{% extends "guides/insar_product_guide_template.md" %}
{% block header %}
This document is a guide for users of Sentinel-1 Burst Interferometric Synthetic Aperture Radar (InSAR) products generated by the Alaska Satellite Facility (ASF).
Single Look Complex (SLC) data from the Sentinel-1 mission that is suitable for use in interferometry has historically been packaged into Interferometric Wide (IW) SLC products. These IW SLC products include three sub-swaths, each containing many individual burst SLCs. The framing of the IW SLCs is not consistent through time, so when using IW SLCs as the basis for InSAR, scene pairs do not always fully overlap.
In contrast, working at the burst level of the Sentinel-1 SLC data provides a couple key benefits:
1. Bursts are consistently geolocated through time
The coverage of a burst is the same for every orbit of the satellite, so you can be confident that every burst with the same Full Burst ID{target=_blank} in a stack of acquisitions will cover the same geographic location.
2. Bursts cover a smaller geographic area
IW SLC products are extremely large, and in many cases, only a small portion of the image is of interest. You can process only the bursts that cover your specific area of interest, which significantly decreases the time and cost required to generate InSAR products.
Refer to the Sentinel-1 Bursts tutorial{target=_blank} to learn more about how ASF extracts burst-level products{target=_blank} from Sentinel-1 IW and EW SLCs.
The Sentinel-1 Burst InSAR products are generated using the Jet Propulsion Laboratory's ISCE2 software{target=_blank}. ASF is committed to transparency in product development, and we are pleased to be able to offer an InSAR product that leverages open-source software for processing.
For those who would prefer to work at the scale of a full IW SLC, our original On Demand InSAR{target=_blank} products are still available. These products have a larger footprint, and are generated using GAMMA software{target=_blank}.
Users can request Sentinel-1 Burst InSAR products On Demand{target=_blank} in ASF's Vertex{target=_blank} data portal, or make use of our HyP3 Python SDK{target=_blank} or API{target=_blank}. Input pair selection in Vertex uses either the Baseline Tool{target=_blank} or the SBAS Tool{target=_blank} search interfaces.
Users are cautioned to read the sections on limitations and error sources in InSAR products before attempting to use InSAR data. For a more complete description of the properties of SAR, see our Introduction to SAR{target=_blank} guide.
{% endblock %}
{% block processing_options %}
There are several options users can set when ordering Burst InSAR On Demand products:
-
The number of looks drives the resolution and pixel spacing of the output products:
Looks Resolution Pixel Spacing 20x4 160 m 80 m 10x2 80 m 40 m 5x1 40 m 20 m Products generated with 10x2 looks have a file size roughly 4 times that of 20x4-look products. Similarly, 5x1-look products have a file size roughly 4 times that of 10x2-look products (or 16 times that of 20x4-look products).
The default is 20x4 looks.
-
There is an option to apply a water mask after phase unwrapping. This mask includes coastal waters and large inland waterbodies. A GeoTIFF of the water mask is always included with the InSAR product package, but when this option is selected, the water mask will be applied to the wrapped interferogram, the unwrapped interferogram, and the browse image. Water masking is turned off by default. Refer to the Apply Water Mask section for more information about the water mask and how it is used.
{% endblock %}
{% block workflow %}
The Burst InSAR workflow used in HyP3 was developed by ASF using ISCE2 software. The steps include pre-processing, interferogram preparation, and product creation. Once these steps are performed, an output product package is created. See product packaging for details on the individual files included in the package.
Pre-processing steps prepare the SAR images to be used in interferometry. The pre-processing steps include downloading the burst granules, downloading the DEM file, and downloading the orbit and auxiliary data files.
The burst InSAR workflow accepts as input two Interferometric Wide swath Single Look Complex{target=_blank} (IW SLC) burst granules with the same burst ID. The bursts are downloaded using ASF's Sentinel-1 Burst Extractor{target=_blank}.
In order to create differential InSAR products that show motion on the ground, one must subtract the topographic phase from the interferogram. The topographic phase, in this case, is replicated by using an existing DEM{target=_blank} to calculate the actual topographic phase. This phase is then removed from the interferogram leaving just the motion or deformation signal (plus atmospheric delays and noise).
The DEM that is used for HyP3 InSAR processing is the 2021 Release of the Copernicus GLO-30 Public DEM{target=_blank} dataset publicly available on AWS{target=_blank}, which provides global coverage at 30-m pixel spacing (except for an area over Armenia and Azerbaijan, which only has 90-m coverage). For more information about the 2021 updates, see the 'Releases' section of this article{target=_blank}.
The portion of the DEM that covers the input bursts is downloaded.
For Sentinel-1 InSAR processing, ISCE2 requires additional satellite orbit and calibration metadata files. The orbit files are downloaded from the Copernicus Sentinels POD Data Hub{target=_blank}. The calibration auxiliary data files are downloaded from the Sentinel-1 Mission Performance Center{target=_blank}.
The ISCE2 InSAR processing this product uses follows the workflow in topsApp.py{target=_blank} from steps startup
through geocode
. These steps perform the following processing:
- Extract the orbits, Instrument Processing Facility (IPF) version, burst data, and antenna pattern if it is necessary.
- Calculate the perpendicular and parallel baselines.
- Map the DEM into the radar coordinates of the reference image. This generates the longitude, latitude, height and LOS angles on a pixel by pixel grid for each burst.
- Estimate the azimuth offsets between the input SLC bursts. The Enhanced Spectral Diversity (ESD) method is not used.
- Estimate the range offsets between the input SLC bursts.
- Co-register the secondary SLC burst by applying the estimated range and azimuth offsets.
- Produce the wrapped phase interferogram.
- Apply the Goldstein-Werner{target=_blank} power spectral filter with a dampening factor of 0.5.
- Unwrap the wrapped phase interferogram using SNAPHU{target=_blank}'s minimum cost flow (MCF) unwrapping algorithm to produce the unwrapped phase interferogram.
- Geocode the output products.
A water mask identifying coastal waters and major inland waterbodies is generated using the Global Self-consistent, Hierarchical, High-resolution Geography Database (GSHHG){target=_blank} dataset. This water mask raster is always included with the Burst InSAR products for reference, but is not applied to the interferometry products by default.
Users can optionally choose to apply the water mask to output products, which affects the wrapped interferogram, the unwrapped interferogram, and the browse image. Areas covered by the water mask in these output images are set to NoData.
Applying a water mask to an interferogram is only supported after phase unwrapping. Note that applying the mask after phase unwrapping does not prevent unwrapping errors caused by the inclusion of water pixels as valid data during the phase unwrapping process. When phase unwrapping occurs over large expanses of water, it can lead to unexpected deformation signals or phase jumps in the unwrapped outputs, and the current masking approach does not correct for these impacts.
Image files are exported into the widely-used GeoTIFF format in a Universal Transverse Mercator (UTM) projection. Images are resampled to a pixel size that reflects the resolution of output image based on the requested number of looks: 80 meters for 20x4 looks, 40 meters for 10x2 looks, and 20 meters for 5x1 looks.
Supporting metadata files are created, as well as a quick-look browse image.
{% endblock %}
{% block packaging %}
HyP3 Burst InSAR output is a zip file containing various files, including GeoTIFFs, a PNG browse image, a metadata file, and a README file.
The Burst InSAR product names are packed with information pertaining to the processing of the data, presented in the following order, as illustrated in Figure 3.
- The imaging platform name, always S1 for Sentinel-1.
- Relative burst ID values assigned by ESA. Each value identifies a consistent burst footprint; relative burst ID values differ from one sub-swath to the next.
- The imaging mode, currently only IW is supported.
- The swath number, either 1, 2, or 3, indicating which sub-swath the burst is located in.
- The acquisition dates of the reference (older) scene and the secondary (newer) scene
- The polarizations for the pair, either HH or VV.
- The product type (always INT for InSAR) and the pixel spacing, which will be either 80, 40, or 20, based upon the number of looks selected when the job was submitted for processing
- The filename ends with the ASF product ID, a 4 digit hexadecimal number
Figure 3: Breakdown of ASF Burst InSAR naming scheme.
All of the main InSAR product files are 32-bit floating-point single-band GeoTIFFs. The exceptions to this are the connected components and the water mask, which are both 8-bit unsigned-integer single-band GeoTIFFs.
- The coherence file pixel values range from 0.0 to 1.0, with 0.0 being completely non-coherent and 1.0 being perfectly coherent.
- The unwrapped phase file shows the results of the phase unwrapping process. Negative values indicate movement towards the sensor, and positive values indicate movement away from the sensor. This is the main interferogram output.
- The wrapped phase file indicates the interferogram phase after applying the adaptive filter immediately before unwrapping. Values range from negative pi to positive pi.
- The connected components file delineates regions unwrapped as contiguous units by the SNAPHU unwrapping algorithm.
- The look vectors theta (θ) and phi (φ) describe the elevation and orientation angles of the look vector in radians. The look vectors refer to the look direction back towards the sensor.
- The lv_theta (θ) file indicates the SAR look vector elevation angle (in radians) at each pixel, ranging from -π/2 (down) to π/2 (up). The look vector elevation angle is defined as the angle between the horizontal surface and the look vector with positive angles indicating sensor positions above the surface.
- The lv_phi (φ) file indicates the SAR look vector orientation angle (in radians) at each pixel. The look vector orientation angle is defined as the angle between the East direction and the projection of the look vector on the horizontal surface plane. The orientation angle increases towards north, with the North direction corresponding to π/2 (and south to -π/2). The orientation angle range is -π to π.
- The DEM file gives the local terrain heights in meters, with a geoid correction applied.
- The water mask file indicates coastal waters and large inland waterbodies. Pixel values of 1 indicate land and 0 indicate water. This file is in 8-bit unsigned integer format.
If the water mask option is selected, the water mask is applied after phase unwrapping to exclude water pixels from the output. The water mask is generated using the GSHHG{target=_blank} dataset. To compile the reference shapefile, the full-resolution L1 dataset (boundary between land and ocean) and L5 dataset (boundary between Antarctic ice and ocean) were combined. The L3 dataset (boundary between islands and the lakes they are within) was removed from the L2 dataset (boundary between lakes and land), and this combined dataset was removed from the combined L1/L5 dataset. The GSHHG dataset was last updated in 2017, so there may be discrepancies where shorelines have changed.
A browse image is included for the unwrapped (unw_phase) phase file, which is in PNG format and is 2048 pixels wide.
The tags and extensions used and example file names for each raster are listed in Table 2 below.
{% set base_name = 'S1_136231_IW2_20200604_20200616_VV_INT80_12E3' %}
Extension | Description | Example |
---|---|---|
_conncomp.tif | Connected Components | {{ base_name }}_conncomp.tif |
_corr.tif | Normalized coherence file | {{ base_name }}_corr.tif |
_unw_phase.tif | Unwrapped geocoded interferogram | {{ base_name }}_unw_phase.tif |
_wrapped_phase.tif | Wrapped geocoded interferogram | {{ base_name }}_wrapped_phase.tif |
_lv_phi.tif | Look vector φ (orientation) | {{ base_name }}_lv_phi.tif |
_lv_theta.tif | Look vector θ (elevation) | {{ base_name }}_lv_theta.tif |
_dem.tif | Digital elevation model | {{ base_name }}_dem.tif |
_water_mask.tif | Water mask | {{ base_name }}_water_mask.tif |
_unw_phase.png | Unwrapped phase browse image | {{ base_name }}_unw_phase.png |
Table 2: Image files in product package
The product package also includes a number of metadata files.
Extension | Description | Example |
---|---|---|
.README.md.txt | Main README file for Burst InSAR products | {{ base_name }}.README.md.txt |
.txt | Parameters and metadata for the InSAR pair | {{ base_name }}.txt |
Table 3: Metadata files in product package
The text file with extension .README.md.txt explains the files included in the folder, and is customized to reflect that particular product. Users unfamiliar with InSAR products should start by reading this README file, which will give some background on each of the files included in the product folder.
The text file with extension .txt includes processing parameters used to generate the InSAR product as well as metadata attributes for the InSAR pair. These are detailed in Table 4.
Name | Description | Possible Value |
---|---|---|
Reference Granule | Granule name for reference burst (of the two scenes in the pair, the dataset with the oldest timestamp) | S1_136231_IW2_20200604T022312_VV_7C85-BURST |
Secondary Granule | Granule name for secondary burst (of the two scenes in the pair, the dataset with the newest timestamp) | S1_136231_IW2_20200616T022313_VV_5D11-BURST |
Reference Pass Direction | Orbit direction of the reference scene | DESCENDING |
Reference Orbit Number | Absolute orbit number of the reference scene | 30741 |
Secondary Pass Direction | Orbit direction of the reference scene | DESCENDING |
Secondary Orbit Number | Absolute orbit number of the secondary scene | 31091 |
Baseline | Perpendicular baseline in meters | 58.3898 |
UTCTime | Time in the UTC time zone in seconds | 12360.691361 |
Heading | Spacecraft heading measured in degrees clockwise from north | 193.2939317 |
Spacecraft height | Height in meters of the spacecraft above nadir point | 700618.6318999995 |
Earth radius at nadir | Ellipsoidal earth radius in meters at the point directly below the satellite | 6370250.0667 |
Slant range near | Distance in meters from satellite to nearest point imaged | 799517.4338 |
Slant range center | Distance in meters from satellite to the center point imaged | 879794.1404 |
Slant range far | Distance in meters from satellite to farthest point imaged | 960070.8469 |
Range looks | Number of looks taken in the range direction | 20 |
Azimuth looks | Number of looks taken in the azimuth direction | 4 |
InSAR phase filter | Name of the phase filter used | yes |
Phase filter parameter | Dampening factor | 0.5 |
Range bandpass filter | Range bandpass filter applied | no |
Azimuth bandpass filter | Azimuth bandpass filter applied | no |
DEM source | DEM used in processing | GLO-30 |
DEM resolution | Pixel spacing in meters for DEM used to process this scene | 30 |
Unwrapping type | Phase unwrapping algorithm used | snaphu_mcf |
Speckle filter | Speckle filter applied | yes |
Table 4: List of InSAR parameters included in the parameter text file
{% endblock %}
{% block line_of_sight %}
When looking at a single interferogram, the deformation measurements in the line-of-sight orientation of the sensor indicate relative motion towards or away from the sensor. InSAR is not sensitive to motion in the azimuth direction of the satellite, so motion that occurs in the same direction as the satellite's direction of travel will not be detected.
A single interferogram cannot be used to determine the relative contributions of vertical and horizontal movement to the line-of-sight displacement measurement. To determine how much of the signal is driven by vertical vs. horizontal movement, you must either use a time series of interferograms, or use reference measurements with known vertical and horizontal components (such as GNSS measurements from the region of deformation) to deconstruct the line-of-sight displacement.
{% endblock %}
{% block reference_point %}
The reference point for phase unwrapping is set automatically by the topsApp.py script. It may not be an ideal location to use as a reference point for phase unwrapping. If it is located in an area undergoing deformation, or in an area with low coherence, the unwrapping may be of lower quality than if the reference point was in a more suitable location.
Even when there are no phase unwrapping errors introduced by phase discontinuities, it is important to be aware that unwrapped phase differences are calculated relative to the reference point. The phase difference value of the reference point is set to 0 during phase unwrapping, so any displacement values will be relative to that benchmark.
If you are interested in the amount of displacement in a particular area, you may wish to choose your own reference point. The ideal reference point would be in an area of high coherence beyond where deformation has occurred. The unwrapped phase measurements can be adjusted to be relative to this new reference point. To adjust the values in the unwrapped phase GeoTIFF, simply select a reference point that is optimal for your use case and subtract the unwrapped phase value of that reference point from each pixel in the unwrapped phase raster:
ΔΨ* = ΔΨ - Δψref
where ΔΨ* is the adjusted unwrapped phase, ΔΨ is the original unwrapped phase, and Δψref is the unwrapped phase value at the new reference point.
In general, calculating displacement values from a single interferogram is not recommended. It will be more robust to use a time series approach to more accurately determine the pattern of movement. When using SAR time-series software such as MintPy{target=_blank}, you have the option to select a specific reference point, and the values of the input rasters will be adjusted accordingly. {% endblock %}