Take a regular sample

Take a spatial sample from a SpatRaster, SpatVector or SpatExtent. Sampling a SpatVector or SpatExtent always returns a SpatVector of points.

With a SpatRaster, you can get cell values, cell numbers (cells=TRUE), coordinates (xy=TRUE) or (when method="regular" and as.raster=TRUE) get a new SpatRaster with the same extent, but fewer cells.

In order to assure regularity when requesting a regular sample, the number of cells or points returned may not be exactly the same as the size requested unless you use exact=TRUE.

# S4 method for class 'SpatRaster'
spatSample(x, size, method="random", replace=FALSE, na.rm=FALSE, 
    as.raster=FALSE, as.df=TRUE, as.points=FALSE, values=hasValues(x), cells=FALSE, 
    xy=FALSE, ext=NULL, warn=TRUE, weights=NULL, exp=5, exhaustive=FALSE, 
  exact=FALSE, each=TRUE, ...)

# S4 method for class 'SpatVector'
spatSample(x, size, method="random", strata=NULL, chess="")

# S4 method for class 'SpatExtent'
spatSample(x, size, method="random", lonlat, as.points=FALSE, exact=FALSE)

Arguments

x

SpatRaster, SpatVector or SpatExtent

size

numeric. The sample size. If x is a SpatVector, you can also provide a vector of the same length as x in which case sampling is done separately for each geometry. If x is a SpatRaster, and you are using method="regular" you can specify the size as two numbers (number of rows and columns). Note that when using method="stratified", the sample size is returned for each stratum

method

character. Should be "regular" or "random", If x is a SpatRaster, it can also be "stratified" (each value in x is a stratum), "weights" (each value in x is a probability weight), or "spread" (an approximately regular sample, using compact zones generated with k_means clustering of the raster cell locations)

replace

logical. If TRUE, sampling is with replacement (if method="random")

na.rm

logical. If TRUE, NAs are removed. Only used with random sampling of cell values. That is with method="random", as.raster=FALSE, cells=FALSE

as.raster

logical. If TRUE, a SpatRaster is returned

as.df

logical. If TRUE, a data.frame is returned instead of a matrix

as.points

logical. If TRUE, a SpatVector of points is returned

values

logical. If TRUE raster cell values are returned

cells

logical. If TRUE, cell numbers are returned. If method="stratified" this is always set to TRUE if xy=FALSE

xy

logical. If TRUE, cell coordinates are returned

ext

SpatExtent or NULL to restrict sampling to a subset of the area of x

warn

logical. Give a warning if the sample size returned is smaller than requested

weights

SpatRaster. Used to provide weights when method="stratified"

lonlat

logical. If TRUE, sampling of a SpatExtent is weighted by cos(latitude). For SpatRaster and SpatVector this done based on the crs, but it is ignored if as.raster=TRUE

exp

numeric >= 1. "Expansion factor" that is multiplied with size to get an initial sample used for stratified samples and random samples with na.rm=TRUE to try to get at least size samples

exhaustive

logical. If TRUE and (method=="random" and na.rm=TRUE) or method=="stratified", all cells that are not NA are determined and a sample is taken from these cells. This is useful when you are dealing with a very large raster that is sparse (most cells are NA). Otherwise, the default approach may not find enough samples. This should not be used in other cases, especially not with large rasters that mostly have values

exact

logical. If TRUE and method=="regular", the sample returned is exactly size, perhaps at the expense of some regularity. Otherwise you get at least size many samples. Ignored for lon/lat rasters

each

logical. If TRUE and method=="stratified", the sample returned is size for each stratum. Otherwise size is the total sample size

...

additional arguments passed to k_means when method="kmeans"

strata

if not NULL, stratified random sampling is done, taking size samples from each stratum. If x has polygon geometry, strata must be a field name (or index) in x. If x has point geometry, strata can be a SpatVector of polygons or a SpatRaster

chess

character. One of "", "white", or "black". For stratified sampling if strata is a SpatRaster. If not "", samples are only taken from alternate cells, organized like the "white" or "black" fields on a chessboard

Value

numeric matrix, data.frame, SpatRaster or SpatVector

Examples

f <- system.file("ex/elev.tif", package="terra")
r <- rast(f)
s <- spatSample(r, 10, as.raster=TRUE)
spatSample(r, 5)
#>   elevation
#> 1       380
#> 2        NA
#> 3       495
#> 4       413
#> 5       341
spatSample(r, 5, na.rm=TRUE)
#>   elevation
#> 1       250
#> 2       260
#> 3       332
#> 4       462
#> 5       318
spatSample(r, 5, "regular")
#>   elevation
#> 1       479
#> 2       NaN
#> 3       NaN
#> 4       419
#> 5       290
#> 6       306
#> 7       281
#> 8       286
#> 9       NaN

## if you require cell numbers and/or coordinates
size <- 6
spatSample(r, 6, "random", cells=TRUE, xy=TRUE, values=FALSE)
#>      cell        x        y
#> [1,] 5743 6.095833 49.68750
#> [2,] 3037 6.504167 49.92917
#> [3,] 5898 5.804167 49.67083
#> [4,] 7943 6.220833 49.49583
#> [5,] 2926 6.370833 49.93750
#> [6,] 2015 5.904167 50.01250

# regular, with values 
spatSample(r, 6, "regular", cells=TRUE, xy=TRUE)
#>   cell        x        y elevation
#> 1 7458 6.137500 49.53750       264
#> 2 7505 6.529167 49.53750        NA
#> 3 7411 5.745833 49.53750        NA
#> 4 5368 6.137500 49.72083       289
#> 5 5415 6.529167 49.72083        NA
#> 6 5321 5.745833 49.72083        NA
#> 7 3183 6.137500 49.91250       322
#> 8 1093 6.137500 50.09583        NA

# stratified
rr <- rast(ncol=10, nrow=10, names="stratum")
set.seed(1)
values(rr) <- round(runif(ncell(rr), 1, 3))
spatSample(rr, 2, "stratified", xy=TRUE)
#>         x   y stratum
#> [1,] -162 -81       1
#> [2,]  -54  45       1
#> [3,] -126 -27       2
#> [4,]   90 -81       2
#> [5,] -162   9       3
#> [6,]   54  27       3

s <- spatSample(rr, 5, "stratified", as.points=TRUE, each=FALSE)
plot(rr, plg=list(title="raster"))
plot(s, 1, add=TRUE, plg=list(x=185, y=1, title="points"), col=rainbow(5))

 
# spread 
s <- spatSample(r, 10, "spread", as.points=TRUE)
plot(r); points(s)

  
## SpatExtent 
e <- ext(r)
spatSample(e, 10, "random", lonlat=TRUE)
#>              x        y
#>  [1,] 6.060745 49.45733
#>  [2,] 6.389876 49.58025
#>  [3,] 6.345150 50.02372
#>  [4,] 5.901934 50.08592
#>  [5,] 6.307978 49.52574
#>  [6,] 6.234556 49.74633
#>  [7,] 6.037959 50.11587
#>  [8,] 5.892183 49.49884
#>  [9,] 5.856140 49.48130
#> [10,] 6.134349 49.84738

## SpatVector
f <- system.file("ex/lux.shp", package="terra")
v <- vect(f)

# sample the geometries 
i <- sample(v, 3)

# sample points in geometries
p <- spatSample(v, 3)