SNAP reprojection issue with SLSTR L1 datasets

snap
#1

Hi,
I have some issues in reprojecting SLSTR L1 dataset (S3A_SL_1_RBT).
(Sorry, I wasn’t sure if that topic is a generic SNAP one or a S3TBX instead. I have used SNAP since it is involving reprojection)

When using a graph which does a single full dataset read + reprojection, the output image visually looks like made of different stripes with a visible offset/shift between them.
See this image:

Reprojection.png921×879 302 KB

If I split the work in multiple pieces, by doing several operations composed of
read+subset (Y offset, 400 pixels height, fullswath) + reprojection
things are looking better.

This image depicts the same dataset being reprojected in chunks and finally reassembled with gdalbuildvrt and translated to tiff.

seamlessreprojection.png914×874 313 KB

What do you think about that?
Is it a known issue? Do you have any feedback or suggestion?

Cheers,
Daniele

#2

Dear Daniele,
I haven’t seen this outcome before, thanks for reporting. Can you tell me which reader you used?
Cheers,
Tonio

#3

Dear Tonio,
I have used the SlstrLevel1B500mProductReader (I have temporarily disabled the 1km reader to make sure it use the 500m. Thanks to your reply in the other topic, I will force the formatName to 500m).

In case you need it, this is the graph I have setup:

<graph>
<version>1.0</version>
<node id="read1">
    <operator>Read</operator>
    <sources/>
    <parameters>        
        <file>/data/SENTINEL/SLSTR/S3A_SL_1_RBT____20160426T010814_20160426T011314_20160426T025123_.../xfdumanifest.xml</file>
    </parameters>
</node>
<node id="subset">
    <operator>Subset</operator>
    <sources>
        <source refid="read1"/>
    </sources>
    <parameters>
        <fullSwath>true</fullSwath>
        <bandNames>S1_radiance_an</bandNames>
    </parameters>
</node>
<node id="reproject">
    <operator>Reproject</operator>
    <sources>
        <source refid="subset"/>
    </sources>
    <parameters>
        <wktFile/>
        <crs>EPSG:4326</crs>
        <resampling>Nearest</resampling>
        <noDataValue>-9999</noDataValue>
        <includeTiePointGrids>false</includeTiePointGrids>
        <addDeltaBands>false</addDeltaBands>
        <pixelSizeX>0.003</pixelSizeX>
        <pixelSizeY>0.003</pixelSizeY>
    </parameters>
</node>
<node id="write1">
    <operator>Write</operator>
    <sources>
        <source refid="reproject"/>
    </sources>
    <parameters>
        <file>/data/SENTINEL/SLSTRL1_S1_radiance_an_979_1462451633126_.tif</file>
        <formatName>GeoTIFF-BigTIFF</formatName>
        <deleteOutputOnFailure>true</deleteOutputOnFailure>
    </parameters>
</node>
</graph>

PS: Sorry I didn’t understand how to indent the code. When I have pasted it it was containing 4 spaces indentations.

#4

Hi Daniel,

I did reformat your post.

#5

Hi Devs,
did you have any chance to take a look on that issue?

Cheers,
Daniele

#6

Hi Daniele,
sorry for letting you wait. We’ll investigagte and come back to you about this.

#7

Hi Daniele,
we now investigated this in detail and found that the shifts after reprojection in SLSTR are caused by approximations (to find the geo-position) which where not properly initialised. This is now fixed.

Thanks for this report
Marco

1 Like