Soil phosphorus fractions across contrasting cropping systems and sampling stages: Chemometric screening with excitation‐emission matrix fluorescence

Abstract

Seasonal redistribution of soil phosphorus (P) among labile, moderately retained, and residual pools is difficult to monitor at the scales needed to evaluate fertilizer use efficiency, soil fertility, and P‐loss risks associated with long‐term soil P accumulation. This study evaluated excitation‐emission matrix (EEM) fluorescence spectroscopy combined with chemometrics for screening preplant‐to‐postharvest variation in Hedley inorganic P (Pi) fractions under contrasting cropping systems: annual corn (prevailing practice), corn with cover crops (alternative practice 1), and a perennial mixture (alternative practice 2 [AP2]). Sequential fractionation showed declines in Resin‐Pi and NaHCO ₃ ‐Pi across systems, whereas residual P increased most in AP2. Parallel factor analysis of EEMs resolved protein‐like and humic‐like fluorescence components and indicated a postharvest shift toward more humic‐like fluorescence associated with Ca, Mn, and greater representation of NaOH‐Pi, HCl‐extractable Pi, and residual P pools. Prediction was evaluated using partial least‐squares regression (PLSR), N‐way PLSR (NPLS), and PLSR using PARAFAC scores plus Rayleigh‐region descriptors (PLS‐PARAFAC‐R). Fraction specific performance indicates that each P fraction requires separate calibration rather than a universal model. PLS‐PARAFAC‐R performed best for Resin‐Pi ( R ² Pred = 0.775; ratio of performance to deviation [RPD] = 2.169), Conc‐HCl‐Pi ( R ² Pred = 0.755; RPD = 2.078), and residual P ( R ² Pred = 0.829; RPD = 2.489). PLSR of unfolded EEMs performed best for NaOH‐Pi ( R ² Pred = 0.862; RPD = 2.766) and Dil‐HCl‐Pi ( R ² Pred = 0.662; RPD = 1.769), whereas NPLS was marginally best for NaHCO ₃ ‐Pi ( R ² Pred = 0.781; RPD = 2.199). Overall, EEM fluorescence with chemometric modeling may complement wet‐chemical fractionation by supporting screening‐level assessment of seasonal and management‐related variation in soil P partitioning.