基于多尺度遥感测量的区域土地覆盖格局研究

STUDY ON REGIONAL LAND COVER PATTERNS DERIVED FROM MULTI- SCALE REMOTELY SENSED DATA

利用1 km、4 km和8 km 3种空间分辨率的NOAA/AVHRR数字影像,对中国NECT样带西部地区进行了土地覆盖分类及其景观特征的比较研究。重点比较了几种空间分辨率遥感数据分类结果边界的一致性和空间差异,以及影像所记录的景观格局的差异。为进一步在不同尺度上研究景观变化过程以及尺度转换研究奠定了基础。研究表明:3种空间分辨率的遥感影像所反映的区域土地覆盖的宏观空间格局是一致的,但类型的边界、每一类型斑块的形状和数量均产生较大的差异;经过对反映景观空间结构的4种指标(分维数、破碎度、多样性、优势度)的比较显示出随着遥感影像空间分辨率的变化,影像所反映的景观结构发生了较大的变化。其中,各覆盖类型的分维数表现出最大差异,表征着空间分辨率的变化对斑块复杂程度的影响最大。

We estimated quantitatively the effect of spatial scale magnification on land cover monitoring and landscape characteristics at regional scale, using different spatial remotely sensed data from RS ( Remote Sensing) , GIS (Geographic Information System) and GPS (Global Positioning System) , and taking the west of the Northeast China Transect (NECT), one of the IGBP international transects in China, as the study area. A comparison study on land cover classification and landscape characteristics, using NOAA/AVHRR digital images of 3 different spatial resolutions, was performed in this paper in the study area. USGS EROS (Earth Resources Observation System) Data Center (EDC) provided the monthly remote sensing images as data resources , which were in uniform tempo and region at three different spatial resolutions: 1 km × 1 km, 4 km × 4 km and 8 km x 8 km, spanning a 12-month period (April 1992-March 1993). Firstly, unsupervised classification algorithm ISODATA was employed to process 3 different spatial resolution images and obtained land classification information. Secondly, information transfer among different resolution images was calculated using a map algebraic algorithm. Finally, regional landscape characteristics of land cover types at different spatial scales were compared, taking landscape spatial structure indices such as fractal dimension, fragmentation, diversity and dominance as standards. Particular emphasis was put on the comparison of boundary consistency and spatial difference of classification result from remotely sensed data at 3 different spatial resolutions, and the difference of landscape pattern revealed by the images. The study results showed: 1) 9 land cover types obtained from remote sensing data were: temperate deciduous broadleaf forest, temperate upland deciduous leaflet forest, single crops + temperate typical steppe, temperate deciduous scrub and low forest, temperate meadow steppe, temperate typical steppe, temperate desert steppe, meadow + graminaceous swamp and water body; 2) the macro spatially pattern of regional land cover revealed by images of 3 different spatial resolutions was consistent, but the boundary of land cover type, the shape and the number of patches of each type differed greatly; 3) comparison of four indices (Fractal dimension, Fragmentation, Diversity, and Dominance) showed that the structure of the landscape changed greatly with change of spatial resolution as revealed by images. Fractal dimension of each cover type showed the most obvious variation, which indicated the change of spatial resolution contributed most to the patch complexity. This paper sets up a basis for further study at different scales on landscape change processes.