Useful R Packages for Digital Soil Mapping

In addition to base R, a wide range of contributed packages are essential for implementing Digital Soil Mapping (DSM). Below, packages are grouped by key tasks:

📌 Note: This list is not exhaustive. It highlights packages that are commonly useful for DSM, spatial analysis, and soil modelling — but many others exist and continue to emerge.

  1. Soil science and pedometrics
  2. GIS integration and spatial processing
  3. Model calibration and prediction
  4. Mapping and visualisation

1. Soil Science and Pedometrics

  • ithir
    Core datasets and functions used throughout this course for soil data handling and modelling.

  • aqp
    Algorithms for Quantitative Pedology — soil profile analysis, aggregation, and visualisation.

  • soilDB
    Interfaces with USDA soil data sources (SSURGO, NASIS, SoilWeb) — useful for U.S.-based DSM work.

  • sharpshootR
    Companion to aqp — utilities for soil classification and interpretation, including taxonomy and land capability.

  • Additional packages authored or used by the developer are listed on the software page.

Back to top

2. GIS Tools and Spatial Processing

  • terra
    Modern raster handling and spatial analysis — fast, memory-efficient replacement for raster.

  • sf
    Simple Features for vector data — tidyverse-compatible spatial dataframes.

  • spdep
    Tools for analysing spatial autocorrelation, neighbourhoods, and dependencies — important for diagnostics and spatial structure.

  • whitebox
    Access to WhiteboxTools, a rich set of terrain and hydrological analysis functions.

  • RSAGA
    Interface to SAGA GIS — for DEM processing, hydrology, and terrain metrics.

  • qgisprocess
    Use QGIS processing algorithms directly from R.

  • geodata
    Download and prepare global raster datasets (climate, elevation, landcover) for spatial modelling.

Back to top

3. Model Calibration and Prediction

  • caret
    Unified interface for model training, tuning, and validation — supports dozens of algorithms.

  • mlr3
    Modern, modular machine learning framework with resampling, benchmarking, and workflow pipelines.

  • Cubist
    Rule-based regression models with instance-based corrections.

  • C50
    Classification trees and rule-based learners (C5.0) for pattern recognition.

  • gam
    Generalised Additive Models — good for flexible trend modelling.

  • nnet
    Single-layer neural networks and multinomial regression.

  • gstat
    Geostatistics — kriging, variogram modelling, and conditional simulation.

  • automap
    Automatic variogram fitting and kriging using gstat.

  • randomForest
    Ensemble learning via Breiman’s random forest algorithm.

  • xgboost
    Fast and scalable gradient boosting — popular for predictive soil modelling.

Back to top

4. Mapping and Visualisation

  • ggplot2
    Flexible, layered graphics system — ideal for publication-ready visualisation.

  • viridis
    Colourblind-friendly and perceptually uniform palettes for mapping.

  • tmap
    Thematic map creation using a grammar of graphics — supports both static and interactive maps.

  • leaflet
    Create interactive web maps directly from R.

  • mapview
    Quick, exploratory spatial visualisation — supports raster, sf, and SpatRaster.

  • ggspatial
    Adds scale bars, north arrows, and coordinate reference system annotations to ggplot2 maps.

Back to top