Mismatch in raster width after slice assembly and multilooking

Microwave Toolbox
1

I’ve encountered mismatching raster dimensions after slice assembling and multi-looking Sentinel-1 GRD acquisitions.

Starting with the three products

S1A_IW_GRDH_1SDV_20250416T154949_20250416T155014_058784_074866_7CDF.SAFE
S1A_IW_GRDH_1SDV_20250416T155014_20250416T155039_058784_074866_4586.SAFE
S1A_IW_GRDH_1SDV_20250416T155039_20250416T155104_058784_074866_23CC.SAFE

I applied the operators Calibration and ThermalNoiseRemoval on each using SNAP GPT and then applied SliceAssembly, Multilook with a factor of 8 and finally LinearToFromdB.

In the resulting BEAM-DIMAP file S1A_IW_GRDH_1SDV_20250416T154949_20250416T155104_058784_074866_7CDF_Cal_tnr_Asm_ML_dB.dim there are mismatching values for the raster width:

...
    <Raster_Dimensions>
        <NCOLS>3217</NCOLS>
        <NROWS>6241</NROWS>
        <NBANDS>2</NBANDS>
    </Raster_Dimensions>
...
    <Image_Interpretation>
        <Spectral_Band_Info>
            <BAND_INDEX>0</BAND_INDEX>
            <BAND_DESCRIPTION />
            <BAND_NAME>Sigma0_VV_db</BAND_NAME>
            <BAND_RASTER_WIDTH>3216</BAND_RASTER_WIDTH>
            <BAND_RASTER_HEIGHT>6241</BAND_RASTER_HEIGHT>
            ...
        </Spectral_Band_Info>
        <Spectral_Band_Info>
            <BAND_INDEX>1</BAND_INDEX>
            <BAND_DESCRIPTION />
            <BAND_NAME>Sigma0_VH_db</BAND_NAME>
            <BAND_RASTER_WIDTH>3216</BAND_RASTER_WIDTH>
            <BAND_RASTER_HEIGHT>6241</BAND_RASTER_HEIGHT>
            ...
        </Spectral_Band_Info>

So the whole product is supposed to have 3217 columns, but the bands both actually only have a width of 3216 pixels.

In SNAP, this is also shown accordingly:

grafik
grafik1114×713 172 KB

This tripped the BEAM-DIMAP reader I wrote for further processing SNAP data in Python using xarray and I guess it is an error somewhere along the processing chain and shouldn’t happen.

2

Now I have a different example, where the effect of mismatching raster dimensions are becoming problematic:

After Calibration, ThermalNoiseRemoval, SliceAssembly and Multilook, I end up with a product with the above mentioned problem (NCOLS is 3320, but all BAND_RASTER_WIDTHs are 3319).

Now I performed some BandMaths with two different expressions:

    • band name: incidence_angle
    • expression: incident_angle
    • band name: noise_gap
    • expression: (NESZ_VH == log(0)) || (NESZ_VV == log(0))

The result is a product with two bands of different dimensions:

SNAP screenshot
SNAP screenshot1178×653 114 KB

My guess is that for the incidence angle interpolation the wrong number of columns from <Raster_Dimensions> is used, while for the second expression the correct dimensions of the individual bands are kept.

Since the resulting bands are of different dimensions, they can’t be bundled into an xarray Dataset or other netCDF-like data structures.

Please have a look at this problem, there’s probably some rounding error in the Multilook operator that finds its way into the NCOLS tag in Raster_Dimensions.

3

@jun_lu could you please take a look at this issue?

4

A Jira ticket (Jira) has been created to track the issue. We will look into it. Thank you.

5

The problem has been fixed and the fix should be available in the next release. Thank you.

2 Likes