GLAS HDF5 products are in Hierarchical Data Format 5 format, also known as netCDF-4/HDF5. The intent with the HDF5 format data is to make the Release 33 data available in the same format as several future NASA missions, including ICESat 2 and Soil Moisture Active Passive (SMAP), and to make ICESat 1 Release 33 data more accessible to a broader user community.

Header

The header description information found in the binary files is embedded in the HDF5 files.

Invalid values

Not all data from GLAS are suitable for science processing. Many parameters have invalid values or, in CF terms, "_FillValues". An invalid value means that the measurement is not valid for that element of data.

Each data granule has associated browse products that users can quickly view to determine the general quality of the data in the granule. Browse products consist of image plots of key parameters and statistics.

The file naming convention is as follows:

GLAHxx_mmm_prkk_ccc_tttt_s_nn_ffff.h5

Where:

Algorithms that generate atmospheric products were continually improved as limitations became apparent in early versions of data. Release 33 is the final release of ICESat/GLAS data.

Note: Beginning with Release-28, a new convention is used for the release number (mmm) in file names.

Please see the following for more information:

• ICESat/GLAS YXX Release Numbers
• ICESAT/GLAS CSR SCF Release Notes for Orbit and Attitude Determination (PDF file)

Table 4 lists approximate file sizes for each product.

Total volume of the HDF5 Atmospheric data products is approximately 3 TB.

GLAS/ICESat coverage is global between 86°N and 86°S with occasional off-nadir pointing to the poles or other targets of opportunity. GLA02 and GLA07 files span two orbits. GLA08-11 files span 14 orbits (1 day).

Spatial searching is disabled for GLA02. The orbit for GLA02 is a predicted orbit and does not show any target-of-opportunity pointing. Spatial searching for GLA02 is not enabled in Reverb because of a potentially large number of false negatives. If you order GLA05 to GLA15 data granules (which support spatial searching) and you require matching GLA02 granules, please note the data times and/or file names from the GLA05 to GLA15 granules; use these times to perform your search for GLA02 granules.

The atmospheric channel of GLAS measures the vertical structure of backscatter intensity (with a laser footprint of 60 m), from -1.5 km to a height of about 41 km, with 76.8 m vertical resolution. Horizontal resolution of the raw data is a function of height. Between 10.5 km and 20.5 km, eight laser shots are summed, producing a horizontal resolution of 1.4 km (5 Hz sampling rate). Between 20.5 km and 41 km, 40 shots are summed, providing a horizontal resolution of about 7.5 km (1 Hz sampling rate).

Please refer to the Data Release Schedule for the temporal coverage of specific products and releases.

The ICESat/GLAS Data Releases page describes characteristics of each release.

Also visit the official ICESat Web site (follow the "Time-Track Conversion" link) to see the Pass ID for a user-specified year, day, and time.

Atmospheric data are sampled 40 times per second, but not all products record data at this rate. Sampling rates include 4 seconds, 1 second, 5 Hz, or 40 Hz, depending on the product.

Please see the following tables of data records for each product:

• GLAH02 Records
• GLAH07 Records
• GLAH08 Records
• GLAH09 Records
• GLAH10 Records
• GLAH11 Records

See the GLAS Atmospheric Products User Guide (PDF, 213 KB) for details of working with the GLAS atmospheric parameters.

See Ordering ICESat/GLAS Products from NSIDC for a list of order options.

The following external links provide access to software for reading and viewing HDF5 data files. Please be sure to review instructions on installing and running the programs.

HDFView: Visual tool for browsing and editing HDF4 and HDF5 files.

HDF Explorer: Data visualization program that reads Hierarchical Data Format files (HDF, HDF-EOS and HDF5) and also netCDF data files.

Panoply netCDF, HDF and GRIB Data Viewer: Cross-platform application. Plots geo-gridded arrays from netCDF, HDF and GRIB datasets.

For additional tools, see the HDF-EOS Tools and Information Center.

For tools related to the original binary GLAS data, see Tools for working with ICESat/GLAS Data.

GLAS atmospheric data are prone to the following errors:

Normalized LIDAR signal

• Limited knowledge of the laser energy and performance
• Inaccurate dead time (photon coincidence) correction factors for the 532 nm channel, and digital-to-analog conversion factors for the 1064 nm channel
• Boresite inaccuracy
• How well the etalon filter is tuned to the laser frequency (532 nm channel)

Attenuated backscatter

• Presence of aerosol or clouds in the portion of the atmosphere used to calculate the calibration constant
• Error involved with computing the molecular backscatter cross section at the calibration height (errors in knowledge of atmospheric temperature and pressure)

Cloud layer height and Earth surface height

• Multiple scattering may cause the cloud bottom to appear somewhat lower than it actually is
• Low signal-to-noise ratio may cause false positives

PBL and elevated aerosol layer height

• Low signal-to-noise ratio may cause false positives
• Accuracy of satellite altitude and time of laser fire in the absence of a detectable ground return
• Sampling frequency (bandwidth, which determines the vertical resolution of the data)
• Number of LIDAR shots averaged together (horizontal resolution)
• Optical depth of the PBL

Optical depth

• Error in optical depth increases with layers of decreasing geometrical depth

Note that Level-0 data from 19 November to 14 December 2003 have a time stamp error, where dates are reset to 01 January 2000. This passes through to all higher-level products.

Browse products contain quality information. These browse products are available along with corresponding data granules.

Normalized LIDAR signal

The most practical way to measure data quality is to integrate the entire LIDAR signal from 20 km to the end of the profile and compute the mean signal and standard deviation. Both should fall within known limits if the data are good. A histogram of the LIDAR return is also constructed. The degree to which the histogram deviates from a Poisson distribution indicates how much of the signal is contained in the return. The "i_g_TxNrg_qf" quality flag is set for each 40 Hz shot to characterize the laser energy:

1 = full laser energy (within 90% of expected maximum value)
2 = marginal laser energy (between 70-90% of expected maximum value)
3 = deficient laser energy (less than 70% of expected maximum value)

Boresite accuracy is assessed by integrating the 532 nm return signal from 20 km to 41 km. The expected number of integrated photons (I_s) in this region is about 800. A quality flag is set, depending on the magnitude of the integrated return:

1 = excellent signal strength (I_s > 800)
2 = good signal strength (600 < I_s < 800)
3 = marginal signal strength (400 < I_s < 600)
4 = poor signal strength (200 < I_s < 400)
5 = bad data (I_s < 200)

Attenuated backscatter

Confidence flags include a measure of the variability of the calibration constant (for both channels) as a function of time, and the quality of the attenuated backscatter profile. The laser energy flag and the integrated return flag should be evaluated to eliminate bad shots. Quality of the calibration constant is assessed by evaluating its variability with time and the difference between the constants calculated at two different heights. The 532 nm attenuated backscatter cross-section profiles are checked by normalizing them by the attenuated molecular profile. This should produce a profile that ranges between 0.9 and about 10.0. This test could only be applied to data with a ground return, as the values below thick clouds will approach zero. Another test is to integrate the attenuated backscatter from 20 km to 40 km and divide by the integrated attenuated molecular backscatter to form a ratio very close to unity. A major deviation from "1" would indicate a problem (Palm et al. 2002).

Cloud layer height and Earth surface height

Quality of this product is judged by how successful it is at finding all detectable cloud layers and locating their boundaries in the atmospheric profile. The best approach is to plot the computed cloud boundaries on top of image segments constructed from LIDAR profiles. These should reveal systematic and random faults in the results of the procedure. If shortcomings exist, parameters used in the computation of thresholds will be adjusted to fix the discrepancies.

The following confidence tests are used for the layer boundary results of each profile:

• For each layer detected, a flag to indicate high, medium, or low confidence
• For the top and bottom of each layer, a single number indicating a number of sample bins within which the boundary exists at a specified probability
• For each profile, a single number representing the probability that an undetected layer exists
• For a positive ground signal, a flag to indicate a high, medium, or low confidence
• For a positive ground signal, the number of sample bins within which the actual ground height exists at a specified probability
• For negative ground signal detection, a probability that a detectable ground signal actually exists but the algorithm fails to calculate a ground signal

PBL and elevated aerosol layer height

Confidence of height determination is measured by the difference between the average signal levels outside and inside of the PBL. Confidence is also measured by the standard deviation of the heights for a given segment.

Optical depth

• Layers will be screened so they don't overlap or become embedded.
• Visual screening with imagery will occur to ensure layers are labeled "cloud" or "aerosol" or "polar stratospheric cloud" correctly.
• As transmission profiles are processed, transmission calculations will be tested for out-of-bounds sitations such as increasing transmission with range or large negative transmission.
• Confidence flags will be produced for each particulate layer or profile to determine the number and type of suspect input parameters, and whether the transmission profiles passed their tests. This information will be passed to each of the output parameter confidence flags.

A formal calibration-validation (CV) plan is pending. More information will be available in the near future.

5. References and Related Publications

Related Data Collections

Related Web Sites

6. Document Information

Acronyms and Abbreviations

The acronyms used in this document are listed in Table 5.

Document Creation Date

31 December 2012

Document Revision Date

12 April 2013

Document URL

http://nsidc.org/data/docs/daac/glas_icesat_l1_l2_global_atmosphere-hdf5-gd.html