[Imports] About a proposed import in Turkey using Copernicus metadata

[Christoph Hormann]

> ahmetlii _ <mappinglander at gmail.com> hat am 15.12.2020 14:16 geschrieben:
> 
> Hi. I plan to make an import in Turkey using Copernicus's CORINE landcover metadata (https://land.copernicus.eu/pan-european/corine-land-cover/clc2018) since since it's compatible with ODbL (https://land.copernicus.eu/pan-european/corine-land-cover/clc2018?tab=metadata). [...]

Maybe it would be a good idea to create kind of a FAQ for imports with this as a frequently recurring question:

Q: I want to import some remote sensing based automated land cover classification data set.  What do you think about it and how should i proceed?

A: This is almost certainly a bad idea for a number of reasons:

1) The idea of mapping of landcover in OSM is fundamentally different from closed landcover classification systems based on remote sensing data.  In OSM we have a number of established tags for landcover that are used for areas positively matching the meaning of the tag while other areas remain unclassified due to the lack of an established tagging until someone develops a tagging for them.  Many established landcover tags differentiate also cultural semantics and not just physical appearance.  Automated landcover classification systems however assign every point on the earth surface to the least unlikely of a set of classes typically based exclusively on physical appearance.
2) classification accuracy of automated landcover classification systems is typically low for cartographic standards even within the specification and paradigm of the classification (often in the range of 90-95 percent at best).
3) with widespread availability of high resolution imagery spatial resolution of such data sets tends to be low in comparison hence for someone manually mapping an area such data tends to be an obstacle rather than help.
4) most populated parts of the planet meanwhile have significant landcover mapping in OSM already so the task of conflating data to be imported with pre-existing data without creating a mess is difficult and work intensive.

--
Christoph Hormann
https://www.imagico.de/