contour problems at antimeridian line

[PBrockmann]

Starting from GeoTIFF Contours II (https://bl.ocks.org/mbostock/83c0be21dba7602ee14982b020b12f51), I have tested direct reading from netCDF files (the standard of all Climate Models from IPCC and all Earth System Modeling community). See https://github.com/PBrockmann/D3_netcdfjs and try netcdfjs_05.html

It sounds very promissing. So a great great thank for this d3.contour module !
In my dreams, I would also like to get a raster pixels rendering as in http://geoexamples.com/d3-raster-tools-docs/code_samples/raster-pixels-page.html but that is another story. Is there an official d3 to do that BTW ?

The problem with d3.countour in a geo context is that some contours at antemeridian line are shifted either up or down.
See:

The map get from netcdfjs_05.html is zoomable so the problem can clearly be seen.

I haven't been able to find where this shift comes from ?

Any help welcomed on this topic that could help a huge community.

[mbostock]

My guess is that the shift comes from beveling at the edges #4 because the contour algorithm must contain the polygon with the grid. In GeoTIFF Contours II, I map x = [0, n] in the source raster to [-180, +180]. Is it possible you’re mapping from x = [0, n - 1] instead?

[PBrockmann]

No I don't think so and moreover I have also noticed the problem directly from the GeoTIFF Contours II example.
See d3_contours_01.html in my repository that is a copy of this example and rotate the projection to see the problem.
Thank for your interest in this issue.

I have just updated my repo because a file was missing to test the netcdfjs_05.html

[PBrockmann]

Here is a capture from your GeoTIFF Contours II example with contours set with
var contours = d3.contours().thresholds(d3.range(-0, 40, 1)).size([n, m]);
Here zoomed on contours where the problem occurs ie lines have oddly unsmooth path.

[PBrockmann]

Try to analyze this problem. The geoStitch algorithm does correct the fact that the 2 polygons are separated by the antimeridian line and should be considered as one unique polygon but it leaves 2 small up and down stitches that should not be there.

A solution could be to run d3-contour with duplicates columns at start and end of the array but my tries are not successfull for now.

[Fil]

my solution is to remove the shared points from the stitched polygons
PBrockmann/D3_netcdfjs#1

[mbostock]

Updated example: https://observablehq.com/@d3/geotiff-contours-ii