Package index • autoSpectra

OSSL Data

Download and manage Open Soil Spectral Library data

ossl_cache_dir(): Default local cache directory for OSSL data
ossl_download(): Download all three OSSL components (VisNIR, MIR, soillab)
ossl_download_file(): Download one OSSL component file to the cache directory
ossl_join(): Join OSSL spectra with soil lab data for model training
ossl_l1_labels: Fancy display labels for OSSL L1 properties
ossl_l1_properties: All OSSL Level-1 harmonized soil property variable names (34 targets)
ossl_load_raw(): Load a cached OSSL CSV (gzipped) as a data.table
ossl_mir_instruments: MIR instruments contributing to OSSL v1.2
ossl_mir_matrix(): Extract OSSL MIR spectra as a matrix aligned to the OSSL standard grid
ossl_prepare(): Convenience: load and join all OSSL data for a given sensor type
ossl_soillab(): Extract OSSL Level-1 soil lab data for a set of properties
ossl_visnir_instruments: VisNIR instruments contributing to OSSL v1.2
ossl_visnir_matrix(): Extract OSSL VisNIR spectra as a matrix aligned to the OSSL standard grid

Download pre-trained soilVAE models from Zenodo (one-time setup)

download_ossl_models(): Download pre-trained OSSL soilVAE models from Zenodo
models_available(): Check whether pre-trained models are available locally

Train soilVAE models on OSSL or custom datasets

build_soilVAE(): Build the soilVAE asymmetric autoencoder model
train_soilVAE(): Train a soilVAE model for a single soil property
train_ossl_models(): Train soilVAE models for all properties of one model family using OSSL data
load_soilVAE(): Load a saved soilVAE model and its scaler
save_scaler(): Save a z-score scaler to RDS and JSON
save_metrics_json(): Save per-model metrics JSON (RMSE, R2, RPIQ, conformal quantiles, latent stats)
metrics_from_y(): Compute prediction metrics (RMSE, R², RPIQ)
split_idx(): Split indices into train / calibration / test sets

Load models once per session and reuse them without disk I/O

get_cached_model(): Retrieve a soilVAE model from the in-memory cache
preload_ossl_models(): Pre-load all soilVAE models for a family into memory
list_cached_models(): List models currently held in the in-memory cache
clear_model_cache(): Clear the in-memory model cache

Predict soil properties and assess applicability domain

predict_soil(): Predict soil properties from a spectral data frame using soilVAE models
predict_applicability(): Compute applicability-domain scores for new samples
format_predictions(): Round predictions and rename columns to display labels
get_family(): Lookup a model family from the registry
family_matches(): Test whether a family matches a sensor type
property_label(): Get a display label for a soil property key

Spectral preprocessing functions

preprocess_visnir(): Canonical two-step SG preprocessing for VisNIR (reflectance input)
preprocess_mir(): Canonical two-step SG preprocessing for MIR (absorbance input)
reflectance_to_absorbance(): Convert reflectance to absorbance
apply_pipeline(): Apply a spectral preprocessing pipeline
apply_sg_matrix(): Apply Savitzky-Golay filter row-wise to a spectral matrix
apply_snv(): Apply Standard Normal Variate (SNV) row-wise
parse_sg(): Parse SG parameter string of the form "SG(m,p)", "SG(m,p,d)", etc.
sg_safe_params(): Compute safe SG parameters given available number of columns
resample_to_grid(): Resample a spectral matrix to a target wavelength grid via linear interpolation
get_wavelengths(): Extract wavelength/wavenumber positions from a data frame's column names

Plotting functions

plot_spectra(): Plot one or more spectra from a data frame
plot_mean_spectrum(): Plot a mean spectrum with ±1 SD ribbon from a spectral matrix
plot_predictions(): Plot predicted vs observed values for a soil property
plot_applicability(): Plot latent-space applicability domain scores

Model registry and metadata

model_registry: Official autoSpectra model family registry
ossl_l1_properties: All OSSL Level-1 harmonized soil property variable names (34 targets)
ossl_l1_labels: Fancy display labels for OSSL L1 properties
ossl_visnir_instruments: VisNIR instruments contributing to OSSL v1.2
ossl_mir_instruments: MIR instruments contributing to OSSL v1.2

Launch the interactive application

Internal helper functions

feature_minmax(): Compute per-band min/max feature range from a preprocessed matrix
dir_create(): Create a directory if it does not already exist
vneed(): Safe shiny::validate wrapper