Accuracy of root modelling and its impact on simulated wheat yield and carbon cycling in soil

Abstract Accurate modelling of root biomass and root distribution of crop plants has become increasingly important in order to address issues related to carbon sequestration in soil and resource use efficiency of crops under different environmental and management conditions. However, the performance of crop models for simulating crop root system has been rarely tested in many environments due to lack of detailed data on roots. In this paper, we present detailed measurement data on root biomass and distribution in 0–200 cm soil profile at key developmental stages of wheat crop at Wuqiao in the North China Plain, and compare them with the root dynamics simulated by the agricultural systems model APSIM. The objectives are to test the model performance for simulating root biomass and distribution, and to investigate the potential impact of errors in root modelling on simulated wheat yield and soil carbon change. Compared to the measurements at the study site, APSIM version 7.5 underestimated the rooting front advance and final rooting depth of winter wheat, but overestimated the root biomass and root–shoot ratio at maturity by 100–200%. Modifications to root growth parameters in the model led to improved simulations of root depth and biomass dynamics. The modifications had little impact on simulated shoot biomass and grain yield under conditions of sufficient nitrogen supply, but led to higher simulated grain yield when nitrogen was deficient. They also resulted in reduction in simulated soil organic carbon in the top (0–20 cm) soil layer by 0.02%. A meta-analysis combining wheat root and shoot data from literature indicated that the APSIM model simulated a much tighter root–shoot relationship than that observed in the data, and that both versions of the model simulated the root–shoot relationships within the measured range. The relationship simulated from the original model better matched the data points from dry climate and dryland conditions. While the modified model based on Wuqiao data simulated a relationship better matching the data points from wet climate and irrigated conditions.