Development and application of multiscale object-based approaches for remote sensing landcover analysis and resource management

这个星球及其支撑的景观是一个复杂的系统，由大量异质成分组成，这些成分以非线性方式相互作用，具有层次结构，空间相关性和规模依赖性。遥感技术是大面积土地覆盖数据的主要来源。因此，为了充分理解、监测、建模和管理我们在景观中的相互作用，我们需要能够整合遥感图像和适当的多尺度理论和方法，以特定的表达尺度定义景观组件（即场景对象），并在相应的层次尺度上将它们与相关对象链接/评估。然而，遥感图像中没有物体，只有像素，“景观”的概念因用户而异，没有“比例尺科学”，也没有固定的比例尺定律或规则来转换不同比例尺之间的数据。在克服这些挑战的共同努力中，我的科学计划的目标是开发和应用基于遥感、地理信息系统、高性能计算和地理可视化的创新多尺度对象方法。这将增强对多尺度景观结构和动态的理论和理解，以及创新的新工具和方法，以更好地评估和管理复杂的环境。具体项目侧重于森林景观和开发基于多尺度物体的工具，用于(i)改进森林生物量/碳估算；（ii）分层森林抽样方案，将昂贵的激光雷达（光探测和测距）估算扩展到更大的区域；（iii）计算机辅助森林清查，从单个树木更新到林分和区域。该研究计划将利用高知名度的遥感数据集、高性能计算以及与国家和国际项目（即EOSD和NFI）专家的合作研究。将培养大量高素质人才（1名PDF， 4名博士，8名硕士学生和5名实地/研究助理）。

This planet, and the landscapes it supports are complex systems composed of a large number of heterogeneous components that interact in non-linear ways, are hierarchically structured, spatially correlated and scale dependent. Remote sensing technologies represent the primary source for large area landcover data. Therefore, to fully understand, monitor, model, and manage our interaction within landscapes, we require approaches capable of integrating remote sensing imagery and appropriate multiscale-theory and methods to define landscape components (i.e., scene objects) at their specific scales of expression, and link/evaluate them with related objects at corresponding hierarchical scales. However, there are no objects in remote sensing images, only pixels, the concept of 'landscape' varies between users, there is no 'science of scale', nor are there fixed scaling laws or rules for translating data between scales. In a concerted effort to overcome these challenges the goal of my scientific program is to develop and apply innovative multiscale object-based approaches using remote sensing, geographic information systems, high-performance computing and geo-visualization. This will result in enhanced theory and understanding of landscape structure and dynamics at multiple scales, and innovative new tools and methods to better assess and manage complex environments. Specific projects focus on forest landscapes and the development of multiscale object-based tools for (i) improved forest biomass/carbon estimates, (ii) stratified forest sampling schemes to extend expensive LiDAR (Light Detection and Ranging) estimates over larger areas, and (iii) computer assisted forest inventory updating from individual trees, to stands and regions. This research program will take advantage of high-profile remote sensing data sets, high-performance computing, and collaborative research with experts in national and international projects (i.e., EOSD, and NFI). Considerable training of highly qualified personnel will be achieved (1 PDF, 4 PhD, 8 MSc students and 5 Field/Research Assistants).