Winter Wheat Freeze Injury Quantitative Research Based On Multivariate Analysis And Hyper-spectral Remote Sensing Data

In this paper, a controlled field experiment of a strong winterness winter wheat Jintai 182 was conducted. The mobile refrigeration equipment was used to simulate the freeze injury in jointing stage. The canopy spectral reflectance and related agronomy parameters of winter wheat were measured and t analyzed after different freezing days. The quantitative monitoring model of winter wheat under different freezing injury treatments was constructed by principal component analysis, the partial least squares and stepwise multiple linear regression analytical method (PLS-SMLR). The results showed as follows:1?The agronomic parameters of winter wheat presented notable change under different freezing injury treatments. Compared to CK, the POD, SOD and MDA of winter wheat under different freezing injury treatments increased, while other agronomic parameters decreased. Differences gradually reduced with increasing days after freezing.2?Correlation analysis among agronomic parameters indicated that the information of freezing injury overlapped. Therefore, four uncorrelated components were extracted by principal component analysis, which contained 88% of the original information. The freeze injury comprehensive evaluation index (FICEI) was obtained by the formula.3?The freezing injury degree was divided into four categories by cluster analysis, which were no damage, slight damage, moderate and serious frost damage. The results indicated that the degree of freeze injury of winter wheat could be evaluated according to the FICEI value. When the FICEI>1.578, winter wheat was not affected by freeze injury. When 0.388< FICEI< 1.578,winter wheat suffered -2?or -4? short-term damage and it was affected by slight damage. When -0.128< FICEI<0.388, winter wheat was suffered from -6? short-term damage or -4? long-term damage and it was affected by moderate damage. When the FICEI<-0.128, winter wheat was suffered from -6? long-term damage and it was affected by serious injury.4?The spectrum variation curves had almost the same change trend under different freezing injury treatments. The reflectance in the visible region were firstly decreased and later increased, while the near-infrared region showed an opposite trend. During the same period, the hyperspectral reflectance of winter wheat was not the significant results in the visible region under different freezing injury treatments, and significantly increased and positively correlated with degree of freezing injury in the near-infrared region. First derivative spectral reflectance of winter wheat had two reflection peaks and one absorption valley. In addition, the degree of freezing injury and the peak was positive correlation. Furthermore, red edge position showed a trend to shift to short-wave band, and shift from originally red to shorter blue.5?The spectrum sensitive bands of winter wheat FICEI were extracted based on PLS analysis method. They were 429-473 nm,519-584 nm,658-750 nm and 1139-1350 nm, respectively. And the quantitative monitoring model of winter wheat FICEI was constructed using SMLR analysis method. The R2 of model was 0.677, and the root mean square error was only 0.337. In addition, The R2 of verification model was 0.804, and the root mean square error was 0.127. It indicated that the monitoring model had better prediction effect based on the PLS-SMLR.