Regional Prediction of Soil Nutrient Availability and Management Using Portable X‐Ray Fluorescence Spectrometry

ABSTRACT Portable x‐ray fluorescence (pXRF) spectroscopy has emerged as a rapid and cost‐effective technique for soil elemental analysis; however, its application for predicting available nutrients and supporting agronomic decision making remains insufficiently evaluated under field conditions. This study assessed the potential of pXRF to predict soil nutrient availability and inform fertilizer recommendations in dryland cropping systems of the eastern Pacific Northwest, USA. A total of 113 soil samples collected from 31 field locations and four depth intervals (0–120 cm) were analyzed using conventional laboratory methods and pXRF under multiple moisture conditions (oven‐dry, field‐moist, saturated paste, and after removal of excess water). Predictive models were developed using multiple linear regression (MLR), partial least squares (PLS), and random forest (RF), and evaluated using cross‐validation metrics and agronomic classification agreement. Results demonstrated strong predictive performance for exchangeable base cations, particularly calcium (Ca) and magnesium (Mg), with low classification mismatch (< 3%), indicating high reliability for fertility mapping. Zinc (Zn) showed moderate predictive capability, while potassium (K) and copper (Cu) exhibited intermediate performance, suggesting their suitability for preliminary screening. In contrast, soil pH and Olsen phosphorus (P) showed limited predictive accuracy and high misclassification rates (up to 59% and 50%, respectively), particularly near agronomic thresholds critical for lime and fertilizer decisions. Among moisture conditions, saturated paste provided the most consistent predictive performance across variables. These findings highlight that pXRF is suited as a complementary tool for rapid spatial assessment, rather than a replacement for conventional laboratory analyses.