Data Description

Parameter

Each data file contains gridded Snow Water Equivalent (SWE) estimates for the Northern Hemisphere (SWE_NorthernDaily) and Southern Hemisphere (SWE_SouthernDaily). The SWE values have a scale factor of 1 for the Northern Hemisphere and a scale factor of 2 for the Southern Hemisphere. This data set also includes an ancillary quality assurance text file that provides summary statistics for the values included in this field.

Valid parameter values include:

• 0-240: SWE values in millimeters (mm)
• 247: Incorrect spacecraft attitude
• 248: Off-earth
• 252: Land/Snow impossible
• 253: Ice
• 254: Water
• 255: Missing or out-of-bounds data

Each data file also includes a gridded parameter flag field for the Northern Hemisphere (Flags_NorthernDaily) and Southern Hemisphere (Flags_SouthernDaily). The values included in this field are the same as the values in the SWE parameter field with the exception of value 241: Snow Possible. The Snow Possible flag shows all grid cells containing values within the valid SWE data range; 0 and 240 mm.

Valid flag values include:

• 241: Snow possible
• 247: Incorrect spacecraft attitude
• 248: Off-earth
• 252: Land/Snow Impossible
• 253: Ice
• 254: Water
• 255: Missing or out-of-bounds data

File Information

Format

Data are provided in HDF-EOS5 (.he5) format and are stored as 8-bit unsigned integers. For software and more information, visit the HDF Group website.

File Contents

As shown in figure 1, each data file includes two data fields for the Northern Hemisphere and Southern Hemisphere, and two metadata fields (CoreMetadata and StructMetadata).

Ancillary Data

There are two ancillary text files (.qa and .ph) and one ancillary .xml file included with each day of data. The .qa text file provides summary quality statistics for each data parameter. The .ph text file provides a list of the input data files. The .xml file provides file level metadata.

Naming Convention

Files are named according to the following convention and as described in Table 1.

File Name Convention
AMSR_U2_L3_DailySnow_X##_yyyymmdd.he5

File Name example
AMSR_U2_L3_DailySnow_B01_20180625.he5

Spatial Information

Coverage

The total coverage for this product is the Northern and Southern Hemisphere, from 89.24° S to 89.24° N and from 180° E to 180 W. 

Projection

Data are provided in Northern Hemisphere and Southern Hemisphere EASE-Grid projections. For details, please see the NSIDC EASE-Grid: Projection web page.

Grid Description

Grids are 721 rows x 721 columns. For details about the EASE-Grid projections, related products, and tools, see the NSIDC EASE-Grids Overview website.

Spatial Resolution

The nominal spatial resolution is 25 km.

Geolocation

Tables 3 and 4 below provide projection and grid details for this data set.

Geolocation Tools

For this EASE-Grid product, the tar files Nl_geolocation.tar and Sl_geolocation.tar contain geolocation tools. These tools include map projection parameters (.mpp files), grid parameter definitions (.gpd files), latitude/longitude binary files, and conversion software such as C, FORTRAN (FORmula TRANslation), and IDL (Interactive Data Language). These tar files are available via FTP.

Land Masks

A 25 km Northern Hemisphere land mask called amsr_gsfc_25n.hdf and a 25 km Southern Hemisphere land mask called amsr_nic_25s.hdf are available for use with this product. These masks are available via FTP.

Temporal Information

Coverage

The temporal coverage of this data set extends from 01 Dec 2012 to the present.

Resolution

Daily

Sample Data Image

Data Acquisition and Processing

Background

This AMSR Unified SWE product is produced using data collected by the Advanced Microwave Scanning Radiometer (AMSR-E) on board the Aqua satellite from June 2002 to October 2011, and from the Advanced Microwave Scanning Radiometer 2 (AMSR2) on board the JAXA GCOM-W1 satellite from July 2012 to the present. The purpose of the AMSR Unified Data Set is to provide the science community with intercalibrated climate products from both the AMSR-E and the AMSR2 instruments. To accomplish this task, JAXA is providing AMSR-E L1R Brightness Temperatures (T_bs), equivalent in content to the AMSR2 L1R T_bs.

The Level-1R input data consist of resampled T_bs. The T_b sensor footprints (instantaneous fields of view) vary with frequency. Resampling remaps the T_bs to sets of consistent footprint sizes using a Backus-Gilbert method. Each resampled set corresponds to the footprint of one frequency and contains that frequency plus all higher-resolution frequencies. Therefore, the number of channels in each resampled set of T_bs varies. See JAXA Level 1R documentation or Maeda et al. (2016) for more information.

Derivation Techniques for SWE Grids

SWE grids are derived from JAXA AMSR2 L1R resampled T_bs. These observations are equivalent to JAXA AMSR-E L1R T_bs, thus enabling consistent T_b measurements to be used across the JAXA AMSR2 and JAXA AMSR-E SWE products. The use of swath T_bs instead of averaged T_bs is important because atmospheric influence on T_bs is nonlinear, and the use of averaged T_bs would dilute the atmospheric signal.

Algorithms

The current operational SWE algorithm, also known as the Columbia AMSR Snow Water Equivalent (CASWE) algorithm is being used to estimate SWE data for the Northern Hemisphere. The AMSR-E Snow Depth Algorithm described in Kelly (2009) is being used to estimate SWE data for the Southern Hemisphere.

SWE Processing for the Northern Hemisphere

The operational SWE algorithm (Tedesco-Jeyaratnam, 2016) utilizes climatological data, an electromagnetic model; also known as a snow emission model, and artificial neural networks for estimating snow depth as well as a spatial-temporal dynamic density scheme to convert snow depth to SWE. A summary of the processing steps is provided below.

1. Create a training dataset for use with Artificial Neural Networks (ANNs)
A training dataset is required for tuning the Artificial Neural Network (ANN). The training dataset is obtained using the inputs and outputs of the TKK electromagnetic model (Pulliainen, 2006) and consists of simulated T_bs and the corresponding snow depth, snow density, and near-surface temperature inputs. Training is performed using a back propagation algorithm (Tedesco, 2004).

2. Estimate snow grain size using Artificial Neural Networks (ANNs)
Snow grain size estimates are obtained from two Artificial Neural Networks (ANNs) trained with the TKK electromagnetic model; one ANN for grain size estimates using 36.5 GHz T_b values and a second ANN for grain size estimates using both 18.7 and 36.5 GHz T_b values. Two different grain size values are used to account for the different penetration depths of the microwave frequencies within the snowpack and the vertical distribution.

The ANN used here is a feed forward network (Haykin, 1999) with one hidden layer containing four neurons. This optimal ANN architecture was chosen by comparing the performance of different architectures using the root mean square error (RMSE) between measured and estimated surface T_bs at 18.7 GHz and 36.5 GHz, using both horizontal and vertical polarization channels.

3. Compute retrieval coefficients
Retrieval coefficients relate T_b at different frequencies to snow depth and are computed using estimates of the effective grain size obtained from the ANNs. The retrieval coefficient values are calculated using the equations specified in section 3.2.1 of Tedesco-Jeyaratnam (2016).

4. Convert snow depth to SWE
Snow depth estimates from retrieval coefficients are converted to SWE using temporally and spatially varying snow density maps (Sturm et al, 2010).

5. Create SWE grids
After the algorithm is run on the AMSR2 JAXA L1R T_b data, daily SWE estimates are mapped to a 25 km EASE grid. The gridding is performed using the drop-in-the-bucket process.

SWE Processing for the Southern Hemisphere

The AMSR-E Snow Depth Algorithm estimates SWE using a three-step process: first, the presence of snow is detected; second, the depth of snow present is estimated; and third, snow depth estimates are converted to SWE. This is achieved through the construction of a mean January to March global snow density map, which was created using the Canadian data of Brown and Brazen (1998), and the Russian snow survey data of Krenke (1998), and interpolated to the seasonal snow classifications system of Sturm et al. (1995). SWE is then the product of depth and density. Both SWE and depth processing streams require calibrated T_b measurements at 10 GHz, 18 GHz, 23 GHz, and 89 GHz, as well as ancillary land cover data. See Kelly (2009) for a detailed description of the AMSR-E snow depth algorithm. 

Assessment

Each HDF-EOS5 data file contains core metadata with Quality Assessment (QA) metadata flags that are set by the operational processing code run by the AMSR Science Investigator-led Processing System (SIPS) prior to delivery to NSIDC. A separate metadata file in .xml format is also delivered to NSIDC with the HDF-EOS5 file.

This file contains the same quality assessment (QA) metadata flags as the core metadata contained in the HDF-EOS5 file. Three levels of QA are applied to AMSR2 files: automatic, operational, and science QA. If a file/granule passes automatic QA and operational QA, the file is forwarded to NSIDC for archive and distribution. If not, the issue is resolved and the file is reprocessed. Science QA is performed automatically during nominal processing, but only reviewed (closely, after-the-fact) in conjunction with questions that arise after processing is complete. The three QA stages are described in nore detail below.

Automatic QA

Out-of-bounds L1R T_bs are screened out before T_bs are interpolated to the 25 km grid.

Operational QA

AMSR2 L1R data are subject to operational QA by JAXA prior to arriving at the AMSR SIPS for processing to higher level products. Operational QA varies by product, but it typically checks for the following criteria in a given file (Conway 2002):

• File is correctly named and sized
• File contains all expected elements
• File is in the expected format
• Required EOS fields of time, latitude, and longitude are present and populated
• Structural metadata are correct and complete
• File is not a duplicate
• HDF-EOS5 version number is provided in the global attributes
• Correct number of input files were available and processed

Science QA

In the SIPS environment, as part of the processing code, the science QA includes checking the maximum and minimum variable values, and the percentage of missing data and out-of-bounds data per field. At the Science Computing Facility (SCF), co-located with the SIPS, post-processing science QA involves reviewing the operational QA files and browse images, and performing the following additional QA procedures (Conway 2002):

• Comparisons with historical data
• Detection of errors in geolocation
• Verification of calibration data
• Detection of trends in calibration data
• Detection of large scatter among data points that should be consistent.

Several tools have been developed to aid in the QA process of the Level 3 AMSR2 products.  The AMSR SIPS provides software that creates a QA browse image in Portable Network Graphics (.png) format that can be used for visual QA.  The team lead SCF (TLSCF) provides metadata and QA software specific to each product; these software generate the metadata files discussed above and a QA summary report in text format. The products of these tools are provided to NSIDC along with each data granule.

Anomalies

Refer to the AMSR2 LANCE Anomalies Page web page for information regarding data anomalies or gaps in coverage.

Instrument Description

For a detailed description of the AMSR2 instrument, refer to the AMSR2 Channel Specification and Products web page.

Software and Tools

For tools that work with HDF-EOS data, see the HDF-EOS web page and the NSIDC HDF-EOS web page. This data set is compatible with ArcGIS.

Contacts and Acknowledgments

Marco Tedesco
Lamont-Doherty Earth Observatory of Columbia University
New York, USA 

Jeyavinoth Jeyaratnam
The City College of New York
New York, USA 

References

Tedesco, Marco, and Jeyavinoth Jeyaratnam. 2016. A New Operational Snow Retrieval Algorithm Applied to Historical AMSR-E Brightness Temperatures (ATBD). Remote Sensing, 8(12) 1037. (PDF)

Pulliainen, J. 2006. Mapping of snow water equivalent and snow depth in boreal and sub-arctic zones by assimilating space-borne microwave radiometer data and ground-based observations. Remote Sensing Environment, (101) 257–269.

Maeda, T., Taniguchi, Y., and K. Imaoka. 2016. GCOM-W1 AMSR2 Level 1R Product: Dataset of Brightness Temperature Modified Using the Antenna Pattern Matching Technique. IEEE Transactions on Geoscience and Remote Sensing, 54(2) 770-782. https://doi.org/10.1109/TGRS.2015.2465170.

Kelly, Richard. (2009). The AMSR-E Snow Depth Algorithm: Description and Initial Results. Journal of The Remote Sensing Society of Japan, (29) 307-17. https://doi.org/10.11440/rssj.29.307.

Tedesco, et al. 2004. Artificial neural network-based techniques for the retrieval of SWE and snow depth from SSM/I data. Remote Sensing of Environment, 90(1) 76-85.

Haykin, S. 1999. Neural networks: A Comprehensive Foundation. Prentice Hall: Upper Saddle River, NJ, USA.

Backus, G. E. and J. F. Gilbert. 1967. Numerical Applications of a Formalism for Geophysical Inverse Problems. Geophysical Journal International 13(1-3), 247–276. https://doi.org/10.1111/j.1365-246X.1967.tb02159.x.

Brown, R. D. and R. O. Braaten. 1998. Spatial and temporal variability of Canadian monthly snow depths. Atmosphere-Ocean, 36(1), 37-54.

Krenke, A. 1998, updated 2004. Former Soviet Union Hydrological Snow Surveys, 1966-1996. Edited by National Snow and Ice Data Center. Boulder, Colorado USA: National Snow and Ice Data Center. http://dx.doi.org/10.7265/N58C9T60.