景观指数抽样土地覆盖验证点（中国）与GlobeLand30（2010）数据产品比对数据集

为评估土地覆盖数据分类精度，采取一定量的验证点加以评估是目前常用的一种办法。本研究采取景观指数方法（Landscape Shape Index，LSI）在中国境内抽样选取1000个土地覆盖验证点。在这些验证点中，包括耕地类型135个验证点、森林类型142个验证点、草地类型146个验证点、灌木地类型116个验证点、湿地类型62个验证点、水体类型96个验证点、人造地表类型100个验证点、裸地类型124个验证点、冰川和永久积雪类型79个验证点。利用Google Earth高分历史影像判读验证样本，与对应的GlobeLand30-2010数据集进行了比对，并建立比对数据库。比对结果表明，中国区GlobeLand30-2010产品的总体精度为77.90%。该数据集包括：（1）1000个验证点的空间分布数据和属性数据。属性数据中，包括GlobeLand30(2010)产品对应的土地覆盖类型、30m X 30m尺度上的判定类别和300m X 300m尺度上的判定类别等。（2）9种土地覆盖类型的1000个验证点与GlobeLand30 (2010)数据产品对应点比对表，包括300m X 300m Google Earth 遥感影像图。该数据集由18个数据文件组成，存储为.shp、.kmz和.xlsx三种数据格式，数据量为90.2 MB（压缩为2个文件，88.4 MB）。Currently, a commonly used method for evaluating the classification accuracy of land cover data is to utilize a certain number of validation points. In this study, we selected 1000 land cover validation points through sampling across China using the Landscape Shape Index (LSI) method. Among these validation points, there are 135 for cropland, 142 for forest, 146 for grassland, 116 for shrubland, 62 for wetland, 96 for water bodies, 100 for artificial surfaces, 124 for bare land, and 79 for glaciers and permanent snow cover. Using high-resolution historical imagery from Google Earth, we interpreted the validation samples, compared them with the corresponding GlobeLand30-2010 dataset, and established a comparison database. The comparison results indicate that the overall accuracy of the GlobeLand30-2010 product for the Chinese region is 77.90%. This dataset includes: (1) Spatial distribution data and attribute data for the 1000 validation points. The attribute data covers the land cover type corresponding to the GlobeLand30 (2010) product, the classification results at the 30m × 30m scale, and the classification results at the 300m × 300m scale, etc. (2) Comparison tables between the 1000 validation points of the 9 land cover types and their corresponding points in the GlobeLand30 (2010) data product, including 300m × 300m Google Earth remote sensing imagery. This dataset consists of 18 data files, stored in three formats: .shp, .kmz, and .xlsx, with a total uncompressed size of 90.2 MB (compressed into 2 files, 88.4 MB).

搜集汇总

背景与挑战

背景概述

该数据集是一个用于评估Globeland30（2010）数据产品在中国区域土地覆盖精度的验证数据集。它基于景观指数抽样方法，包含了1000个验证样本点，覆盖9种土地覆盖类型，并计算出Globeland30（2010）的总体精度为77.90%。数据集提供样本点的地理属性数据和与Globeland30的对比数据，格式多样，总大小为90.2 MB。