Predictive geochemical mapping using machine learning in western Kenya

Digital soil mapping is a cost-effective method for obtaining detailed information regarding the spatial distribution of chemical elements in soils. Machine learning (ML) algorithms such as random forest (RF) models have been developed for such tasks as they are capable of modelling non-linear relationships using a range of datasets and determining the importance of predictor variables, offering multiple benefits to traditional techniques such as kriging. 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 auxiliary environmental covariates. The model was applied to predict the total concentration (mg kg-1 ) of 56 elements, soil pH and organic matter content, as well as to assess prediction uncertainty using 466 soil samples in western Kenya (Watts et al 2021). The results of iodine (I), selenium (Se), zinc (Zn) and soil pH are highlighted in this work due to their contrasting biogeochemical cycles and widespread dietary deficiencies in sub-Saharan Africa, whilst soil pH was assessed as an important parameter to define soil chemical reactions. Algorithm performance was evaluated to determine the 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 the out-of-bag predicted values. The IDW predictor variables had the greatest impact on assessing the distribution of soil properties in the study area, however, the inclusion of auxiliary values did improve 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 agricommunity utilise the geochemical maps presented here, and the webtool, for assessing the relationship between environmental geochemistry and endemic diseases and preventable micronutrient deficiency.