Modelling Potassium Properties in East African Soils Using Infrared Spectroscopy

Potassium (K) is an essential plant nutrient and is often in short supply in the highly weathered soils of East Africa. The development of effective and sustainable fertilisation strategies could be helped by developing rapid and relatively cheap methods that provide initial assessments of inherent nutrient stocks and soil K behaviour. Mid‐infrared (MIR) and near‐infrared (NIR) models are presented covering a range of soil K properties and K‐bearing mineralogy. Initially, 90 soils from Kenya, Tanzania and Uganda were analysed for total K, SrCl 2 exchangeable K, K‐release rates ( β ) and quantitative X‐ray diffraction (XRD). Using MIR and NIR scans for each sample, as well as an additional 446 soil samples from the Winam Gulf (WG) in Kenya, key spectral components were identified using a ‘Single Mixture Modelling Resolution’ analysis, which were used to parameterise models using either stepwise or censored regression techniques for the measured soil K properties and K‐bearing minerals in the 90 samples analysed. Stepwise regression models were fitted for K Total ( R 2 = 0.90), SrCl 2 extractable K ( R 2 = 0.68), the release constant β ( R 2 = 0.53), Kaolinite ( R 2 = 0.79) and the Chemical Index of Alteration ( R 2 = 0.93), with all models being highly significant ( p < 0.001). Censored regression models were fitted for the K‐containing minerals K‐feldspar, mica and illite, along with quartz. The logSigma parameter was highly significant for each ( p < 0.001). Models were applied to the WG soil samples, and estimates were plotted using different lithologies as variables. Potential exists for these models to act as a toolbox to assess inherent K stocks and likely K dynamics as an initial assessment in developing strategies to improve K sustainability and fertilisation practices. Models could be rerun to include newly generated MIR and NIR scans from soil surveys.