大地域、长时间序列的森林变化类型、方向和强度时空意义明确的信息对于生物多样性丰富度维持、气候调节、碳蓄积和水分供给等科学应用具有基础性数据支撑作用，及时、准确地采集森林变化信息已成为当前自然资源管理者、全球变化研究科学家及政府政策制定者等的重点关注。项目的总体目标是基于长时间序列LandsatTM/ETM+/OLI数据，发展一个综合的、基于遥感的连续变化检测和分类算法实现森林急剧变化、季节性变化和缓慢变化的类型分离和制图。借助傅里叶思想，采用稳健迭代重复加权最小二乘方法拟合像元的反射率时间序列谐波模型，通过比较模型预测值与遥感实际观测值差异并阈值化（因像元而异、数据驱动的阈值）来识别断点(急剧变化),通过分段线性拟合求取斜率来定义森林渐变并累积渐变总量，评估拟合模型的峰值和谷值差异来定义季节性变化特征,为在30米空间、16或8天时间尺度上实现森林变化类型、方向和强度监测提供模型和方法。

Large-scale, long time series and spatially explicit forest change information on change type, direction and magnitude plays an underlying role in supporting diverse scientific applications including maintaining biodiversity richness, regulating climate, sequestering carbon and supplying water etc., thus, timely and accurate aquisition of such information has become the major concern of current natural resources managers, global change investigators and governmental policy makers. The major objective of the study is to develop an integrated and continuous change detection and classification algorithm to separate and map forest abrupt change, seasonal change and gradual change events based on long time series Landsat Landsat TM/ETM+/OLI observations. The Fourier idea is introduced into our methodological development. Specifically, a harmonically related weighted sum model consisting of sins and consines, constant and trend terms is proposed to represent the aggregated contributions of forest abrupt change, seasonal change and gradual change to the spectral reflentance change of a pixel in the long time series, and this model is fitted by using the robust iteratively reweighted least squares method. Through comparing the modeled reflectances and the actual Landsat observations, applying a pixel-specific and data-driven thresholding technique, forest abrupt change events (represented as breaks in the modeled time series reflectances) are identified, simultaneously, the piecewise linear fitting based on the observations between the two identified breaks is used to derive the slope to characterize forest gradual change event and obtain the accumulated graudal magnitude. And assessing the difference between the modeled peak values and valley values is applied to define forest seasonal change characteristics. As a result, an integrated mapping model to capture forest change types, directions and magnitudes in the 30 meters spatial resolution, and 16 days or 8 days period cycle manner is developed to satisfy multi-discinplinary scientific data demands for forest change information.