Diagnosis Study On Later Spring Frost Of Winter Wheat Based On Hyperspectral Remote Sensing

The North China Plain is the main production zone of winter wheat (Triticum aestivum L.), where the winter wheat is affected seriously by later spring frost. In some serious years, the affected area and the loss of yield were tremendous. The later spring frost often occurs after the jointing stage of winter wheat. If the warming temperature decreases sharply and suddenly in later spring day, the winter wheat will suffer from frost damage easily. This frost injury is so called later spring frost in agriculture. The sudden of occurrence and the concealment of harmness are the principal characteristics of this kind of freeze injury. As a result, the traditional diagnosis methods of agricultural diseaster can not diagnosis the later spring frost quickly, precisely and efficiently.Hyperspectral remote sensing is an accurate and high efficiency technology, which is one of the significant tools to realize precision agriculture. Real-time information of the crop growth can be obtained by this technique quickly and objectively. According to the merits of hyperspectral remote sensing, it was used in this paper as the new diagnosis method of later spring frost. We conducted a controlled field experiment with the artificial frost apparatus to simulate the frost damage in2011-2012, measured the canopy hyperspectral reflectance of different cultivars of winter wheat after jointing stage under different freezing temperature treatments at different dates, analysed the characteristics of canopy original spectrum and1st derivative of canopy spectrum. Addtionally, we researched the yield factors per plant of winter wheat under different freezing treatments, to investigate the influence of freezing temperature and freezing dates. Analysing the degrees of correlation between canopy hyperspectral data and yield factors of winter wheat, we selected the sensitive spectrums and hyperspectral parameters of later spring frost. Based on the linear regression models between hyperspectral sensitive indices and yield factors per plant, we researched the accuracy of stimulation and caculated the sensitive coefficients of different indices. After all, we constructed the hyperspectral diagnosis indices of later spring frost of winter wheat. The main results and conclusions of this paper are as follows: 1. Later spring frost influenced the hyperspectral data remarkably and the canopy hyperspectral reflectance changed notably on1d after freeze injury. As the decrease of freezing temperature, the’green peak’(550nm spectrum) canopy reflectance curves of frost group became plane, and the canopy reflectance at’red well’(670nm spectrum) increased gradually. At the same time, the original canopy reflectance curves in530-680nm also became plane, and the canopy reflectance at’near infrared flat’(760-925nm spectrum) decreased overall. Additionally, the slope of canopy reflectance curves in760-925nm increased remarkably. As the treatment temperature decreased, the1st derivative curve became plane at visible band, and the’red edge’(680-760nm spectrum)1st derivative also decreased. Meanwhile, the’red edge position’ moved to the short wave direction. All of these above indicated that the growth of winter wheat had been impacted seriously. Furthermore, as the change of freezing date, the divergence between canopy hyperspectral data of different cultivars and spectra was also notable.2. Analysing the relationship between the canopy hyperspectral reflectance and yield factors per plant of winter wheat, we found that the correlation coefficients between original canopy reflectance and yield factors were mainly negative under different freezing temperature treatments, while the transition wavebands between positive and negative coefficients at different freezing dates often appeared in680-760nm(’red edge’spectrum). Under freezing treatments at different dates, the correlation coefficients between1st derivative spectrum and yield factors at visible light are often higher than those at’near infrared flat’(760-925nm spectrum), and the significant level above0.05mostly appeared in350-700nm. Under both freezing treatments (different freezing temperature and different freezing dates), the degrees of correlation between1st derivative spectrum and yield factors per plant of winter wheat are higer than those between original spectrum and yield factors at most wavebands, and the undulatory property of correlation coefficients became lager simultaneously.3. Under different freezing temperature treatments, the1st derivative spectrum diagnosis indices of yield factors per plant are mainly at’near infrared flat’(760‖925nm spectrum), while this kind of diagnosis indices under freezing treatments at different dates appeared at visible light mostly. Among the diagnosis indices based on hyperspectral parameters, the performance of SDR/SDy and λR is very remarkable, so we can consider using them as majoy hyperspectral diagnosis indices in further study. Most of the hyperspectral diagnosis indices of later spring frost in this study are based on the hyperspectral data at ’red edge’(680-760nm spectrum), which illustrated that the ’red edge’(680-760nm spectrum) was very sensitive to later spring frost damage just like other agricultural diseasters of winter wheat. According to this, the deeper investigation on the characteristics of ’red edge’(680-760nm spectrum) will be needed in further relative study.