Hi, I had some intensive learning during the last days, and thanks again for your help.
After all it turns out that it is something Nicolas wrote, it is a matter of the projection. The Japanese software just exported to a projected format, but the original data seem to be in lat long. I found a way to get the unprojected data and now can create a beautiful hillshade in an unprojected lat long layer. And it even looks good in a projected project (EPSG:6670) with on the fly projection. For hydrological analyses I need to use the projected data, but the artifacts do not matter here. While binge-watching YouTube videos I realized that these artifacts occur with the pros as well when they use the projected layers for their analyses. Now everything much better and "in place". Best, Maria > Am 15.05.2020 um 13:21 schrieb Nicolas Cadieux <nicolas.cadi...@archeotec.ca>: > > > > Nicolas Cadieux > Ça va bien aller! > >> Le 14 mai 2020 à 23:12, Nicolas Cadieux <nicolas.cadi...@archeotec.ca> a >> écrit : >> >> Hi, >> >> >> See below for comments. >> >> Nicolas Cadieux >> Ça va bien aller! >> >>> Le 14 mai 2020 à 22:21, Priv.-Doz. Dr. Maria Shinoto >>> <maria.shin...@zaw.uni-heidelberg.de> a écrit : >>> >>> Hi again, >>> >>> and sorry for the ongoing discussion. >>> >>> Today I exported a selection of the DEM data to a shapefile, just 9MB for >>> the main file, and this makes testing very fast. >>> >>> (A) TINs did not work. >> >> TIn interpolation has memory problems with large data sets. Same problem >> since QGIS 2x at least. It was cool features but is not made to handle >> today’s data sets. >>> >>> (B) I tried all steps carefully again, but even the GDAL raster is horrible >>> now. >>> >>> Here are some screenshots with my explanation and the protocol for >>> rasterization and filling nodata. >>> >>> It seems that the artifacts are due to no data fields that evolve during >>> rasterization as a pattern. These nodata fields may be due to a slight >>> inclination of the grid from the export of the data with the Japanese >>> software. >>> >>> 1) The point grid, one can see the inclination >>> >> <01.jpeg> >>> >>> >>> 2) The raster of the same area, one can see the points of the vector point >>> grid along the white empty space; this is NODATA. >>> >> <02.jpeg> >>> >>> >> I would use gdal_grid not rasterize. Use Gdal grid with a larger search >> circle will solve this problem. Use nearest neighborhood with a search >> radius larger than the pixel (like 7m). That will reduce the no data. Click >> on the help or go to the gdal website. That will help you add the missing >> parameters like the -txe and -tye. (The extent) and the -outsize for the >> number of pixels. >> >>> I add the protocol >> <2020-05-15-rasterize-protocol-for-selection.txt> >>> >>> >>> >>> 3) Using the Fill NODATA from the Raster menu makes a beautiful looking >>> raster, there seem to be no flaws. >>> >> <03.jpeg> >>> >>> >> >> That fixes things but adds new data to the raster. This may be unwanted. >> >>> I add the protocol. >>> >> <2020-05-15-fill-nodata-protocol-for-selection.txt> >>> >>> >>> 4) This is the same area as in (3), but instead of a pseudocolor ramp shown >>> as hillshade. >>> >> <04.jpeg> >>> >>> >> This is normal if you select a bad z factor (probably not the case here). >> You will have the same thing if you zoom in and have nearest neighbour in >> the “zoomed in” under “resampling“ in the hillshade symbology window. >>> >>> 5) This is the impression from a larger area. >>> >> <05.jpeg> >>> >>> >>> >>> 6) This is the same small area hillshaded with the GDAL tools. Looks good, >>> but suffers from the same artifacts. >>> >> >> No this is way it should look like (Image under). You can see the pixels >> because you are zoomed in. Again, select the correct z factor (if x,y are >> in long -lat and z is in meters or feet.) (probably ok here). >> >> <06.jpeg> >>> >>> >>> >> Play with the resampling zoomed out parameters in symbology >> >> >>> 7) The larger area from hillshade in GDAL tools. >>> >> <07.jpeg> >>> >>> >>> >>> >>> >>> I sorry to be so insisting on the problem, I think it is not the problem of >>> QGIS, but perhaps there are solutions to such a case. -- The projection is >>> OK, and the base map fits perfectly. >>> >>> Best and Thanks to anyone trying to help, >>> Maria >>> >>> >>> >>> _______________________________________________ Qgis-user mailing list Qgis-user@lists.osgeo.org List info: https://lists.osgeo.org/mailman/listinfo/qgis-user Unsubscribe: https://lists.osgeo.org/mailman/listinfo/qgis-user
