2

I downloaded Sentinel-2 stack from the Google Earth Engine. For the export I used resolution (scale: 10), while some bands there should have resolution 20 m (SWIR) or 60 m(e.g. Cirrus):

Export.image.toDrive({
  image: exportImgSentinel, 
  description: 'Sentinel', 
  folder: 'Sentinel', 
  fileNamePrefix: 'Sentinel', 
  region: region, 
  scale: 10, 
  crs: 'EPSG:4326'
});

Opening it as a RasterStack in R, I checked resolution of the stack:

res(Sentinel_weird_1)
[1] 8.983153e-05 8.983153e-05

And the same for each separated band within. OK, I tried to make some Tasseled Cap bands, typing formula using respective bands. I have got raster with only 2 different values, that is apparently wrong. What to do in such cases? Should I:

  • download separate bands from GEE with different resolution, and then disaggregate in R bands with 20 or 60 m resolution to the 10 m?

  • or it will not work, since some data accuracy from 20-60 m resolution will be lost, and I should aggregate bands to the worst (60 m) resolution? That is, I can only make TCT maps with such bad resolution from Sentinel-2 images?

For Landsat data everything is clear. res function shows 30 m scale resolution. But I wanna have better map with 10 m resolution. Can I?

Maksym Matsala
  • 57
  • 9

2 Answers2

4

I had a similar problem. The TC is developed for Landsat Thematic Mapper products. You have to adjust the coefficients used for the band transformation. There is a paper from Shi & Xu (2019) where they adjusted the coefficients to the S2 MSI at sensor reflectance.

There is also a database for indices, but there they simply use the Landsat coefficients for the S2 MSI data.

Jeremy Caney
  • 7,102
  • 69
  • 48
  • 77
Marcel Mohr
  • 41
  • 2
0

You will have to use the 20m resolution Sentinel. The 10m does not have the bands necessary for tasseled cap transformation. Coefficients apply to each of the 7 non-thermal bands.

DrMaynard
  • 1