RADARSAT-2 Applications
Gordon Staples MDA Richmond, BC, CANADA [email protected]
0 10 Quad Quad (A) hh−hv hh−vh hh−vv vv−vh hh−hv(A) hh−vv(A) vv−vh(A) −1 10 hh vv hv vh
−2 10 Probability of missed detection
−3 10 −7 −6 −5 −4 −3 −2 −1 0 10 10 10 10 10 10 10 10 False alarm probability Acknowledgements
Heather McNairn, AAFC
Ridha Touzi, CCRS
Paris Vachon, DRDC-O
Bernd Scheuchl, Bernhard Rabus, MDA
Dean Flett, CIS
©MDA From Science to Operational Applications
Adapted from van der Sanden, CJRS, June 2004 ©MDA Applications
Agriculture
Wetland Mapping
InSAR
Ship Detection
Ship-Iceberg Discrimination,
Sea Ice
©MDA Crop Classification Using Integrated Optical and SAR Data
Agriculture and Agri-Food Canada is developing a method to deliver an annual inventory of crops using a multi-sensor solution.
This classified map was derived using multi- temporal optical (Landsat and SPOT) and SAR (RADARSAT-1 and Enivsat ASAR) acquired throughout the 2005 growing season. A decision-tree classifier was used.
Once available, the method will be migrated to use RADARSAT-2 dual-polarization data.
Classification Accuracies
Cereals 76.4% Corn 86.3% Pasture 92.9% Potato 87.7% Soybean 92.3%
Overall 88.5% Kappa 0.84
©MDA S6A HH S5A HH S3A HH S7A HH Identifying Tillage Events Using RADARSAT
F11-D F11-A
F11-C F11-B
17-Oct-2005 24-Oct-2005 31-Oct-2005 03-Nov-2005
Average dB Tilled -7.0 dB 0 Average dB Untilled -11.5 dB -1 Diff. Tilled vs Untilled 4.5dB Identifying the timing of tillage -2 events is important in assessing -3 best management practices and in -4 Oct 17-24 Oct 24-31 erosion and carbon modeling. -5 -6 Oct 31-Nov 3 -7
° (dB) -8
σ -9 -10 -11 -12 -13 -14 Ref F11-A F11-B F11-C F11-D Field ID
©MDA ©MDA ©MDA InSAR and PS-InSAR
RADARSAT-2 for InSAR – On-board GPS will determine sensor’s location +-60m – Post processing (Precise Orbit Determination software) will improve this to +-15m – SLC data processed to zero Doppler and fully compatible for InSAR/PSInSAR processing squint spectrum [degrees] squint spectrum latitude [degrees] RADARSAT-1/2 Cross-InSAR/PS- InSAR issues: – data acquisition difference of ~ 30 min Æ reduced temporal Almost no azimuth spectral overlap (except in immediate vicinity of perigee correlation, but…frequency o difference of 5.3000 GHz vs. near 80 latitude) 5.441 GHz – Yaw-steering implemented to minimize UltraFine mode range cell migration – range/azimuth spectral overlap
©MDA RADARSAT-1/2 InSAR and PS-InSAR
Distributed Targets – No 'natural' azimuth spectrum overlap if RADARSAT-2 is yaw-steered to zero Doppler – Range spectrum overlap requires large baseline to compensate (15 km) and decorrelation with slope will be a serious issue – Cross InSAR mission similar to ERS/ENVISAT is not possible operationally – At best cross InSAR limited to (large) corner reflector like point targets
Point Targets – Range phase offset has sensitivity comparable to ERS-ENVISAT – Azimuth phase offset is very sensitive => PS-InSAR time series could only be combined if RADARSAT-2 yaw steering mode is switched off
Solution – RADARSAT-1/2 overlap – RADARSAT-2 UltraFine (3 m resolution) for the identification of smaller point targets
©MDA Potential of RADARSAT-2 POLinSAR
Challenges – 24-day repeat-period leads to low-coherence – C-band frequency provides less penetration than the L-band data used in initial PolinSAR demonstrations
Mitigating factors – Single-polarization repeat-pass RADARSAT-1 interferometry is proven – RADARSAT-2 will have improved baseline control and knowledge – Spaceborne PolinSAR will provide better registration accuracy thus significantly reducing mis-registration artefacts
©MDA Marine Surveillance Ocean Intelligence
Cg A
B
Canadian Forces requirements and RADARSAT-1 image showing internal operations, with SAR potential highlighted waves (A & B). Use of HH or VV in yellow. All SAR-derived requirements polarization provides optimal information. can be extracted from a single image.
©MDA Ship Detection
0 10 Quad Quad (A) hh−hv hh−vh hh−vv vv−vh hh−hv(A) hh−vv(A) vv−vh(A) −1 10 hh vv hv vh
CFAV Quest
−2 10 76 m × 12.6 m Probability of missed detection
CFAV Quest 6-Oct-03, single look
−3 θinc = 42° 10 −7 −6 −5 −4 −3 −2 −1 0 10 10 10 10 10 10 10 10 φasp = 0° False alarm probability Vt = 3 kt Wind speed = 6.0 m/s Detection Performance – probability of missed Wave height = 1.6 m detection vs. probability of false alarm
©MDA Dual Co-Polarized RADARSAT-2 Mode
16 m x 5 m Possible dual co-polarized mode parameters without relative phase optimized for balanced performance.
Possible dual co-polarized mode parameters with relative phase optimized for swath width.
©MDA Ship-Iceberg Discrimination
Ship
Iceberg
Identification of scattering mechanism based on the Symmetric Scattering Characterization Method (SSCM) and mapped to Cameron elemental scatterers.
Example of target Poincare longitude for ship and iceberg targets derived from SSCM. Regions of no coherence (NC) identified. ©MDA SAR Data for Operational Sea Ice Monitoring
Cooperation between Canada (CSA) and Finland (TEKES)
Objectives – Operational demonstration of dual- polarization SAR to improve sea ice monitoring CV-580 Simulated RADARSAT-2 – Demonstrate RADARSAT-2 polarimetric SAR for enhanced tactical sea ice classification.
Project overview – Classification Incidence angle – Data fusion (SSM/I, SAR, Ice Chart) correction – Automated ice/no ice classification – Compensation of incidence angle variation
MYI frozen melt pond 0 MYI hummock FYI with thin snow – Fieldwork for validation -5 -10
-15 Backscatter (dB) Backscatter – -20 Data delivery -25
-30 15 25 35 45 55 65 Incidence angle Fieldwork
©MDA Proposed Process
ASAR AP Ancillary Datasets CIS Analyst CIS Ice Chart HV+HH Image
Inc. angle compensated
Classifier (Wishart)
Automatically generated binary Ice /No ice Product
Ancillary Datasets Shafer Dempster Reasoning
Ice Chart
} Pure Ice/Water } SSM/I
©MDA Looking Ahead
SOAR Program will provide validation of RADARSAT-2 applications potential, especially quad-polarized data
Based on experience with ENVISAT data, use of dual-polarized RADARSAT-2 for operational applications expected early in the mission
Use of quad-polarized data will take longer, and operational use will require the translation quad-polarized data “output” into user-friendly information products
Quad-polarized data sets for research purposes: – sites are TBD, but may include Vancouver, San Francisco, Oberphaffenhofen, Flevoland, ocean/coastal area, others? – possible temporal acquisitions and co-ordination with other sensors (multi-frequency) – available after RADARSAT-2 calibration
©MDA