Can reader 'modis_l1b' correct MODIS Bow Tie Effect?

[yhyxzx]

Can reader 'modis_l1b' correct MODIS Bow Tie Effect?

[djhoese]

The MODIS bow tie is "corrected" by resampling data using the Scene.resample interface. The bow tie is "detected" by recognizing the fill values associated with those pixels and then they are ignored when they are resampled.

If you are looking to keep the data in the native swath structure but want to some how exclude the bow-tie pixels let me know. That is a much different conversation though.

[yhyxzx]

@djhoese Does Scene.resample work like Geometric Correction > Reproject GLT with Bowtie Correction in envi? Is the result similar to the third figure in the link, "Georeferenced image with bow tie correction"https://www.nv5geospatialsoftware.com/docs/ApplyingCorrectionForMODISBowTieEffect.html

[djhoese]

I have no experience with ENVI, but yes I think so. It is strange to me that they have a resampling without bow tie correction since treating the data as it is meant to be treated (knowing which pixels are fill values and ignoring them during resampling) gets you this bow tie correction for free.

Satpy uses pyresample under the hood for resampling. Here is pyresample's description of resampling:

https://pyresample.readthedocs.io/en/latest/concepts/resampling.html

[yhyxzx]

In this process, is a more advanced product MOD03 necessary? @djhoese

[djhoese]

MOD03 is the geolocation product, right? While you might get "OK" results using the geolocation that comes in the data files, I will always recommend using the higher resolution geolocation information in the MOD03 files. It should also speed up processing as you shouldn't need to interpolate the geolocation as much.

[yhyxzx]

After my experiments and comparisons, I found that scn.resample is indeed effective in removing the bow-tie effect, thanks, I will close this issue.