[R-sig-Geo] Create a simple spatial grid

Barry Rowlingson b.rowlingson at lancaster.ac.uk
Thu May 22 12:46:07 CEST 2014


On Wed, May 21, 2014 at 5:14 PM, Remi Genevest <rgenevest at free.fr> wrote:
> Thanks a lot Barry. Your post perfectly matches what I was looking for.
>
> Well now if I want to do a coarser grid covering Europe, what I should do is
> doing more or less the same thing, just to change the resolution, right?
>
> res(r)=c(100,100)  # an approximate 100meters squared cells
>
> Then I got my raster in a EU coordinates system, and I would like to
> transform it into a WGS 84 (so as adding other layers would become easier).
> I tried this way, but it doesn't work :
>
> # change the CRS into WGS84 (ll)
> r.new<-projectRaster(r,crs="+init=epsg:4326")
> # adapt the extent to Europe
> ext = extent(-10.417,31.917,34.083,71.083)
> extent(r.new)<-ext
> # plot the map
> plot(r.new,axes=TRUE)
> # Here I get a plot that is not projected in a proper way.

 Oh its very proper. Its just that you've changed the raster between
systems that don't have parallel axes.

 Rasters can only be defined by: origin x,y; cell-size x,y; number of
cells in x,y; the coordinate system for x and y.

 A system like the EU projection maps the sphere onto a cone, and
unwraps the cone to a flat surface (topologically speaking a cone is
flat - you can make a cone from a flat sheet unlike a sphere). If you
have a grid on that basis and want to convert to cells aligned in
lat-long, you'll get what looks like a rotated grid.

 I suspect what you might want is an approximate 100m grid over
Europe, aligned on lat-long lines. In which case, you can have 100m
cells in the middle of Europe, slightly more closely spaced cells in
the north, and slightly further spaced cells in the south - because
the grid has to span the same number of degrees over its entire
latitude.

 To make that, work out what 100m is in degrees longitude at your
study are mid-latitude, and create a raster from those parameters. You
can compute the distance distortion by computing great-circle
distances from grid cell coordinates.

Barry



More information about the R-sig-Geo mailing list