Calculate percentage of a class in a raster within a buffer in moving windows in R

Question

I would like to calculate the percentage of a class (for example forest) in a classified raster image inside a buffer (for example 1000m).

Speciffically, I have a 30x30m classified raster and for each pixel I'd like to know the percentage of forest in a buffer of 1000m for the entire image. The result should be a raster where each pixel has the percentage of the class in that buffer.

I think one way could be using "window_lsm" from the package "landscapemetrics", but I'm not sure.

Answer 1

The landscapemetrics::window_lsm is very slow. You would be better served just using a focal function in raster or terra. Here is an example in terra, you can read in your raster using terra::rast.

Add package and create an example 4 class raster.

library(terra)
r <- rast(nrow=1000, ncol=1000)
  r[] <- sample(1:4, ncell(r), replace=TRUE)

Write a function that returns percent of class(s). We set the default to 2,3 as to illustrate that there may be more than one class of interest (eg., 2 forest classes). You can simply change the default value(s) to suit your data.

pclass <- function(x, y=c(2,3)) {
  return( length(which(x %in% y)) / length(x) )
}

Now, we pass our pclass function to the focal function thus, returning class percent within a 9x9 window.

( pf <- terra::focal(r, w=matrix(1, 9, 9), pclass) )

par(mfrow=c(2,1))
  plot(r)
  plot(pf)

If your raster is projected and you want a circular focal window you can use the focalMat function to create an appropriate matrix (in this example 90m radius). The one catch is that the function returns weights, which can be dealt with by replacing non-zero values with 1 (as shown). This matrix would then be passed to the w argument in terra::focal.

f <- terra::focalMat(r, 90, "circle") 
  f[f > 0] <- 1