Soil parent material prediction for Brazil via proximal soil sensing

摘要: Abstract Parent material (PM) is key in the thorough understanding of soils. However, complexity PM distributions and difficulty reaching deep soils prevent its detailed assessment. Proximal sensors, such as portable X-ray fluorescence spectrometer (pXRF), might ease this process. This work attempts to prove potential pXRF predict different PMs from analyses soil samples. The study encompassed five Brazilian states representing 1,541,309.409 km2, where 310 samples various classes derived 12 were collected analyzed by PXRF. Support Vector Machine (SVM) Random Forest (RF) algorithms used for modeling. Modeling comprised three datasets: one containing all data (310 samples), a dataset with younger (151 samples) older soils, conceptually less influenced their (159 understand how soil-PM chemical proximity affects prediction performance, assessed via overall accuracy Kappa coefficient. Data distribution showed can discriminate types resulting regardless degree weathering. Prediction results prominent: RF SVM achieved roughly 0.9 predicting data. For remaining datasets, 0.96 nearly 0.92 0.87 0.9, respectively, confirming that are slightly easier predict, but even heavily altered pedogenetic processes be accurately predicted. Results confirm data, which help mapping consequent activities tropical conditions.