Rice Yield Estimation Based On The Simulation Of Yield Components

Accurate rice yield estimation is of great significance to ensure food security in China. The rice yield is composed of four yield components, i.e., effective panicle,1000-grain weight, seed setting rate and grain number per spike. Analyzing the response mechanism of four yield components to light and temperature, and the contribution of four yield components to the yields is important for the development of high-yield cultivating measures. Therefore, two-year field-seeding experiments (2012-2013) using two indica two-line hybrid rice cultivars, Lingliangyou268 and Liangyoupeijiu, were conducted in Nanjing to investigate the light and temperatures effects on four yield components as well as yields, and try to establish the light and temperature effect functions to simulate the changes of rice yields and yield components with environmental elements.To achieve this, tillering data and meteorological data were used to analyze the relationship between tiller dynamic and average light and temperature conditons during the growing periods for two rice cultivars. And, co-effect equations of light and temperature were established to substitute the corresponding equations in a widely used tillering model to simulate the dynamic tillering in tillering growing period. To simulate the dynamic tillering in tillering extinction stage, the current model was coupled with ORYZA2000 using a one-way coupling method. With this method, the current model accepted outputs from ORYZA2000. Based on the improved tillering model, the dynamic tillering was simulated and the number of effective panicle was obtained.Then based on observed rice grain filling data and meteorological data, Richards equations was extended by a light and temperature correction function, which consists a set of light and temperature characteristic parameters. SCE-UA was applied to estimate all the parameter values in the response equations, and the rice grain filling model was established to simulate 1000-grain weight.And then, the relationships between sterile grain rate and grain number per spike and main climatic variables from panicle initiation to grain filling were analyzed to extraction main affecting factors. Principal Component Regression Analysis was used to establish regression equation, and based on this simulation models were established for grain number per spike and sterile grain rate respectively, and grain number per spike and seed setting rate were obtained. Finally, rice yield were estimated by multiplying the values of effective panicles,1000-grain weight, seed setting rate and grain number per spike. In addition, the relationships between rice yield and the four yield components were investigated. The summarized results and conclusions are listed as follows:1. Light and temperature affect crop growth in tillering period. In tiller growing stage, the growth rate of tillers and the actual maximum tiller density have significant positive correlation with the two meteorological factors, and it was well reflected in the light and temperature response curves obtained based on the co-effect equations of light and temperature. In tiller extinction stage, the coupled model shows good performance in simulating daily tiller density and linking daily tiller density with development rate and biomass. The results showed that light and temperature could affect the tillering extinction directly as well as affect it by affecting the biomass. In conclusion, the improved model effectively improved the simulation accuracy of tiller dynamic.2. The light threshold R0, maximum temperature Tmax, minimum temperature Tmin, optimum temperature Top for Lingliangyou268 in grain filling period were 18.94 MJ·m-2·d-1,33.29℃, 6.81℃,30.28℃, respectively, and 21.71 MJ·m-2·d-1,33.74℃,6.10℃,24.16℃ for Liangyoupeijiu. The light and temperature response curves showed an evident difference between the two rice cultivars. The rice filling model established on light and temperature correction function has performance in simulating the grain filling process of the two rice varieties and improved the simulation accuracy of 1000-grain weight.3. The statistical model showed a good performance in simulating grain number per spike and sterile grain rate. The correlation coefficients for simulated and observed rain number per spike and sterile grain rate for Lingliangyou268 were 0.83 and 0.96, respectively,0.85 and 0.77 for Liangyoupeijiu, respectively. All the correlation coefficients were significant at the 0.05 probability level.4. The simulated and observed rice yield were closed to each other. The correlation coefficients for simulated and observed rice yield for Lingliangyou268 and Liangyoupeijiu were 0.83 and 0.96, respectively, and significant at the 0.01 probability level for both two rice cultivars. The RMES of simulated and observed rice yield for Lingliangyou268 and Liangyoupeijiu were 106.2 g·m-2 and 40.3 g·m-2, respectively.In conclusion, the yield estimation method proposed in this paper shows higher precision in rice yield simulation than traditional method adopted in ORYZA2000 model. This could provide a scientific reference for the study of yield estimation and a practical reference for improving rice production.