Mapping tropical forest cover and deforestation using synthetic aperture radar (SAR) images
Mapping tropical forest cover and deforestation using synthetic aperture radar (SAR) images
复制标题
DOI：
10.1007/s12518-010-0026-9
复制
发表时间：
2010-07
影响因子：
2.7
通讯作者：
M. Rahman;J. Sumantyo
中科院分区：
文献类型：
--
作者：
M. Rahman;J. Sumantyo
世界上许多地区森林覆盖的变化导致大气碳的积累增加，从而加速了全球变暖的进程。光学遥感已被用于绘制毁林图和量化毁林情况，但由于图像上有云覆盖，其应用受到限制。最近几次星载合成孔径雷达飞行任务的可用性扩大了利用雷达图像监测森林覆盖变化的范围。这项调查的目的是审查合成孔径雷达数据评估和测绘毁林情况的能力。研究区位于孟加拉国东南部的热带森林地区。本研究使用了1994年航天飞机成像雷达-C（SIR-C）数据和2007年高级陆地观测卫星（ALOS）PALSAR数据。利用航天飞机雷达地形使命数字高程模型数据对ALOS PALSAR数据进行了正射校正。在两个SAR场景之间进行图像到图像的几何配准。修剪研究区域并将其作为子集分离。SIR-C数据（L和C波段）为双极化（HH和HV），PALSAR（L波段）为四极化（HH、HV、VH和VV）。在两种合成孔径雷达场景中，都识别出五种不同的土地覆盖类型（森林、高地土壤/灌木、低地土壤、住区和水/湿地）。一个额外的类代表森林再生长只能在SIR-C图像上识别。这两个图像进行分类使用最大似然算法。从随机选择的独立验证像素计算分类精度。除PALSAR图像的用户精度外，森林的精度都在83%以上。在研究期间，该地区的森林面积从18 000公顷减少到13 800公顷。这项研究的结果将有助于了解SAR的适用性地图和量化的热带地区的森林覆盖变化。
The changes in forest cover in many parts of the world lead to increase the accumulation of atmospheric carbon and thus accelerate the process of global warming. Optical remote sensing has been used to map and quantify deforestation but its application is limited because of the presence of cloud coverage on the images. Recent availability of several space-borne synthetic aperture radar (SAR) missions has widened the scope of utilizing radar images for monitoring of forest cover change. The objective of this investigation is to examine the capability of SAR data to assess and map deforestation. The study area is located at the tropical forest region of southeastern Bangladesh. Shuttle Imaging Radar-C (SIR-C) data of 1994 and Advanced Land Observation Satellite (ALOS) PALSAR data of 2007 were used in this study. ALOS PALSAR data were orthorectified with Shuttle Radar Topographic Mission digital elevation model data. Image to image geometric registration was done between the two SAR scenes. Study area was clipped and separated as subsets. SIR-C data (L- and C-bands) was in dual polarization (HH and HV) and PALSAR (L-band) was in quad-polarization (HH, HV, VH, and VV). Five different categories of land covers (forest, upland soil/shrubs, lowland soil, settlements, and water/wetlands) were recognized on both SAR scenes. An additional class representing forest re-growth could be identified only on SIR-C image. Both the images were classified using maximum likelihood algorithms. The classification accuracy was computed from the randomly selected independent validation pixels. The accuracy for forest is more than 83% except users accuracy computed for PALSAR image. Forest was reduced from 18,000 to 13,800 ha in the region during the study period. The results of this study will be useful for understanding the applicability of SAR to map and quantify forest cover changes in the tropics.