Skip to contents

Change the cell size of a GRaster

Source: R/resample.r
resample.Rd

resample() changes the cell size (resolution) of a GRaster using either another raster as a template or a user-defined resolution. Note that the extent of the output raster may be expanded to accommodate an integer number of cells. The function is not guaranteed to recreate the same output as terra::resample(), even when the same resampling method is used.

Usage

# S4 method for class 'GRaster,GRaster'
resample(x, y, method = NULL, fallback = TRUE)

# S4 method for class 'GRaster,numeric'
resample(x, y, method = NULL, fallback = TRUE)

Arguments

x

The GRaster to resample.

y

Either a GRaster to serve as a template, or a numeric vector with two or three values. If a numeric vector, the values represent east-west and north-south resolution for 2D rasters, or east-west, north-south, and top-bottom resolution for 3D rasters.

method

Character or NULL: Method to use to assign values to cells. Partial matching is used.

  • NULL (default): Automatically choose based on raster properties (near for categorical or integer rasters, bilinear for continuous data).

  • "near": Nearest neighbor. Best for categorical data, and often a poor choice for continuous data. If nlevels() is >0, this method will be used regardless of the value of method. If you still want to use a different method, coerce the raster to a different type using as.int(), as.float(), or as.doub().

  • "bilinear": Bilinear interpolation (default for non-categorical data; uses weighted values from 4 cells).

  • "bicubic": Bicubic interpolation (uses weighted values from 16 cells).

  • "lanczos": Lanczos interpolation (uses weighted values from 25 cells). Note that methods that use multiple cells will cause the focal cell to become NA if there is at least one cell with an NA in the cells it draws from. These NA cells can be filled using the fallback option.

fallback

Logical: If TRUE (default), then use "lower" methods to fill in NA cells when a "higher" method is used. For example, if method = "bicubic", NA cells will be filled in using the bilinear method, except when that results in NAs, in which case the near method will be used. Fallback causes fewer cells to revert to NA values, so can be better at resampling the edges of rasters. However, fallback does increase processing time because each "lower" method must be applied, then results merged.

Value

A GRaster.

See also

terra::resample(), modules r.resample and r.resamp.interp in GRASS

Examples

if (grassStarted()) {

# Setup
library(terra)

# Elevation raster
madElev <- fastData("madElev")
elev <- fast(madElev)

### Resample raster to 120 x 120 m
elev120 <- resample(elev, c(120, 120), method="bilinear")
elev
elev120

### Resample using another raster as a template
###############################################

template <- aggregate(elev, 4)

nearest <- resample(elev, template, method = "nearest")

bilinear <- resample(elev, template, method = "bilinear")
bilinearNoFB <- resample(elev, template, method = "bilinear", fallback = FALSE)

bicubic <- resample(elev, template, method = "bicubic")
bicubicNoFB <- resample(elev, template, method = "bicubic", fallback = FALSE)

lanczos <- resample(elev, template, method = "lanczos")
lanczosNoFB <- resample(elev, template, method = "lanczos", fallback = FALSE)

# rasters resampled without fallback have fewer non-NA cells
resampled <- c(nearest, bilinear, bilinearNoFB, bicubic, bicubicNoFB, lanczos,
    lanczosNoFB)
names(resampled) <- c("nearest", "bilinear", "bilinearNoFB", "bicubic",
    "bicubicNoFB", "lanczos", "lanczosNoFB")
ones <- resampled * 0 + 1
global(ones, "sum") # number of non-NA cells
global(resampled, c("mean", "sd", "min", "max")) # other statistics

# Compare fallback to no fallback
frLanczos <- rast(lanczos)
frLanczosNoFB <- rast(lanczosNoFB)

plot(frLanczos, col = "red",
    main = "Red: Cells in fallback not non-fallback", legend = FALSE)
plot(frLanczosNoFB, add=TRUE)

# Compare fasterRaster with terra
coarserTerra <- aggregate(madElev, 4)
terraLanczos <- resample(madElev, coarserTerra, method = "lanczos")

frLanczos <- extend(frLanczos, terraLanczos)
frLanczosNoFB <- extend(frLanczosNoFB, terraLanczos)

frLanczos - terraLanczos
frLanczosNoFB - terraLanczos

plot(frLanczos - terraLanczos, main = "Difference")
plot(frLanczosNoFB - terraLanczos, main = "Difference")

plot(terraLanczos, col = "red",
    main = "Red: Cells in terra not in FR", legend = FALSE)
plot(frLanczos, add=TRUE)

plot(frLanczos, col = "red",
    main = "Red: Cells in FR not in terra", legend = FALSE)
plot(terraLanczos, add=TRUE)

}