基于全球遥感产品的森林概念协同  

Forest definitions collaboration based on global remote sensing data products

作　　者：胡涛 彭建[1] 董建权 肖善才 夏沛 HU Tao;PENG Jian;DONG Jianquan;XIAO Shancai;XIA Pei(Laboratory for Earth Surface Processes,Ministry of Education,College of Urban and Environmental Sciences,Peking University,Beijing 100871,China;College of Landscape Architecture,Beijing Forestry University,Beijing 100083,China)

机构地区：[1]北京大学城市与环境学院地表过程分析与模拟教育部重点实验室,北京100871 [2]北京林业大学园林学院,北京100083

出　　处：《地理学报》2024年第5期1115-1128,共14页Acta Geographica Sinica

基　　金：宁夏回族自治区重点研发计划(2021BEG03019)[Foundation:Key Research and Development Program of Ningxia,No.2021BEG03019]。

摘　　要：森林的准确界定和高精度森林遥感数据产品是监测和评估森林变化的重要基础。目前存在多种森林定义,基于不同定义的全球森林遥感产品亦层出不穷。本文基于土地利用和土地覆被视角,梳理不同国际环境及林业组织对森林的定义,从生长方式、林龄和郁闭度3个方面开展森林分类,梳理从单一传感器到光学与微波遥感融合的森林遥感数据集演化。从森林要素阈值、分类详尽程度和空间分辨率3个方面对比不同遥感数据集的森林定义差异,并基于验证样本、区域精度差异及数据一致性总结产品精度验证不足。认为现有森林遥感数据集的定义及精度差异制约了数据使用者对于森林变化监测。未来应基于“视角—要素—阈值”森林定义框架进一步协同森林定义,减少由于森林定义要素阈值差异导致的面积估计偏差;结合多源遥感数据及深度学习方法开发精细化森林遥感数据,并关注不同林种识别;建设森林遥感数据共享平台,通过地图可视化方式向数据使用者展示全球森林遥感数据定义、时空分辨率及在不同地区的数据精度,明晰多源森林遥感产品适用范围,提升数据可使用性。Forest definitions and remote sensing datasets provide a conceptual basis for monitoring forest change.In this study,we present an overview of forest definitions from the views of land use and land cover,introduce the forest categorization from three aspects:growth mode,forest age and canopy density,and review the evolution of forest remote sensing dataset from single sensor to optical and microwave remote sensing.Additionally,the differences in forest definition between various remote sensing data sets were compared from three perspectives:the threshold for forest elements,the level of classification and the spatial resolution.The shortcomings in product accuracy verification were summarized using data consistency,validation samples,and regional accuracy differences.In the future,the forest definition should be further coordinated based on the forest definition framework of"perspective-factor-threshold",and the area estimation bias caused by the different factor thresholds in the forest definition should be minimized.Meanwhile,deep learning and multisource remote sensing data should be applied to produce accurate forest remote sensing data sets,especially for identifying various forest species.Finally,platforms for forest remote sensing datasets sharing need to be built to clarify the forest definition,spatio-temporal resolution,and data accuracy of the datasets.

关 键 词：森林定义 森林遥感数据产品 精度验证 适用范围 

分 类 号：S771.8[农业科学—森林工程] P237[农业科学—林学]