Abstract: 【Objective】The aim of this study is to analyze the variance of a variety of vegetation indices derived from HJ-1B (One of the Small Satellite Constellations for Environment and Disaster Monitoring and Forecasting, HJ-1A/1B/1C) CCD multi-temporal images in pre- and post-lodging, in order to achieve the objective for measuring the area of maize lodging and evaluation of disaster rank. 【Method】This paper took Gaocheng, Hebei Province as a research area where suffered a heavy rainstorm which led to a large-scale maize lodging on August 3rd, 2015. A variety of vegetation indices derived from HJ-1B CCD multi-temporal images in pre- and post-lodging period were adopted to conduct the sensitivity analysis among various vegetation index variations (?VIi, the value of vegetation index before lodging minus that of after lodging) and the lodging characteristic values measured in field (Lodging proportion for research areas) to seek the preferable vegetation indices for maize lodging monitoring. The double threshold partition strategy of normal statistical theory was adopted to determine the thresholds and a remote sensing monitoring model for maize lodging was built based on the analysis of the difference of vegetation indices pre- and post-lodging to evaluate the disaster degree of the maize lodging. The accurate assessment of the model was conducted by comparing the in-field measured data and predicted results from the built model at last. 【Result】Based on the features of HJ–1B CCD2 multispectral reflectance curve of the town, water, lodging and no lodging maize, there was a difference between maize and building and water in special reflectance. In the visible spectrum, spectral reflectance of lodging maize was higher than that of no lodging maize, therefore, it was opposite in near infrared wave. From the analysis of RVI difference image, it nearly conform to normal distribution (skewed normal distribution). The result of correlation analysis suggested that RVI (ratio vegetation index) had the best correlation with the proportion of maize lodging (R=0.9377). It means that the more serious of the lodging, the higher the RVI difference. It showed that the maize growth status was obviously affected by lodging, the more severe of the lodging, and it was more difficult for maize to restore to its normal state before lodging and thus causing bigger values of the RVI differences. Generally, RVI differences for different small areas showed a normal (or skew) distribution in general from the statistical analysis of the image of RVI difference. The data measured infield about lodging proportion and the values predicted from the model were adopted to carry out confusion matrix analysis, it showed that a good overall classification accuracy up to 85.7% while the Kappa coefficient was 0.804. The spatial distribution of maize lodging area and the degree of disaster could be mapped via the constructed maize lodging remote sensing model based on the differences of vegetable index derived from the double threshold partition strategy which based on normal (skew) statistical theory. The result from remote sensing mapping was basically consistent with the monitoring data from the local agricultural technical extension station. 【Conclusion】After occurrence of lodging of maize, there was a big difference between the maize growth and the resumption of maize after disaster. Therefore, it was concluded that the remote sensing monitoring model based on the differences of RVI measured in pre- and post-lodging period of maize is a feasible method for reflecting the information which indicate the changes of canopy groups described by different lodging degrees and can satisfy the need for measuring the area of maize lodging and evaluation of disaster rank.

Key words: maize, lodging, vegetation index, change analysis, multi-temporal, monitoring