Remove redundant boundaries?

posted 28 Nov 2011, 17:46 by David Pairman   [ updated 1 Jul 2012, 22:06 by David Pairman ]
Should boundaries between identical classes be removed?  

Our compilation to date has maintained an absolute respect for earlier mapping by not removing any existing boundaries, but inserting new boundaries along new or corrected land cover edges. As we do this, segments of former polygons have their attributes corrected so they often become described identically to their former neighbours. The old boundaries between neighbouring polygons therefore ceases to delineate any real difference between one side and the other so we’re contemplating a pre-release dissolve of these now-spurious boundaries to ‘clean up’ the database. What do you think?

Collaborating agencies: Please leave your views as comments below, along with your name/organisation.
Click on image to to expand.

Near Bay of Islands. The top two numbers represent original LCDB-1 and -2 classifications and the bottom three represent the current LCDB-1, -2 and -3 classifications. The cyan polygons above are now the same exotic class (54) at all there dates - should we remove the boundary? Our compilation has not removed any boundaries, yet… 


At the April 2012 Technical Advisory Group meeting it was resolved to dissolve out redundant boundaries in the main LCDB V3.0 database. 

However, it was realised that doing so could make it more difficult to change due to correction of past errors rather than actual land-cover change - in particular if these boundaries are dissolved we couldn't tag a part of a polygon ad being corrected or the authority for the correction. To address this it was agreed that a layer of corrected errors should be produced whenever a new version of the LCDB is released. Where a part of a polygon is partitioned off as a previous error now corrected, only that part is put in the “corrected“ layer, not the parent polygon (which has also obviously been modified).