Question on slope computation

[laetitialalla]

Hello,

i have (again ;) ) another question about data pre-processing in your dataset : How do you compute the slope from the elevation data ?

I have an implementation of Horn algorithm (with a Sobel convolution), as described here in the "planar method" section. I tried it on your elevation data (to be precise, on scene 6) but I cannot reproduce exactly your results.

Here is an histogram of the distribution of slope values in :

• your data _sentinel12_copdem30_6slope.tif (blue)
• the data I computed (orange) : a planar method on _sentinel12_copdem30_6elevation.tif In X : the value of the slope. In Y : the % of pixels that have that value . With your method, the slope has more 0s and less high values.
  

maybe you use a different algorithm ? Or perhaps you do some post-processing on the slope data ?

Again, thanks a lot for your help :) This dataset is VERY useful !

All the best, Laetitia

[MWieland]

Hi Laetitia, Many thanks for your continued interest in the dataset. We are happy that it is useful for you. We actually used the same method to compute the slope as you referenced above. No post-processing was applied. Based on this it is hard to identify the actual cause of the mismatch. My recommendation would anyway be to start from the original elevation data and compute the slope using your own method. This way you ensure higher consistency between training and inference stages of your machine learning project. Cheers, Marc

[MWieland]

Closing this issue due to inactivity. Please feel free to reopen, should you have further questions regarding slope computation.