I have a problem with processing S1 GRD products with SNAP. I believe there is an error in the way resampled DEM values are calculated in the Terrain-Flattening algorithm.
I did a simple test with the Terrain Flattening operator:
I preprocessed a whole S1 product, and some area inside that product with different bounding boxes around it (200m, 5 km, 5.1km, and 6km).
After that, I collocated these terrain-flattened products and checked the value of gamma0 for some pixels in the center of the selected area. To my surprise, the pixel values were all different (for the whole product and for the subsets with different boxes).
I know that Terrain-Flattening operator is based on this paper:
, and that it uses bilinear DEM resampling (which is the default for TF operator also). So my question is - how is it possible that the gamma values for pixels in the center of the area are affected by the size of the image processed (different bounding boxes)?
Between 5km and 6km bounding box, nothing has changed for that pixel - it still corresponds to one or more( or less) pixels in the DEM, the same DEM pixels (should) influence the sigma0 output for that pixel. But surely pixels far away from that location shouldn’t affect the output value.
Is there a bug in the Terrain-Flattening algorithm or am I missing something?