Linking measurable and conceptual phosphorus pools (in APSIM) enables quantitative model initialisation

Phosphorus (P) is an essential plant macro-nutrient, yet it is deficient in 65 % of agricultural soils worldwide. Agricultural systems models enable the integration of plant-soil-climate-management interactions to investigate crop responses to P fertilisation and improve P use efficiency. However, current models cannot align their modellable P pools with values obtained from soil tests. This limits their applicability since soil testing is the most widely used tool to assess soil P status, which is then used to predict fertiliser P requirements based on assumed crop P demand for optimal growth in the field. Our study introduces a modelling framework akin to inversely modelling in the Agricultural Production Systems sIMulator (APSIM) to quantitatively derive the most likely P modelling parameters for different soils and empirically link them to common soil P test values. The methodology was first tested using data from an 8-year alfalfa (syn. lucerne) experiment (1997–2004) on two soil types in the mid-west of the United States to establish the adequacy of the P modelling framework in APSIM. We then extended this approach to eight Australian soil types using a simulation study based on known wheat yield response curves to soil P tests to derive empirical relationships between the labile P values in APSIM and common soil test P values (Bray-2 P and Colwell P) for the soils studied. Cross-validation yielded an average R 2 of 0.98 and an average Lin’s Concordance Correlation Coefficient (CCC) of 0.92. Our work thus enables the initialisation of the labile P pool in APSIM using Bray-2 P and Colwell P data, enhancing the usability and accuracy of agricultural systems models in predicting crop P requirements and optimising P fertiliser use across diverse soil types in different agro-climatic regions. • Process-based models can be useful for analysing crops’ response to P fertilisation. • Current models lack the ability to match soil P tests to conceptual soil labile P. • We used inverse modelling to quantitatively derive P parameters in APSIM. • Empirical relationships were derived between APSIM-labile P and measured soil P fractions. • Enables initialisation of labile P pool in APSIM using common soil P test values.