基于GEE的东南亚主产区橡胶林分布遥感提取

GEE-based Extraction of Rubber Forest Distribution in Main Producing Areas of Southeast Asia

为遥感监测全球橡胶主产区橡胶的长势,选取橡胶产量最大的泰国、马来西亚、印度尼西亚三国,开展基于Google Earth Engine(GEE)的橡胶林分布遥感提取研究。通过目视解译选择典型样本区,根据Landsat 7多波段光谱特征、MODIS NDVI反映的植被物候特征建立分类回归树CART分类模型提取橡胶林分布。精度评价显示模型总体分类精度为95.8%,Kappa系数为0.94,生产者精度达到94.8%,用户精度为88.2%,达到较高水平。模型提取结果显示:橡胶林在泰国中部、南部半岛,马来半岛的东部和南部地区,苏门答腊岛分布较为集中,而泰国北部、加里曼丹岛及其他岛屿橡胶林相对稀疏。

In order to monitor the growth of rubber forest in the main rubber foresttree producing areas of the world by remote sensing, we carried out the remote sensing extraction of rubber forest distribution based on Google Earth engine(GEE) in Thailand, Malaysia and Indonesia, which are the three largest rubber producing countries in the world. The typical sample areas were selected through visual interpretation, and the classification regression tree CART classification model was established based on the multi-band spectral albedo of Landsat 7 and the vegetation phenological characteristics reflected by MODIS NDVI. The overall classification accuracy of the model was 95.8%, the Kappa coefficient was 0.94, the producer accuracy was 94.8%, and the user accuracy was 88.2%, reaching a high level. The extraction results show that rubber plantations are concentrated in central and southern peninsula of Thailand, the eastern and southern parts of Malay Peninsula and Sumatra Island, while rubber plantations are relatively sparse in northern Thailand, Kalimantan Island and other islands.