Sampling within a region

Draws a simple random sample of the rows whose coordinates fall inside a region. It decides where sampling happens, not how evenly it covers the ground: for a sample spread evenly across space — usually far more precise when what you measure varies smoothly — use design_spread(), on the frame restricted to the region if need be.

Usage

design_spatial(coords, region, n, crs = 4326, na_rm = FALSE)

Arguments

coords

Two column names, c(x, y) — longitude then latitude.

region

An sf or sfc geometry, or a list of them, which is unioned. Reprojected to crs when it differs.

n

Number of rows to draw from inside the region.

crs

Coordinate reference system of the coordinate columns. Defaults to EPSG:4326.

na_rm

Drop rows with missing coordinates instead of raising an error.

Value

A design object, for use with draw().

Coordinate order

coords is c(x, y) — longitude first, then latitude — matching sf::st_as_sf(). Getting this backwards puts your points somewhere else entirely, so it is worth checking against a known landmark once.

Regions that cross the antimeridian

With spherical geometry enabled (sf::sf_use_s2(), the default), consecutive polygon vertices are joined by the shortest great-circle path. An edge from longitude -179 to +179 therefore spans the 2 degrees across the antimeridian, not the 358 degrees the coordinates suggest. A "whole world" rectangle collapses into a 2-degree-wide pole-to-pole strip of 2.8 million km2 – against the roughly 510 million km2 of the globe – and contains almost nothing.

Ring orientation is not the cause and reversing it does not help: sf normalises winding, so both directions give the same strip. Adding intermediate vertices does not reliably help either, because the intermediate edges still bow along geodesics.

draw() warns whenever any edge of region spans more than 180 degrees of longitude. If you hit it, either split the region at the antimeridian into two polygons, or switch to planar interpretation with sf::sf_use_s2(FALSE).

See also

draw()

Other designs: design_bootstrap(), design_certainty(), design_cluster(), design_multistage(), design_reservoir(), design_simple(), design_spread(), design_stratified(), design_systematic(), design_temporal(), design_weighted()

Examples

box <- sf::st_sfc(
  sf::st_polygon(list(cbind(c(0, 10, 10, 0, 0), c(0, 0, 10, 10, 0)))),
  crs = 4326
)
df <- data.frame(id = 1:20,
                 lon = seq(1, 9, length.out = 20),
                 lat = seq(1, 9, length.out = 20))
draw(df, design_spatial(c("lon", "lat"), region = box, n = 5), seed = 1)
#>   id      lon      lat
#> 1  1 1.000000 1.000000
#> 2  2 1.421053 1.421053
#> 3  4 2.263158 2.263158
#> 4  7 3.526316 3.526316
#> 5 13 6.052632 6.052632