基于生态效应的水稻籽粒蛋白质含量预测模型研究

在中国、日本、泰国不同生态环境下进行多品种籼型和粳型水稻(Oryza sativa)的区域种植试验,通过分析水稻籽粒蛋白质含量与纬度、海拔、抽穗后温度和太阳辐射等气候生态因子的相互关系,确立了影响水稻籽粒蛋白质积累的主要气候生态因子函数,并使用权重系数来进一步修订各气候生态因子对水稻籽粒蛋白质的作用,构建出基于生态效应(主要气候生态因子函数)的水稻籽粒蛋白质含量预测模型。利用不同年份、不同生态点、不同品种类型的试验资料对所建模型进行了检验,籼稻和粳稻籽粒蛋白质含量的预测误差RMSE平均分别为0.27%和0.24%;籼稻试验点和粳稻试验点的预测误差平均为0.25%和0.22%,表明模型总体上具有较好的预测性和实用性。

AN ECOLOGICAL MODEL FOR PREDICTING PROTEIN CONTENT IN RICE GRAINS

Grain protein content in rice (Oryza sativa) is an important indication of rice nutritional quality and is determined by both genotype and environmental factors. Much attention has been paid to the relationships between protein accumulation in rice grains and environmental factors. Modeling changes of protein content in rice under different environmental conditions is of great help for predicting rice quality and ecological zoning. In this study, Indica and Japonica rice varieties grown under different ecological environments in China (Jiangsu), Japan (Iwate, Shimane, Kyoto) and Thailand (Changmai) were grown for two years, and the relationships between grain protein content and environmental factors, including latitude (x1), altitude (x2), ecological height, latitude, altitude, (x3), and average temperature (x4), lowest temperature (x5), highest temperature (x6) and solar radiation during the grain filling period were analyzed. Environmental factor-driven equations were established for Indica rice: f (x1) =8.4E-04x12 +0.049 05x1+0.282 1, f(x2) =2.82E-07x22-3.8E-04x2+1.128 6, f (x3) = 9.2E-10x32-3.1E-05x3+1.255　6, f (x4) =-0.007　02x42+0.303　7x4-2.285, f (x5)=-0.009　18x52+0.313　9x5-1.684　6, f (x6)= -0.006　99x62+0.372　8x6-3.971　7; and for Japonica rice: f (x1)=-0.000 695x12+0.038 51x1+0.466 1, f (x3)= 4.02E-10x32-1.2E-05x3+1.090　4, f (x4)= -0.004　16x42++0.185　9x4-1.077　4, f　(x5)=-0.006　54x52+0.228　6x5-0.996　8, f (x6)=-0.004　12x62+0.226x6-2.099　1. By using a weighted method to modify the impacts of environmental factors on grain protein accumulation, an ecological model was developed for predicting grain protein content in Indica rice: PC=PC0×［0.138　9×f(x1)+0.152　6×f(x2)+0.187　4×f(x3)+0.180　2×f(x4)+0.168　5×f(x5)+0.173　4×f(x6)］; and PC=PC0×［0.262　3×f(x1)+0.09×f(x3)+0.202　3×f(x4) +0.246　1×f(x5)+0.198　4×f(x6)］ for Japonica rice, where PC0 is the specific protein content of cultivar. The model was validated using the data sets for different years, eco-sites and varieties. The root mean square error (RMSE) was 0.27% and 0.24% for Indica rice and Japonica rice, respectively, and the RMSE was 0.25% and 0.22% for the growing areas of Indica and Japonica rice types, respectively. The results indicated that the model was accurate and applicable for predicting protein content in rice under different conditions. Yet, more experimental data in different eco-environments are required for wider testing of the present model.