[GRASS-user] Removing small areas using v.clean

[Markus Metz]

On Thu, Sep 22, 2011 at 5:26 AM, Pierre Roudier
<pierre.roudier at gmail.com> wrote:
> BTW I confirm I got similar problems with the 6.4 series,
>
Thanks for the input vector, I could reproduce the effect. But I am
not sure if this is really a problem. The method to remove small areas
finds the neighbouring area with which the longest boundary is shared.
This neighbouring area can also be the surrounding "nothing" in which
case the area is not merged with a real area but erased. That is, my
first comment was wrong, areas are not always merged with a
neighbouring real area, only if this neighbouring real area is the one
with the longest shared boundary. The cleaning tool assumes that small
areas are spurious and removes them by merging or by erasing. In case
of ties, when there are two neighbours with equally longest shared
boundary, as in your case, whichever neighbour is checked first will
be used. That neighbour can also be the empty space around.

Markus M


>
> 2011/9/22 Pierre Roudier <pierre.roudier at gmail.com>:
>> Thanks Markus,
>>
>> Here's an archive containing the input vector layer, along with
>> v.clean output:
>> https://df.bestgrid.org/quickshare/fb0bb3af778c8a03/vectors.tar.gz
>>
>> Here's the result of v.info -h on the output:
>>
>> COMMAND: r.to.vect input="c_means" output="foo" feature="area"
>> GISDBASE: /data/GRASSDATA
>> LOCATION: NZTM2000 MAPSET: BH USER: roudierp DATE: Thu Sep 15 17:29:53 2011
>> ---------------------------------------------------------------------------------
>> COMMAND: v.clean input="foo" layer="1"
>> type="point,line,boundary,centroid,area,face,kernel" output="bar"
>> tool="rmar
>> ea" thresh=900
>> GISDBASE: /data/GRASSDATA
>> LOCATION: NZTM2000 MAPSET: BH USER: roudierp DATE: Thu Sep 15 17:31:45 2011
>> [Raster MASK present]
>>
>> Cheers,
>>
>> Pierre
>>
>> 2011/9/21 Markus Metz <markus.metz.giswork at googlemail.com>:
>>> On Wed, Sep 21, 2011 at 12:51 AM, Pierre Roudier
>>> <pierre.roudier at gmail.com> wrote:
>>>> Markus,
>>>>
>>>> Sorry for this late reply as I was away from keyboard the last few days,
>>>>
>>>> I encountered this problem while testing the latest grass_svn  (grass
>>>> 7). I can give it a shot using a stable version if you want,
>>>
>>> You can try, but I would be surprised if the result is any different.
>>> Can you make the vector available for testing, or can you reproduce
>>> this with a vector of the sample datasets?
>>>
>>> Markus M
>>>>
>>>> 2011/9/17 Markus Metz <markus.metz.giswork at googlemail.com>:
>>>>> On Fri, Sep 16, 2011 at 6:28 AM, Pierre Roudier
>>>>> <pierre.roudier at gmail.com> wrote:
>>>>>> Dear list,
>>>>>>
>>>>>> I got a problem as I'm trying to remove small areas of a vector layer
>>>>>> using v.clean with the rmarea tool: some of these small areas that are
>>>>>> on the boundary of the layer are getting erased rather than being
>>>>>> merged with the closest polygon.
>>>>>>
>>>>>> Is that an expected behaviour?
>>>>>
>>>>> No, they should always be merged with a neighbouring area along the
>>>>> shared boundary. Areas are supposed to be erased only if they are
>>>>> isolated, i.e. do not share a boundary with another area. What GRASS
>>>>> version are you using? There was a bug which was fixed 2 years ago.
>>>>>
>>>>> Markus M
>>>>>
>>>>
>>>>
>>>>
>>>> --
>>>> Scientist
>>>> Landcare Research, New Zealand
>>>>
>>>
>>
>>
>>
>> --
>> Scientist
>> Landcare Research, New Zealand
>>
>
>
>
> --
> Scientist
> Landcare Research, New Zealand
>