NaN values in slave images with back-geocoding

[oe_mjelle]

Dear developers,

While applying back-geocoding for a stack of 5 Sentinel 1 IW SLC products, NaN values pop up for certain slave images. The problem does not seem to lie in the products itself, as interchanging the master and slave product, and then starting the back-geocoding, generates an acceptable result. The difference between the master and slave image here is that the master product has only a small land portion (and a large ocean part), whereas the slave product covers a bit more of the land area. Could somebody please clarify if this bug is known or what is causing this effect?

Thanks in advance for your help.

Products used in example, taking subswath 2:

S1B_IW_SLC__1SDV_20211220T221744_20211220T221811_030115_03988D_390F
S1A_IW_SLC__1SDV_20211214T221822_20211214T221850_041011_04DF2A_DB23
S1B_IW_SLC__1SDV_20211208T221745_20211208T221812_029940_039300_72B6  
S1A_IW_SLC__1SDV_20211202T221822_20211202T221850_040836_04D91F_639F
S1B_IW_SLC__1SDV_20211126T221726_20211126T221753_029765_038D78_4A3A

Graph file settings used for back-geocoding:

<node id="2-Back-Geocoding">
    <operator>Back-Geocoding</operator>
    <sources>
      <sourceProduct refid="1-ProductSet-Reader" />
    </sources>
    <parameters class="com.bc.ceres.binding.dom.XppDomElement">
      <demName>Copernicus 30m Global DEM</demName>
      <demResamplingMethod>BISINC_11_POINT_INTERPOLATION</demResamplingMethod>
      <externalDEMFile />
      <externalDEMNoDataValue>0.0</externalDEMNoDataValue>
      <resamplingType>BISINC_11_POINT_INTERPOLATION</resamplingType>
      <maskOutAreaWithoutElevation>false</maskOutAreaWithoutElevation>
      <outputRangeAzimuthOffset>false</outputRangeAzimuthOffset>
      <outputDerampDemodPhase>true</outputDerampDemodPhase>
      <disableReramp>true</disableReramp>
    </parameters>
  </node>

[joaquimrosa]

Would @lveci @marpet have any suggestion? Thanks

[joaquimrosa]

Or @jun_lu perhaps?

[ABraun]

@oe_mjelle

The difference between the master and slave image here is that the master product has only a small land portion (and a large ocean part), whereas the slave product covers a bit more of the land area.

can you please give a visual example?
Surely, when the images differ too much, it can make a difference which is set as the reference product.

[Orland]

@ABraun

[quote=“MyCCPay, post:1, topic:34474”]
The difference between the master and slave image here is that the master product has only a small land portion (and a large ocean part), whereas the slave product covers a bit more of the land area.

can you please give a visual example?
Surely, when the images differ too much, it can make a difference which is set as the reference product.
[/quote]

Facing same issue but no response from anyone and couldn’t find this topic troubleshooting in google.

[oe_mjelle]

Hi all,

Thanks for your reply.

Below the quick-look of the master product and the first slave product. We have used the middle subswath and our area of interest lies in the top part of the image.

This is the master product with less land area:

S1B_IW_SLC__1SDV_20211220T221744_20211220T221811_030115_03988D_390F

And this is the first slave image with more of the land cover:

S1A_IW_SLC__1SDV_20211214T221822_20211214T221850_041011_04DF2A_DB23

So to state the problem again. We get correct values for back-geocoding if we switch around this master and slave. If we do not switch around and use the order as mentioned, Nan values arise for the slaves that are not aligned.