Zoom 2 and 3 relief is too generalized

tilezen / joerd

Joerd can be used to download, merge and generate tiles from digital elevation data

MIT License

315 stars 48 forks source link

Zoom 2 and 3 relief is too generalized #165

Open nvkelso opened 6 years ago

nvkelso commented 6 years ago

Whatever source we're using at zoom 4 we should also use at zoom 2 and 3?

http://tangrams.github.io/walkabout-style/#3/73.80/-75.32

v1.0

v1.1

iandees commented 6 years ago

When I tried adjusting GMTED to show a zoom earlier, nothing changed. I looked at the footprints database and noticed that GMTED has no footprint over Greenland:

The bit in the middle there is the southern tip of Greenland. Seth pointed out that it was the part south of 60° so we went looking for other data at the GMTED viewer. The 7.5 arcsec tile in that area only includes the southern tip of Greenland that was covered by SRTM (which is what we have). But the 30 arcsec data tile includes Greenland:

This tile and most of the other 30 arcsec tiles (including ones north of Antarctica) don't seem to be in the elevation-sources-prod bucket, so they weren't transcoded. I'm going to transcode them now and see if this helps with Greenland.

iandees commented 6 years ago

I transcoded GMTED from here, but the resulting transcoded images don't mark zero as nodata, so we end up with stuff like this (with gmted blocking etopo):

And even with the 30 arcsec GMTED Iceland doesn't look as good as it should:

icelandtoggle

Still not sure what we're missing here. It looks like it's bad data, not bad software.

nvkelso commented 6 years ago

@zerebubuth Digging back in the archives can you sleuth out the data difference here?

iandees commented 6 years ago

When I loaded in some missing 30 arcsec GMTED to solve the problem mentioned in the comment above, the DEM included water with a value of 0 (zero), so I re-processed them with a VRT that included zero as a nodata value. Normally this would be a quick project, but as I was processing the data the website that we were downloading from stopped working (https://topotools.cr.usgs.gov/GMTED_viewer/viewer.htm - still has a bad redirect or CDN config). This caused the transcoder batch process to delete all the existing GMTED we had (https://github.com/mojodna/marblecutter/issues/88) so I had to manually download it from the terrible USGS EarthExplorer site. Once that was mirrored (thank goodness for gigabit at home) I transcoded through the existing process and we've got better-looking GMTED now. But it's still not as good as what's in v1 for Greenland.

nvkelso commented 6 years ago

I'm seeing a relief detail offset in Norway, too. I suspect the code that adjusts the footprint min_zoom for high latitudes is off by 1 or 2 – that we should be seeing the better resolution stuff earlier than we are. Zooms 7, 8, 9 seems like they should all have the content from zoom 10 generalized down in this area of Norway.

http://tangrams.github.io/walkabout-style/#9/62.6316/6.8170

v1.0

v1.1