Package index • soilFlux

Data preparation

Standardise raw soil data for modelling

prepare_swrc_data(): Prepare a soil data frame for SWRC modelling

Model fitting

Train the CNN1D-PINN and swrc_fit S3 methods

fit_swrc(): Fit a physics-informed CNN1D SWRC model
build_swrc_model(): Build the CNN1D monotone-integral SWRC model
norouzi_lambdas(): Return default Norouzi et al. (2025) loss weights (lambdas)
build_residual_sets(): Build physics residual point sets (S1 – S4)
print(<swrc_fit>): Print method for swrc_fit
summary(<swrc_fit>): Summary method for swrc_fit

Prediction

Predict water content at specific or dense pF grids

predict_swrc(): Predict water content at specific pF or matric-head values
predict_swrc_dense(): Predict dense SWRC curves for a set of soil profiles
predict_theta_s(): Extract saturated water content (theta_s) from covariates
predict(<swrc_fit>): Predict method for swrc_fit

Evaluation & metrics

Performance metrics and evaluation

evaluate_swrc(): Compute metrics from a swrc_fit on new data
swrc_metrics(): Compute regression metrics for SWRC predictions
swrc_metrics_by_group(): Compute regression metrics by group

Visualisation

Publication-quality plots

plot_swrc(): Plot soil water retention curves (SWRC)
plot_pred_obs(): Plot predicted vs. observed water content
plot_swrc_metrics(): Plot model performance metric comparison
plot_training_history(): Plot training loss history

Texture

USDA texture classification and triangle

classify_texture(): Classify soil texture according to the USDA system
add_texture(): Add texture classification column to a data frame
texture_triangle(): Plot a USDA soil texture triangle (ternary diagram)

Save / Load

Persist and reload trained models

save_swrc_model(): Save a fitted SWRC model to disk
load_swrc_model(): Load a previously saved SWRC model from disk
swrc_model_exists(): Check whether a model directory contains a valid saved model

Internal helpers

Scaling, physics, utilities

fit_minmax(): Fit a min-max scaler from a training data frame
apply_minmax(): Apply a fitted min-max scaler to a data frame
invert_minmax(): Invert a min-max scaling transformation
pf_from_head(): Convert matric head (cm) to pF
head_from_pf(): Convert pF to matric head (cm)
pf_normalize(): Normalise pF values to [0, 1]
head_normalize(): Normalise matric head (cm) to the pF domain
parse_depth(): Parse a soil depth string into midpoint and label
parse_depth_column(): Parse depth column in a data frame
fix_bd_units(): Detect and correct bulk-density units
theta_unit_factor(): Detect theta unit scale factor
compute_physics_loss(): Compute the physics-informed residual loss (Norouzi et al. 2025)
residual_to_tensors(): Convert residual point sets to TensorFlow tensors
make_obs_matrices(): Build observation matrices for the CNN1D model
make_profile_array(): Build a sequence array for one or more soil profiles
make_train_step(): Create an eager-mode train step function
safe_grad(): Safely compute a TensorFlow gradient
data_prep: Data preparation for CNN1D SWRC modelling
io: Save and load fitted SWRC models
metrics: Performance metrics for SWRC models
model: CNN1D monotone-integral model architecture
physics: Physics-informed constraints for SWRC modelling
plots: Publication-quality plots for SWRC analysis
predict: Prediction from fitted SWRC models
scale: Min-max feature scaling
texture: USDA soil texture classification
train: Training the CNN1D SWRC model
utils: Utility functions for soilFlux

Data

Example datasets

swrc_example: Example soil water retention dataset