Longitude / Latitude banding in L2 MODIST data

[csprh]

Hi All

I have been accessing L2 granules obtained using fd_matchup.py and MODIST.  The internals of fd_matchup uses the CMR search of OB_DAAC L2 granules (e.g. url: https://cmr.earthdata.nasa.gov/search/granules.json).

The metadata suggests this was created using L2gen.

The thing is that the Longitude and Latitude arrays have some serious banding issues.  Each 10 pixels or so there is a discontinuity creating a "sawtooth" type of response.



This corresponds to very visible "banding" in the products (e.g. chor_a etc.) where strips are repeated due to the positional sawtooth function described above.


Is this a common / fixable phenomenon?  I'm using this data  for machine learning and so the banded redundancy is not ideal.  If I reproject using the lat, lon arrays this creates missing stripes of data....which is also not ideal.

Anyone got any ideas?  Is this a specific sensor (MODIST) problem or something that is common across all sensors....or possibly a reprocessing error?

Thanks

Paul

[fredoceansips]

Paul,

What you are seeing is the MODIS "bowtie", which results in increasing overlap between pixels from successive scans toward the edges of the swath.  The bowtie results from the MODIS viewing geometry, in which 10 detectors view the Earth along-track during a scan.  This is shown in more detail here:
https://oceancolor.gsfc.nasa.gov/docs/bowtie/modis/

The only solution is to map the data.  There should be no gaps when this is properly performed; the data from each scan abut the neighboring scans near the center of the swath and overlap everywhere else.

Fred

[csprh]

Hi Fred

Thanks so much.  Being a newbie in this area, it's really difficult to find this information without such key words as "bowtie".  Having said that, reading various references, it seems the effect is a fairly obvious side effect of the capture process.

Thanks for putting me on the path so quickly. 

regards

Paul