中文摘要：

以南京紫金山国家森林公园为研究区域,以2002年100块样地调查数据、免费开放的2001—2010年Landsat时间系列堆栈（LTSS）数据为主要信息源,分别采用原始波段、辐射校正、光谱指数＋地形因子3种策略建立2002年风景林生物量K最邻近算法估测模型。在精度验证基础上,分森林公园综合整治前（2001—2004年）、综合整治中（2005—2007年）、综合整治后（2008—2010年）3个时段,对研究区域2001—2010年的风景林生物量变化及其驱动因素进行了时间轨迹分析。研究表明：1）在3种KNN建模策略当中,光谱指数＋地形因子的综合策略相关系数最高,标准误差、平均相对误差最低,预测精度最高;2）2001—2010年,研究区域风景林生物量分别呈现缓慢下降、先降后升、缓慢上升的复杂变化趋势;3）2001—2010年,研究区域的森林干扰经历了上下小幅波动、缓慢增强、迅速下降3种不同的变化趋势;4）在空间分布格局上,风景林生物量在3个时段分别经历了聚集性缓慢增强、破碎化加剧、趋向稳定3种变化趋势。本文的研究成果可以为区域、长时间尺度上包含森林干扰与恢复因子的高精度碳计量模型的建立提供科学依据。

英文摘要：

There is a large amount of carbon stored in scenic forest biomass, which plays an important role in regional carbon cycle. The spatial effect of forest disturbance and recovery on the release of carbon should be taken into account for accurately measuring scenic forest biomass. An estimation model using K nearest neighbor （ KNN） algorithm for the biomass of scenic forest was built based on the survey data from 100 sample plots in 2002 at Zijin Mountain National Forest Park of Nanjing, and data of free Landsat time series stacks （ LTSS） during 2001 to 2010 by using three approaches of original band, radiometric correction, and spectral index plus terrain factors. The time trajectory analysis of the biomass of scenic forest and drivers was conducted with three periods, i. e. , before the renovation （2001--2004）, during the renovation （2005--2007） and after the renovation （2008--2010） based on the precision validation. Study results showed that：1） among the three approaches, comprehensive model of spectral index plus terrain factors outperformed others with the highest correlation coefficient and prediction accuracy, and the lowest standard error and the minimum mean relative error;2 ） during 2001 to 2010 , the biomass in scenic forest showed a complex trends of slow decline, down then up and slowly rising;3） during 2001 to 2010, forest disturbance in the study area had three different trends of slight fluctuation, slow increase, and rapid decline;4） on the spatial distribution, the biomass in scenic forest experienced the trend of slow increase of aggregation, intensified fragmentation and stabilization in three periods respectively. Results of this study could provide a scientific basis for the establishment of long term high-precision carbon accounting models covering forest disturbance and recovery factors at regional scale.