2 Objectives
The Soils4Africa Digital Soil Mapping Manual has the following objectives:
- Standardize Data Preparation
- Demonstrate loading, cleaning, and processing of soil point data (Section 3.2).
- Illustrate assembly and masking of raster covariates along with automated covariate selection (Section 3.3).
- Show how to build a regression matrix linking soil observations to covariate values (Section 3.4).
- Demonstrate loading, cleaning, and processing of soil point data (Section 3.2).
- Guide Model Fitting and Selection
- Provide code examples for fitting a Random Forest model (Section 4.2).
- Explain methods for feature (covariate) importance analysis and recursive feature elimination (Section 4.3).
- Provide code examples for fitting a Random Forest model (Section 4.2).
- Assess Model Accuracy
- Describe techniques for extracting accuracy statistics and visualizing model performance (Section 4.4).
- Enable Spatial Prediction
- Demonstrate how to apply the fitted model to predict soil properties across a study area and export the resulting maps (Section 4.5).
- Quantify Prediction Uncertainty
- Show how to fit a Quantile Random Forest model to derive 90 % prediction intervals and create uncertainty layers (Section 4.6).
- Promote Reproducibility and Adaptability
- Use parameterized scripts and well-commented R code, allowing users to adapt the workflow for different soil properties, regions, or modelling approaches.
By achieving these objectives, this manual ensures that users—regardless of prior DSM experience—can produce consistent, transparent, and reproducible soil property maps to support sustainable land management and decision-making across Africa.