soilsampling: Soil Sampling Design Methods

The soilsampling package provides methods for designing soil sampling schemes including spatial coverage sampling, simple random sampling, stratified random sampling, and composite sampling. The package uses sf for spatial operations and does not require Java or the sp package.

Details

The package implements three main sampling approaches:

Simple Random Sampling (ss_random()): Selects sampling locations uniformly at random within the study area.
Stratified Random Sampling (ss_stratified()): Divides the study area into compact strata and samples randomly within each stratum.
Spatial Coverage Sampling (ss_coverage()): Places samples at the centroids of compact geographical strata for optimal spatial coverage.

The stratification is based on k-means clustering as described in Walvoort et al. (2010). Two k-means algorithms are available:

Transfer algorithm: Standard k-means that creates compact strata of potentially unequal sizes.
Swop algorithm: Creates compact strata of equal size, useful for composite sampling.

Main Functions

Stratification:

ss_stratify() - Create spatial strata using k-means

Sampling:

ss_random() - Simple random sampling
ss_stratified() - Stratified random sampling
ss_coverage() - Spatial coverage sampling (centroids)
ss_coverage_equal_area() - Coverage sampling with equal-area strata
ss_composite() - Composite sampling from equal-area strata

Visualization:

ss_plot() - Plot stratification and samples
ss_plot_samples() - Plot sampling points only

Utilities:

ss_to_sf() - Convert results to sf object
ss_to_data_frame() - Convert results to data frame
ss_summary() - Get summary statistics
ss_n_strata() - Get number of strata
ss_n_samples() - Get number of samples
ss_area() - Get stratum areas
ss_relative_area() - Get relative stratum areas

Workflow

A typical workflow for spatial coverage sampling:


library(soilsampling)
library(sf)

# Load or create study area
study_area <- st_read("study_area.shp")

# Option 1: Direct coverage sampling
samples <- ss_coverage(study_area, n_strata = 50, n_try = 10)

# Option 2: Create strata first, then sample
strata <- ss_stratify(study_area, n_strata = 50, n_try = 10)
samples <- ss_coverage(strata)

# Visualize
ss_plot(strata, samples = samples)

# Export coordinates
coords <- ss_to_data_frame(samples)
write.csv(coords, "sampling_points.csv")

When to Use Each Method

Spatial coverage sampling (ss_coverage()): Best for model-based inference (kriging). Provides optimal spatial coverage for interpolation.
Stratified random sampling (ss_stratified()): Best for design-based inference. Provides valid probability samples while ensuring coverage.
Simple random sampling (ss_random()): Use when no prior stratification is desired. Valid for design-based inference.
Composite sampling (ss_composite()): Use when samples will be physically combined before analysis to reduce costs.

References

de Gruijter, J.J., Brus, D.J., Bierkens, M.F.P., and Knotters, M. (2006). Sampling for Natural Resource Monitoring. Springer, Berlin.

Walvoort, D.J.J., Brus, D.J., and de Gruijter, J.J. (2010). An R package for spatial coverage sampling and random sampling from compact geographical strata by k-means. Computers & Geosciences 36, 1261-1267. doi:10.1016/j.cageo.2010.04.005

Author

Maintainer: Carlos Carbajal ccarbajal@educagis.com