College of forestry, Nanjing Forestry University,College of forestry, Nanjing Forestry University
基于1986年到2011年的Landsat影像，以南方人工林分布区域广东省佛冈县为例，运用Landsat生态系统自适应处理系统（LEDAPS）预处理生成标准的地面反射率数据构建Landsat时间序列堆栈（LTSS）用于LandTrendr算法监测人工林森林干扰与恢复的长时间序列变化，分析了连续24a森林干扰的年份变化、干扰量以及干扰持续的时间，验证了算法识别干扰的精度，并探讨了人工林干扰的驱动力。结果表明佛冈县的森林干扰较为剧烈，一般都在1000 hm2。而1987、2002、2004、2005、2006、2007和2009年的干扰面积均超过2000 hm2，其中1987、2007年两年的干扰面积达到6000 hm2以上。相比森林干扰的变化，佛冈县的森林恢复面积随时间的变化相对平稳。通过对佛冈县森林干扰和恢复面积的趋势分析，发现20世纪80年代末到90年代森林干扰和恢复的面积基本少于2000年以后的变化面积，变化趋势比2000年以后的显得平缓；从2000年开始，森林干扰面积逐渐上升，总体面积变化趋势高于森林的恢复，但森林的恢复面积仍有所提升。其中，佛冈县的森林干扰持续1a时间的面积比例约38%，持续2a时间约28%，持续3a时间约25%，持续4a时间约7%，主要为短期急剧的干扰事件。另外，持续时间为4a以上的森林干扰和恢复的面积在佛冈县不超过100hm2。2000年之前持续干扰和急剧干扰面积相当，变化比较平缓；到2000年之后，急剧干扰的面积远大于持续干扰，最高约达2800 hm2，但两者都呈现波动上升的变化趋势。在选取的两个4km2的样方中，基于影像光谱识别以及通过比对干扰资料的可视化验证方法表明算法结果与真实地表的解译信息较吻合，误差约为0.1km2。利用长时间序列遥感影像进行森林干扰的自动化监测十分必要，导出的定性、定位与定量信息，一方面为可持续的森林经营奠定基础，另一方面为评价森林生产力与森林碳储量提供有效的数据支撑。
Yearly Landsat imagery from 1986 to 2011 of a typical plantation region in Fogang County, Guangdong Province, southern China, was used as a case study. The pre-processing Landsat Ecosystem Disturbance and Adaptive Processing System (LEDPAS) algorithm was implemented to generate standard surface reflectance images to construct a Landsat time series stack (LTSS). The LTSS was fed to the Landsat-based detection of trends in disturbance and recovery (LandTrendr) algorithm to monitor long-term changes in plantation disturbance and recovery, followed by an intensive validation and a continuous 24 years change analyses on annual change, and disturbance amount and duration. Validations derived from two chosen sample plots of 4 km2 indicated that the LandTrendr-based mapped disturbance results strongly agreed with those derived from the visual interpretation of the pre-and post-disturbance multispectral images and visualization of the local disturbance documents, with an error of 0.1 km2. Results indicated that the forest disturbances that occurred in Fogang County were relatively drastic. An annual disturbance of 1000 hm2 was witnessed for most years of the study, and an annual disturbance of over 2000 hm2 occurred in 1987, 2002, 2004, 2005, 2006, 2007, and 2009. Particularly, the disturbance of 1987 and 2007 exceeded 6000 hm2. In comparison to forest disturbance, forest recovery areas were relatively stable. Through a trend analysis of forest disturbance and recovery in Fogang County, forest disturbance and recovery areas mapped in the late 1980s through 1990s were less than those mapped after 2000, and the trend was lower than that after 2000. Since 2000, the forest disturbance areas have gradually increased, with a slight increase in forest recovery, but the overall magnitudes of forest disturbance exceeded those of forest recovery. The area of forest disturbance with a duration of 1 year accounted for 38%, 28% for a duration of 2 years, 25% for a duration of 3 years, and 7% for a duration of 4 years; these disturbances were classified as abrupt and short-term disturbance events. Gradual forest disturbance and recovery events for a duration over 4 years existed, but the overall areas were less than 100 hm2/a, and were highly different from the areas of abrupt disturbance events. Prior to 2000, abrupt and gradual disturbance areas were almost equal, with a gentle change. After 2000, abrupt disturbance areas were greater than those of gradual disturbances, with a maximum of 2800 hm2, and both abrupt and gradual disturbances showed an undulatory increasing trend. Based on the history and status of forest disturbances in Fogang County, the factors contributing to the environmental disturbance of forest plantations were analyzed to develop effective forest management strategies and countermeasures. The current study demonstrated the need to use dense time series images to map forest disturbance and recovery events in plantation forests. This approach could provide qualitative, locational, and quantitative forest change results for the land use decision-makers and conservation communities, enabling the strategic development of sustainable forest management and provide effective data support to evaluate forest productivity and carbon storage.