Change detection of the effects of a wind downburst

Hi all,
I am struggling to work on images stacks in order to perform change detection. The purpose of my analysis is to find the effects of a wind downburst that damaged some roofs and trees some months ago in a small area near Turin (Italy). From the survey on the field I know some locations of the damages and I would expect to find them on the change detection results (besides I would like to find some damages not seen on the field…). However it doesn’t seem there is a clear correlation between the results and the effects found by the survey. So I am wondering if the scale of phenomenon can’t be detected by S1 images or I did something wrong in the process I followed. An example of the effects on the roofs can be found here:

I considered two GRD S1 images taken at different dates (13th June-7th July 2016). Then I followed this work flow:
Apply Orbit files --> Subset --> Radiometric Calibration --> Terrain Flattening --> Terrain Correction --> Create Stack --> Change Detection

This is the RGB stack (the yellow polygon is the Area of Interest):

Change Detection results (VH polarization):

Zoom on the AoI

Playing with the threshold of the ratio values doesn’t show me any interesting feature related to damages…

Maybe I should apply speckle filtering before any other step in my work flow? Any suggestion is welcomed.
Thank you in advance


Hi Gabriele

Very interesting your survey!!! :smiley:

@gabbro: Consider integrating some more images for your analysis - maybe 3 before and 3 after the incident. Make a stack of them and apply a multi-temporal filter.
It really can increase the image quality, especially in urban areas. See my example here.

1 Like

Hello @ABraun, thanks a lot, your suggestion sounds very interesting. I’ll try and keep you informed.

Many thanks

Hi, I have 3 before and 3 after images but when I want to select them for the Change Detection, I can only select them 2 by 2 and in this way I cannot have one unique file. How can I fix it?

change detection always compares pairs of images. You cannot display changes between three images within one raster. So you would need to compare A and B, then A and C and then B and C. You can stack all three results into one file with the Create Stack module and maybe make an RGB composite or add them up to one final image.