[NUUG kart] Corine Import Vandalism (ex: problem with imported corine data)

Arne Johannessen arne at thaw.de
Wed Apr 13 18:50:49 CEST 2011

Peter Svensson wrote on 7 March:
> To vibrog and kart mailing list:
> I've found a problem with data corresponding to Corine Land Cover dataset
> import.
> [...]
> However, after finding this, i guess the rest of the import could have
> similiar errors and needs to be validated.

I can confirm that there are many errors similar to the one you described in vibrog's import. However, the biggest error by far is the import itself.

In fact, to make a point, I'd go as far as to call this thing not an import, but sheer vandalism.


- The added data is inconsistent with itself (as Peter already pointed out).

- The added data is inconsistent with data from earlier imports. In particular, the Corine data intersects the existing PGS coastline all over the place. This rather obvious kind of error should have been corrected before importing any data.

- The added data appears to be less accurate than existing automatically derived data. In particular, the existing data on lakes was derived automatically from Landsat using the "lakewalker" JOSM plugin (or similar tools) in many cases. It is my impression that the existing data was somewhat more accurate than the Corine import.

- The added data has been imported with complete disregard of existing land cover data that was manually surveyed and added by mappers in some cases. This is _totally_ unacceptable!

I refer you to the discussion "Zero Tolerance on Imports" on [OSM-talk] in late February. This reckless behaviour obliterates many hours, probably even days of work. It frustrates the OSM mappers who have been doing that work up to the point where some lose interest in OSM altogether.

Hopefully everybody will agree that if just one mapper gets discouraged by imports destroying his work it is the absolute worst possible outcome.


* * *

Here's what should have happened, and what I hope will happen in the future in cases where an import looks like a good idea here in Norway:

- The data inconsistencies mentioned above and any others that may be present should have been fixed.
- Areas with existing land cover data should have been identified, and generously removed from the data to be imported.
- The result should have been double-checked not to conflict with existing data in any way.
- The import should have been announced on the [Imports] mailing list.

Then -- and only then! -- should the import have taken place, using a new user account specifically created for this import.

* * *

Alternatively, one might take a look at another solution to this problem: Some groups tend to refrain from doing any automated imports at all. Instead, they set up a WMS server that provides images of the data to be imported, and can easily be used as background reference in JOSM or other mapping software.


This way, the strength of OSM -- which is the individual mappers doing the ground work -- can be utilised much better than by doing any automated import.

This way, we can add actual value to the map instead of adding the mere result of automated scripts.


Arne Johannessen

More information about the kart mailing list