Modeling root biomass of maize in Northeast China

It is of great significance to explore root biomass and establish a universal root biomass model for the monitoring and evaluation of the ecosystem biomass. In order to get the real-time information of a root system, the biomass and ecological indexes of maize root system were collected using soil sampling method and root scanning systems at the agricultural meteorological experiment station in Jinzhou City, Liaoning Province in September 2016. The vertical distribution characteristics of root biomass of maize were analyzed and simulation equations were established based on the relationships between root biomass and root ecological indexes. The results showed that the maize root biomass decreased with increase of soil depth. The root biomass of maize mainly concentrated at the soil depth from 0 cm to 30 cm, which accounted for 94.44% of the total root biomass. The simulation accuracy of exponential and power functions of root biomass with root diameter as independent variable established by ordinary least square method was low, and R² was 0.10 and 0.12, respectively. The biomass model constructed by adding root length as an independent variable significantly improved the simulation accuracy, with R² reaching above 0.91. There was heteroscedasticity issue in the models of root biomass established by the ordinary least square method inducing less stable and inaccurate prediction results. This issue could be eliminated by using logarithmic transformation. The biomass model of maize root system with root length and root dimeter together as variables (D²H) had a better simulation effect and a good prediction accuracy, with the coefficient of determination (R²) as 0.90 and the evaluation indexes MAE, SEE and MPE as 4.38 g, 18.68 g and 16.09%, respectively. The correlation coefficient between simulated and measured values was 0.92 (P < 0.01), indicating that this model could be used to simulate the biomass of maize root system in Northeast China. The study results indicated that the difficulty in observing root biomass in real time could be resolved by using root biomass model combined with the minirhizotron method.