Predict soil properties from spectra

Ergonomic, named entry point for the OSSL-backed predictive pipeline. Accepts either a PedonRecord or a numeric spectra matrix, applies the same preprocessing used at training time (recorded on each model), and returns predictions in the canonical long-form schema.

Usage

predict_from_spectra(
  pedon_or_spectra,
  models = NULL,
  properties = NULL,
  overwrite = FALSE,
  verbose = TRUE,
  ...
)

Arguments

pedon_or_spectra

A PedonRecord (predictions merged into the pedon) OR a numeric matrix / vector of raw Vis-NIR spectra (rows = horizons, columns = wavelengths).

models

A named list of soilKey_pls_model objects (output of train_pls_from_ossl). Required.

properties

Character vector of property names to predict. Defaults to all properties in models.

overwrite

Passed to fill_from_spectra when pedon_or_spectra is a PedonRecord.

verbose

Verbosity passed downstream.

...

Ignored (reserved for future backends).

Value

Either the mutated PedonRecord (invisibly) or a data.table with columns horizon_idx, property, value, pi95_low, pi95_high, n_neighbors.

Details

When pedon_or_spectra is a PedonRecord, this function delegates to fill_from_spectra with method = "pretrained" and the predictions are written back to the pedon (with source = "predicted_spectra" provenance). When pedon_or_spectra is a numeric matrix or vector, this function returns the prediction data.table directly without touching any pedon.

Examples

if (FALSE) { # \dontrun{
lib <- download_ossl_subset(region = "south_america")
models <- train_pls_from_ossl(lib,
                                properties = c("clay_pct", "ph_h2o"))
predict_from_spectra(my_pedon, models = models)
} # }