2025.1.0-build2

Bug Fix:

• Fixed a bug in the shadow detector: an offset needed to be added to indices when projecting in the ground geometry to avoid kernel crash.

Details:

• This is the second build of version 2025.1.0. The first build (2025.1.0-1) involved an update to the Conda recipe without code changes.

2025.1.0

New features

• Radiometric terrain correction (or flattening) is now available in the SLC processor
• Our algorithm differs from the SNAP implementation. This is actually a modified version of the radiometric normalization algorithm described in SNAP's terrain correction documentation (see https://step.esa.int/main/wp-content/help/versions/9.0.0/snap-toolboxes/org.esa.s1tbx.s1tbx.op.sar.processing.ui/operators/RangeDopplerGeocodingOp.html). Two things are different:
  • Instead of the sine of the projected incidence angle, the tangent is computed to comply with the gamma nought convention (as in Small's algorithm).
  • The simulated backscatter is regridded and accumulated in the SAR geometry to account for the many-to-one and one-to-many relationships.
• Shadows are detected and set to NaNs in the normalized image.
• Added tutorial notebooks to demonstrate feature usage.

Bugfixes and improvements

• Add libgdal jpeg dependency to work with the new version of rasterio.
• Temporary fix: latest pyproj version was introducing a bug in the computation of the DEM XYZ coordinates. Force to use the previous version.
• Refactored both InSAR and SLC processors: we do not use child processes anymore and set the GDAL_CACHEMAX value to a small number. Benchmarks showed smaller memory usage than previously and no performance loss.
• Fixed some exceptions in the processors (the raise keyword was missing leading the exception to not be triggered).

2024.10.1

New features

• Better S1 coherence estimator
  • Multilooking is performed prior to coherence computation
  • It is applied to Each individual term in the coherence expression
  • This way larger sample sizes can be collected even with small boxcar windows
  • It avoids averaging the coherence itself (theoretically sounder)
  • Binary erosion is applied to mitigate discontinuities at subswath borders
  • Default multilook is still [1, 4] but coherence window size is now [3, 3]

Bugfixes

• Fixed an error with DEM download due to Planetary Computer now requiring signed urls

Other

• Upgraded dependencies

2024.10.0

New features

• Sentinel-1 zipped products are now handled (updated all notebooks accordingly).
• InSAR processing of partially overlapping products, for instance between S1A and S1B, as long as share the same orbit.

Other

• Refactored block processing utility to handle larger overlap.
• Goldstein filter now uses different block size (32) and overlap results in smaller final block size.
• Added Etna notebook to illustrate partial overlap in InSAR (docker version only)
• Simple script that shows all S1 products located in a directory on a map (not part of the conda package)

2024.9.2

New features

• New function S1.process.goldstein: the Goldstein interferometric filter is an adaptive method that allows to reduce noise on phase.
• Tutorial notebooks on how to use the function
• All file names like "ifg*" are now treated as interferograms by S1.process.geocode_and_merge_iw. This way users can compare several phase filtering methods easily.

Bugfixes

• Correct options for orbit retrieval ['POE', 'RES'] using pyrosar
• In S2.process_tiles, baseline > 4 replaced by baseline >= 4
• In S1.process.coherence output is tiled like input

Misc

• S1.process.coherence with manually fixed chunk size is faster.

2024.9.1

New feature -- Custom Sentinel-1 Pipelines

• New S1.process.apply_multilook function to apply multilook on a tiff in the SAR geometry.
• Multilooking may now also be applied in S1.process.coherence, S1.process.amplitude and S1.process.interferogram rather than in S1.process.geocode_and_merge_iw.
• These functions change the transform in the GeoTIFF files. S1.process.sar2geo is now aware of this and applies automatic rescaling. This allows the user to call S1.process.sar2geo after multilooking an image and define custom processing chains.
• In this spirit, two new helper functions allow to apply any user defined function to all subswaths and polarizations present in the output product directory.
• Therefore, custom processing chains of type S1.process.prepare_insar -> custom processing -> S1.process.geocode_and_merge_iw can be easily defined by the user.
• The same can be applied to S1.process.prepare_slc for single images.
• Notebooks showing an example of such custom pipeline implementing both coherent and incoherent change detection have been added to the docs and the example folders.

Bugfixes

• Corrected the S1.util.presum function to avoid overwriting the input array, reformatted docstrings

Other

• Added test script for S1.process.preprocess_insar_iw
• Automatically downloaded DEM (for S1 processing) file names are now based on unique hashes instead of being based on product/subswath/polarization.
• This has several advantages:
  • It makes filenames shorter
  • A new file is created every time a DEM related parameter changes (e.g. dem_upsampling) instead of overwriting the old one
  • It enables re-using DEM files across functions
  • When processing both polarizations, the DEM is not downloaded twice under a different name anymore
  • The parameter force_dem_download is now set to False by default

2024.9.0

New features

• New SLC processor to compute and geocode amplitude from a single SLC product.
• Calibration can now be either set to sigma or beta nought in both InSAR and SLC processors. For the InSAR processor it is useful when amplitude is computed.
• Both beta and sigma nought calibration factors are computed by the new S1IWSwath.calibration_factor function. Beta normalization is not applied anymore by S1IWSwath.read_burst.
• Processing functions process_* now have a cal_type parameter to control the calibration type.
• The docker version can now be built with or without SNAP (connect to the folder with SSH and select rebuild without cache).

Bug fixes

• dem_force_download warning: it is now displayed when the value is set to False.
• the warp_kernel parameter is now propagated in all processing functions.

Other

• Unit tests for S1IWSwath.calibration_factor and S1IWSwath.read_burst.
• Test script and example notebook for process_slc.
• More consistent logging messages.

2024.8.1

• Refactored the Sentinel-1 _process_bursts function of the InSAR processor
• Geocoding lookup tables (LUT) now uses a common DEM
• No more LUT merging after processing, safer handling of burst overlap to avoid gaps
• The processor uses a new function S1IWSwath.fetch_dem that downloads the DEM for all processed bursts
• The DEM is now upsampled after download, not at terrain correction stage
• Burst geometries are computed at S1IWSwath initialization and stored as a class member
• Burst geometries are now used instead of GCPs
• DEM cropping uses a buffer around the burst geometry
• Fixed a bug in the load_metadata function: rfi files are filtered out from the metadata file search
• Handling of non-float NaNs in DEM (at coordinate reading)
• range_doppler function skips NaN values
• Optimization: Using cache=True and nogil=True in coregister_fast and range_doppler
• Optimization: avoid the creation of multiple small arrays by coregister_fast results in speedup
• Bug fix: in the s1-core-demo notebooks, burst index testing has been corrected in the interferogram warping cell
• Better test data
• Wrote a test for the insar processor using this data

2024.8.0

This version contains bugfixes. rioxarray was not producing valid output COGs because it required extra options at raster creation.

Rendering of geocoded products on a map should be much faster.

Change list:

• Fix invalid COGs: use COG driver after merging
• Fix the geocoding of real-valued data by also using COG in sar2geo
• Using better COG creation options
  • ZSTD compression instead of DEFLATE
  • overview and projections with nearest neighbor resampling
  • using num_threads="all_cpus" to make compression faster

2024.7.0

Code improvements and some bugfixes:

• Burst Ids are checked only when available in the product metadata
• Improved code clarity by renaming some variables, explicitly naming positional parameters
• Added user friendly exceptions to check product validity
• Cleaned up some old comments
• Added numexpr to environments for upcoming optimizations
• Test data for full processor
• Optional directory for orbit files (default is \tmp)
• Added and improved existing logging messages
• Refactored resampling to use remap
• Fixed bug in resampling: using output profile of upsampled DEM instead of original DEM
• Changed example notebooks accordingly