2套30 m分辨率土地覆被遥感数据类别精度及空间一致性特征研究

The Category Accuracy and Spatial Consistence Characteristic Analysis for Two 30 m Resolution Land Cover Products in China

摘要: 土地覆被数据是自然资源管理、城市规划、气候模拟、环境保护建模等领域的重要信息，对其进行精度评价并揭示其类别混淆特征对于众多科学领域具有重要意义。本文以中国研发的2套30米分辨率全球尺度土地覆被数据Globeland30和FROM-GLC-Hierarchy数据为研究对象，以国际组织发布的验证数据为参考，从国家、区域及类别尺度进行了比较和评价，探索了空间一致性随地形的变化特征。结果表明：2套数据的林地、耕地、水体三种类别均具有较高的分类精度，Globeland30的草地具有较高的用户精度和制图精度，林业、农业及水资源领域研究可以选择二者作为基础数据源，草地科学研究可以选择Globeland30作为基础数据源。2套数据的类别空间不一致性与高程具有典型的正相关，其不一致比例随着高程的增加明显增大，在1500 m以上区域，二者的类别不一致性比例最大，达到了42.91%；二者类别不一致性区域主要位于2~15度之间，其面积接近占60%，2度以下区域不一致性最低，约占7.09%。研究表明地形条件是影响大尺度土地覆被制图的重要因素，因此，未来应进一步加强高海拔及景观异质性区域土地覆被分类算法的研发。

Abstract: Land cover map is vital for research and applications concerning natural resource and environ-mental modeling, so assessment of their category accuracy and category confusion is very important for some specific applications. Using the existing referenced data from international organization, we compared the category accuracy and spatial consistency from national and regional scale, and explore the change characteristic of spatial heterogeneity. The results show that, forest, cropland and water category all have the high Producer's Accuracy and User’s Accuracy, so these two data can be as input land cover data for forest, cropland and water scientific regions. The category heterogeneity shows a significant positive correlation with the elevation; above 1500 m, the percent of category inconformity is up to 42.91%, and inconformity mainly appears in 2~15 degrees, with areas occupying nearly 60%; below 2 degrees, the percent of category inconformity is minimum, only 7.09%. In the future, we should put more attention on the classification algorithm in heterogeneous areas.