window_lsm for large raster files

[Gangrelino]

Hi all,
I have a large binary raster that I am trying to calculate some landscape metrics (lsm_l_area_mn, more specifically) using moving windows (window_lsm). The issue I am facing is that the raster might be too large and window_lsm is not only taking forever to run but also spitting the following error later:

Error in .focal_fun(as.vector(t(v)), w, paddim, runfun, NAonly) : 
  Evaluation error: object 'f' not found.
invalid option "error"

Here is some details of my raster:

class      : RasterLayer 
dimensions : 24674, 13797, 340427178  (nrow, ncol, ncell)
resolution : 50, 50  (x, y)
extent     : 171030.3, 860880.3, 5425628, 6659328  (xmin, xmax, ymin, ymax)
crs        : +proj=tmerc +lat_0=0 +lon_0=-115 +k=0.9992 +x_0=500000 +y_0=0 +datum=NAD83 +units=m +no_defs 
source     : mosaic.tif 
names      : mosaic 
values     : 1, 1  (min, max)

Does anyone recognize the above error? Also, is there a different or better approach to deal with large raster files that could speed the processing time?

The raster was initially a 0/1 raster, but then I realized that window_lsm will calculate the chosen metric for all classes (including the 0 class), so I reclassified the raster to have only one class (1), while leaving the other cells as NA.

Here is an example of the approach I am taking.

library(landscapemetrics)
library(landscapetools)

#converting to binary (1/NA)
example_landscape<-landscape %>% reclassify(rcl = c(-1,1,NA,
                                                    1,2,1,
                                                    2,3,1))

window_w2w<-window_lsm(example_landscape,matrix(1,5,5), what = "lsm_l_area_mn" ) 

The example does not produce any errors and is slightly different from what I am doing, the only difference being the size of the window (21x21 in my case)

[bitbacchus]

Hm, your raster doesn't seem to be that big... Can you maybe share the tif file so we can try to reproduce the issue?

[Gangrelino]

For sure, thanks for your willingness to help. The raster can be accessed here

[Gangrelino]

I also tried to GitHub version and it gives the same error

[bitbacchus]

OK, I tried to run this, and you were right, the raster is rather big for window_lsm and it takes ages.

The plan that should kinda solve this is now to re-implement window_lsm with internals from terra (#245) so that it can hopefully be run parallelized in future (rspatial/terra#36).