Demonstration of SAR Interferometry under Crossing Orbits using TerraSAR-X and TanDEM-X P aco López-Dekker , Pau Prats, Francesco De Zan, Steffen Wollstadt, Daniel Schulze, Gerhard Krieger and Alberto Moreira Microwaves and Radar Institute, German Aerospace Center (DLR)
Theory: InSAR under crossing orbits
Consider a pair of SAR acquisitions with crossing ground-tracks
In general, different observation geometry implies that different portions of the 2-D ground spectra are sampled
For small crossing angles, common spectrum can be achieved/increased by squinted acquistions
Ground projected squint angle Spectral overlap achieved by applying opposed squints i.e. a Doppler Centroid offset is required!!!
First TanDEM-X DEM
Acquisition while TDX was chasing TSX (before pursuit monostatic commissioning phase)
Crossing tracks due to Earth rotation
Small crossing angle -> small squints (< beam width)
Motivation: scientific impatience and proof of concept
Repeat-pass InSAR using different (crossing) tracks
Large crossing angles (1° - 2°) -> large squints (>> beam width)
Allows 1,5 and 6 day repeat-pass acquisitions with 11 day repeat cycle
for restricted regions (where ground tracks converge)
Motivation: proof of concept, scientifically relevant data
Time series being acquired
TanDEM-X Commissioning Phase First bi-static DEM 8 October (MET +113) June‘10 July‘10 Aug‘10 Sep‘10 Oct‘10 Nov‘10 Dec‘10 Early Orbit Phase Grg Segment Checkout Bi-static Commissioning Phase 20 km Formation TDX Orbit Drift 16.000 km 20 km Close Helix-Formation 300-400 m Launch 21 June DEM Acquisition First SAR Image 24 June (MET +3.6) First DEM 16 July (MET +25) TDX Monostatic Comm. Phase 6 Months Commissioning Phase First close formation DEM 19 October (MET +124) First bi-static SAR image 8 August (MET +48)
TanDEM-X first DEM
Along track (~200 km) separation results in crossing ground-tracks due to Earth Rotation
0.13° relative squint required
~2 km resulting effective baseline
h amb = 3.8 m
Almost out of the box processing with TAXI
TanDEM-X first DEM (October Revolution Island)
DEM Error/Performance High-pass filtered DEM Coherence to point-to-point error Scaling to restore white noise powe
Repeat-pass using different tracks (AKA 1/5-day repeat-pass)
Closest (non identical) tracks after (approximately) whole number of days
No! Same fringe pattern appears also in zero-squint 11-day repeat.
11 day coherence is very bad.
Since baseline is small and constant -> high temporal decorrelation
Ronne #1, 300 MHz strip-map (experimental) 11 day repeat 11d
Fringe rate is proportional to time-lag
Can be explained by surface velocity
Fringes indicate velocity gradients (velocity itself is ambiguous).
Velocity can be estimated from radargrammetric shifts
Coupling between height uncertainty and varying baseline introduces azimuth phase ramp
5d 1d Ronne #1, 300 MHz strip-map (experimental) Interpretation Zero baseline: Only noise and temporal decorelation Large baseline: Large volume decorrelation (geometric decorrelation addressed by azimuth adaptive range spectral filtering) Zero baseline: Only noise and temporal decorelation
Velocity measurements on the Filchner/Ronne Ice Shelf (taken from http://nsidc.org) The velocity contours are from Vaughn and Jonas, 1996 (units of m/yr). (Vaughan, D.G., and M. Jonas, 1996. Measurements of velocity of Filchner-Ronne Ice Shelf. Filchner-Ronne Ice Shelf Programme (FRISP) Report No. 10. Ed. H Oerter. Pub. AWI, Bremerhaven. 111-116.)
Other things to do: exploit varying baseline Ronne #1 150 MHz strip-map (3rd cycle) 1 day repeat Homogeneous scene Rapidly varying baseline Rapidly varying kz
Joint histogram (pdf) of k z and coherence shows well defined relation.
Due to homogeneity of scene, mean coherence vs. kz can be directly related to volume decorrelation.
Mean SNR and temporal decorrelation introduce a constant factor.
The feasibility of InSAR under crossing orbits has been experimentally validated
TanDEM-X experiment shows potential for XTI under very long baselines
Repeat-pass acquisitions under crossing tracks demonstrated
1/5/6-day “repeat-pass” acquisitions possible
Possible scenes constrained to very narrow band of latitudes…
… yet scientifically very interesting areas can be addressed
Longer time-series over Ronne Ice Shelf and other regions in Antarctica
Some InSAR processing challenges remain (improve coregistration under low coherence areas, eliminate some systematic phase errors).
Study of Ice structure exploiting variation of baseline.