Continental and global mapping of forest biomass by fusion of InSAR, lidar and stereophotogrammetric satellite data

拟议的研究计划寻求开发新的3D遥感方法，在大陆和全球范围内绘制森林生物量以及高度和蓄积量的地图。对这三个高度相关的结构参数在空间上的连续估计将极大地提高我们对碳循环的了解，同时为大规模生境碎片化研究和全州范围内对不断增长的种群数量的估计提供关键数据。 要对总生物量作出大陆或全球的无偏估计，就需要我们根据区域内不同的森林和地形类型(即从闭塞的阔叶林到疏林，从平坦的地形到崎岖的地形等)，并根据多云条件和植被物候，选择区域内最适当的遥感工具，并且我们寻求最佳融合方法来合并所选的数据集。 我们的具体目标是：1)评估在大陆和全球尺度上生成冠层高度模型(CHM=DSM-DTM)的最佳策略；2)开发从CHM生成空间连续和无偏的森林结构属性估计(生物量、高度和蓄积量)的统计方法；3)详细阐述通过极化相干层析成像(PCT)直接提取生物量估计的方法，可选择以DSM和DTMS为界限。将测试以下数据源：对于立体声DSM：ASTER、SPOT等；对于InSAR DSM：Tandem-X；对于DMS：机载激光雷达、滤波InSAR DSM和基于PCT的DMS。区划将基于森林类型、地形、云量以及每个遥感数据集相对于采集图像时植被物候状态的充分性。各种统计技术将被用来根据CHM或直接根据断层图像来预测生物量和体积。大片区域(1000平方公里)的机载激光雷达数据与实地数据相结合，将用于分析误差源，并创建生物量验证地图。

The proposed research program seeks to develop novel 3D remote sensing methods for mapping forest biomass, as well as height and volume, at continental and global scales. Spatially-continuous estimates of these three highly correlated structural parameters will greatly improve our knowledge of the carbon cycle while providing critical data for large scale habitat fragmentation studies and statewide estimates of growing stock volume. Producing continental or global unbiased estimates of total biomass will require that we choose the most appropriate remote sensing tools regionally as a function of the very contrasted forest and terrain types encountered over very wide areas (i.e. from closed broadleaf to open taiga forests, from flat to rugged topography, etc.), as well as according to cloudiness conditions and vegetation phenology, and that we seek the best fusion methods for combining the chosen datasets. Our specific objectives are to 1) evaluate the best strategies for generating canopy height models (CHM = DSM - DTM) at continental and global scales, 2) develop statistical methods for producing spatially-continuous and unbiased forest structural attribute estimates (biomass, height and volume) from the CHMs, and 3) elaborate methods for extracting biomass estimates directly by polarimetric coherence tomography (PCT), optionally bounded by DSMs and DTMs. The following data sources will be tested; for stereo DSMs: ASTER, SPOT, etc.; for InSAR DSMs: Tandem-X; for DTMs: airborne lidar, filtered InSAR DSMs, and PCT-based DTMs. Regionalization will be based on forest type, topography , cloudiness, and the adequacy of each remote sensing dataset relatively to the phenological state of the vegetation at the time of image acquisition. Various statistical techniques will be employed to predict biomass and volume from the CHMs, or directly from the tomograms. Airborne lidar data over extensive areas (several 1000 sq km), combined to field data, will be used to analyze the error sources and to create validation maps of biomass.