[R-sig-Geo] merge sliver polygons

[Murray Richardson]

Hi Roger

I'm just getting around to trying this out. 

I must say, I'm not clear on how to work with the neighbour list object 
to accomplish this.  Specifically, how would I identify the largest 
neighbours of the sliver candidates?

Furthermore, do I first subset the polygons according to the area 
threshold, and then construct the neighbour list for only these 
candidate slivers?  I'm not clear on how to identify neighbours for a 
subset of polygons.

Sorry to bother you with this again.  It will be very useful if it works 
as I would also like to use it for region merging based on other attributes.

Still waiting on that book...!

Murray



Roger Bivand wrote:
> On Wed, 13 Aug 2008, Murray Richardson wrote:
>
>> Hello again r.sig.geo list,
>>
>> Thanks Roger, for help on my previous question regarding iterating 
>> through a shapefile.
>>
>> I'm sure once I receive my copy of  "Applied Spatial Data Analysis 
>> with R" I will find answers to simple questions like this on my own, 
>> but in the meantime....
>>
>> Is it possible to merge sliver polygons that fall below a certain 
>> threshold area with adjacent neighbours (e.g. perhaps using  
>> unionSpatialPolygons but without aggregating any  polygons?).  If a 
>> sliver shares edges with more than one polygon, it doesn't really 
>> matter which one it merges with, but if I had to choose a rule I 
>> would have it merge with the largest one.
>
> Not such a simple question ...
>
> Both the Polygon and Polygons objects in the SpatialPolygons object 
> have "area" slots, with different roles. The Polygon objects have a 
> correct naive area in the geometry of the coordinates taken as planar. 
> The Polygons objects use the "gross" area of Polygon objects belonging 
> to them, but "only" to provide the plot order (plot from largest to 
> smallest to avoid over-painting).
>
> If you "trust" the area slot of the Polygons objects (beware of hole 
> Polygon objects), you can first find your candidate slivers by 
> retrieving the areas by:
>
> Polygons_areas <- sapply(slot(SPobj, "polygons"),
>   function(x) slot(x, "area"))
>
> and set a cutoff. Then use poly2nb(SPobj, queen=FALSE) in spdep to 
> find the neighbours (rook criterion). Next use the output object to 
> identify the largest neighbours of the sliver candidates, and build a 
> "new Polygons" ID vector. Finally, use unionSpatialPolygons(). I'm 
> assuming you wouldn't have asked if there was useful data in the slivers!
>
> Hope this helps,
>
> Roger
>
>>
>> Thanks in advance,
>>
>> Murray Richardson
>>
>> _______________________________________________
>> R-sig-Geo mailing list
>> R-sig-Geo at stat.math.ethz.ch
>> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>>
>>
>