Study On Model Improving Of Grain Protein In Rice Based On Plant Nitrogen Flow

It is of great significance to predict the total content and dynamic process of grain protein accumulation to ensure accurate simulation of quality model in RiceGrow and promote sustainable development of rice quality in the future.This study aimed to the improvement of grain protein accumulation model in RiceGrow,quantitative methods and mathematical algorithm were improved according to the results of model calibration and validation based on nitrogen flow in rice organs.Experimental data including different varieties,years and treatments were used by model testing on ORYZA2000 and improved RiceGrow.A comparative analysis was conducted with the test results.Consequences will be useful to further verification and evaluation.RiceGrow-Grain Protein model was divided to three components as N uptake before anthesis,N absorption after anthesis and N transportition of vegetative organs according to plant physiological activities;every component was tested respectively.In model calibtation,parameters were determined and results showed a high conformity between simulated values and observed values.The NRMSE(Normalized Root Mean Square Error)of grain protein total content and protein accumulation dynamic of two varieties in each test under differents treatments of were less than 20%.Under the same setting of parameters for model validation,results showed NRMSE on grain protein content and grain nitrogen accumulation dynamic were 23.6%,34.8%respectively;NRMSE for N uptake before anthesis,N absorption after anthesis and N transportition were 60.3%,69.6%and 60.3%respectively;RMSE of model simulation displayed a significant difference between differernt varieties;at the same time,the simulation showed no difference between nitrogen levels on plant nitrogen absorption and transporation.Parameter settings,quantitative mode of nitrogen flow dynamic,N factors,and time scale of simulation were suggested to be further improved to enhance predictive and adaptive capacity of simulation model.Dynamic changes of nitrogen content of leaves,leaf sheaths along with the growth period were quantified to predict N accumulation in grains.Parameter and N factor setting,methods and quantitative mode,time scale for simulation were improved respectively.Ability of N uptake and N transporation were defined as variety parameters to measure the relative differences on N uptake and transporatin in physiological processes.The original linear response model was replaced by quadratic curve model to simulate nitrogen dynamic.N impact factors acted differently on different growth period and plant nitrogen physiological activity;plant N uptake was mainly affected by the growth of the vegetative organs before anthesis,nitrogen absorption and transfer process mainly depended on nitrogen accumulation of vegetative organs at flowering after anthesis.PDT(physiological development time)was used for simulated time scale.Different experimental data were used to test the model.The validation results showed that the model has a good performance and fitted well between predicted values and observed values.The improved model displayed a good predictability on grain nitrogen accumulation between different varieties.Using the experimental data of different N application levels on validation of ORYZA2000 and RiceGrow-grain protein accumulation model,comparative analysis showed the NRMSE of ORYZA2000 prediction was 35.2%on average;RMSE for nitrogen accumulation of leaf,stem sheath and panicle were 41.3%,28.3%,29.5%,respectively;however a good behavier on simulation of plant total nitrogen accumulation with RMSE to be 19.0%.RiceGrow-grain protein accumulation model showed a good conformity and consistency between the predicted and observed values,RMSE for nitrogen dynamic of individual organ and plant total accumulation were all below 20%.ORYZA2000 acted poorer in NO and N1 and N4 interchange conditions when simulating between different varieties and different organs;although RiceGrow-grain protein accumulation model owed a good predictivity overall,but also appeared similar situation.