基于时间序列MODIS EVI数据的森林生长异常监测

摘要

Forest growth is mainly currently monitored using in-situ measurements in northeast of China. To effectively monitor forest growth disturbance at large scale, we attempted to use remote sensing technique, particularly, time series MODIS data from 2004 to 2006. The annual time series of 8-day enhanced vegetation index ( EVI) dataset was generated and smoothed using a Savitzky-Golay filter. The EVI trajectory during growth season was simulated using a logistic model. From the simulated trajectory, the EVI area of growth season and annual EVI entropy were calculated. These two factors were combined to map the disturbance regions of forest growth. Finally, the disturbance regions were verified using a set of random samples. The result indicates that the disturbance points have distinctively higher entropy and lower peak. Some of these points also show abrupt EVI decline during the midseason of the peak phases or double peaks. This approach is demonstrated to be feasible for disturbance monitoring of forest growth.%利用遥感技术,基于时间序列的MODIS数据对2004-2006年东北三省的林区进行森林生长异常监测.首先利用MODIS数据时间分辨率高的特点,采用Savitzky-Golay滤波函数平滑8天合成的EVI,计算生长季面积和年EVI曲线熵值,两指标联合得到3年间变化量大的像素点,定义为森林生长异常点;然后抽取异常点的时间序列曲线进行分析,并结合森林灾害事件进行比较验证.结果表明:异常点曲线的熵值明显大于正常年,生长季峰值低,并且在生长旺季会出现峰值突然持续下降或双峰等异常现象,这与该区域森林生长异常发生时的植被反射率表征一致,说明用该法对森林生长异常进行监测是基本可行的.