Forests are dynamic ecosystems, subject to both natural and anthropogenic agents of change. Wildfire, harvesting and other human activities alter the tree-covered area present in forests. From national and international reporting perspectives, forests include areas currently treed, as well as those disturbed forest areas that are not currently treed but will be, given time for regeneration and the advancement of natural successional processes. As a consequence, forest area can be depicted at a particular point in time, informed by a retrospective temporal context. Using time series of Landsat imagery, annual land cover maps can be generated that are informed by knowledge of past disturbance history (such as wildfire and harvesting). In this research, we use over three decades of annual land cover data generated from Landsat time series to generate a spatially explicit estimate of the forest area of Canada in 2010. We demonstrate how land cover and disturbance information can be combined to map the area of ‘forest’, as defined by the Food and Agricultural Organization of the United Nations (FAO), within Canada’s 650 Mha of forested ecozones. Following this approach, we estimated Canada’s total forest area in 2010 to be 354.5 Mha. This estimate includes 324.5 Mha of current forest cover in 2010, plus an additional 33.2 Mha (or 9.4 per cent) of temporally informed forest area where tree cover had been temporarily lost due to fire or harvest, less 3.3 Mha that were removed to meet a definitional minimum size (0.5 ha) for contiguous forest area. Using Canada’s National Forest Inventory (NFI) as an independent reference source, the spatial agreement between the two estimates of forest area was ~84 per cent overall. Aspatially, the total area of the Landsat-derived estimate of 2010 forest area and the NFI baseline estimates differed by only 3 per cent, with notable regional differences in the wetland-dominated Hudson Plains Ecozone. Satellite-derived time series land cover and change information enable spatially explicit depictions of forest area (distinct from representations of forest cover) in a robust and transparent fashion, producing information of value to science, management and reporting information needs.

中文翻译：

---

基于卫星的时间序列土地覆盖和变化信息，以绘制符合国家和国际报告要求的森林面积

森林是动态的生态系统，受自然和人为因素的影响。野火，砍伐和其他人类活动改变了森林中树木覆盖的区域。从国家和国际报告的角度来看，森林包括当前已被树木覆盖的区域，以及那些当前尚未被树木覆盖但将要被赋予再生和自然演替过程进展的受干扰森林区域。结果，可以通过回顾性的时间背景在特定的时间点描绘森林面积。使用Landsat影像的时间序列，可以生成年度土地覆盖图，并根据过去的干扰历史（例如野火和收割）来获知。在这项研究中 我们使用从Landsat时间序列生成的过去三十年的年度土地覆盖数据来生成2010年加拿大森林面积的空间显式估计。我们演示了如何将土地覆盖和干扰信息结合起来绘制“森林”面积，根据联合国粮食及农业组织（FAO）的定义，在加拿大650 Mha的森林生态区内。按照这种方法，我们估计2010年加拿大的森林总面积为354.5 Mha。该估计数包括2010年当前的森林覆盖率324.5 Mha，以及因火灾或砍伐而暂时失去树木覆盖率的33.2 Mha（占9.4％）的临时森林面积，为满足森林覆盖率而减少了3.3 Mha。连续森林面积的最小定义尺寸（0.5公顷）。使用加拿大国家森林清单（NFI）作为独立参考源，两次森林面积估算之间的空间一致性约为84％。从地域上看，Landsat得出的2010年森林面积估算总面积与NFI基准估算值之差仅3％，在以湿地为主的哈德逊平原生态区中存在明显的区域差异。卫星衍生的时间序列土地覆盖和变化信息能够以健壮和透明的方式在空间上对森林区域进行清晰的描述（与森林覆盖的描述不同），从而产生对科学，管理和报告信息需求有价值的信息。Landsat得出的2010年森林面积估算的总面积与NFI基准估算值的仅相差3％，在以湿地为主导的哈德逊平原生态区中存在明显的区域差异。卫星衍生的时间序列土地覆盖和变化信息能够以健壮和透明的方式在空间上对森林区域进行清晰的描述（与森林覆盖的描述不同），从而产生对科学，管理和报告信息需求有价值的信息。Landsat得出的2010年森林面积估算的总面积与NFI基准估算值的仅相差3％，在以湿地为主导的哈德逊平原生态区中存在明显的区域差异。卫星衍生的时间序列土地覆盖和变化信息能够以健壮和透明的方式在空间上对森林区域进行清晰的描述（与森林覆盖的描述不同），从而产生对科学，管理和报告信息需求有价值的信息。