Soil Properties and Phosphorus Isotopic Exchangeability in Cropped Temperate Soils

Isotopically exchangeable phosphate (P) is the main source of P for most crops. The amount of P that is located on the solid phase of a soil and that can exchange with P in the soil solution can be calculated knowing the concentration of water-extractable P (C-P ) and two kinetic parameters: (i) the ratio between the total introduced radioactivity added as P-33 in a soil/water system and the radioactivity remaining after 1 min in this solution (R/r(1minute) ) and (ii) the rate of disappearance of P-33 from the solution with time (n). The aim of this article was to propose statistical models describing R/r(1minute) and n as a function of C-P and other soil properties. By doing so we made the hypothesis that it is possible, knowing C-P and selected soil properties, to calculate the amount of P located on the solid phase of the soil that remains isotopically exchangeable after 1 min and 60 min (P-r1minute , P-r60minutes ). This work was done in two steps. First, isotopic exchange kinetic experiments were carried on nine soil samples (set A) sampled in 1998 that had been incubated in the presence of increasing concentrations of water-soluble P. Results from these experiments allowed development of models to calculate R/r(1minute) as a function of C-P , soil pH, and dithionite-extractable Fe (Fed) and n as a function of C-P . In the second step, we compared R/r(1minute) , n, P-r1minute , and P-r60minutes values obtained from the isotopic exchange kinetic experiments conducted on an independent set of 27 soil samples (set B) to the values computed from our statistical models. Modeled and experimental values were highly significantly correlated to each other, showing that the three parameters C-P , soil pH, and Fed explained more than 75% of the variability of the experimental results. Experimental and modelled values of R/r(1minute) and P-r1minute were not different from each other, whereas experimental values of n and P-r60minutes were statistically significantly different from the modeled values. Finally, the limits and applicability of our approach for the development of a pedotransfer function to predict P-r are discussed.