Exploring Sentinel-1 Bursts and Phase Information with ASF and Python (Colab)

Or How to Assess the Value of Polarimetric Phase Without Downloading Everything

Introduction

Working with Sentinel-1 SLC data is often heavy: SAFE products are large (several gigabytes), and burst-level analysis usually requires downloading everything. With the Alaska Satellite Facility (ASF) burst service, we can now directly access burst-level GeoTIFFs, making it much easier to quickly explore VV/VH intensity and phase.

If we only need already-prepared datasets, cloud platforms like Google Earth Engine, Planetary Computer, or Sentinel Hub are fantastic. They give fast, processed access to GRD and analysis-ready products.
But here is the limitation: they still don’t provide SLC bursts, which means the phase information is missing.

That’s frustrating not only for interferometry (InSAR), but also for more general polarimetric exploration: we often forget that the relative phase between VV and VH carries rich information about scattering mechanisms.

The catch is: downloading a full Sentinel-1 SLC product just to explore this is overkill — too big, too heavy.

Luckily, ASF now provides direct burst downloads. This notebook shows how to quickly fetch a burst, and take a first look at what the polarimetric phase reveals — without headaches, and without terabytes of data.

Polarimetry with Sentinel-1

Sentinel-1 Interferometric Wide (IW) mode provides dual polarization data (VV + VH or HH + HV), not full quad-pol acquisitions.
This limitation means we cannot directly reconstruct the full scattering matrix, but it is still possible to describe the polarization ellipse of the scattered wave from the dual-pol channels.

From this polarization ellipse, two key parameters can be derived:

• Orientation of the ellipse (related to the dominant scattering direction).
• Ellipticity of the ellipse (related to the type of scattering, e.g. surface vs. volume).

Beyond these basic descriptors, we can also look at the statistical stability of the ellipse. This requires second-order estimations, i.e. computing the covariance matrices built from Jones vectors.

There are two main strategies:

• Temporal averaging:
  Recommended in most cases, because it avoids spatial resolution loss. However, it requires a time series of coregistered data, which is not always available.
• Spatial averaging:
  This is the approach we use in this notebook for simplicity. We compute local covariance (Jones) matrices within a spatial neighborhood. The neighborhood is not square, but aligned with the sensor’s resolutions: typically, there is a factor of about 4 between the azimuth and range resolutions.

This way, even with dual-pol Sentinel-1 bursts, we can already explore meaningful polarimetric parameters — at least qualitatively — and highlight the additional information hidden in the phase between VV and VH.

Color composition (HSV)

To make the polarimetric information more intuitive, we propose a color composition in HSV space:

• Hue → orientation of the polarization ellipse (angle of the symmetry axis of the scattering target, projected in the wave plane).
• Saturation → degree of polarization (how stable the polarization state is).
• Value (intensity) → SPAN (total backscattered power).

Interpretation:

• For flat surfaces, the average orientation angle is close to zero, which corresponds to cyan in this representation
• When the structure of the target modifies the scattering symmetry (e.g. relief, oriented structures), the hue shifts accordingly.
• For random volumes, the depolarization is high → saturation is low → the color tends toward gray.
• Temporal averaging usually gives a more reliable estimate of stability and saturation than spatial averaging, but even with local spatial windows we can already explore meaningful contrasts.

This color representation helps us see at a glance what the phase between VV and VH reveals — a piece of information that is still ignored by most current deep learning pipelines applied to Sentinel-1.

Notebook outline

This Colab notebook is structured into the following steps:

1. Data access via ASF
   Request and download a burst-level SLC GeoTIFF directly from the Alaska Satellite Facility, without downloading the full SAFE product.
2. Amplitude quick look
   Display intensity images for VV and VH, and compare their backscatter characteristics.
3. HSV composition
   Build and display the color-coded polarimetric image. This acts as a first visual exploration of what phase adds to burst-level SLC data.

Examples and tips

1. Create an account

Access to ASF burst data requires a NASA Earthdata Login (EDL):
🔗 https://urs.earthdata.nasa.gov/

2. Search for bursts and get download link

Use the ASF Search interface:
🔗 https://search.asf.alaska.edu/

• In Dataset, select S1 Burst.
• Define your area of interest, date range, and other filters.
• Add products to the Download Queue.
• The download links appear in the list (⚠️ note: the link is an HTTPS URL, not the ASF product ID).

Example link (direct TIFF burst file, Paris area):
https://sentinel1-burst.asf.alaska.edu/S1B_IW_SLC__1SDV_20210420T094051_20210420T094119_026549_032B99_9B8D/IW1/VH/4.tiff

And here is the corresponding image (Paris, France)

3. Common issues

• HTTP 202 (Accepted)
  The request was received but the file is not ready yet.
  ASF generates burst files on demand: the first request triggers the server-side processing. This can take a few seconds (sometimes minutes). If you hit a 202 response, wait and retry the download.
• HTTP 502 (Bad Gateway)
  This usually indicates either a temporary ASF service issue, or a request for a burst/polarization that does not exist for that product. Verify the burst index and polarization in the ASF interface before retrying.

Tip: You can implement retries in your download script to handle these cases gracefully.

To conclude…

👉 The main goal is to give a quick look at what the phase adds in Sentinel-1 SLC bursts, without the burden of downloading full products.

🔗 Notebook available on GitHub:

Tutorials_Notebooks/S1_burst_ASF_to_numpy_VV_VH.ipynb at main · elisecolin/Tutorials_Notebooks

For readers new to SAR polarimetry, I also recommend this introduction:

What is a Jones vector and a polarization ellipse

And:
Explore VV, VH or HH, HV SAR time series

Here is a result on Chamonix Valley:

A result extract on Himalaya: