(bug?) Reprojection of Sentinel-1 file causes values amendments

I noticed that when I run Reprojection on following file: ‘subset_calibrated_S1A_IW_GRDH_1SDV_20160512T161044_2_TC.dim’ (which I placed on my Google Drive
(https://drive.google.com/drive/folders/0B0VV80YQj6G2Smw4YlJNUHU5LTQ).
values in output raster are significantly different than in input raster

Before reprojection:

After reprojection:

I use default parameters:

My input file is Terrain-Corrected, subset of Sentinel-1 GRDH file saved in Beam-DIMAP.

Can’t you do Terrain Correction directly into the projection you want?

Before reprojection 0 is defined as no data value.
After reprojection nan is defined as no data value.

Many zeroes in your dataset significantly affect mean and median.

But you are right, the changes are too large to be caused by nodata only. Did your pixel size change?

I can’t open the file on your Google Drive because the dim file alone is just some of the metadata. You also need to upload the folder with the same name:
subset_calibrated_S1A_IW_GRDH_1SDV_20160512T161044_2_TC.data
This stores the actual rasters.

Thank you for quick response.
Actually all needed data are uploaded/available under my link.
Hope that following image will be helpful:

sorry, maybe it wasn’t uploeaded when I checked it first.
Now I can see both.

I tried it and can confirm the mismatch, but only as first sight:

If you uncheck the nodata value used option you will get the same statistics as before:

Thank you for another quick response.
You’re right that > nodata value used option affects significantly statistics.
However it seems that regardless of selected option they’re not exactly the same as before the reprojection.

Let me provide print screen with nodata value used option

selected:

and not selected:

in both cases mean is not equal to 0.0269 (as it was in file before reprojection)

could it be that it simply results from the resampling during the reprojection?

Almost each value gets computed newly due to the small shift of the pixel locations:

Good point, if one wants to preserve the old statistics as closely as possible one should pick nearest neighbor for the interpolation.

Could you tell me if you’re referring to the following option:

As you can see on my previously attached images, exactly this option has been selected.

Anyway, this is not my real problem that statistics are not exactly the same.
My issue is that when I try to set different size pixels in the output, I always receive NaN raster:
http://forum.step.esa.int/t/reprojection-of-sentinel-1-file-to-lower-resolution-always-generates-nan-raster
(it was tested on exactly the same data set).

Since I need to accomplish calculations on SMOS and Sentinel data sets, I would like to:

  • change resolution to lower on Sentinel data
  • change resolution to higher on SMOS data (so that I will have both products in the same resoluton)
  • (I guess?) apply collocation operator.

However, as I mentioned in different thread Resampling operator is not working for me and Reprojection is producing NaN raster (as mentioned above). Could you suggest anything, please?

Why don’t you Terrain Correct directly into the projection you need?

Could you tell me if you’re suggesting me to set:

  • mapProjection parameters
  • pixelSpacingInMeter to destination pixel size, so that I don’t need to use Resampling and/or Reprojection operators
    parameters directly in Terrain-Correction operator, please?

It should probably work for Sentinel GRDH, but I will need to change resolution of the SMOS data.
Will it be enough to apply Raster > Geometric Operations > Collocation to both (Terrain Corrected in proper pixel size and SMOS) to have SMOS resolution changed?

In the Range-Doppler Terrain Correction in the GUI you can set the output projection & pixel size, so you should not have to reproject afterwards.

Collocation should work between SMOS and SAR but I don’t think it has been tested, you could be the 1st one…