Winter Wheat Freeze Injury Research Based On Multi-sources Remote Sensing Data

Recently, with the frequent appearance of extreme global climate, winter wheat freeze injury monitoring is becoming more difficult, unpredictable and complex. With the rapid development of satellite technology, geographical information technology, global positioning system technology, and computer network technology, monitoring crop disaster with remote sensing technology has been widely applied in the research field of agriculture. Besides, the enhanced spatial resolution or spectral resolution of domestic satellite has driven the progress of related researches. However, crop freeze injury monitoring with remote sensing is not systematic enough. And, it is an important problem to effectively utilize the multi-source remote sensing information and extract the characteristic parameters at different periods of freeze injury.In this study, firstly, the responses mechanism of winter wheat freeze injury was analyzed with the hyperspectral data obtained in wintering stage and regreening stage at leaves scale and canopy scale under control. Then, models and methods for winter wheat freeze injury monitoring were constructed, based on the field experimental data and multi-source remote sensing data in Hebei province. The major contents and results in this research are as follows:(1) In the field, different anti-freeze varieties of winter wheat were sowed. We acquired the hyperspectral data to extract sensitive bands of freeze injury in wintering stage at leaf and canopy scales. At the leaf scale, we first analyze the original and first-order differential spectral data of winter wheat freeze injury to extract sensitive bands. And then, we establish the relationship between freezing injury degree DAI and hyperspectral characteristics, and the results showed that the first-order differential value of777nm, SDr/SDb, and (SDr-SDb)/(SDr+SDb) were taken as the variables for DAI estimation with R2≥0.750, where SDr/SDb has the best estimation precision as R2=0.799. At the canopy scale, principal component analysis method was used extract6principal components for SSR estimation, with the precision as R2=0.6309. We selected20vegetation indices (Ⅵ) sensitive to moisture content and chlorophyll of winter wheat, and16VIs have the significant correlation with canopy SSR. At last, we analyzed the indicator of yield though ANOVA, and4indicators were strongly correlated with SSR, which means that winter wheat freeze injury plays a role in yield.(2) We conducted the different temperatures deals though artificial equipment of winter wheat in the northern (in Beijing), and obtain hyperspectral data at the same time. The spectral response and mechanism of winter wheat freeze injury were analyzed with the leaf and canopy spectroscopy data, first-order hyperspectral data and the "red edge" under four temperature deals (T0, T1, T2, T3) of winter wheat. Moreover, the continuous wavelet analysis (CWA) was selected for SPAD estimation, taken the highest significance correlation feature E as the independent variable, with the accuracy as R2=0.786. Furthermore, we analyzed the relationship in the LAI, height of wheat, indicators of yield and the low temperature though ANOVA, and the results showed that winter wheat freeze injury in regreening stage is more extremely affected.(3) At large scale, we selected the analysis of change vector which based on multi-temporal vegetation indexes to improve the monitoring freeze injury accuracy. Winter wheat freeze injury of Gaocheng as study object, various vegetation indexes were extracted from multi-temporal HJ-CCDs data, change vector was built and the trend of dynamic changing was analyzed, combined with the sensitivity analysis of winter wheat freeze injury spectral character, the model of monitoring freeze injury situation disaster remote sensing was built, and monitoring the degree of growth recover. The result showed that the change vector analysis could reflect the distribution and degree of winter wheat freeze injury and recovery. Meantime the change vector model which based on the structure insensitive pigment index had the highest accuracy during based on the other vegetation indexes model, in addition, the model verification results were83.3%,88.9%, respectively.(4) In order to evaluate the severity of freeze injury on winter wheatsystematically and screen effective evaluation indicators severely affecting winter wheat freeze injury, This chcapter has combined the systematic knowledge models of grey theories with remote sensing technology (RS) and Geographic Information System (GIS) technology to establish a large scale multisource information fusion freeze injury overall evaluation model based on the principle and requirements of the evaluating indicator system. It gives examples of wheat freeze injury monitoring applications in Gaocheng and Jinzhou City of Hebei Province to carry out a quantitative evaluation method study on the severity of winter wheat freeze injury. Firstly, a grey relational analysis (GRA) has been conducted on the freeze injury index-stem survival rate and spatial data information such as surface temperature, nutrient content in the soil, thermal inertia of soil and soil water content, etc., which were measured during the period from2009to2010in the wheat of Gaocheng and Jinzhou of Hebei Province. At the same time, the weights of the data were also determined. Then a wheat freezing injury stress multiple factor spatial matrix has been constructed using spatial interpolation technology. Finally, a winter wheat freeze damage evaluation model was established through grey clustering algorithm evaluation and analysis, classifying the study area into3sub-areas, affected by severe disaster, medium disaster or light disaster/unaffected. The results showed that the severe, medium and the light/unaffected areas account for23.9%,40.71%and35.39%of the total wheat growing area respectively in Gaocheng; the severe, medium and light/unaffected areas account for17.12%,41.12%and41.76%of the total wheat growing area respectively in Jinzhou. After the evaluation results were verified by the Kappa Model, the overall accuracy reached78.82%and the kappa coefficient was0.6754. Therefore, through effective integration of grey cluster analysis mathematical models withRS as well as GIS special analysis, quantitative evaluation and study of winter wheat freeze disasters can be conducted objectively and accurately, making the evaluation model more scientific.