Available with Image Analyst license.
A four-image multitemporal coherence (MTC) composite workflow creates a three-band visualization that highlights temporal patterns of surface stability and sudden change using synthetic aperture radar (SAR) data. It compares coherence values from three consecutive image pairs, from four different times.
This method is useful for detecting deformation events in areas that are normally stable. Since coherence is inherently low for vegetation, water, and shadows, a drop in coherence in an otherwise stable area provides evidence of significant change. This method uses images from four different times so that the output composite can detect which time frame that changes have occurred.
Data prerequisites
The following data is needed for this workflow:
- Four SAR datasets covering the same area of interest.
- All four inputs must have the same sensor mode, track (ascending or descending), and polarization. If the inputs are Sentinel-1 TOPS, the same subswath must be used for all inputs.
- A DEM that overlaps with the SAR images, which will be used to process and create accurate coregistration and terrain correction data.
Processing overview
To prepare the data for the coherence time series composite, coherence pairs must be generated for each of the image pairs using the coherence workflow. This will estimate the interferometric coherence between the images. If the inputs are Sentinel-1, the Sentiniel-1 coherence workflow must be followed.
Three coherence pairs must be created, using the coherence workflows:
- Coherence pair for time 1 and time 2
- Coherence pair for time 2 and time 3
- Coherence pair for time 3 and time 4
Create the MTC composite
The Generate Multitemporal Coherence tool can create the four-image MTC RGB composite by combining the three coherence pairs that were created.
| Band | Acquisition | Processing | Range |
|---|---|---|---|
Red band | Time 1 and Time 2 pair | 0 -1 | |
Green band | Time 2 and Time 3 pair | 0 -1 | |
Blue band | Time 3 and Time 4 pair | 0 -1 |
Interpretation
Each color in the RGB composite reflects how coherence changes across the time intervals. Interpretation depends on comparing coherence changes against areas of stability.
The table below explains how to interpret the MTC RGB composite output.
| Color | Band values | Interpretation |
|---|---|---|
White | Red—high backscatter Green—high backscatter Blue—high backscatter | Persistent coherence across all time intervals. |
Yellow | Red—high backscatter Green—high backscatter Blue—low backscatter | Change occurred between image 3 and image 4 |
Cyan | Red—medium to high backscatter Green—very high backscatter Blue—medium to high backscatter | Change occurred between image 1 and image 2. |
Magenta | Red—very high backscatter Green—medium to high backscatter Blue—medium to high backscatter | Change occurred between image 2 and image 3. |
Dark / Black | Red—low backscatter Green—low backscatter Blue—low backscatter | Consistently low coherence, such as water, dense vegetation, and radar shadow. |
Note:
Interpretation can vary in some cases, and more complex patterns may emerge, such as coherence changes across multiple image pairs. This can indicate several disturbances separated by periods of stability, however, interpretation should be approached cautiously and supported by additional information.
Processing considerations
Here are a few key processing considerations to ensure accurate results and proper visualization:
- Optimize coregistration results by using the highest-resolution DEM available. This helps ensure accurate coregistration. Poor coregistration will reduce coherence and introduces false changes.
- Use the same DEM resolution for processing all image pairs. The DEM source can be a different file, but the resolution should be consistent.
- Focus interpretation on regions that are normally stable. Low coherence in inherently unstable surfaces, such as water, vegetation, or shadow, has little analytical value.
- If the time between images is too long, coherence may drop due to natural decorrelation, such as vegetation growth or moisture variation, rather than a true change event.
- Use a layover mask or shadow mask to avoid misinterpreting radar geometry artifacts as change.
- Increase the coherence estimation window size if any band appears noisy.