get_patches function

get_patches

Connected components labeling to derive patches in a landscape.

get_patches(
  landscape,
  class = "all",
  directions = 8,
  to_disk = getOption("to_disk", default = FALSE),
  return_raster = TRUE
)

Arguments

• landscape: A categorical raster object: SpatRaster; Raster* Layer, Stack, Brick; stars or a list of SpatRasters.
• class: Either "all" (default) for every class in the raster, or specify class value. See Details.
• directions: The number of directions in which patches should be connected: 4 (rook's case) or 8 (queen's case).
• to_disk: Logical argument, if FALSE results of get_patches are hold in memory. If true, get_patches writes temporary files and hence, does not hold everything in memory. Can be set with a global option, e.g. option(to_disk = TRUE). See Details.
• return_raster: If false, matrix is returned

Returns

List of SpatRaster

Details

Searches for connected patches (neighbouring cells of the same class i). The 8-neighbours rule ('queen's case) or 4-neighbours rule (rook's case) is used. Returns a list with raster. For each class the connected patches have the value 1 - n. All cells not belonging to the class are NA.

Landscape metrics rely on the delineation of patches. Hence, get_patches is heavily used in landscapemetrics . As raster can be quite big, the fact that get_patches creates a copy of the raster for each class in a landscape becomes a burden for computer memory. Hence, the argument to_disk allows to store the results of the connected labeling algorithm on disk. Furthermore, this option can be set globally, so that every function that internally uses get_patches can make use of that.

Examples

landscape <- terra::rast(landscapemetrics::landscape)

# check for patches of class 1
patched_raster <- get_patches(landscape, class = 1)

# count patches
nrow(terra::unique(patched_raster[[1]][[1]]))

# check for patches of every class
patched_raster <-  get_patches(landscape)

References

Vincent, L., Soille, P. 1991. Watersheds in digital spaces: an efficient algorithm based on immersion simulations. IEEE Transactions on Pattern Analysis and Machine Intelligence. 13 (6), 583-598