SAR Remote Sensing for Change Detection Training Course

Participants can select suitable SAR acquisitions and document a data plan for a change-detection project.

• Sentinel-1 acquisition modes, swaths, polarisations and revisit intervals
• ALOS-2 PALSAR-2 and TerraSAR-X key mission selection criteria
• STAC metadata fields used to screen candidate SAR scenes
• Plan acquisition pairs around event dates, seasons and tidal stage
• Query ASF Vertex and Copernicus Data Space catalogues
• Document licensing, orbit direction and acquisition constraints in project registers

Participants can convert raw SAR scenes into calibrated inputs suitable for quantitative analysis.

• Apply precise orbit files and thermal noise removal
• Calibrate digital numbers to beta nought, sigma nought or gamma nought
• Remove border noise using Sentinel-1 acquisition-specific masks
• Subset scenes using analysis areas, buffers and common extents
• Record processing parameters in a reproducible workflow log
• Inspect histograms and quicklooks before downstream transformations

Participants can select and assess speckle filters without obscuring material change features.

• Select Lee, Refined Lee and Gamma MAP speckle filters
• Set filter window size against feature scale and texture
• Assess edge preservation, radiometric bias and residual speckle
• Use multi-temporal filtering without contaminating event-related change signals
• Convert calibrated backscatter to decibels for visual comparison
• Build RGB composites from polarisation, dates and derived indices

Participants can produce geocoded SAR products with appropriate terrain treatment and raster settings.

• Perform range-Doppler terrain correction using a suitable DEM
• Choose Copernicus DEM resolution for terrain and target size
• Resample raster outputs with nearest neighbour or bilinear methods
• Verify projected coordinates, pixel spacing and map alignment
• Mask radar shadow and layover using incidence-angle geometry
• Export Cloud Optimized GeoTIFF products with complete metadata

Participants can prepare multi-date SAR scenes for reliable pixel-level comparison.

• Co-register multi-date scenes to sub-pixel positional tolerance standards
• Measure tie-point residuals over stable built structures
• Separate true change from geolocation, incidence-angle and seasonal effects
• Normalise backscatter using reference areas and relative radiometric adjustment
• Create analysis masks for water, terrain distortion and unreliable pixels
• Specify minimum detectable change relative to noise and registration error

Participants can create and interpret backscatter change maps using established image comparison methods.

• Calculate log-ratio and log-difference change images from calibrated pairs
• Apply image differencing thresholds using stable-reference distributions
• Use Change Vector Analysis across dual-polarisation feature space
• Generate ratio composites for rapid flood and disturbance screening
• Map positive and negative backscatter changes with directional classes
• Compare threshold outputs against optical imagery and known control sites

Participants can use coherence change to identify surface disturbance while accounting for decorrelation.

• Form interferometric pairs using baseline, temporal separation and Doppler criteria
• Estimate coherence to detect surface disturbance and structural change
• Interpret decorrelation from vegetation, water, snow and long baselines
• Apply Goldstein filtering before phase unwrapping and displacement estimation
• Differentiate coherence loss from true displacement in incident assessments
• Produce coherence-change maps with confidence masks and explanatory notes

Participants can produce and qualify line-of-sight displacement estimates from differential interferograms.

• Build differential interferograms from master, slave and reference DEM inputs
• Remove topographic phase and flatten interferometric phase surfaces
• Unwrap phase using SNAPHU and inspect unwrapping discontinuities
• Convert line-of-sight phase to displacement with wavelength scaling
• Assess atmospheric delay, orbital ramps and unwrapping error sources
• Report displacement direction, uncertainty and temporal interpretation limits

Participants can derive polarimetric features and relate them to physical scattering changes.

• Represent scattering using Sinclair matrices and covariance matrices
• Interpret Pauli RGB decomposition for double-bounce, volume and surface scattering
• Apply Freeman-Durden decomposition to separate dominant scattering mechanisms
• Extract entropy, anisotropy and alpha-angle polarimetric features
• Select polarimetric features for urban damage and vegetation change
• Account for orientation effects and polarimetric calibration limitations

Participants can identify persistent, seasonal and abrupt change from SAR time-series data.

• Assemble Sentinel-1 time series with consistent geometry and preprocessing
• Calculate seasonal baselines using median composites and anomaly scores
• Apply CUSUM change detection to persistent backscatter shifts
• Use BFAST-style decomposition for trend, seasonality and break detection
• Identify acquisition gaps, outliers and anomalous processing artifacts
• Summarise parcel-level trajectories with dates, magnitude and confidence

Participants can train and assess supervised classifiers for mapped change categories.

• Design training samples representing change, no-change and confounding classes
• Extract backscatter, coherence, texture and topographic predictor variables
• Train Random Forest classifiers with stratified reference samples
• Tune support vector machine kernels and class-weight settings
• Separate training, validation and test data by spatial blocks
• Diagnose feature importance, class confusion and spatial overfitting

Participants can combine complementary geospatial data sources to improve change-detection decisions.

• Align SAR and optical grids before pixel-level feature fusion
• Combine Sentinel-1 and Sentinel-2 indicators for flood extent mapping
• Fuse DEM derivatives with SAR features for landslide susceptibility change
• Use object-based image analysis to aggregate segmented target features
• Resolve cloud, shadow and temporal mismatch between sensor observations
• Select early, feature-level or decision-level fusion architectures

Participants can quantify map accuracy and communicate the principal limitations of a change product.

• Construct probability samples using stratified random and systematic designs
• Create reference labels from field records, imagery and authoritative datasets
• Calculate confusion matrices, user's accuracy, producer's accuracy and F1 scores
• Estimate area-adjusted accuracy and confidence intervals for mapped change
• Set detection thresholds from receiver operating characteristic curves
• Document positional, thematic and temporal uncertainty in final outputs

Participants can package SAR change analysis into reproducible workflows and usable operational reports.

• Design reproducible SNAP graphs, GDAL commands and parameter files
• Organise inputs, intermediate rasters and outputs using project naming conventions
• Publish map layers with symbology, scale ranges and metadata
• Write change-detection findings for technical reviewers and decision makers
• Apply data governance rules for sensitive locations and incident imagery
• Define update frequency, alert thresholds and analyst review procedures

Participants can build, review and present a complete SAR change-detection product for a defined operational decision.

• Frame a client-relevant change question, area of interest and decision use
• Select SAR acquisitions and complementary data against stated evidence requirements
• Process calibrated, terrain-corrected and co-registered analysis-ready raster inputs
• Build and validate a change product using documented methods
• Prepare map layouts, accuracy evidence and operational recommendations
• Present findings, limitations and next-step monitoring actions for peer review