Does the APSIM model capture soil phosphorus dynamics? A case study with Vertisols

Abstract

Crop production in the tropics and sub-tropics increasingly relies on phosphorus (P) fertilisers to maintain crop yields. As a result, there is a need to better understand the effectiveness of P management strategies, especially in Vertisols that represent a large proportion of the cropped area. The complexity of soil P dynamics and the interaction with environmental factors and farming practices make this challenging but the use of agricultural systems models such as the Agricultural Production Systems sIMulator (APSIM) can potentially improve our understanding of these interactions. This study (1) gives the state of knowledge of P behaviour in Vertisols and identifies which soil P processes are simulated in the APSIM Soil-P module, (2) tests the ability of APSIM as a tool to simulate soil P dynamics and crop responses in two long-term P studies in Vertisols, and (3) discusses future work needed to improve modelling of soil P dynamics. Soil P dynamics in Vertisols are not well understood, but evidence suggests that precipitation and dissolution of calcium-P minerals may play a significant role in plant P availability. These characteristics are substantially different to the strongly P-sorbing soils on which the APSIM model has previously been tested, so the ability of the model to simulate soil P dynamics and crop responses on these soils is untested. The APSIM model was able to predict the variation in crop yields due to seasonal variation of moisture availability within an acceptable modelling performance but was less able to predict P responses within seasons. This weakness was consistent with the lack of linear correlation between the plant-available P pools predicted by the P module and the measured Colwell-P that was correlated with crop responses. The non-linear relationships also differed between the two Vertisols. The poor understanding of P dynamics in Vertisols and the inability to partition soil P into measurable P pools is hindering the development of a mechanistic P module. The resulting lack of an appropriate mechanistic approach to soil P dynamics questions the predictive ability of the model. Further fundamental work is needed to predict the fate of applied P fertilisers in soils with contrasting physical and chemical characteristics, with this information integrated into an improved P modelling framework.