| Abstract |
- NASA’s ICESat-2 satellite, which launched in 2018, carries the Advanced Topographic Laser Altimeter System (ATLAS), a green-wavelength, photon-counting lidar. While ICESat-2’s primary mission focuses on measurement of elevation of ice sheets, glaciers, sea ice, and vegetation, ATLAS has also proven remarkably effective at measurement of bathymetry, or water depths. However, there is currently no along-track, dedicated ICESat-2 bathymetry data product. To address this need, our University of Texas at Austin (UTA) – Oregon State University (OSU) research team is currently working with the ICESat-2 Project Science Office on a new ICESat-2 Level 3 bathymetric data product, which will be called ATL24.
A key step in the ATL24 workflow involves automatically detecting returns from the seafloor and sea surface. In order to train and test algorithms for automatically detecting sea surface and seafloor returns, a preliminary step was to develop a robust training and testing database, including data from coastal locations around the globe, including varying morphological characteristics, substrate types, and cover types (e.g., seagrass, macroalgae, and coral). Our project team generated the ICESat-2 training and testing database provided through this ScholarsArchive@OSU publication. OSU graduate students, Ellery Ohlwiler, Charles Ghartey, and Ruth McCullough led the generation of the included datasets.
Areas with bathymetry were determined though OpenAltimetry. Track ID, Beam, and date were collected to download the full ATL03 Release 5 (Rel 005) data granules from the NASA National Snow and Ice Data Center (NSIDC) database. ICEVis, a UT-Austin proprietary software developed by Matthew Howell, was used for classifications. This project specifically focused on bathymetric bottom and water surface. H5 files were then converted to the final CSV format, where the latitude, longitude and label for each photon are stored. The classification scheme used is consistent with up-to-date ASPRS standards. The bathymetric bottom is labeled as class 40, water surface as class 41, and the remaining points are left as zero or unclassified.
In order to use this database for training and testing algorithms for auto-classification of sea surface and bathymetric bottom photons, it is typically necessary to cross-reference individual photon returns in two different data sets (e.g., the test data set and the reference data set). This can be achieved using the photon indices. The labeled H5 files have a key that corresponds to the groundtrack, and within that key, there are two arrays: photon_index and photon label. To obtain the relevant photon parameters such as latitude, longitude, elevation, etc., the labeled H5 file must be used in conjunction with the original granule. This can be done by reading the whole source granule and matching up photons from that with the photon labels from the labeled H5 file based on the photon index.
Each folder included in this dataset corresponds to one site (i.e., one geographic location). Within each folder is an image of the photon returns, and a satellite photo of the beam track locations. The additional files include the full beam H5 granules, the ICEVis labeled H5 granules, and the CSV file. The supplemental documents include a coastal classification as well as information about each site. The coastal classification scheme used in this work was developed with input from Larry Ward, a coastal geomorphologist at the Center for Coastal and Ocean Mapping (CCOM) Joint Hydrographic Center (JHC) at the University of New Hampshire.
While we developed this database for our work on ATL24, we are making it available through this ScholarsArchive publication in hopes that the datasets will benefit other researchers, including those interested in developing and testing their own algorithms for auto-segmentation of seafloor and sea surface returns. These datasets are provided “as is” and no claims are made regarding their suitability for any particular purpose.
Please note that refraction correction has not been performed on these datasets. The reason is that these data are intended for use in training and testing algorithms for automatically detecting seafloor and sea surface returns in ATL03 datasets—steps which are performed before refraction correction. For users who would like to apply refraction correction, Python and MATLAB scripts are available on this GitHub repository: https://github.com/ICESat2-Bathymetry/Information
Note: The CSV files have orthometric heights listed as EGM08. These have not been converted from the original granule format of WGS84. The team is working on a solution to make the labels and values consistent.
|
|---|
