Predictive geochemical mapping using machine learning in western Kenya - Article abstract - Random Forest spatial prediction

Aug 5, 2026

Digital soil mapping techniques enable cost-effective estimation of the spatial distribution of chemical elements in soils by learning quantitative links between georeferenced observations and environmental covariates. A random forest framework combines inverse distance weighted predictors with ancillary environmental data to predict total concentrations and quantify uncertainty for 56 elements, soil pH, and organic matter using 466 samples across western Kenya. Results focus on iodine, selenium, zinc, and pH, using partial dependence and out-of-bag validation.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages13

Reading Time33 min

AccessFree to download

Introduction

Digital soil mapping (DSM) framework

Relevance of geochemical maps to nutrient dynamics and disease

Geoderma Regional 35 (2023) e00731 Contents lists available at ScienceDirect Geoderma Regional journal [homepage:](homepage: www.elsevier.com/locate/geodrs)[ www.elsevier.com/locate/geodrs](homepage: www.elsevier.com/locate/geodrs) | Predictive geochemical mapping using machine learning in western Kenya
Olivier S. Humphrey a, *, Mark Cave a, Elliott M. Hamilton a, Odipo Osanob, Diana Menyac, Michael J. Watts a
a Inorganic Geochemistry, Centre for Environmental Geochemistry, British Geological Survey, Nottingham, UK b School of Environmental Sciences, University of Eldoret, Eldoret, Kenya
c School of Public Health, Moi University, Eldoret, Kenya | | | | --- | --- | --- | | A R T I C L E I N F O | | A B S T R A C T | | Keywords: Random Forest Machine learning Soil Geochemistry Uncertainty Kenya | | Digital soil mapping techniques represent a cost-effective method for obtaining detailed information regarding the spatial distribution of chemical elements in soils. Machine learning (ML) algorithms using random forest (RF) models have been developed for classification, pattern recognition and regression tasks, they are capable of modelling non-linear relationships using a range of datasets, identifying hierarchical relationships, and determining the importance of predictor variables. In this study, we describe a framework for spatial prediction based on RF modelling where inverse distance weighted (IDW) predictors are used in conjunction with ancillary environmental covariates. The model was applied to predict the total concentration (mg kg − 1) and assess the prediction uncertainty of 56 elements, soil pH and organic matter content using 466 soil samples in western Kenya; the results of iodine (I), selenium (Se), zinc (Zn) and soil pH are highlighted in this work. These elements were selected due to contrasting biogeochemical cycles and widespread dietary deficiencies in sub-Saharan Africa, whilst soil pH is an important parameter controlling soil chemical reactions. Algorithm performance was evaluated determining the relative importance of each predictor variable and the model's response using partial dependence profiles. The accuracy and precision of each RF model were assessed by evaluating out-of-bag predicted values. The models R2 values range from 0.31 to 0.64 whilst CCC values range from 0.51 to 0.77. The IDW predictor variables had the greatest impact on assessing the distribution of soil properties in the study area, however, the inclusion of ancillary environmental data improved model performance for all soil properties. The results presented in this paper highlight the benefits of ML algorithms which can incorporate multiple layers of data for spatial prediction, uncertainty assessment and attributing variable importance. Additional research is now required to ensure health practitioners and the agri-community utilise the geochemical maps presented here for assessing the relationship between environmental geochemistry, endemic diseases and preventable micronutrient deficiency. | 1. Introduction Digital soil mapping (DSM) employs a generic framework to predict a target variable or class at an unobserved location based on the quantitative relationship between georeferenced observations and one or more environmental covariate which is likely to impact the variable or class of interest within a defined area (Asgari et al., 2020; Lagacherie et al., 2006; McBratney et al., 2003; Sylvain et al., 2021; Wadoux et al., 2019; Zeraatpisheh et al., 2019). The spatial distribution of chemical elements in soils, originating from geogenic and anthropogenic sources, can provide critical information for assessing mineral exploration, environmental monitoring, and assessing nutrient dynamics (Hengl et al., 2015; Johnson and Ander, 2008; Sylvain et al., 2021; Wadoux et al., 2019). The need for comprehensive, accurate and up-to-date soil information maps is an essential component for the formulation of agricultural

FAQs

What is the core goal of the study in western Kenya?

To build a machine-learning framework that predicts the spatial distribution of soil chemical elements, soil pH, and organic matter while also evaluating prediction uncertainty.

Which modeling approach is used and how are covariates incorporated?

Random forest models are developed using inverse distance weighted predictors together with ancillary environmental covariates to improve spatial prediction.

How is prediction accuracy and uncertainty assessed?

Model performance is evaluated using relative predictor importance, partial dependence profiles, and out-of-bag predicted values for accuracy metrics such as R2 and CCC.

Aug 5, 2026

Upload by Lucas Martin