I am trying to do a simulation, but a continent I created using a topology to join multiple guidelines together refuses to be coloured with its proper raster.
All other continents keep their raster and when I draw in a new continent it keeps its raster as well. It is a supercontinent though. Is there some kind of upper limit in dimensions where the raster doesn’t draw in anymore?
I don’t know what to try, nor what to give to explain my case better, so any question or suggestion is welcome
By raster you might be meaning polygon fill (a single solid colour). Because topologies cannot reconstruct rasters (images) but they can be filled with a solid colour.
If so, then there can be issues with extremely huge polygons that cover around half the globe’s area (or more). For these polygons what’s inside and outside can become inverted if the centroid of the boundary outline goes outside the polygon - as can happen when the polygon outline appears to be a smaller polygon on the other side of the globe, rather than a larger polygon consuming more than half the globe (in which case who’s to say it’s not actually the smaller polygon).
My process is as follows.
I create guidelines of the features I want to keep. These are all lines and saved as unclassified features. I then create a topology with all these lines, “melting them together” in one feature, saved as a topological plate. Lastly, I copy the points of this feature, delete the doubles if needed, and save the new polygon shape as continental crust
This saved feature does not have a polygon fill, but I imported a separate raster connected to different layers to fill, among which is the continents layer. So to summarise, the end effect would be an outline in the colour dictated by the plate ID and a filling dictated by the raster.
I started with roughly 25-30% of the planet as continental crust, I don’t think I grew that to the point it doubled, nor do I see the other half being coloured instead.
My hypothesis comes from the fact that when I create a new smaller continent, even using the process I described, it is filled normally. The opposite happens with this continent or any other big continent, whether it wraps around the “back” of the planet or not.
The polygon fill feature works normally for the problem-continent as well, so I don’t think it’s a matter of rendering, but again I’m thinking out loud more than anything else
Thanks for clarifying that. So you start with a topology but end up with a regular polygon that you then attach a raster (image) to.
Raster reconstruction also has an issue with really large polygons, but it manifests in a different way (since it was implemented differently), which would explain why polygon-fill works for your problem continent (but not an attached raster).
I think you could have a large polygon that is relatively thin but wraps around the globe more than 180 degrees (think of a thin polygon following a vertical meridian and extending past both the North and South poles). This polygon might not render a raster but it would still render a fill colour. This has always been a limitation of rendering rasters on polygons (due to polygon not being projectable onto a plane). But I think this might be the first time (or one of the few times) it’s been encountered (that I’ve heard about). In a future version of GPlates (with Vulkan rendering, instead of OpenGL) I am considering an alternative approach that will avoid this issue.
Feel free to send me your problem polygon (to confirm this is actually the cause).
Hello again. I apologise for the delay, my work just pushed me down hard and had almost no time for side projects.
Following your explanation I tried an experiment moving the continent to the “center” so it wouldn’t wrap around the edges, but no dice.
How can I share the polygon?
To speed up the process I exported the coordinates in a txt file and copied under here, do tell me if there’s another procedure to follow. Sorry in advance for the long message
Actually that was just an example of a polygon that would fail to render a raster (when I mentioned “think of a thin polygon following a vertical meridian and extending past both the North and South poles”).
GPlates is generally agnostic to the poles and the dateline (in its internal operations). So any rotation/reorientation of a problematic problem will still have the same issue in GPlates.
Thanks for that. I converted that output to a GPML file using PlateTectonicTools with python -m ptt.convert_xy_to_gplates -l -p polygon-breaks-raster.txt (after removing the last line and adding an empty line beginning with > at the top). Also note that PlateTectonicTools will soon be part of GPlately.
After loading it into GPlates I can see that while it can be filled (see image below) it cannot render a raster for the reason I stated in a previous post. It exceeds the internal limit of 162 degrees (it’s actually not 180 degrees as I mentioned earlier). And your polygon spans about 170 degrees. You can measure this with the Measure tool on the 3D orthographic globe by clicking two points near each end of the polygon (and then read the line distance).
In a future version of GPlates I am considering an alternate approach to avoid this issue.
