Combine observational data and modelling to quantify cultivar differences of soybean.

• Observed phenology, LAI, biomass & yield of 72 soybean cultivars are presented. • APSIM captured LAI & biomass dynamics of multiple cultivars across different years. • APSIM improvement need to include simulation of grain setting & its impact on yield. • We provided the first data set in APSIM for modelling soybean genotypes in China. This paper combines the analysis of soybean growth data with modelling to quantify cultivar differences and to evaluate the ability of the APSIM-Soybean model for simulation of soybean biomass and grain yield. Crop data include main developmental stages, leaf area index (LAI), above ground biomass, grain yield and grain number of 72 soybean cultivars in China. We derived the parameters controlling soybean flowering and maturity time in the lack of data from contrasting environments, then tested APSIM to simulate LAI, biomass and grain yield. Our results show that once phenology is correctly simulated, APSIM was able to capture the LAI and biomass growth dynamics of multiple cultivars and across different years. Our data showed a strong correlation of grain yield with grain number, but not with grain size. Future model improvement could include the simulation of grain setting (grain number and grain filling rate) and their difference between cultivars so that yield difference can be simulated. Further scenario modelling demonstrated that APSIM also captured the genotype by environment interactions as they impact on yield. The derived parameters for 72 soybean cultivars provide the first data set in APSIM for improved modelling of soybean genotypes in China and beyond, adding to the current APSIM validation database of soybean. This also formed the basis for future modelling of intercropping systems with soybean, e.g. maize/soybean intercropping. [ABSTRACT FROM AUTHOR]