[R-sig-Geo] problem with rasterToPolygons x worldclim

[Michael Sumner]

On Thu, 7 Apr 2016 at 00:22 Michael Sumner <mdsumner at gmail.com> wrote:

> On Wed, 6 Apr 2016 at 23:45 Karla Shikev <karlashikev at gmail.com> wrote:
>
>>
>> Hi Michael,
>>
>> Thanks for the tips. I tried the help for rasterToPolygons, but none of
>> the options (e.g. dissolve=TRUE) made any difference. I tried gPolygonize
>> and it worked, except for (as you predicted), if the edges meet the raster
>> extents - but that definitely was an advance!
>>
>>
> I see the issue, the fun argument to rasterToPolygons is masking out
> values from that interval, but it's not setting them all to the same value
> - so you still get individual pixel polygons unless you mask on presence in
> the interval or not (as a binary):
>
> library(raster)
>
> r<-getData('worldclim', var='bio', res=10)$bio1
>
> e <- extent(-140,-100, 50, 60)
>
> xx <- crop(r,e)
> ## mask out pixel values first
> xx <- xx > 40 & xx < 60
>
> ## 576 polygons
> slice1<- rasterToPolygons(xx, fun = function(x) {x == 1})
> ## 1 polygon
> slice2 <- rasterToPolygons(xx,  fun = function(x) {x == 1}, dissolve =
> TRUE)
>
>
> Still it's not a very nice polygon, if I can I'll try a different way.
>
> Cheers, Mike.
>
>
>
>

You can use tricks to fill the outer edge so you get a valid contour line
all the way around (it's not always going to work - see here for the
explanation by whuber:
http://gis.stackexchange.com/questions/61550/colouring-areas-between-vector-contours
)

Is this better? Maybe - there are lots of other ways - maybe a buffer on
this gives roughly what you need. Maybe you only need a total area so
summing pixels after filtering is better?

Would be nice to have contouring that reliably produced polygons, but it's
a tricky problem. Other kinds of shape-finding might be better, with
alphahull on the pixel points perhaps, but you quickly get into "geometry
fudger" territory here rather than objective measures.

HTH




>
>>
>> I really appreciate your help and if you just point the way I can try to
>> go after what needs to be done, but right now I´m stumped.
>>
>> best, Karla.
>>
>> On Mon, Apr 4, 2016 at 5:38 PM, Michael Sumner <mdsumner at gmail.com>
>> wrote:
>>
>>> Read the help for rasterToPolygons, alternatively try chaining
>>> rasterToContour and rgeos::gPolygonize.
>>>
>>> The latter may need coaxing if your edges meet the raster extents.
>>>
>>> Cheers, Mike
>>>
>>> On Tue, 5 Apr 2016, 05:32 Karla Shikev <karlashikev at gmail.com> wrote:
>>>
>>>> Dear all,
>>>>
>>>> Here's an issue that is related to the previous one. In the commands
>>>> below
>>>> I'm trying to make a polygon for all the regions within a range of mean
>>>> annual temps. However, rasterToPolygons will draw a square around each
>>>> value in the original raster, rather than providing me the actual
>>>> polygon,
>>>> given that resolution. Any hints? Again, any help will be greatly
>>>> appreciated.
>>>>
>>>> Karla
>>>>
>>>> _____
>>>>
>>>> library(raster)
>>>>
>>>> r<-getData('worldclim', var='bio', res=10)$bio1
>>>>
>>>> e <- extent(-140,-100, 50, 60)
>>>>
>>>> xx<-crop(r,e)
>>>>
>>>> slice1<-rasterToPolygons(xx, fun = function(x) {x > 40 & x < 60})
>>>>
>>>> plot(slice1)
>>>>
>>>>         [[alternative HTML version deleted]]
>>>>
>>>> _______________________________________________
>>>> R-sig-Geo mailing list
>>>> R-sig-Geo at r-project.org
>>>> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>>>>
>>> --
>>> Dr. Michael Sumner
>>> Software and Database Engineer
>>> Australian Antarctic Division
>>> 203 Channel Highway
>>> Kingston Tasmania 7050 Australia
>>>
>>>
>> --
> Dr. Michael Sumner
> Software and Database Engineer
> Australian Antarctic Division
> 203 Channel Highway
> Kingston Tasmania 7050 Australia
>
> --
Dr. Michael Sumner
Software and Database Engineer
Australian Antarctic Division
203 Channel Highway
Kingston Tasmania 7050 Australia

	[[alternative HTML version deleted]]