Predicting available phosphate ions from physical-chemical soil properties in acidic sandy soils under pine forests

For economic and environmental reasons, and for biomass production, appropriate concepts and diagnostic systems based on relevant processes are required to assess the phosphorus (P) supply capacity of the soils in the long term and to adapt P fertilization accordingly in forests. The amount of available phosphate ions (iP) can be quantified using an isotopic dilution procedure. However, this method is difficult to apply since it requires the use of radioactivity (P-32 or P-33). Our objective was thus to build pedotransfer functions for the prediction of available iP from physical-chemical soil properties. Using an isotopic dilution procedure, we assessed isotopic dilution parameters and calculated isotopically exchangeable iP (E) and diffusive iP at the solid-to-solution interface (Pr). Following previous studies, E and Pr were considered as available iP. Pedotransfer functions were developed for forest podzols and arenosols (litter + 0-120 cm) with wide ranges of soil organic matter and poorly crystalline aluminum (Al) and iron (Fe) oxides. Using Pearson's correlation coefficients, we selected the physical-chemical properties that were the most closely correlated with the isotopic dilution parameters. Then, we used linear and nonlinear regressions between these soil properties and the isotopic dilution parameters to build pedotransfer functions. Pedotransfer functions were built using a calibration dataset (144 soil samples) and validated using another data set (34 soil samples). For the sandy and acidic forest soils studied here, the soil reactivity inferred with the isotopic dilution parameters was mainly determined by poorly crystalline Al and Fe oxides and to a lesser extent by organic matter. Soil organic matter probably competed with iP for the reactive surfaces, i.e., the surfaces of Al and Fe oxides. The relationships between these soil properties and the isotopic dilution parameters led to pedotransfer functions enabling the correct prediction of soil reactivity and E as a function of time, independently to the soil depth and at the scale of a forest ecosystem. The adjusted pedotransfer functions could however not be used to precisely predict Pr. As an alternative, other pedotransfer functions were adjusted to more precisely predict an indicator of Pr (Pr in 1,000 min). Our results revealed the determinant role of poorly crystalline Al and Fe oxides in the dynamics of iP at the soil-to-solution interface and these results were related to other results on P biogeochemistry at the ecosystem scale. They also indicated that generalized pedotransfer functions-which already include the main explanatory variables-would be improved by integrating the effect of depth.