Sampling

Jakub Nowosad

2025-04-07

Let’s start by attaching the necessary packages and setting the seed.

library(simsam)
library(terra)
#> terra 1.8.42
set.seed(2024-11-20)

Next, we define the raster grid to sample points/locations from.

rast_grid = terra::rast(xmin = 0, xmax = 200, ymin = 0, ymax = 200,
                        ncols = 200, nrows = 200)

Alternatively, we can read a raster from a file.

Sample

The simsam package supports several types of sampling, including jittered, clustered, and random with the sam_field() function. All these types of sampling require the rast_grid object, the number of samples to create (size), and the type of sampling (type).

The jittered sampling is a regular sample moved by an amount of noise. The value parameter controls the amount of jitter. Importantly, this sampling type ensures that all the samples fall within the sampling window.

sample_j1 = sam_field(rast_grid, size = 100, type = "jittered", value = 5)
plot(sample_j1)

sample_j2 = sam_field(rast_grid, size = 100, type = "jittered", value = 20)
plot(sample_j2)

The clustered sampling creates samples forming groups of points with similar locations. The outcome is controlled by the nclusters and value parameters, where the former controls the number of clusters and the latter the amount of noise (location variance) within each cluster.

sample_c1 = sam_field(rast_grid, 100, "clustered", nclusters = 5, value = 5)
plot(sample_c1)

sample_c2 = sam_field(rast_grid, 100, "clustered", nclusters = 5, value = 10)
plot(sample_c2)

Random sampling creates samples with no spatial structure.

sample_r = sam_field(rast_grid, 100, "random")
plot(sample_r)

Moreover, the sam_field() function can be used with additional arguments passed to the terra::spatSample() function, when the type is not specified. In such cases, the method argument can be used to control the sampling method, e.g., method = "regular" or method = "random" and the exhaustive argument to sample only from not NA areas in the input raster.

sample_u1 = sam_field(rast_grid, 100, method = "regular")
plot(sample_u1)

sample_u2 = sam_field(rast_grid, 100, method = "random", exhaustive = TRUE)
plot(sample_u2)