i.sentinel_2.ndvidiff

[griembauer]

This PR adds an Addon i.sentinel_2.ndvidiff which

• loads sentinel-2 data from two time intervals using t.sentinel.import
• optionally applies cloud masking using t.sentinel.mask
• aggregates all bands to one band per time interval using t.rast.mosaic

(----> perspectively, these three first tasks could be out-sourced to an own module to further increase reusability)

• calculates the aggregated NDVI for each of the two time intervals and difference between the two time intervals
• extracts areas of significant NDVI loss via threshold
• optionally exports result data to target folder

Note:

• This new addon calls various submodules, where small changes are implemented in the following PRs. Hence, these PRs need to be merged first:
  • t.sentinel.import: Add option to add a band offset after importing S-2 data t.sentinel#23
  • t.sentinel.import: specify cloud STDS import t.sentinel#24
  • [general_cleanup]: Add STVDS deletion and switch logic of group and STDS deletion grass-gis-helpers#34
  • cleanup: remove base rasters t.rast.mosaic#7

[linakrisztian]

Very nice Addon, with a lot of functionalities, thanks!
Looks good to me, even though I did not tested it. I just have some very minor comments.

Maybe in another step, it would be nice to create some more functions, to make the code a little more structured and/or as you already mentioned out-source few parts to an own module to further increase reusability.
But nevertheless very nice addon:)

[linakrisztian]

That is interesting. I always had the information, that maximum is best for reduction of cloud effects. But if you tested it, it's good to know that minimum is better

[griembauer]

I think it depends what it is aggregated. In this case, it is reflectance values - they are typically high in clouded areas so taking the minimum minimizes the risk of still having a cloud in there. If we aggregate NDVI values, it might make sense to take the maximum as clouds typically have low NDVI values.