EKATERINA TYMOFYEYEVA GMTSAR BATCH PROCESSING

Size: px

Start display at page:

Download "EKATERINA TYMOFYEYEVA GMTSAR BATCH PROCESSING"

Brooke Craig
5 years ago
Views:

1 EKATERINA TYMOFYEYEVA GMTSAR BATCH PROCESSING

2 THANK YOU! Xiaopeng Tong Xiaohua (Eric) Xu David Sandwell Yuri Fialko

3 OUTLINE Batch processing scripts in GMTSAR (focus on Sentinel-1) SBAS: a method for calculating InSAR time series. Atmospheric correction method for InSAR time series.

4 WHY BATCH PROCESSING? Automated processing of large (~TB) volumes of SAR data to generate hundreds to thousands of interferograms. Analyze interferograms to estimate ground deformation and how it varies in time: Earthquake cycle deformation Volcano unrest Glacier flow Urban and infrastructure Hydrology Landslides

5 Batch processing for time series

6 OVERVIEW Organize your data and decide processing strategies. Choose a master image. Align a stack of SAR data. Form interferograms. Post-process (SBAS time series, Atmospheric correction) Examples available on GMTSAR website: Envisat, Sentinel-1A

7 STEP ZERO: ORGANIZE YOUR DATA Directory tree (just a suggestion): ERS/ENVISAT/ALOS-1 SENTINEL-1A (TOPS) TRACK# TRACK# RAW TOPO INTF TOPO RAW F1 F2 F3 RAW INTF RAW INTF RAW INTF

8 STEP ONE: CHOOSE YOUR MASTER IMAGE SENTINEL-1A (TOPS) preproc_batch_tops.csh (mode 1) required inputs: data.in, dem.grd, mode Pre-process a stack of SAR data using default parameters (Earth radius, Doppler centroid, near range) Generate baseline-time plot to choose master images, alignment strategy, and interferometric pairs.

9 STEP ONE: CHOOSE YOUR MASTER IMAGE SENTINEL-1A (TOPS) preproc_batch_tops.csh (mode 1) input file data.in: S1A-IW1-SLC-VV-**D1T1**-001:S1A-IW1-SLC-VV-**D1T2**-001:S1A_***EOF S1A-IW1-SLC-VV-**D2T1**-001:S1A-IW1-SLC-VV-**D2T2**-001:S1A_***EOF S1A-IW1-SLC-VV-**D3T1**-001:S1A-IW1-SLC-VV-**D3T2**-001:S1A_***EOF

10 STEP ONE: CHOOSE YOUR MASTER IMAGE (EXAMPLE) A good master image 0 S1A _ALL_F1 covers your area of interest and minimizes perpendicular baselines baseline (m) S1A _ALL_F1 S1A _ALL_F1 between interferometric pairs. 50 Baseline-time plot: baseline.ps year

11 STEP ONE: CHOOSE YOUR MASTER IMAGE (EXAMPLE) A good master image , 11/10 S1A _ALL_F1 covers your area of interest and minimizes perpendicular baselines between interferometric baseline (m) 2014, 12/ , 12/28 S1A _ALL_F1 S1A _ALL_F1 pairs. 50 Baseline-time plot: baseline.ps year

12 STEP ONE: CHOOSE YOUR MASTER IMAGE (EXAMPLE) A good master image , 11/10 S1A _ALL_F1 covers your area of interest and minimizes perpendicular baselines between interferometric baseline (m) 2014, 12/ , 12/28 S1A _ALL_F1 S1A _ALL_F1 pairs. 50 Baseline-time plot: baseline.ps year

13 STEP TWO: PREPROCESS YOUR BATCH SENTINEL-1A (TOPS) preproc_batch_tops.csh (mode2) Modify data.in: put the master first! Output aligned.slc files that will be combined to form interferometric pairs.

14 STEP TWO: PREPROCESS YOUR BATCH SENTINEL-1A (TOPS) preproc_batch_tops.csh (mode 2) input file data.in: S1A-IW1-SLC-VV-**D2T1**-001:S1A-IW1-SLC-VV-**D2T2**-001:S1A_***EOF S1A-IW1-SLC-VV-**D1T1**-001:S1A-IW1-SLC-VV-**D1T2**-001:S1A_***EOF S1A-IW1-SLC-VV-**D3T1**-001:S1A-IW1-SLC-VV-**D3T2**-001:S1A_***EOF MASTER

15 STEP THREE: MAKE INTERFEROGRAMS! Choose the pairs you , 11/10 S1A _ALL_F1 want to connect. Interferograms in a circle should end up to zero! baseline (m) 2014, 12/ , 12/28 S1A _ALL_F1 S1A _ALL_F1 50 Baseline-time plot: baseline.ps year

16 STEP THREE: MAKE INTERFEROGRAMS! intf_tops.csh intf.in batch_tops.config batch_tops.config parameters: proc_stage master_image threshold_snaphu filter_wavelength range_dec, azimuth_dec, dec_factor switch_land defomax

17 STEP THREE: MAKE INTERFEROGRAMS! intf_tops.csh intf.in batch_tops.config intf.in is a list of pairs you want to connect, e.g., S1A _ALL_F1:S1A _ALL_F1

GMTSAR BATCH PROCESSING OPTIONAL STEP: MERGE THE SUB SWATHS Inside batch_tops.config, set proc_stage to 1, threshold_snaphu to 0, and threshold_geocode to 0.

18 GMTSAR BATCH PROCESSING OPTIONAL STEP: MERGE THE SUB SWATHS Inside batch_tops.config, set proc_stage to 1, threshold_snaphu to 0, and threshold_geocode to 0. Run script merge_unwrap_geocode_tops.csh (on individual pairs) or batch_merge.csh (on a whole batch) to merge the (wrapped) phase, then unwrap and geocode.

19 OTHER SATELLITES (NOT SENTINEL) pre_proc_batch.csh SUPERMASTER! Get a baseline table Choose master(s) and SUPERMASTER! align_batch.csh align slaves to masters and the super master

20 BATCH PROCESSING RESULTS Interferograms are stored in individual folders in intf/ (intf_all/ for Sentinel, separate for each sub swath) The folder is named after the two dates (e.g., _ ) Each interferogram folder contains the following files: Amplitude, phase, correlation, unwrapped phase, filtered phase image files in NetCDF format.grd Corresponding files after geocoding with suffix _ll.grd Postscript plots:.ps Google Earth.kml and.png

21 MANY INTERFEROGRAMS 50 S1A _ALL_F3 S1A _ALL_F3 S1A _ALL_F3 S1A _ALL_F3 S1A _ALL_F3 0 S1A _ALL_F3 S1A _ALL_F3 S1A _ALL_F3 S1A _ALL_F3 S1A _ALL_F3 S1A _ALL_F3 S1A _ALL_F3 baseline (m) S1A _ALL_F3 S1A _ALL_F3 S1A _ALL_F S1A _ALL_F year

22 TIME SERIES METHODS Stacking (average multiple interferograms to get mean LOS rate) Small-Baseline Subset (SBAS) Test case available on the GMTSAR website Persistent Scatterer Insar (PSInSAR) StaMPS (Stanford Methods for PS InSAR - written in Matlab) Giant (Generic InSAR Analysis Toolbox) Multiple time-series analysis tools (written in Python)

23 TIME SERIES METHODS Stacking (average multiple interferograms to get mean LOS rate) Small-Baseline Subset (SBAS) Test case available on the GMTSAR website Persistent Scatterer Insar (PSInSAR) StaMPS (Stanford Methods for PS InSAR - written in Matlab) Giant (Generic InSAR Analysis Toolbox) Multiple time-series analysis tools (written in Python)

24 Small&Baseline&Subset&(SBAS)&methods& W βb βb λ λ λ λ... 0 m 1 m 2... m s Δh = W d 1 d W = diag{γ 1,γ 2,γ 3,!,γ n } d i B i γ i m i Δh?? )Unwrapped)phase)?? )Perpendicular)baseline)?? )Coherence))??))))))Incremental)range)change)??))))))DEM)error) [Berardino)et)al.,)2002;)Schmidt)and)Burgmann,)2003])

25 Comparison between the coherence-based SBAS and the tradi6onal SBAS ﬂight look coherence-based SBAS Tradi6onal SBAS

tab (list of interferograms), orbit numbers, baselines. scene.

26 TIME SERIES EXAMPLE ALOS data over Coachella Valley, California Time series prepared with the script sbas (included in GMTSAR) Input: intf.tab (list of interferograms), orbit numbers, baselines. scene.tab (list of scenes), day numbers N (number of interferograms) S (number of SAR scenes) xdim, ydim (dimensions of the interferograms must be the same for all images)

27 GMTSAR BATCH PROCESSING ATMOSPHERIC CORRECTION

$ATMOSPHERIC CORRECTION Radar waves are refracted Pressure Temperature Water vapor Atmospheric noise is unpredictable Turbulent$

28 ATMOSPHERIC CORRECTION Radar waves are refracted Pressure Temperature Water vapor Atmospheric noise is unpredictable Turbulent component (power law in space, uncorrelated in time) Seasonal weather patterns (correlated in time) Layers (correlated with topography)

29 ATMOSPHERIC CORRECTION Ionosphere Charged particles - radar phase speeds up. Ionosphere is dispersive. Long spatial wavelengths

30 ATMOSPHERIC CORRECTION Options: Weather models TEC models Filtering (in time or space or both) Topography-correlated models

31 ATMOSPHERIC CORRECTION METHOD Interferograms that share a common scene also share the same atmospheric contribution. Subtracting pair 2-3 from pair 1-2 will amplify the atmospheric signal, α2

32 ATMOSPHERIC CORRECTION METHOD For constant deformation and equal time spans between acquisitions, increasing the length of the averaging stencil increases the accuracy of retrieved αi For non-linear deformation and irregular acquisitions, we adopt an iterative procedure.

33 ATMOSPHERIC CORRECTION METHOD Calculate atmospheric contribution, αi, for all pixels for each of the scenes in the catalog. Compute the Atmospheric Noise Coefficient (ANC, below) for each scene. Obtain independent estimate for deformation by computing time series and fitting smoothing spline. Remove the estimate of deformation from interferograms that go into subsequent calculations. Recalculate αi, starting with scenes with the highest ANC. Recalculate ANC for each scene. Recalculate deformation signal, using estimates of αi to correct for atmospheric contribution. Repeat Steps 4-7 until convergence.

34 ATMOSPHERIC CORRECTION EXAMPLE Atmospheric phase maps, ERS1-2 and ENVISAT Track 170

35 ATMOSPHERIC CORRECTION EXAMPLE May 8, 2010 Sept. 2, 2000 July 8, 2006 April 26, 1997

36 ATMOSPHERIC CORRECTION ATM. CORRECTION EXAMPLE

37 ATMOSPHERIC CORRECTION VALIDATION: SYNTHETIC TESTS LOS displacements (mm) (a) -30 Input signal Uncorrected time series Corrected time series (d) -30 Input signal Uncorrected timeseries Corrected timeseries Results of the synthetic data test. (a-c) Regular acquisitions with small baselines. (d- f) Set of interferograms that mimics the ERS/ENVISAT baseline distribution (Figure 4). (a,d) No LOS displacements (mm) (b) -30 Input signal Uncorrected time series Corrected time series (e) Input signal Uncorrected timeseries Corrected timeseries deformation. (b,e) Constant velocity of 1 mm/ yr. (c,f) Variable velocity. Blue dots denote prescribed deformation signal, red dots and crosses denote recovered time series with LOS displacements (mm) (c) -30 Input signal Uncorrected time series Corrected time series Time (years) (f) -30 Input signal Uncorrected timeseries Corrected timeseries Time (years) and without atmospheric corrections, respectively.

38 ATMOSPHERIC CORRECTION VALIDATION OF CORRECTED TIME SERIES (CGPS) LOS displacements (mm) Time (years)

39 ATMOSPHERIC CORRECTION EXAMPLE: DEFORMATION AT CERRO PRIETO GEOTHERMAL FIELD Xu et al., 2017

40 ATMOSPHERIC CORRECTION EXAMPLE: DEFORMATION AT CERRO PRIETO GEOTHERMAL FIELD Xu et al., 2017

41 DEM: 0.5 GB 30 SAR IMAGES: 2-3 GB EACH INSAR TIME SERIES: PRICELESS.

Terrain Motion and Persistent Scatterer InSAR

Terrain Motion and Persistent Scatterer InSAR Andy Hooper University of Leeds ESA Land Training Course, Gödöllő, Hungary, 4-9 th September, 2017 Good Interferogram 2011 Tohoku earthquake Good correlation