Estimation of the stand ages of tropical secondary forests after shifting cultivation based on the combination of WorldView-2 and time-series Landsat images
Estimation of the stand ages of tropical secondary forests after shifting cultivation based on the combination of WorldView-2 and time-series Landsat images
复制标题
DOI：
10.1016/j.isprsjprs.2016.06.008
复制
发表时间：
2016-09-01
影响因子：
12.7
通讯作者：
Kitayama, Kanehiro
中科院分区：
文献类型：
--
作者：
Fujiki, Shogoro;Okada, Kei-ichi;Kitayama, Kanehiro
在婆罗洲山区，当地人进行传统的轮作耕作，保护区周围是不同年龄的恢复次生植被斑块，研究了一种估算次生植被林龄的新方法。在景观水平上确定林分年龄对改善保护政策至关重要。我们将高分辨率卫星图像（WorldView-2）与时间序列陆地卫星图像相结合。我们从Landsat时间序列图像的变化检测分析中提取了林分年龄（从最近一次砍烧以来经过的时间），并将所得的林分年龄叠加到使用WorldView-2进行基于对象的图像分析分类的片段上。我们将林分年龄作为响应变量，将基于对象的指标作为自变量，建立解释林分年龄的回归模型。随后，我们利用回归模型和线性判别分析将目标区域的植被划分为6个年龄单元和1个橡胶林单元（1-3年、3-5年、5-7年、7-30年、30-50年、bb0 -50年和“橡胶林”）。验证表明准确率为84.3%。我们的方法在将小于7年的高动态先锋植被划分为2年的间隔方面特别有效，这表明可以高精度地检测到植被冠层的快速变化。结合光谱时间序列分析和基于高分辨率影像的物象指标，实现了集约轮作下动态植被的分类，并生成了基于林龄的信息丰富的土地覆盖图。(C) 2016国际摄影测量与遥感学会（ISPRS）Elsevier B.V.版权所有。
We developed a new method to estimate stand ages of secondary vegetation in the Bornean montane zone, where local people conduct traditional shifting cultivation and protected areas are surrounded by patches of recovering secondary vegetation of various ages. Identifying stand ages at the landscape level is critical to improve conservation policies. We combined a high-resolution satellite image (WorldView-2) with time-series Landsat images. We extracted stand ages (the time elapsed since the most recent slash and burn) from a change-detection analysis with Landsat time-series images and superimposed the derived stand ages on the segments classified by object-based image analysis using WorldView-2. We regarded stand ages as a response variable, and object-based metrics as independent variables, to develop regression models that explain stand ages. Subsequently, we classified the vegetation of the target area into six age units and one rubber plantation unit (1-3 yr, 3-5 yr, 5-7 yr, 7-30 yr, 30-50 yr, >50 yr and 'rubber plantation') using regression models and linear discriminant analyses. Validation demonstrated an accuracy of 84.3%. Our approach is particularly effective in classifying highly dynamic pioneer vegetation younger than 7 years into 2-yr intervals, suggesting that rapid changes in vegetation canopies can be detected with high accuracy. The combination of a spectral time-series analysis and object-based metrics based on high-resolution imagery enabled the classification of dynamic vegetation under intensive shifting cultivation and yielded an informative land cover map based on stand ages. (C) 2016 International Society for Photogrammetry and Remote Sensing, Inc. (ISPRS). Published by Elsevier B.V. All rights reserved.