Code and aggregate results for topsoil nutrient prediction and soil information comparisons in West Africa

Analysis code, configuration files, tests, and aggregate results supporting the prediction of topsoil (0–20 cm) total nitrogen, total phosphorus, and exchangeable potassium from legacy soil profiles of the Africa Soil Profiles database (AfSP v1.2) and the WoSIS snapshot of December 2023 in Burkina Faso, Côte d'Ivoire, Ghana, Mali, Nigeria, and Senegal. The central analysis compares four quantile regression forest models for total phosphorus (environmental covariates only; measured clay, sand, organic carbon, and cation exchange capacity; their combination; and organic carbon and cation exchange capacity without texture) on 973 complete-case profiles under six validation designs (random, spatially blocked, kNNDM, rainfall-zone-out, country-out, and survey-out). Additional analyses test the prediction of these soil properties from environmental covariates, a two-stage model, residual associations with soil and terrain properties, and the sensitivity of the results to the exclusion of the AfSP survey report SZ. The archive also includes the original environmental modeling analyses for all three nutrients, the article figures, Supplementary Tables S1–S67, the outer-fold memberships of the complete-case comparison (without measurements, coordinates, or predictions), and 18 national maps from the original environmental models. These maps describe the models and are not validated nutrient products. Original soil observation tables, observation-level model outputs, and model checkpoints are not redistributed. The workflow scripts download the source releases from their providers and verify them against recorded SHA-256 checksums; access and reuse follow the conditions of each source. Licensing: Code is distributed under the MIT License. Original author-generated outputs are distributed under the Creative Commons Attribution-NonCommercial 4.0 International License (CC BY-NC 4.0), subject to applicable third-party rights and attribution requirements (LICENSE-OUTPUTS.md). Integrity verification: Run python3 tools/verify_archive.py to verify archive integrity without the original soil observations. MANIFEST.csv lists the size and SHA-256 checksum of every distributed file.