| The high and stable yield of rice are crucial to ensuring China’s food security.Therefore,high and stable yield is always the main goal for breeders.Since the first green revolution,dwarf breeding has significantly reduced plant height and increased the harvest index,which helps to increase grain production.Harvest index is a comprehensive index to measure and define the relationship between grain yield and aboveground biomass.However,the relationship between harvest index and related agronomic traits is unclear,and the determination of harvest index can only be carried out after harvested.In this study,60 varieties with different harvest index which were bred,collected or preserved by Rice Research Institute Guangdong Academy of Agricultural Sciences were used for the examination and correlation analysis of harvest index and related agronomic traits.Quantitatively obtain the canopy structure spectrum data of important rice development stages,and conduct joint analysis with the harvest index and related agronomic traits,use the random forest algorithm to establish the prediction model of harvest index and yield-related traits,and construct the model of high harvest index varieties,in order to provide a theoretical basis for the high-throughput analysis of field agronomic traits and the cultivation of a new generation of high-harvest index varieties.Below are key research findings:1.Determination of the harvest index of different types of rice varieties.The results showed that the harvest index of the current varieties was between 0.32 and0.60,and the harvest index of Yuexiangzhan was the highest at 0.60 with many tillers,high seed setting rate,and relatively little aboveground biomass;the harvest index of Yuenongsimiao was 0.57,the yield per plant was 22.78±0.46 g,and the aboveground biomass per plant was 39.97±1.71 g.Yuexiangzhan increased significantly;Tetep had the lowest harvest index of 0.32,yield per plant was 13.49±1.12 g,aboveground biomass was 42.15±2.89 g,and plant height was 165.30±3.27 cm,showing more straw and lower yield.The results of correlation analysis showed that the harvest index had a significant positive correlation with the yield per plant,with a correlation coefficient of 0.57,and a significant negative correlation with straw weight and plant height,with correlation coefficients of-0.52 and-0.59,respectively;thousand-grain weight and grain length and width ratio,effective tiller number were significantly negatively correlated,and the correlation coefficients were-0.59 and-0.44respectively.2.The spectral characteristics of different harvest index varieties were obtained by using multi-spectral sensors mounted on UAVs.The results of correlation analysis showed that the amount of straw per plant was significantly negatively correlated with Ex R,Ex GR,and COM,and the correlation coefficients were-0.55,-0.52,and-0.51,respectively;it was significantly positively correlated with NDVI,and GNDVI,and the correlation coefficients were respectively 0.44 and 0.48;on the texture index,the grain weight per plant was significantly negatively correlated with the homogeneity and the second moment of the angle,and the correlation coefficients were-0.48 and-0.43;the correlation between the harvest index and the entropy was0.43.Significant positive correlation,angular second moment correlation-0.48,significant negative correlation.For the model built with RGB,the plant height has the highest prediction accuracy at the full heading stage,with a determination coefficient R~2of 0.88,and the highest aboveground biomass at the mature stage is the R~2of 0.87;the prediction model built with the texture index has the highest prediction accuracy at the full heading stage is the plant height.Height and aboveground biomass,R~2is 0.82,the harvest index is the highest at maturity,and R~2is 0.82;in the prediction model constructed by the multi-spectral vegetation index,the highest prediction accuracy of full heading stage is plant height,R~2is 0.87,and the maturity The agronomic trait with the highest accuracy was the harvest index,with an R~2of 0.84;in the graph-spectrum multi-data fusion prediction model,the highest accuracy at the full heading stage was the harvest index,with an R~2of 0.87,and the highest accuracy at the maturity stage was the aboveground biomass,with an R~2of 0.86.3.According to the analysis results of the agronomic traits of representative varieties with different harvest indexes,the plant type indexes with high harvest index are proposed as follows:high harvest index(0.55-0.60),more grains per panicle(150-250 grains),stronger tillering ability(7-11 panicles/plant),moderate plant height(95-115 cm),strong root system vitality,and coordinated stems and leaves.Representative varieties include Yuexiangzhan,Yuenongsimiao,Yuehesimiao,etc.At the same time,it is proposed that compared with the first-generation high-harvest index variety Yuexiangzhan,Yuenongsimiao and Yuehesimiao,etc.,the aboveground biomass and yield per plant were significantly increased while maintaining a high harvest index,which is a new generation of high-harvest index varieties.In summary,this study combines manual measurement with multispectral data carried by UAV to construct a field agronomic trait prediction model.Based on the above ground agronomic traits and root characteristics of high harvest index varieties,the plant type of high harvest index varieties is constructed,providing a theoretical basis for high-throughput analysis of field agronomic traits and breeding of rice varieties with high harvest index. |
PDF Full Text Request