funcs

A practical R toolbox for digital soil mapping, raster processing, environmental covariates, machine learning, and spatial prediction.

What This Is | Highlights | Quick Start | Function Map | Citation

What This Is

funcs is a personal, research-oriented collection of R functions and workflow scripts developed by Cassio Marques Moquedace to support reproducible work in soil science, digital soil mapping, environmental modelling, raster processing, machine learning, model evaluation, and spatial prediction.

This repository is intentionally lightweight: functions are kept as standalone .R files that can be sourced directly into analysis projects. It is not currently structured as a formal R package.

Highlights

Soil data diagnosis Evaluate completeness, bottlenecks, depth coverage, and attribute trade-offs in harmonized soil databases.

Raster operations Crop, mask, project, tile, resample, write classified rasters, and prepare large spatial domains.

Model workflows Regression and classification templates with repeated runs, LOOCV, grouped validation, RFE, and custom metrics.

Spatial prediction Utilities for quantile random forest prediction, probability rasters, uncertainty outputs, and model diagnostics.

Quick Start

Source only the function you need:

source("https://raw.githubusercontent.com/moquedace/funcs/main/utils/soil_attr_balance.R")

res <- soil_attr_balance(
  data = soil_data,
  attrs = soil_attributes,
  unit_cols = "coord_id",
  depth_cols = c("upper_depth_cm", "lower_depth_cm"),
  required_attrs = "c_gkg",
  target_attrs = "c_gkg",
  min_pct = 0.50,
  ranking_metric = "weighted_score",
  return_selected = TRUE
)

For balanced raster processing:

source("https://raw.githubusercontent.com/moquedace/funcs/main/geospatial/balanced_raster_tiles.R")

tile_result <- balanced_raster_tiles(
  base_raster = terra::rast("path/to/base_raster.tif"),
  n_tile_rows = 10,
  n_tile_cols = 10,
  output_dir = "output/balanced_tiles",
  diagnostic_plots = TRUE
)

Function Map

Area	Main files
Soil database diagnostics	soil_attr_balance.R, soil_attr_balance_documentation.md, soil_attr_balance_examples.md
Balanced raster tiling	balanced_raster_tiles.R, balanced_raster_tiles_documentation.md, balanced_raster_tiles_examples.md, check_tiles.R
Raster processing	crop_mask_project.R, standardize_crop_mask_raster.R, change_resolution.R, focal_resample.R, tile_raster.R, tile_raster_path.R, writeRaster_factor.R, rst_class_by_value.R
Environmental covariates	soilgrids_raster.R, download_febr_soildata.R, process_landsat_indices.R, calc_index_sentinel.R, morphometry_saga.R
Regression modelling	regression_modeling_single_run.R, regression_modeling_repeated_runs.R, regression_modeling_loocv.R, regression_modeling_leave_one_group_out.R
Classification modelling	classification_modeling_single_run.R, classification_modeling_repeated_runs.R, classification_modeling_loocv.R, classification_modeling_leave_one_group_out.R
Prediction writers	predict_qrf_raster.R, pred_writer_qrf_raster.R, pred_writer_raster_prob_raw.R, pred_writer_raster_prob_raw_resample.R
Model evaluation	pst_res_mqi.R, pst_res_class.R, pst_res_class_multiclass.R, calcular_metricas_raster.R, partial_dependence.R, aoa_meyer.R, explain_future.R
Data preparation	remove_outliers.R, points_to_class.R, spl.R, bivariate_map.R, app_vect.R
Utilities	install_load_pkg.R, copy_with_robocopy.R, gcs_download.R, gcs_upload.R, gbm_custom.R, caret_rfe_functions.R

Recommended Use

1. Read the relevant .md documentation when available.
2. Source the specific .R file needed for your workflow.
3. Keep project-specific paths, input data, and output folders in your own analysis repository.
4. Check package dependencies inside each function or workflow before running.

Repository Philosophy

This repository favors practical reuse over package ceremony. Many routines were created inside research projects and later generalized when they became useful across multiple workflows.

The main design goal is to make recurring geospatial and modelling tasks easier to inspect, adapt, and reuse.

Citation

If this repository supports your work, cite it as:

Moquedace, C. M. (2026). funcs: R functions and computational routines for
digital soil mapping, raster processing, environmental modelling, and spatial
prediction. GitHub repository. https://github.com/moquedace/funcs

Machine-readable citation metadata is available in CITATION.cff.

License

This repository is made available under the MIT License. See LICENSE.