Rubber Plantations in Xishuangbanna: Remote Sensing Identification and Digital Mapping
Rubber Plantations in Xishuangbanna: Remote Sensing Identification and Digital Mapping
复制标题
DOI：
--
复制
发表时间：
2012
期刊：
影响因子：
--
通讯作者：
Zhang Jinghua
中科院分区：
文献类型：
--
作者：
Zhang Jinghua
橡胶园是西双版纳的主要人工景观。橡胶林的提取和动态监测对该地区的经济发展和生态保护具有重要意义。基于MODIS-NDVI数据，分析了橡胶林植被覆盖类型的物候特征，确定了橡胶林遥感检测的时间窗口。根据橡胶林不同生长阶段的光谱差异，采用面向对象的分类方法，利用TM图像进行空间格局判别。我们发现，橡胶检测的最佳时间窗口是从1月至2月或从6月上旬至10月下旬。利用时间序列NDVI分析发现，成熟橡胶林与其他土地覆盖类型差异明显，幼龄橡胶林常与休闲地、茶园混淆。橡胶林的NDVI特征不是唯一的分类特征。基于GPS和视觉采样的总分类精度为85.20%，比基于像元分类的决策树方法高出约5.20%。10年生幼龄橡胶林的准确率为92.50%，10年生成龄橡胶林的准确率为76.42%，两者之比为1.04 ∶ 1。遥感结果与真实的情况非常接近，即民营橡胶园面积大于国有种植园面积。我们提出了一种新的方法来验证和提高橡胶分类的准确性。变化分析表明，成熟胶林的提取方法比幼龄胶林的提取方法更可靠，成熟胶林很少转化为其他土地利用类型。橡胶林采伐与动态监测对橡胶产量估算、空间布局、环境保护和地方政府决策具有重要意义。这是一种动态监测橡胶林的有效方法，可推广到其他多年生植被的提取。
Rubber plantations are the dominant artificial landscape across Xishuangbanna. Extraction and dynamic monitoring of rubber plantations is important to the region's economic development and ecological protection. Based on MODIS-NDVI data we analyzed phenological characteristics of vegetation cover type and determined the temporal window for rubber forest detection. According to spectral differences between rubber forests at different growth stages we used object-oriented classification to discriminate spatial patterns with TM imagery. We found that the optimum temporal window for rubber detection was from January to February or from early June to late October. Using times series NDVI analysis we found that mature rubber forest has obvious differences compared with other land cover types and that young forest rubber was often confused with fallow farmland and tea garden. The NDVI characteristic of rubber forests is not the sole classification feature. Based on GPS and visual sampling total classification accuracy was 85.20%, about 5.20% higher than the decision tree method based on pixel classification. The accuracy of young rubber forest (10 years old) was 92.50% and mature rubber forest (10 years old) was 76.42%, a ratio of 1.04[∶]1. Remote sensing results were very close to the real condition, namely the area of private rubber plantation more than state-owned plantation. We propose a new method for the validation and improvement of rubber classification accuracy. Change analysis showed that the extraction method of mature rubber forest was more credible than for young rubber forest; mature rubber forest seldom transformed into other land use types. Rubber forest extraction and dynamic monitoring is important for rubber yield estimation, spatial distribution, environmental protection and local government decision-making. This is an effective method for monitoring rubber plantations dynamically and can be extended to other perennial vegetation extraction.