SoilGrids spatial prior

Reads a categorical raster of dominant Reference Soil Groups around the pedon's site, buffers the point in metric coordinates, extracts all pixel values within the buffer, and returns the empirical class frequency as a probability distribution over RSG codes.

Usage

spatial_prior_soilgrids(
  pedon,
  system = c("wrb2022", "usda"),
  buffer_m = 250,
  source_url = NULL,
  n_classes_top = 10,
  lut = NULL,
  ...
)

Arguments

pedon

A PedonRecord with non-NULL site$lat and site$lon.

system

Classification system; "wrb2022" (default) maps SoilGrids integer codes through the WRB lookup table. "usda" is reserved for a future SoilGrids-USDA layer.

buffer_m

Buffer radius in metres around the point (default 250 m, i.e. one SoilGrids pixel).

source_url

Optional. A path or URL accepted by terra::rast. If NULL, falls back to getOption("soilKey.test_raster").

n_classes_top

Keep only the top N classes by frequency (default 10). Set to Inf to keep all.

lut

Optional named integer vector mapping raster values to RSG codes. Default is soilgrids_wrb_lut; pass a custom one if your raster uses different codes.

...

Reserved for future use.

Value

A data.table with columns rsg_code, probability.

Data source

For real use, pass source_url pointing at a SoilGrids "MostProbable WRB" GeoTIFF / COG, e.g. one of the regional cuts published at https://files.isric.org/soilgrids/latest/data/wrb/. For tests, set options(soilKey.test_raster = "/path/to/syn.tif") to point at a local synthetic raster – this avoids network access in CI.

Coordinate handling

We use sf::st_transform when sf is available; otherwise we fall back to terra::project on a single-point SpatVector. The buffer is constructed in metric (UTM) coordinates so buffer_m is in metres regardless of the pedon CRS. The raster itself is queried in its native CRS via terra's automatic reprojection.

See also

spatial_prior, soilgrids_wrb_lut.