Earthdata Forum

Hi All,

Looks like the default land mask for Sentinel processing is too course. Some of the pixels have rectangular/square shape along the edges as opposed to the exact shape of land, thus covering some water pixels. Is there a way to correct this, please? Same applies to the default cloud mask.

Thanks.

Yes,  the default landmask is a 15arc-second resolution, so ~500m, which is much more coarse than the MSI (or Landsat's OLI) instrument.  As described in my response to the OLI question about this issue, you can pass l2gen a null landmask (you could always apply a landmask after the processing).  Or you can define your own mask - sure a more difficult task.   To to that you'd have to match the format of our 15arc landmask, but at a higher resolution - this can even be localized - assuming you don't need the whole globe at 2om...

The cloud mask is based solely on measure radiances.  It is what it is, but you can change the theshold on which the flag is set with the cloud_thresh option.

Sean

Thanks Sean!

Note: post l2gen processing (if you don't apply the land mask there), you can apply a 50m land mask using the SeaDAS LandWaterMask tool (GUI: Layer > Coastline, Land & Water; GPT LandWaterMask).  But this tool is outside of the OCSSW processing stream so you could not continue with the resultant file back into the OCSSW stream (for instance l2bin).

Danny

Sean's suggestion to create your own landmask should allow you to use the full set of OCSSW tools.   Another option is to add or modify a product in a level 2 file.   It is really not very difficult to add level 2 products to SeaDAS 7 NetCDF4-CF format files, and examples using Python and R have appeared in this forum, see:

adding a product to L2 in Python
gpt merge

Note that some programs look at the products.xml file for information about the product, so you will need to add an entry for your new product to that file.

Thanks to you all!