[R-sig-Geo] Unexpected behaviour of "area" slot in sp Polygons

[Gregory Duveiller]

Dear All,

I noticed something unusual when trying to get the areas of some polygons I
extracted from a raster.

In the script below, I generate a random raster with 4 classes, which I
then turn to a "SpatialPolygonsDataFrame". I then try to calculate areas
per class in two different ways.

########

require(sp)
require(raster)
require(rgeos)

nrows=100; ncols=100

rdum<-raster(nrows=nrows,ncols=ncols,xmn=0,ymn=0,ymx=100,xmx=100)
rdum[]<-round(runif(nrows*ncols,1,4))

vdum <- rasterToPolygons(rdum,dissolve=T)

# calculate areas
df <- data.frame(cls=1:4)
for(i in 1:4){
  ddd <- sapply(vdum at polygons[[i]]@Polygons, function(x) x at area)
  df$area1[i] <- sum(ddd)
  df$area2[i] <- sum(as.vector(rdum==i))
}
#########

When I try to calculate the area per class from the attributes of the
Polygons, I do not always get the same number as when I simply count the
number of pixels of each class in the raster. In one iteration, for
example, I get the following number in which the areas in the Polygons is
higher by 52 units:

> df
  cls area1 area2
1   1  1670  1670
2   2  3384  3352
3   3  3355  3335
4   4  1643  1643

Is there a reason one would not expect these to be always equal?

Thanks in advance...

Gregory Duveiller

	[[alternative HTML version deleted]]