DEM sub setting and projection issues

Lisa Hecker

unread,

Oct 8, 2021, 8:16:15 AM10/8/21

to RSGISLib Support

Hi,

We recently discovered artefacts in our S1 ARD which are due to a pixel shift that occurs when sub setting the DEM. We have found a way to ‘trim off’ the affected areas but wondered if you could let us know which function exactly ARCSI uses to subset the DEM?

On another note, when processing data from the Western Isles of Scotland with OSGB projection we see a gap between two adjacent granules, for example T29VND and T29VPD.

The same granules do not show this gap when processed with TM65 or WGS84 projection. Equally, granules directly adjacent to the East do not have the problem when processed in OSGB. I wondered if perhaps you had any idea why this happens (wkt file we use attached)?

Many thanks,

Lisa

BritishNationalGrid.wkt

Peter Bunting

unread,

Oct 8, 2021, 8:31:09 AM10/8/21

to Lisa Hecker, RSGISLib

Hi Lisa,

Sounds strange. Subsetting the DEM uses a couple of steps rsgislib.imageutils.createCopyImage is used to create the image file which the DEM values will be polluted and then the gdal.ReprojectImage function is used to populate that image. See lines 983 - 990 in arcsirun.py.

Don’t know about the gap without having the data and trying to run it through myself. I’m afraid I don’t have time right now to debug this but I could look in more detail in a couple of weeks.

Cheers, Pete

On 8 Oct 2021, at 13:16, Lisa Hecker <lisa....@jncc.gov.uk> wrote:

Hi,
We recently discovered artefacts in our S1 ARD which are due to a pixel shift that occurs when sub setting the DEM. We have found a way to ‘trim off’ the affected areas but wondered if you could let us know which function exactly ARCSI uses to subset the DEM?
On another note, when processing data from the Western Isles of Scotland with OSGB projection we see a gap between two adjacent granules, for example T29VND and T29VPD.

<WesternIsles_gap.png>

The same granules do not show this gap when processed with TM65 or WGS84 projection. Equally, granules directly adjacent to the East do not have the problem when processed in OSGB. I wondered if perhaps you had any idea why this happens (wkt file we use attached)?
Many thanks,
Lisa

--
You received this message because you are subscribed to the Google Groups "RSGISLib Support" group.
To unsubscribe from this group and stop receiving emails from it, send an email to rsgislib-suppo...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/rsgislib-support/ec9aadf9-c234-4815-8f29-15f6a62c825fn%40googlegroups.com.
<WesternIsles_gap.png><BritishNationalGrid.wkt>

Lisa Hecker

unread,

Oct 8, 2021, 9:05:43 AM10/8/21

to RSGISLib Support

Many thanks Pete!

The gap issue is there for the granules 29VNE, 29VPE, 29VND and 29VPD from at least 2018 onward. All orbits that cover those granules produce the same gap. When you get round to look at this and need any more information just let me know!