Changelog

Source: NEWS.md

soilFlux 0.1.5

Bug fixes (CRAN reviewer requests)

• DESCRIPTION: removed single quotes around acronyms (CNN1D-PINN, SWRC, pF) — CRAN policy requires single quotes only for package/software names.
• DESCRIPTION: added <doi:10.1029/2024WR038149> link to the Norouzi et al.
  2025. reference; corrected journal name from Journal of Hydrology to Water Resources Research.
• Documentation: added missing \value tags to predict.Rd, predict.swrc_fit.Rd, print.swrc_fit.Rd, and summary.swrc_fit.Rd.
• Examples: replaced all \dontrun{} with \donttest{} throughout the package, as the examples can in principle be executed (they require TensorFlow, but are not inherently unrunnable).
• io.R: removed the default dir = "./models/swrc" path from save_swrc_model() to comply with CRAN policy against writing to the user’s home filespace by default; updated examples to use tempdir().

soilFlux 0.1.4

Bug fixes

• Updated cran-comments.md to reflect version 0.1.4 submission.
• Minor whitespace normalization in project configuration files.

soilFlux 0.1.3

Bug fixes

• Added prebuilt vignette index (inst/doc/index.html) to fix CRAN NOTE “VignetteBuilder field but no prebuilt vignette index”.
• Quoted technical acronyms ('CNN1D-PINN', 'SWRC', 'pF') in DESCRIPTION to reduce spell-check NOTEs on CRAN incoming checks.

soilFlux 0.1.2

Bug fixes

• Fixed LaTeX PDF build error caused by Unicode character ≈ (U+2248) in .Rd documentation. Replaced with \eqn{\approx} in make_obs_matrices and swrc_example man pages.
• Added inst/WORDLIST with domain-specific terms (SWRC, PINN, pF, Norouzi, matric, convolutional, positivity) to suppress spell-check NOTEs on CRAN incoming checks.

soilFlux 0.1.1

Bug fixes

• Pre-built vignette output added to inst/doc/ to avoid LaTeX/Pandoc WARNINGs during CRAN checks.
• Updated .Rbuildignore accordingly.

soilFlux 0.1.0

• Initial CRAN submission.

New features

• build_swrc_model() — CNN1D monotone-integral architecture (Norouzi et al. 2025).
• fit_swrc() — Full training loop with physics constraints, early stopping, and configurable loss weights.
• predict_swrc() — Point predictions at arbitrary pF / matric-head values.
• predict_swrc_dense() — Dense continuous SWRC curves for any number of soil profiles.
• predict_theta_s() — Extract modelled saturated water content (θs).
• evaluate_swrc() — Compute R², RMSE, MAE on a test set.
• swrc_metrics() / swrc_metrics_by_group() — Regression performance metrics.
• norouzi_lambdas() — Loss-weight configurations from Norouzi et al. (2025) Table 1.
• build_residual_sets() / residual_to_tensors() — Physics collocation points (S1–S4).
• compute_physics_loss() — Evaluate the four physics constraints.
• fit_minmax() / apply_minmax() / invert_minmax() — Column-wise min-max scaling.
• prepare_swrc_data() — Standardise raw soil data (units, depth parsing, theta normalisation).
• make_obs_matrices() / make_profile_array() — Build 3-D sequence arrays for the CNN1D.
• parse_depth() / parse_depth_column() — Parse depth strings to midpoint and label.
• pf_from_head() / head_from_pf() / pf_normalize() / head_normalize() — pF / matric-head conversion helpers.
• fix_bd_units() / theta_unit_factor() — Automatic unit correction.
• save_swrc_model() / load_swrc_model() / swrc_model_exists() — Model persistence (weights + metadata).
• plot_swrc() — Publication-quality SWRC curve figure.
• plot_pred_obs() — Predicted vs. observed scatter plot with metrics.
• plot_swrc_metrics() — Bar chart comparing model performance.
• plot_training_history() — Training / validation loss curves.
• classify_texture() / add_texture() / texture_triangle() — USDA texture classification and ternary diagram.
• swrc_example — Example soil characterization dataset.