Estimating Forest Above ground biomass from sentinel imagery using gray level co_occurence matrix

seid · October 16, 2018, 9:30am

hi, would you help me how to estimate above ground biomass from sentinel imagery using gray level co_occurrence matrix

mengdahl · October 16, 2018, 9:33am

I am not aware of studies where that was shown to be possible.

mfitrzyk · October 16, 2018, 9:35am

There are some papers showing positive correlation between texture and biomass.
eg. https://ieeexplore.ieee.org/document/1294615
The GLCM calculation is possible in SNAP from Raster->Image analysis->Texture analysis

ABerni · October 16, 2018, 11:46am

something similar but also including Sentinel: https://www.mdpi.com/2072-4292/10/6/831

seid · November 14, 2018, 2:33pm

Hi is there any one who help me to correlate forest above ground biomass with vegetation indices

seid · January 19, 2019, 2:15pm

Hi how to do gray level co occurrence matrix from sentinel 1 GRD imagery using SNAP software

ABraun · January 19, 2019, 2:30pm

You find it under Raster > Image Analysis > Texture Analysis > Grey-Level Co-Occurence Matrix

seid · January 19, 2019, 2:35pm

Thank you very much, I get it.But it unable to do the processing,what preprocessing steps i have to do

ABraun · January 19, 2019, 2:47pm

Preprocessing of Sentinel-1 is described here: Radiometric & Geometric Correction Workflow
GRD data is sufficient for the derivation of textures. Don’t apply speckle filtering, because it reduces image texture.

seid · January 19, 2019, 2:50pm

Thank you very much!

seid · January 19, 2019, 3:12pm

I have followed all the steps but the final result after doing Range Doppler Terrain Correction is smooth.So is there a way to change to normal image or any other steps to follow?

ABraun · January 19, 2019, 4:10pm

Can you give an example of what you mean by smooth, maybe with a screenshot?

Which steps exactly did you apply?

Terrain Correction includes resampling of the pixels which slightly changes their values (two neighboring pixels might get changed to one mean value). You can select nearest neighbor resampling during the TC step but this should make not a too big difference.
Otherwise you can increase the radius in the GLCM to derive textures at different scales.

seid · January 21, 2019, 6:48am

seid · January 21, 2019, 6:53am

I have followed these steps;
With out (Speckle filtering),and calibrated to sigm0

Thermal noise removal
Apply Orbit file
Calibration to Beta0
(Speckle filtering)
Radiometric terrain flattening
Range Doppler Terrain Correction

But the result is black(smooth) as you can see from the figure.

ABraun · January 21, 2019, 7:36am

How does it look after Terrain Flattening?

seid · January 21, 2019, 8:34am

It unable to do Radiometric terrain flattening on sigma0. But it works on Beta0.

The Range Doppler Terrain Correction result from this TF is the following image which is black.

seid · January 21, 2019, 1:22pm

Sorry,Is there a solution for this problem?

ABraun · January 21, 2019, 6:58pm

have you tried switching to the SRTM 1Sec instead of 3Sec?
There were problems reported lately regarding the 90m SRTM.

seid · January 23, 2019, 8:09am

Yes I have tried but still doesn’t work

seid · January 31, 2019, 9:43am

Hi ABraun, The Geometric correction works on Ellipsoid correction(Average Height Range Doppler. As I read from the manual it works on the average height of the backscatter rather than DEM.Is there much difference between this result and the one which is done using Range Doppler correction? and also is the geometric correction has an effect on the backscatter information(pixel value).