Multitemporal analysis of the 3D structure and biomass of subarctic open canopy forests using photogrammetric point clouds

点击翻译按钮获取中文摘要

Subarctic forests are open woodlands sparsely populated by small trees. They cover, in Canada alone, an area of more than 2.5 million km2 (roughly the size of Argentina). Spread between the 50th and 70th northern parallels, these forests fall within the regions that are experiencing some of the greatest climate warming. Such a high degree of change will likely induce forest densification, increase ecosystem productivity, and transform habitats. A sharp increase in the amount of sequestered atmospheric carbon by denser and taller forests over such huge areas is possible, but is still debated, and could go undocumented. Indeed, few resources are allocated to measuring and mapping the height, density, and biomass of these forests due to the high costs of data acquisition in such remote areas. Airborne surveys are logistically difficult, and conventional remote sensing based on 2D satellite images (e.g. Landsat) lack accuracy. New types of remote sensing data and analysis methods are needed to greatly increase our understanding of the dynamics of these forests. Our goals are therefore to a) use new very high resolution satellite stereo images to reconstruct 3D models of subarctic forests at a pixel size of 30 cm, b) use historical stereo aerial photos acquired since the 1950s to create similar 3D models representing the structure of these forests in the past, c) extract measurements of tree height, crown width, number of trees per hectare (density), and biomass density (tons of woody biomass per hectare) from these recent and retrospective 3D reconstructions, d) measure the 3D changes of the forest structure over time, and e) link these changes to the evolving climate. Throughout this endeavour, we will be designing and testing a entirely new approach for assessing the error of the stereo forest attribute measurements. By synthesizing stereo images using computer models that mathematically represent the optics and viewing geometry of the sensor or camera, the sun’s position, the state of the atmosphere, etc., we will assess how well these artificial images allow us to measure forest attributes. This new approach will first need to be validated using data collected in the field. Once proven, we’ll be able to use it to predict the error for any image taken from very high resolution satellites or aerial camera. The proposed research will first have an important impact on our capacity to map subarctic forests with high accuracy and low ambiguity. We will have created the means for extracting detailed forest information for any subarctic location (in Canada, Russia, etc.), based on very high resolution stereo-images without the need for costly field calibration. Our research will lead to new knowledge on the amount and rate of change of forest carbon contents of subarctic forests, will help improve long-term climate model predictions for high latitudes, and will provide a much more precise picture of the habitats of the taiga.