Determination of base saturation percentage in agricultural soils via portable X-ray fluorescence spectrometer

Soil base saturation percentage (BSP) plays an important role in the assessment of soil taxonomic classification and soil fertility. Conventionally, soil BSP measurement methods are fraught with many drawbacks such as being laborious, time consumptive, destructive to the samples, and can lead to the underestimation of true cation exchange capacity (CEC). Recently, proximal sensors such as portable X-ray fluorescence (PXRF) spectrometry have proven to be effective for rapid physicochemical analysis of soils. In this study, we proposed and examined a PXRF-based method to predict BSP using 300 soil samples from the active agricultural lands in six states across the USA; Colorado, California, Minnesota, Nebraska, Oklahoma, and Texas. An Olympus Vanta series PXRF analyzer was employed to measure Mg, Ca, and K for BSP prediction. Results were validated using four different multivariate models [generalized additive model (GAM), multiple linear regression (MLR), random forest (RF), regression tree (RT)] via R 3.5.1. Predictive model performance was assessed via root mean squared error (RMSE), coefficient of determination (R-2), residual prediction deviation (RPD), the ratio of performance to interquartile (RPIQ) range, and bias. While predicting BSP from PXRF quantified elements, models exhibited R-2, RMSE (%), and RPDs as follows: GAM = 0.58, 9.0, 1.6; MLR = 0.45, 10.4, 1.4; RF = 0.62, 8.7, 1.6; RT = 0.68, 7.9, 1.8, respectively. Soil cation exchange capacity was also predicted using a similar approach, with similar and moderate predictive performance; GAM produced R-2, RMSE (cmol(c) kg(-1)), and RPD of 0.69, 5.6, 1.8, respectively, relative to laboratory data. This study showed that the PXRF elements can be used to predict BSP with fair accuracy for the range of agricultural soils examined. As such, further study and enhancement of the approach outlined herein on a wider array of soils is warranted.