DOWNLOADING DATA VIA HTTPS
To learn more about Earthdata Login and register for an account, please see How to Register with Earthdata Login. Once you have logged in, data can be downloaded via a Web browser, command line, or client. For help with downloading data, please see Options Available for Bulk Downloading Data from HTTPS with Earthdata Login.
NOTE: Reverb will be decommissioned in the coming months and replaced with Earthdata Search. All links to Reverb will be removed at that time.
IceBridge DMS L3 Photogrammetric DEM, Version 1
The IceBridge DMS L3 Photogrammetric DEM (IODMS3) data set contains gridded digital elevation models and orthorectified images of Greenland and Antarctica derived from the Digital Mapping System (DMS). The data were collected as part of Operation IceBridge funded campaigns.
|Data Contributor(s):||Ryan Dotson, John Arvesen|
|Metadata XML:||View Metadata Record|
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.Dotson, R. and J. Arvesen. 2012, updated 2014. IceBridge DMS L3 Photogrammetric DEM, Version 1. [Indicate subset used]. Boulder, Colorado USA. NASA National Snow and Ice Data Center Distributed Active Archive Center. doi: http://dx.doi.org/10.5067/39YO5T544XCC. [Date Accessed].
Detailed Data Description
The DEMs are provided as 40 cm resolution gridded, uncompressed 32-bit floating point GeoTIFF files. For a few 23 March 2011 lines, a 20 cm grid was used because of a lower flight altitude.
NOTE: Currently, the TIFF Sample Format tag is not embedded in the images. This results in client programs either auto-detecting the pixel format, or assuming a default. It has been discovered that some programs appear to assume a 32 bit unsigned integer format, while other programs auto-detect the correct 32 bit floating point. In the near future, TIFF headers will be updated to include the TIFF Sample Format tag indicating 32 bit floating point.
Orthorectified color images are provided as 10 to 20 cm gridded compressed 8-bit RGB GeoJPEG files.
XML files contain file level metadata and location, platform, and campaign information.
Data are available on the FTP site in the
https://n5eil01u.ecs.nsidc.org/ICEBRIDGE/IODMS3.001/ directory. Within this directory, the folders are organized by date, for example
Folders contain GeoTIFF, GeoJPG and XML files.
The files are named using the conventions shown below. File name variables are described in Table 1:
Example for a single DMS frame:
||Short name for IceBridge DMS L3 Photogrammetric DEM|
||Four-digit year of survey|
||Two-digit month of survey|
||Two-digit day of survey|
||Hour (00 - 23)|
||Hundredths of Second|
||Sequence number from DMS camera|
||File content type.
DEM: gridded elevation data
RGB: orthorectified visible imagery
.tif = GeoTIFF file
.tfw = TIFF world file
.jpg = JPEG image file
.jgw = JPEG world file
GeoTIFF DEM files range from approximately 4 MB to 8 MB each.
JPEG image files range from approximately 2 MB to 10 MB each.
The entire data set is approximately 9.2 GB.
Spatial coverage for the IceBridge DMS L3 Photogrammetric DEM campaigns include the Arctic and Antarctica from the 2011 to 2013 campaigns.
Arctic / Greenland:
Southernmost Latitude: 60° N
Northernmost Latitude: 90° N
Westernmost Longitude: 180° W
Easternmost Longitude: 180° E
Southernmost Latitude: 90°S
Northernmost Latitude: 60°S
Westernmost Longitude: 180° W
Easternmost Longitude: 180° E
Resolution and coverage of DEMs and orthorectified images depends on the flight elevation and source raw image resolution.
The DEMs are 40 cm resolution grids. For a few 23 March 2011 lines, a 20 cm grid was used because of a lower flight altitude.
Orthorectified images are 10 to 20 cm. For a few lines, the resolution is 5 cm, but in general it is 10 cm for Sea Ice flights.
Projection and Grid Description
Polar Stereographic North Pole, WGS-84
Polar Stereographic South Pole, WGS-84
The projection for all files are in EPSG 3413 for the northern hemisphere and EPSG 3031 for Antarctica.
These data were collected from 18 March 2011 to the present as part of NASA Operation IceBridge funded campaigns.
IceBridge campaigns are conducted on an annual repeating basis. Arctic and Greenland campaigns are conducted during March, April, and May, and Antarctic campaigns are conducted during October and November.
The IceBridge DMS L3 Photogrammetric DEMs are GeoTIFF imagery, in meters and above the WGS-84 ellipsoid.
Sample Data Record
The figures below display the
IODMS3_20130410_14474149_06610_DEM.tif GeoTIFF DEM file, and the
IODMS3_20130410_14474149_06610_RGB.jpg JPEG RGB file.
The listing below shows the information embedded in the GeoTIFF DEM image file in Figure 1. GeoTIFF file information can be displayed using tools such as gdalinfo from the Geospatial Data Abstraction Library (GDAL). Visit the GDAL Web site for more information.
Size is 1070, 1338
Coordinate System is:
Origin = (-114047.830829568890000,-2143173.486468893000000)
Pixel Size = (0.429558153786877,-0.429558153786877)
Image Structure Metadata:
Upper Left ( -114047.831,-2143173.486)
Lower Left ( -114047.831,-2143748.235)
Upper Right ( -113588.204,-2143173.486)
Lower Right ( -113588.204,-2143748.235)
Center ( -113818.017,-2143460.861)
Band 1 Block=1070x1 Type=UInt32, ColorInterp=Gray
The IODMS3_20130410_14474149_06610_DEM.tfw TIFF World file contains the following:
Software and Tools
Software that recognizes the GeoTIFF file format is recommended for these images. See libGeoTIFF.
Along with coordinate and projection information, additional metadata is embedded in the GeoTIFF files. NSIDC provides a MATLAB GeoTIFF reader that reads time and position from the GeoTIFF data files. The additional fields can be extracted using the gdalinfo command line utility available from the Geospatial Data Abstraction Library (GDAL). Visit the GDAL Web site for more information.
Data Acquisition and Processing
The DMS DEM production process consists of several steps and uses a combination of commercial and custom developed software. Due to limitations of the DMS camera and limitations of the IceBridge flight plans, the photogrammetric DEMs were adjusted to an external control data set. We used IceBridge ATM L1B Elevation and Return Strength QFIT and HDF5 data for this purpose.
The DEMs are first generated by purely photogrammetric techniques using key point matching, bundle adjustment, and dense stereo depth map generation. Agisoft PhotoscanTM is used to do this processing. The result of the Agisoft PhotoscanTM processing is a depth map at a 1:4 resolution ratio from the original imagery.
A rigid Helmert transformation (7 parameters: 1 scale, 3 rotational, 3 translational) is then computed for each depth map to match the ATM point cloud. After alignment, the depth map is projected to a polar stereographic projection and the final DEM and orthorectified images are produced.
A quality review is performed on the final frames to identify processing failures.
The altimetry measurements are computed by standard photogrammetric processing techniques with some necessary changes based on the DMS sensor and flight profiles used. The internal algorithmic details used by the Photoscan software package are considered proprietary and are not published by Agisoft. At a high level, the Photoscan processing steps consist of:
- Determine an approximate location for each image
- Match feature points between images (key feature points)
- Compute refined locations and orientations for each image
- Dense stereo matching and filtering
- Generation of triangulated mesh based on dense matches
- Projection of mesh to a depth map for each frame
Two fundamental challenges with the DMS raw imagery required modifications to this typical DEM generation workflow:
- Pre-deployment calibration reports available for the DMS camera were not usable for this processing workflow. The investigators determined that several factors resulted in variation in the actual camera characteristics from the calibration report. Specifically:
- Many data sets had changes to the camera aperture to account for scene brightness variations. Changes in aperture will affect the geometric lens distortion model as the portion of the lens used varies.
- Temperature and vibration variation throughout the flight impacted the effective focal length of the camera, which is a key photogrammetric parameter determining scale and 3D location.
- Focus changes to the camera impact effective focal length and lens distortion. Focus adjustments are required for the DMS camera to acquire high quality imagery, but can result in changes to the effective focal length.
- Lack of control data and crossings.
- IceBridge flights do not normally contain adjacent flight line grids or crossings, which are important for photogrammetric workflows because they lend numerical stability to the bundle adjustment processing. Single, long flight lines processed in Photoscan and other available photogrammetric packages result in slowly varying tilts and vertical shifts of the DEMs along the flight line. The shifts can range from several decimeters to several meters.
As a result of these limitations, raw output DEMs from the Photoscan software package have vertical and horizontal errors on the order of tens to hundreds of centimeters. In most cases, there are significant "tilts" in the output DEM surfaces which cannot be calibrated out due to the lack of adjacent flight lines.
Thus, a correction/calibration process is used to compute and apply a rigid 3D transformation to correct the depth maps prior to generating the final DEMs. The process is based on minimizing the difference between the DEM and the ATM LIDAR point cloud measurements on a per-frame basis.
The resulting singe-frame DEMs have a zero- or near-zero mean difference in absolute altitude from the corresponding LIDAR point cloud. The DEMs contain substantially more elevation detail (resolution) than the LIDAR data. In the event that LIDAR data is not available for a given frame, the frame is not processed.
The input data were acquired from two locations:
- Images and LIDAR measurements were acquired from NSIDC as published data sets: IceBridge DMS L0 Raw Imagery and IceBridge ATM L1B Qfit Elevation and Return Strength data.
- Additional DMS image metadata (timestamps and trajectory info) were acquired from the DMS instrument group at NASA Ames.
LIDAR Alignment Algorithm
In order to reduce the tilting and scaling issues that sometimes occur in the raw depth maps generated by Photoscan, a rigid transform is used to map these to existing ATM LIDAR data. This is done as a 4-step process:
- Identify LIDAR points contained in the image. This is done by estimating an image footprint based on the nominal lens and camera characteristics and time-synchronized location and orientation data from the DMS sensor. This typically results in a set of several thousand LIDAR points. The points are converted to an Earth-Centered, Earth-Fixed (ECEF) coordinate system.
- A scale is determined for the depth map. This is done by computing the average 3D distance from the relevant LIDAR points to the exposure location. For each LIDR point, the point is projected onto the image, determining an X-Y pixel coordinate for that point. The depth (range) for that pixel in the depth map is recorded. A single image-wide scale factor is then computed to convert the Photoscan depth map scale to the LIDAR physical scale based on the parings of LIDAR distances and depth map distances to the exposure location.
- The re-scaled points from Step 2 define a second 3D point cloud from the depth map. This depth map point cloud is used to compute a Helmert transformation for the entire image based on the corresponding LIDAR points. The Helmert transform is computed by a Singular Value Decomposition (SVD) method on the sparse point pairs (Sjöbert 2013).
- The transform and scale factor are applied to the entire depth map, resulting in a "dense" 3D point cloud in ECEF coordinates. The point cloud is converted to Polar Stereographic coordinates, and a Delaunay triangulation of the point cloud is computed. The point cloud is then rasterized based on the triangulation to the final gridded DEM.
Trajectory and Attitude Data
The trajectory and attitude data used in processing were acquired by the DMS Applanix POS/AV 510 system and provided by the DMS instrument group.
An outline of the processing steps follows:
- Correlation of raw imagery with DMS navigation data (Lat/Lon/Alt/Roll/Pitch/Heading/Time)
- Import raw frames and metadata into Photoscan software package on a per-flight basis
- Photoscan processing workflow:
- Group images into blocks of 25 frames
- Match features across image block
- Adjust image alignment (bundle adjustment)
- Dense Point Matching
- Depth Map Generation
- Export Depth Maps
- Correction processing
- Import depth maps and determine ATM data for each frame extent
- Compute rigid transformation for each frame
- Rectify and output final DEM and RGB files
- Quality review
- Automated comparison of output DEMs with LIDAR cloud including elevation offsets and Standard deviation of error.
- Manual inspection of suspect frames, rejecting those with clear problems
The underlying accuracy of the measurements can depend on a number of factors, including:
- Camera calibration and stability. This typically manifests in two forms:
- Focal length errors, which result in scale factor error in the Z-dimension of the raw frame. The LIDAR correction algorithm is designed to correct this.
- Distortion errors. In some cases, the distortion estimation routines in the Photoscan software package result in a poor lens distortion model. This can result in DEMs which are slightly concave or convex on the edges.
- GPS trajectory accuracy for DMS exposure locations. This is generally not a concern due to the correction against LIDAR data.
- Ability to match points between frames, both for sparse and dense matching steps. In some cases, especially with low-contrast images with fog or clouds, the Photoscan software may be not able to match points. This can result in a failure to produce any depth map, or can occasionally produce depth maps with some artificial features. In some cases, low contrast sea ice or sheet ice can exhibit spikes in the final DEM. JPEG compression artifacts can amplify this effect.
- Accuracy of ATM data set. Given the LIDAR correction process in use, any systematic errors in the ATM Level-1B data set will impact the photogrammetric DEMs.
- Accuracy of the alignment process for depth map correction. The correction process depends on having a reasonable estimate for initial alignment between the LIDAR point cloud and the raw image footprint. Large initial displacements could cause incorrect correction processes. In this case, the mean vertical error is typically still near zero.
DMS provides natural color or panchromatic tracking imagery from low and medium altitude research aircraft. The system configuration includes a 21 megapixel Canon EOS 5D Mark II digital camera, computer-controlled intervalometer, and an Applanix POS/AV precision orientation system. In-flight operators maximize image quality with adjustments to exposure and intervalometer settings.
References and Related Publications
Contacts and Acknowledgments
Ryan C. Dotson
Fireball International Services Corp.
Reno, NV 89519
John C. Arvesen
Cirrus Digital Systems
Tiburon, CA 94920
NASA Contract NNH11CC78C. We would like to thank Pete M. Noles, Fireball International Services Corporation for his dedication and many, many long hours processing and verifying these data sets. We would also like to thank the entire DMS instrument group at NASA Ames, and specifically Rose Dominguez, for her support and assistance gathering the DMS raw metadata and her cooperation in modifying DMS operations to help improve the quality of these data sets.
Document Creation Date
03 December 2014
Document Revision Date
26 January 2015