the normal time frame that NSIDC uploads files to our FTP site, you can contact NOHRSC, as they distribute the data for operational users.
NSIDC has not conducted an assessment of these data. For information, please refer to Barrett (2003), or contact NOHRSC.
Data Access and Tools
The following tutorials are avaialbe for the SNODAS product:
Data Acquisition and Processing
NOHRSC supplies NSIDC with files that only have the eight variables contained in this data set. No additional processing is done at NSIDC except for the renaming of the file extension. See the References section for a list of papers on NOHRSC processing.
References and Related Publications
Contacts and Acknowledgments
Thomas Carroll (Retired)
National Weather Service
National Operational Hydrologic Remote Sensing Center
1735 Lake Drive W.
Chanhassen, MN 55317
USA
This data set and documentation were developed with the assistance of NOHRSC Director, Thomas Carroll, and NOHRSC staff as well as NSIDC's Andrew Barrett.
Development and distribution of the data set from NSIDC is supported by funding from NOAA's National Environmental Satellite, Data, and Information Service (NESDIS) and the National Centers for Environmental Information (NCEI).
Document Information
Document Authors
This documentation was written by Keri Webster and Florence Fetterer and is based on the publication National Operational Hydrologic Remote Sensing Center Snow Data Assimilation System (SNODAS) Products at NSIDC.
Creation Date
November 2004
Revision Date
11 Febrary 2020: A. Windnagel updated the Data Description section describing the persistent zero issue in SWE/depth data.
01 July 2019: A. Windnagel updated the documentation to reflect the name change to the header file from .hdr to .txt.
04 January 2019: A. Windnagel updated the documentation with a list of tools to manipulate SNODAS data, updated the Overview section, and updated the Detailed Data Description section to note that the RUP model is used now instead of the RUC2 model.
23 February 2012: A. Windnagel updated the documentation to describe the unmasked files that are now available.
27 June 2011: A. Windnagel updated the File and Directory Structure and File Naming Convention sections to describe the FTP site structure. Also removed the Opening FTP .tar.gz Files with WinZip section since the files are not tarred and gzipped the same way anymore.
05 February 2010: A. Windnagel removed all references to the GISMO subsetting interface because it is being decommissioned.
15 January 2010: A. Windnagel added an SSI about the new Beta Advanced Data Search interface.
18 June 2009: A. Windnagel updated the Grid Description section with information on the order of the array.
14 May 2009: A. Windnagel updated the File Naming Convention section that was missing some information, added information on opening the .tar.gz files with WinZip, added information on obtaining near-real-time data, and added a glossary.
07 August 2008: D. Miller updated guide doc with edits from Florence Fetterer and Andy Barrett.
01 April 2008: D. Miller reformatted and reorganized the guide documentation based on comments from User Services (Kara Gergely) to make the guide documentation easier to use. USO was receiving a lot of questions about this data set.
02 February 2007: F. Fetterer made the following changes: Added link to the Bureau of Reclamation and WWA Web sites, added units and product code to table, added information on using GISMO formerly found in the FAQ.
02 February 2006: F. Fetterer added links to a Frequently Asked Questions page authored by L. Ballagh.
22 December 2005: F. Fetterer added text describing the renaming of .grz files at NSIDC. Renaming was instituted in December 2005 for the following reasons: 1) .grz is not a standard data type or file extension, 2) The compression and storage of the files is accomplished by tarring each set and then compressing them using the gzip compression program. This has several recognized file name extensions, but the most prevalent is .tar.gz. Changing the extension to this more recognized format will help alleviate user confusion while at the same time not altering the actual distributed data files contained within the tarred file.
19 December 2005: F. Fetterer added text advising users needing data on an operational basis to contact NOHRSC.
09 May 2005: F. Fetterer added information on subsetting options.