Hello,
I've converted Sentinel-2 Level-1C scenes into Level-2A bottom of atmosphere (BOA) products using the ESA SNAP Sentinel Toolbox. I would now like to use FME to read the scene classification (SCL-20m) SNAP outputs and create a raster mosaic of these. The attached illustration shows this data in ArcGIS Pro. I've been unable to read the data in FME.
Thanks.
Tony
Hi @aviveiro,
According to this page, the BOA tiles are in the JPEG2000 format, which FME can read. I downloaded a sample and found that the tiles are georeferenced, and contain metadata in the jpeg2000_gml_root_instance attribute.
Hi @aviveiro,
According to this page, the BOA tiles are in the JPEG2000 format, which FME can read. I downloaded a sample and found that the tiles are georeferenced, and contain metadata in the jpeg2000_gml_root_instance attribute.
I'm experiencing the same issue. The Sentinel-2 MSI SAFE reader will accept L1C but not L2A.
While we can dig into the structure and read the various jp2000 and xml files directly, it would be nice if we could just point the reader to a MTD_MSIL2A.xml file instead of a MTD_MSIL1C.xml file.
They are the same format so there's no good reason for the level2A to crash FME.
I'm experiencing the same issue. The Sentinel-2 MSI SAFE reader will accept L1C but not L2A.
While we can dig into the structure and read the various jp2000 and xml files directly, it would be nice if we could just point the reader to a MTD_MSIL2A.xml file instead of a MTD_MSIL1C.xml file.
They are the same format so there's no good reason for the level2A to crash FME.
Hello @jdh,
Sorry to hear you're also having issues with the Sentinel-2 MSI SAFE reader. I have filed an issue, tracked as (FMEENGINE-66168) to improve this. I have made a note to update this post upon its progression.
Sorry about this!
Kailin Opaleychuk
Hello @jdh,
Sorry to hear you're also having issues with the Sentinel-2 MSI SAFE reader. I have filed an issue, tracked as (FMEENGINE-66168) to improve this. I have made a note to update this post upon its progression.
Sorry about this!
Kailin Opaleychuk
@kailinatsafe, if it helps, L2A data is available directly from copernicus, you don't need to process data in SNAP https://scihub.copernicus.eu/
@kailinatsafe, if it helps, L2A data is available directly from copernicus, you don't need to process data in SNAP https://scihub.copernicus.eu/
@jdh, thank you for sharing that resource!
@kailinatsafe, if it helps, L2A data is available directly from copernicus, you don't need to process data in SNAP https://scihub.copernicus.eu/
@jdh, sorry to bother you again. Would you mind sharing what version of FME you are using?
@jdh, sorry to bother you again. Would you mind sharing what version of FME you are using?
For the rest of the week build 20232. I intend to install 2020.1 on monday.
For the rest of the week build 20232. I intend to install 2020.1 on monday.
Hello again @jdh,
When reading in the L2A data, open the Parameters > Feature Type Name(s), and change the field from Bands to Resolution. I have added a photo below for reference. 
If you're willing, give this a try and let me know if you are still having issues!
Thank you for your patience on this matter,
Kailin Opaleychuk
Hello again @jdh,
When reading in the L2A data, open the Parameters > Feature Type Name(s), and change the field from Bands to Resolution. I have added a photo below for reference.
If you're willing, give this a try and let me know if you are still having issues!
Thank you for your patience on this matter,
Kailin Opaleychuk
I can confirm it doesn't crash in Resolution feature type, but that's not ideal for my scenario.
I need to use bands 2,3,8 and 11 on hundreds of images, and that will involve reading in both the 10m and 20m feature types and then removing bands 4 , 5, 6, 7, 8A and 12, and then making sure to get the right list element from sentinel2_band{}.
So this requires reading in more than twice as many raster files as needed, and then performing list manipulation, which is memory expensive.
For what it's worth the error I get when trying to add bands as a feature type is:
|INFORM|SENTINEL2 reader: Opening dataset '{redacted}\\S2B_MSIL2A_20191130T160619_N0213_R097_T17RLM_20191130T201643.SAFE\\MTD_MSIL2A.xml'
|ERROR |SENTINEL2 reader: Unsupported band: 20
|ERROR |SENTINEL2 reader: Failed to open the dataset 'SENTINEL2_L2A:{redacted}\\S2B_MSIL2A_20191130T160619_N0213_R097_T17RLM_20191130T201643.SAFE\\MTD_MSIL2A.xml:20m:EPSG_32617'. Please ensure source data is valid and correct reader is selected
|WARN |A fatal error has occurred. Check the logfile above for details
|ERROR |Failed to obtain any schemas from reader 'SENTINEL2' from 1 datasets. This may be due to invalid datasets or format accessibility issues due to licensing, dependencies, or module loading. See logfile for more information.
I can confirm it doesn't crash in Resolution feature type, but that's not ideal for my scenario.
I need to use bands 2,3,8 and 11 on hundreds of images, and that will involve reading in both the 10m and 20m feature types and then removing bands 4 , 5, 6, 7, 8A and 12, and then making sure to get the right list element from sentinel2_band{}.
So this requires reading in more than twice as many raster files as needed, and then performing list manipulation, which is memory expensive.
For what it's worth the error I get when trying to add bands as a feature type is:
|INFORM|SENTINEL2 reader: Opening dataset '{redacted}\\S2B_MSIL2A_20191130T160619_N0213_R097_T17RLM_20191130T201643.SAFE\\MTD_MSIL2A.xml'
|ERROR |SENTINEL2 reader: Unsupported band: 20
|ERROR |SENTINEL2 reader: Failed to open the dataset 'SENTINEL2_L2A:{redacted}\\S2B_MSIL2A_20191130T160619_N0213_R097_T17RLM_20191130T201643.SAFE\\MTD_MSIL2A.xml:20m:EPSG_32617'. Please ensure source data is valid and correct reader is selected
|WARN |A fatal error has occurred. Check the logfile above for details
|ERROR |Failed to obtain any schemas from reader 'SENTINEL2' from 1 datasets. This may be due to invalid datasets or format accessibility issues due to licensing, dependencies, or module loading. See logfile for more information.
The granule organisation for a L2A images is:
So the problem band is presumably the Scene Classification map (SCL).