TanDEM-X calibrated Raw DEM generation
Introduction
TanDEM-X is a synthetic aperture radar (SAR) mission using two satellites flying in close formation with the generation of a global Digital Elevation Model (DEM) accomplishing the high accuracy HRTI-3 standards as primary objective. A 90% relative point-to-point height accuracy of 2 m is required for moderate terrain at 12 m posting. Moreover the mission has also important secondary objectives such as the generation of local DEMs following the higher standard accuracy HRTI-4, the support to applications based on Along Track Interferometry (ATI) and on PolInSAR and, more in general, to new SAR modes and techniques (Krieger et al., 2007). All of these goals, together with the complexity of the bistatic system and its synchronization, require a robust, flexible, stable and dedicated processor. The Integrated TanDEM-X Processor (ITP) is built to deal with the bistatic processing issues, providing as output a Raw DEM and the operational standard products from experimental modes (Breit et al., 2011).
The Raw DEM is the product of the interferometric chain embedded in ITP (Fritz et al., 2011): it is generated using as input the unwrapped phase, calibrated for a local factor to make it proportional to the real terrain height. Before entering in detail on the topics of the paper, the Raw DEM has to be clearly distinguished from the final calibrated TanDEM-X DEM, main output of the mission. A TanDEM-X bistatic acquisition may reach 300 s depending on the orbit and the actual terrain acquired, which means about 2000 km in along-track. A single processing over such long data take is thus operationally hardly possible and the data take is split in scenes of about 50 km along-track – the across-track extension is about 30 km – with a small overlap among them. The interferometric DEM retrieved from the single scene belonging to the bistatic acquisition is called Raw DEM. Unlike the TanDEM-X DEM, the Raw DEM could be affected by instrument phase drift or baseline errors, causing height tilts or biases (Hueso-Gonzalez et al., 2010). A calibration strategy making use of correction functions and ICESat data as Ground Control Points is implemented in the Mosaicking and Calibration Processor (MCP) (Wessel et al., 2011) for the generation of the global TanDEM-X DEM. Preliminary studies on the accuracy of the TanDEM-X DEM are described in (Gruber et al., 2012). Using an intermediate version of the TanDEM-X DEM, generated during the first coverage, an absolute accuracy below 1 m has been found for flat terrains by comparisons with GPS tracks, laser scanning DEM and ICESat points as ground truth. Thus, the absolute and relative accuracy requirements can clearly be fulfilled.
In the context of SAR interferometry, the coregistration processing step plays an important role. Due to the different viewing geometries, the active and the passive channel focused data are locally shifted in the two-dimensional slant range – azimuth domain. This local shift is proportional to the time delay between the active and the passive channel and consequently to the local height. Using this information it is in principle possible to generate DEMs following a stereo-radargrammetric technique, thus exploiting the distance transformation functions by projections on two antennas in a different spatial position (Leberl, 1990). Besides the geometrical shifts, one can use the unwrapped phase to retrieve the absolute ranging satellite-target. But whereas the geometrical shifts are absolute, i.e. reflecting the real geophysical delay, the unwrapped phase is usually referred to one point and what is missing to the absolute ranging is the so called absolute phase offset, which has to be estimated and added to the unwrapped phase before converting to height and geocoding it. Madsen (1995) firstly compared the stereo-radargrammetric and the interferometric measures for the absolute phase offset estimation with the residual delay method, foreseeing a resampling of the passive channel with a mapping matrix given by the unwrapped phase, in a way that the active and the resampled passive channels result in same delay. A discrepancy between them is an estimate of the absolute phase offset and is measured through a second coregistration. This method, tested also for the Shuttle Radar Topography Mission (SRTM) (Rabus et al., 2003), assumes that the coregistration accuracy is not sufficient for a direct comparison with the unwrapped phase. In Section 2 it is shown how the ITP coregistration algorithm retains a sufficient accuracy for the direct use of the coregistration shifts. Moreover, in the residual delay method, the phase offset estimate results biased in the presence of one or more phase unwrapping errors. In the proposed algorithm the larger portion of the DEM not affected by errors is detected and masked, providing thus a useful indicator to the TanDEM-X Ground Segment for a re-processing of scenes with a percentage of unwrapping errors extending to more than the 3% of the Raw DEM, as in TanDEM-X specifications (Krieger et al., 2007).
In Section 3 the Raw DEMs calibration within a data take is handled. As mentioned before, a TanDEM-X data take is split into several Raw DEMs with a small overlap between them. Since the calibration step implemented in MCP is data take dependent (Gruber et al., 2012), a continuity among Raw DEMs without offsets is mandatory. The phase offset calculated by ITP for each scene is used for two purposes. On the one hand, a phase offset chosen as the integer part modulo 2π of the estimate accomplishes the continuity requirement. On the other hand, the fractional part of the estimate is used as indicator of possible instrument inaccuracies.
Finally, in Section 4 exemplary results are shown.
Section snippets
Raw DEM generation
The processing steps requested for the generation of the two channels focused data are here only briefly introduced. A first fundamental operation is the synchronization pulse evaluation for the phase and timing corrections to be applied to the SAR data, followed by a bistatic focusing replica and the computation of all the focusing parameters (Bamler et al., 2007, Breit et al., 2011). This point is crucial as it defines the phase offsets and delays used in the interferometric processing chain.
Raw DEM calibration
In the first year of the operational stage of the mission, starting from December 12, 2010, over 12,000 data takes were acquired and split in more than 120,000 Raw DEMs. As outlined in Fig. 4, there is a small overlap between the Raw DEMs belonging to one data take. The Raw DEM calibration is performed following an intra-data take strategy, finalized to ensure height continuity among overlapping Raw DEMs. The global TanDEM-X DEM calibration strategy follows instead an inter-data take approach (
Processing example
An operational processing example helps to better understand the algorithms presented in Section 2. The considered acquisition was taken on April 11, 2011, over the Cordillera Central Mountains in Peru. The main parameters are outlined in Table 1.
The SAR amplitude image of the master channel is shown in Fig. 6. The scene is mountainous, composed of several peaks and valleys, presenting all the typical SAR geometric phenomena due to the side looking geometry: layover, shadow and foreshortening.
Conclusions
The generation of the TanDEM-X Raw DEM follows an optimal, operationally efficient and accurate processing chain. The focusing and the interferometric chains are embedded in one single processor, the Integrated TanDEM-X Processor, installed in the central processing and archiving facilities (PAF) in DLR Oberpfaffenhofen, Germany. A novel quality parameter stating the goodness of the overall processing is derived. A quality analysis is fundamental for a possible (re-)planning or (re-)processing
References (18)
- et al.
The Shuttle Radar Topography Mission (SRTM) – a new class of digital elevation models acquired by spaceborne radar
ISPRS Journal of Photogrammetry and Remote Sensing
(2003) - et al.
Accuracy of differential shift estimation by correlation and split-bandwidth interferometry for wideband and delta-k SAR systems
IEEE Geoscience and Remote Sensing Letters
(2005) - et al.
Processing of bistatic SAR data from quasi-stationary configurations
IEEE Transactions on Geoscience and Remote Sensing
(2007) - et al.
TerraSAR-X SAR processing and products
IEEE Transactions on Geoscience and Remote Sensing
(2010) - Breit, H., Younis, M., Balss, U., Niedermeier, A., Grigorov, C., Hueso-Gonzalez J., Krieger, G., Eineder, M., Fritz,...
- Fritz, T., Rossi, C., Yague Martinez, N., Rodriguez Gonzalez F., Lachaise, M., Breit, H., 2011. Interferometric...
- et al.
The wavenumber shift in SAR interferometry
IEEE Transactions on Geoscience and Remote Sensing
(1994) - Gruber, A., Wessel, B., Huber, M., Roth, A., 2012. Operational TanDEM-X DEM calibration and first validation results....
- et al.
Development of the TanDEM-X calibration concept: analysis of systematic errors
IEEE Transactions on Geoscience and Remote Sensing
(2010)
Cited by (154)
HoMeNL: A Homogeneity Measure-Based NonLocal Filtering Framework for Detail-Enhanced (Pol)(In)SAR Image Denoising
2023, ISPRS Journal of Photogrammetry and Remote SensingAutomatic calving front extraction from digital elevation model-derived data
2022, Remote Sensing of EnvironmentPhase unmixing of TerraSAR-X staring spotlight interferograms in building scale for PS height and deformation
2021, ISPRS Journal of Photogrammetry and Remote SensingSpatiotemporal changes of glacier and seasonal snow fluctuations over the Namcha Barwa–Gyala Peri massif using object-based classification from Landsat time series
2021, ISPRS Journal of Photogrammetry and Remote SensingQuantifying the impact of X-band InSAR penetration bias on elevation change and mass balance estimation
2024, Annals of Glaciology