Ingest landsat8 tiffs using EPSG:32650 failed

[tosen1990]

All I want is just to tile the tiff using the original projection which I read from landat8 tiffs like ESPG:32650. And I'v tried many different tiffs of scenes. When ingesting using projection EPGS:32650,the extent error throws out always. But when change it to EPSG:4326 /WebMecaotr,I always success. I am not sure this is a bug? I've never seen anyone use this kind of project for ingesting for now.

val targetLayoutScheme = ZoomedLayoutScheme(CRS.fromName("EPSG:32650"), 256) 
acc :+ rdd.reproject(targetLayoutScheme, bufferSize = 30, Reproject.Options(method = Bilinear, errorThreshold = 0)) Caused by: geotrellis.vector.ExtentRangeError: Invalid Extent: xmin must be less than xmax (xmin=500001.0179734276, xmax=500000.97161722643)

Scenes of tests downloaded from aws are including:

• LC08_L1TP_123032_20180118_20180206_01_T1
• LC08_L1TP_123032_20180102_20180104_01_T1
• LC81220322015019LGN00

[pomadchin]

Hey @tosen1990 thanks a lot for your report!

It looks like indeed we don't support this projection well, it's a proj4 bug. I investigated this morning a lil bit more; it looks like we need to use ExtendedTranverseMercator to handle UTM projections:

• Need to switch utm from tmerc to etmerc OSGeo/PROJ#316
• https://github.com/locationtech/proj4j/blob/master/src/main/java/org/locationtech/proj4j/proj/ExtendedTransverseMercatorProjection.java

Aslo mb it's a good point to start swtitching to locationtech proj4j (if it's published for sure), otherwise it looks like we need to accept this pr and to fix these tests.

However we'll look closer into your particular case, mb it's really related to the way we support utm projections and not to this proj4 bug ._.

[pomadchin]

@tosen1990 are you on 2.1? // just to confirm

[tosen1990]

@pomadchin yes,I'm on gt 2.1.

[pomadchin]

Looked closer into this issue, and yes there is a problem in proj4 projection of some random projections into some of UTM projections.

Here is that situation that all UTM projections are used for some local math (that's why there are so many UTM zones) and it makes no sense as a destination projection for general use. ZoomedLayoutScheme computes some particular tile global location and level basing on the projection WorldExtent information and it's obviously is not supported by UTM projections.

The solution would be probably to implement a different ZonedLayoutScheme, but at this point we don't support UTM zones as target Global projections.

The proper solution would be to use a different layout scheme: FloatingLayoutScheme or LocalLayoutScheme as this CRS is local.

[tosen1990]

@pomadchin Totally understand and thanx for your time.