Allow classes for window_lsm()

[mhesselbarth]

Allow also class level metrics for window lsm.

Either specify focal class beforehand or return RasterStack in which each layer represents a class.

[Nowosad]

I would suggest the second option (RasterStack) or, if possible both options.

(We should also wait for the email reply before committing to any solutions...)

[mhesselbarth]

I think you were not included in the reply, sorry (I just forwarded it to you).

I think something similar to e.g. the class argument in e.g. show_lsm() would be cool. So basically it possible to get all classes or only selected ones. From a computational point, we need to calculate all classes anyways and then either just return one specific class or all.

[bitbacchus]

From a computational point, we need to calculate all classes anyways and then either just return one specific class or all.

Not necessarily, I think. I'm just hacking a class-level moving-window version of ed for myself and I just reclassify everything except the focal class to NA. Then I added an optional count_background_NA option to lsm_l_ed:

library(landscapemetrics)
library(raster)

landscape <- landscapemetrics::landscape
window <- matrix(1, nrow = 5, ncol = 5)
f_class <- 1
classes <- raster::unique(landscape)
bg_classes <- classes[classes != f_class]

for (bg in bg_classes) {
    landscape <- raster::reclassify(landscape, cbind(bg, NA))
}

ed_map <- landscapemetrics::window_lsm(landscape = landscape, 
                                       window = window, 
                                       what = c("lsm_l_ed"))
ed_map <- ed_map[[1]]
landscapetools::show_landscape(ed_map)

^{Created on 2022-07-06 by the reprex package (v2.0.1)}

This required tweaking lsm_l_ed and raster:::focal_fun, however...