spatialEco (version 1.3-7)

polygon_extract: polygon raster extract

Description

Fast method for extracting raster values to polygons

This method for raster extraction uses the raster cell indices and is quite a bit faster with polygon data than other methods. This is especially true with large raster stacks (eg., time-series). The cell indices are returned using cellnumbers. If ids argument is provided a column with values from the associate column are included otherwise "row_names" is returned which corresponds to the rownames in the source polygon object. Please note that if use.terra = TRUE it will coerce to a terra class if not already. With large data the coercion overhead may be worth it, providing speed gains. If the raster is a terra SpatRaster it will operate in its native class. The cells = TRUE argument will return the cell indices which could be used at a later time.

Usage

polygon_extract(
  r,
  p,
  ids = NULL,
  cells = FALSE,
  asList = TRUE,
  use.terra = FALSE
)

Value

A list object containing a data.frame for each polygons raster values, as columns. Additional columns are "row_names" or which ever column is passed to the ids argument and "cells" if cells = TRUE. If asList = FALSE a data.frame will be returned

Arguments

r

RasterLayer, RasterStack, RasterBrick or SpatRaster object

p

sf polygon data

ids

A unique id field contained in p, will be assigned to output otherwise will return rownames

cells

FALSE | TRUE - Return cell index ids

asList

TRUE | FALSE - Output list object

use.terra

FALSE | TRUE - Use terra for extracting indices

Author

Jeffrey S. Evans <[e-mail]>

Examples

 library(sf)
 library(raster)
 
 nc <- sf::st_read(system.file("shape/nc.shp", package="sf"))
   nc <- sf::st_cast(nc, "POLYGON")
 
 #### multi-band 
 i=100; j=100
 r <- do.call(raster::stack, replicate(20, 
 	         raster::raster(matrix(runif(i*j), i, j))))
       names(r) <- paste0("time", 1:nlayers(r))			 
     extent(r) <- extent(nc)
 	  proj4string(r) <- st_crs(nc)$proj4string
 
 plot(r[[1]])
   plot(st_geometry(nc), add=TRUE)
 
 ( e <- polygon_extract(r, nc) )
 ( e <- polygon_extract(r, nc, ids="CNTY_ID") )
 
 # Column means
 lapply(e, function(x) apply(x[,2:ncol(x)], MARGIN=1, FUN=mean) )
 
 #### Single band mean
 ( e <- polygon_extract(r[[1]], nc, ids="CNTY_ID") )
   unlist(lapply(e, function(x) mean(x[,2], na.rm=TRUE) ))

# \donttest{ 
 # Leveraging cell ids, pulls values, calculates   
 # new value, and assigns to source cell using
 # index from cells = TRUE 

 e <- polygon_extract(r, nc, cells=TRUE)
 e <- data.frame(
   cells=unlist(lapply(e, function(x) as.numeric(x[,22]))),
   means=unlist(lapply(e, function(x) apply(x[,2:22], MARGIN=1, FUN=mean))) 
 )
 
 # copy raster and assign NA's
 r2 <- r[[1]]
   r2[] <- rep(NA, ncell(r))

 # assign using cell indices
 r2[e$cells] <- e$means 
   plot(r2)

 # benchmark against raster extract
 system.time({
   e <- polygon_extract(r, nc)
 })
 
 system.time({
   e <- raster::extract(r, nc)
 })

# }