Strategies for Measuring Large Scale Ground Surface...
Transcript of Strategies for Measuring Large Scale Ground Surface...
Strategies for Measuring Large Scale Ground Surface Deformations: PSI Wide Area Product Approaches
J. Duro(1), R. Iglesias(1), P. Blanco-Sánchez(1), D. Albiol(1), T. Wright(2), N. Adam(3), F. Rodríguez(3), R. Brcic(3), A. Parizzi(3), F. Novali(4), P. Bally(5)
(1) (2) (3) (4) (5)
Outline
• Motivation – Terrafirma Project – Wide Area Product (WAP)
• PSI processing for WAP – PSI Processing for each individual track/frame – Data merging – Parallel Tracks Calibration
• Performance Analysis and Discussion – Multi-seed Impact – Phase noise impact – APS impact – Persistent scatterers spatial distribution – Calibration and final results
• Conclusion and future work
Outline
• Motivation – Terrafirma Project – Wide Area Product (WAP)
• PSI processing for WAP – PSI Processing for each individual track/frame – Data merging – Parallel Tracks Calibration
• Performance Analysis and Discussion – Multi-seed Impact – Phase noise impact – APS impact – Persistent scatterers spatial distribution – Calibration and final results
• Conclusion and future work
Motivation Terrafirma Project
Terrafirma services help to identify and mitigate risk
• Terrafirma was started in 2003 by NPA Group. The 3rd Stage of the project was led by Altamira Information.
• Terrafirma is an open service partnership where competitor providers work together to produce a standardised and validated advanced products and services.
• Terrafirma services have been delivered to over 90 users via a Service Level Agreement, which includes User Feedback and Utility reports.
• The Terrafirma service offering includes the wide-area map designed to enable the delivery of seamless motion maps over wide areas utilising the Sentinel-1 satellites.
Terrafirma is a pan-European Terrain Motion Hazard Information Service
Motivation Wide Area Product Possible definitions WAP?
A) Multi PSI
• 1 or + frames at F.R. & F.W. • 1 or + clusters of points linked to a ref. • Delivered as independent DB • PSs repeated at overlapping regions?
B) Wide area PSI • 2 < frames at F.R. & F.W. • Calibrated with or without GPS (offsets?, ramps?) • Unique reference point? • Unique DB (without PSs overlap)?
C) Others ….
GREECE 10 frames (GPS calibrated)
WAP product definition is linked to the objectives
Different options can be found
What’s the difference w.r.t classical PSI approach?
Measure large-scale motions (plate tectonics)?
Outline
• Motivation – Terrafirma Project – Wide Area Product (WAP)
• PSI processing for WAP – PSI Processing for each individual track/frame – Data merging – Parallel Tracks Calibration
• Performance Analysis and Discussion – Multi-seed Impact – Phase noise impact – APS impact – Persistent scatterers spatial distribution – Calibration and final results
• Conclusion and future work
PSI Processing for WAP Large AOI to be covered
Volos Orbit Track - 74 ERS1&2 SAR images - 215 Differential Interferograms
Larissa Orbit Tracks - 64 ERS1&2 SAR images - 180 Differential Interferograms
¡ N 90km 45km 0m
PSI Processing for WAP PSI Processing for each individual track/frame + merging
seed
seed
Calibration by means of
Common area Statistics
Volos Frame
Larissa Frame
Deformation rate [mm/y]
Calibrated Joint Displacement Map
Deformation rate [mm/y]
Altamira Information WAP PSI over Larissa and Volos Tracks
PSI Processing for WAP Possible final product…
• Unique reference point
• Full swath at full resolution
• Millions of measurement points in the DB
• Track merging without overlap
Deformation rate [mm/y]
aquifer
aquifer
landslides
Altamira-Information WAP PSI over Larissa and Volos Tracks
PSI Processing for WAP Can we trust large scale motion trend retrieved by WAP PSI?
Outline
• Motivation – Terrafirma Project – Wide Area Product (WAP)
• PSI processing for WAP – PSI Processing for each individual track/frame – Data merging – Parallel Tracks Calibration
• Performance Analysis and Discussion – Multi-seed Impact – Phase noise impact – APS impact – Persistent scatterers spatial distribution – Calibration and final results
• Conclusion and future work
+ -
Performance Analysis and Discussion Development of a simulator for performance tests
PSI
Single or multi-seed
Temp & Spat Baselines
Phase Noise Impact Residual Orbit Impact APS impact PSC density Network construction
+ Plane Adjust.
Simulated Large-Scale
Velocity Interf
Gen Δvi
αi
Δvo
αo
+ +
+ Residue
αi
Simulated velocity PSI Velocity (1 seed) PSI Velocity (6 seeds) Difference
seed
seed
seed
seed
seed
seed
seed
Differents offsets
between clusters
Inci = 2 cm/year αi = 7 degrees
Performance Analysis and Discussion Multi-seed Impact
-1cm/y 1cm/y -1cm/y 1cm/y -1cm/y 1cm/y -1cm/y 1cm/y
1 seed is mandatory to measure large scale motion trends
Interf Gen
Performance Analysis and Discussion Phase noise impact
Coherence
1 … 0.5 0.4 0.3 0.2 0.1
ramp increment
[cm/y]
2 2.00 2.00 2.00 2.03 2.05 2.53 5.27
1.5 1.50 1.50 1.50 1.53 1.55 2.1 10.72
1 1.00 1.00 1.00 1.03 1.05 0.89 9.47
0.5 0.50 0.50 0.50 0.52 0.56 0.31 7.17
Coherence
1 … 0.5 0.4 0.3 0.2 0.1
ramp orientation
[cm/y]
2 7.00 7.00 7.00 7.41 7.64 13.1 68.9
1.5 7.00 7.00 7.00 7.43 7.72 15.3 95.7
1 7.00 7.00 7.00 7.38 7.77 2.07 109
0.5 7.00 7.00 7.00 7.41 8.83 1.89 91.7
Estimated Plane Characteristics
Simu Velocities
PSI Velocity
Estimated Plane
PSI
Interf Gen
Interf Gen
HIGH AGREEMENT GOOD AGREEMENT
LOW AGREEMENT
+
Δvo
αo
… … … …
Coherence VS
Phase Stdev
Δvi
1.5 cm/y
Δvi
2.0 cm/y
Δvi
0.5 cm/y
Interferometric Phase Noise Values
Strong phase noise is required produce an impact
Performance Analysis and Discussion Residual orbit impact
Simulated Orbital Ramp Residue PSI Velocity Est. Plane
-1cm/y 1cm/y
Simulated Velocity
Δvi 2 cm/year
αi 7 deg.
-1cm/y 1cm/y -0.3cm/y 0.3cm/y -1cm/y 1cm/y
Δvo 2.3 cm/year
αo 8 deg.
Temporal Distribution
Acquiston date
# In
terf
150
0
1994 1996 1998 2000
-1cm/y 1cm/y
orbit inaccuracies only in one SLC appearing in 10 interferograms 3mm/y + 1o error Residual ramp
Orbit state vectors inaccuracies have an important impact on the measurements of large scale motion trends
Possible Strategies
- First order de-trending by means of range and azimuth FFT analysis.
- Second order orbit inaccuracies compensation.
- Phase model fit to carry out a correction of the orbit state vectors
- …
When large-scale motion is expected, distinguishing between displacement
signals and phase patterns due to orbit inaccuracies is a key issue
Performance Analysis and Discussion Residual orbit compensation
Est. Residual Orbit Original Interferf Compensated Interf.
Performance Analysis and Discussion APS impact
-1cm/y 1cm/y
Simulated velocity LOW APS TURBULENCE
0 cm/y 0.3 cm/y -1cm/y 1cm/y
PSI velocity Residue Modulus
-1cm/y 1cm/y
MEDIUM APS TURBULENCE
0 cm/y 0.3 cm/y
PSI velocity Residue Modulus
-1cm/y 1cm/y
HIGH APS TURBULENCE
0 cm/y 0.3 cm/y
Residue Modulus
-1cm/y 1cm/y
Estimated Plane
Perfect Performance! Good Performance! Bad Performance in areas with steep topography
PSI velocity
High residue
Δvi = 2 cm/year
αi = 7 degrees
Δvo = 2.4 cm/year
αo = 5 degrees
Atmospheric signal delays originate important local residuals which decreases the accuracy in trends retrieval
Performance Analysis and Discussion APS compensation
Original Interf. APS model Comp. Interf.
LOW APS TURBULENCE Orig. Interfef. APS model Comp. Interf.
HIGH APS TURBULENCE
• APS model from global meteorological data (ERA-I from the ECMWF) • APS model fitted from interferometric phases • APS estimated inside the PSI processing by the application of several filters
Performance Analysis and Discussion Persistent scatterers spatial distribution
PSI velocity
PSI velocity PSI velocity
-0.3cm/y 0.3cm/y -0.3cm/y 0.3cm/y -0.3cm/y 0.3cm/y
High density Mid density Low density
Δvi = 2 cm/year
αi = 7 degrees
+ APS
-1cm/y 1cm/y 0 cm/y 0.3 cm/y
PSI velocity Residue Modulus
Medium APS turbulences
• Simulations made with medium APS turbulences (gave good performance in previous tests)
Model Coherence
Performance Analysis and Discussion Persistent scatterers spatial distribution
Residue Modulus
Residue Modulus Residue Modulus
0cm/y 0.3cm/y 0cm/y 0.3cm/y 0cm/y 0.3cm/y
Model Coherence
in large links decrease
due to APS
High density
Δvi = 2 cm/year
αi = 7 degrees
+ APS
Mid density Low density
Higher density is required if APS compensation is made inside PSI processing Accurate processing with low density can be only achieved by using external data (APS model, GPS)
Outline
• Motivation – Terrafirma Project – Wide Area Product (WAP)
• PSI processing for WAP – PSI Processing for each individual track/frame – Data merging – Parallel Tracks Calibration
• Performance Analysis and Discussion – Multi-seed Impact – Phase noise impact – APS impact – Persistent scatterers spatial distribution – Calibration and final results
• Conclusion and future work
Summary and conclusions
• Different processing options can be adopted to achieve wide area coverage
– Main difference resides in keeping one unique reference or not for all the PS measurements
– Processing with a unique reference and tracks calibration are compulsories for large scale motion trends retrievals
• Orbit state vectors inaccuracies have the greatest impact on the accuracy of large scale motions
• Atmospheric perturbations originate more “local” errors which can also reduce the accuracy over large scale (depending on the unwrapping strategy)
– High PS density allows to overcome rapid phase variations
• Low density of PSs have an important impact on the measurement of large scale trends:
– Require the use of external data: Numerical models to generate the APS a priori and/or GPS data for cluster calibration
– Advance phase filtering strategies or more redundancy on the PS network configuration are required
• In case of high quality data (high density PSs, a lot of interferograms, good orbit state vectors) large scale motion trends can be measured with good accuracy
Open questions for the RT
• What should be the scope of WAP?
• It is really needed Wide Area Product at full resolution?
• It is required to have a unique spatial reference?
• What is the benefit of Sentniel-1 for WAP?
• What is the accuracy and the robustness of the numerical weather models for based APS compensation on this automatically? Can we achieve a good spatial resolution with this model based data for a proper correction of the local turbulences?
• Can we perform PSI clusters calibration with available GPS data? Can GPS network provide a good density to calibrate PSI WAP data? It is precise enough for LOS calibration?
… question and suggestions are welcome …
THANKS FOR YOUR ATTENTION!