me and my student are using Landsat-7 ETM imagery to create high-resolution ice-water maps.
For this we create a product based on bands 2 to 4 which we read from the *_MTL.txt files
We had problems saving the product and/or a subset of the created bands / products and were told that we need to re-sample the bands first before we do our computations. So even though one does computations based purely on the bands with 30 m pixel size SNAP is mixing up information from all bands - also the panchromatic one, which has 15 m pixel size. Fine, we did the re-sampling and were able to safe the product properly for the Arctic. By properly we mean: the product, latitude, and longitude for each pixel.
Now, the same procedure seems not to work for images from Antarctica. Neither are we able to create a “subset” in the “export” mode of SNAP nor are we able to properly safe the latitude and longitude coordinates along with the product - no matter whether we do or not do the re-sampling. Either the coordinates are not written at all (which happens when we re-sample the bands before we do our computations) or the coordinates are zero everywhere.
What we tried:
Import data -> re-sampling -> our computations -> export into netCDF
Import data -> re-sampling -> our computations -> export as BEAM-DIMAP or geoTIFF -> re-import -> export into netCDF
Import data -> our computations -> deletion of all bands we don’t need -> export into netCDF
Import data -> our computations -> deletion of all bands we don’t need -> re-sampling -> export into netCDF
All these did not work and we don’t understand while it is fine for the Arctic but not fine for the Antarctic.
Perhaps somebody has an idea into what / which information we could look else to solve the riddle? Thank you so much!
Best regards,
seaice
p.s.: As a side note: Even though Landsat-5 TM also has bands with different pixel resolutions we did not need to do any re-sampling and working with the data of bands 2 to 4 with SNAP was simply a pleasure.
There is a know bug in SNAP. Currently it can’t handle the Polar Stereographic projection of Landsat data.
We have this issue already in our Issue tracker.
So I fear there is currently no way to deal with this data in SNAP.
p.s. For Landsat-5, the data was scanned in 120m (thermal) and 30m (other) resolution but only distributed in 30m.
Thank you so much or your reply and for pointing out that this has been a known issue already.
I checked the respective image and indeed the meta data show “PS” for the attribute MAP_PROJECTION.
Can I conclude then that even though the corner coordinates and the true latitude of the used polar-stereographic projection is given in the meta data, one is not able to carry out a re-projection into UTM with SNAP?!? It seems so. I tried - by first re-sampling all bands to either a 30 m resolution band or a 15 m resolution band and then carrying out a re-projection to UTM zone 22 (south), which is where my image is located, but without success. Without the re-sampling it does not work anyways because of the two different pixel sizes and hence dimensions involved which seems to be problematic for SNAP when applied to Landsat-7 data.
I hope that for Landsat-8, where SNAP allows to ingest data at 15 m or 30 m pixel resolution, this problem does not occur.
Indeed the same problem applies to Landsat 8 data South of 60 degree South.
Is the problem with Landsat in polarstereographic projection solved in the new release of SNAP, version 5.0, i.e. can I carry out reprojections and/or store the Landsat data and products in netCDF?