FORCE MDD Wavelength information not recognized

[Fabian Thiel]

It seems that the wavelength information of FORCE BOA outputs is not recognized anymore.

MDD Example

Band 1 Block=1000x100 Type=Int16, ColorInterp=Gray
  Description = BLUE
  NoData Value=-9999
  Metadata (FORCE):
    Date=2020-01-18T10:02:43.0Z
    Domain=BLUE
    Scale=10000.000
    Sensor=LND08
    Wavelength=0.482
    Wavelength_unit=micrometers

Version 3.10.0.20220401T124144.develop c7d679b

[Andreas Rabe]

Hi Fabian, can you send me the file, so that I can place it on the SAN/_EnMAP for future unittests.

[Fabian Thiel]

I can, but you can use probably any tile/image from your datacube over at HU

[Andreas Rabe]

Haven’t used the FORCE server for a long time. For me it’s easier if you just give it to me.

And we are sure, that your problem is reproducable

[Andreas Rabe]

Info is stored here:

‌

[Andreas Rabe]

@Benjamin Jakimow I placed the dataset here, so that you can use it for testing.

\141.20.140.91\san_EnMAP\Rohdaten\EnMAP-Box external Software Formats\FORCE\20200118_LEVEL2_LND08_BOA.tif

[Fabian Thiel]

@jakimowb It seems that the temporal domain does also not work in the Spectral Library View (when switching to e.g. decimal year as X-axis). I remember that this used to work, didn’t it?
Metadata Example for a FORCE TSS output (first two bands):

Band 1 Block=3000x300 Type=Int16, ColorInterp=Gray
  Description = 20170104_SEN2A
  Min=-10000.000 Max=10000.000
  Minimum=-10000.000, Maximum=10000.000, Mean=-1110.412, StdDev=3297.573
  NoData Value=-9999
  Metadata:
    STATISTICS_APPROXIMATE=YES
    STATISTICS_MAXIMUM=10000
    STATISTICS_MEAN=-1110.4124468383
    STATISTICS_MINIMUM=-10000
    STATISTICS_STDDEV=3297.572621395
    STATISTICS_VALID_PERCENT=1.888
  Metadata (FORCE):
    Date=2017-01-04T00:00:00.0Z
    Domain=NDV_TSS
    Scale=10000.000
    Sensor=SEN2A
    Wavelength=2017.008
    Wavelength_unit=decimal year
Band 2 Block=3000x300 Type=Int16, ColorInterp=Undefined
  Description = 20170127_SEN2A
  Min=-7500.000 Max=10000.000
  Minimum=-7500.000, Maximum=10000.000, Mean=4977.401, StdDev=1717.609
  NoData Value=-9999
  Metadata:
    STATISTICS_APPROXIMATE=YES
    STATISTICS_MAXIMUM=10000
    STATISTICS_MEAN=4977.4006202918
    STATISTICS_MINIMUM=-7500
    STATISTICS_STDDEV=1717.608607659
    STATISTICS_VALID_PERCENT=60.23
  Metadata (FORCE):
    Date=2017-01-27T00:00:00.0Z
    Domain=NDV_TSS
    Scale=10000.000
    Sensor=SEN2A
    Wavelength=2017.071
    Wavelength_unit=decimal year

‌

[Andreas Rabe]

I guess this is QPS related, so I hand over to Benjamin.

Benjamin, note that I placed a test dataset on the SAN for you.

[Benjamin Jakimow]

@Fabian Thiel do you still have access to SAN EnMAP?

@Andreas Janz please add example files to the repository. Files stored in SAN files are not accessible to anyone outside of the HU VPN. For many tests routines small images that consist of 5x5 pixel only should be sufficient. In case of the FORCE examples it’s probably sufficient to create a VRT only that provides the typical FORCE metadata.

[Benjamin Jakimow]

Let’s say a band metadata defines
a) a wavelength according to ENVI (wavelength=350, wavelength unit = nm), and
b) a date-time stamp (date = 2017-01-27T00:00:00.0Z)
What would you expect to be used for the spectral profile?

‌

[Fabian Thiel]

@Fabian Thiel do you still have access to SAN EnMAP?

No not really. I think I would have to ask Stephan again to reactivate my account

[Fabian Thiel]

What would you expect to be used for the spectral profile?

Maybe if metadata information for a temporal and spectral x axis is given, use band index? @Andreas Janz what do you think?

[Andreas Rabe]

Let’s say a band metadata defines
a) a wavelength according to ENVI (wavelength=350, wavelength unit = nm), and
b) a date-time stamp (date = 2017-01-27T00:00:00.0Z)
What would you expect to be used for the spectral profile?

FORCE creates single-date spectral BOA raster or single Band/Index timeseries stacks. In both cases, one of the dimensions is constant (single time or single index). We should just skip the constant dimension, which results in either a spectral or a temporal profile.

[Andreas Rabe]

@Andreas Janz please add example files to the repository. Files stored in SAN files are not accessible to anyone outside of the HU VPN. For many tests routines small images that consist of 5x5 pixel only should be sufficient. In case of the FORCE examples it’s probably sufficient to create a VRT only that provides the typical FORCE metadata.

I always use original external data for testing. Any manipulation (e.g. making it smaller) will potentially remove some characteristics from the external product. I find that dangerous and wouldn’t recommend it.

If Fabian needs the SAN data, I can upload it to my GoogleDrive.

[Benjamin Jakimow]

spectral property key search is now case-insensitive, resolves #1203

Signed-off-by: Benjamin Jakimow [e-mail] [e-mail]

→ <<cset 875392d0f28c>>