Prediction of total silicon concentrations in French soils using pedotransfer functions from mid-infrared spectrum and pedological attributes

Silicon (Si) is the second most abundant element of the Earth's crust, and its terrestrial cycle depends on soil, vegetation, and human activities. The spatial extent of terrestrial Si perturbation is poorly documented since maps of Si concentration in soils are rare. In addition, Si content is rarely measured in non-paddy soil databases. Here we demonstrate that pedotransfer functions based on either pedological attributes (particle size fraction, pH, organic carbon, cation exchange capacity, calcium carbonate and parent material) or mid infrared spectra (MIRS) can be used to accurately predict total Si concentration. In this research, we utilised a unique dataset from the French monitoring network of soil quality (RMQS - Reseau de Mesures de la Qualite des Sols) database. Pedotransfer functions were built using a regression tree model on a subset of the data for which total Si concentration was measured. To compare the relative performance of the models obtained for the two different sources of data, a suite of performance indicators were calculated. Our results showed that PTF based on MIR spectra produces highly accurate and precise estimates of the total Si concentration for French soils. The pedological PTF is less accurate, but still provides a good estimation of the Si concentration. The pedological PTF provides an alternative method when only basic soil data are available, and an approximate estimation of Si concentrations is sufficient. These PTFs can be readily applied at the European scale except on a few soil groups not represented in France.