Journal of Geo-information Science >
Quality Assessment of Multi-resolution Remote Sensing Land Cover Data: A Case Study in Taoyuan County of Hunan Province
Received date: 2015-04-16
Request revised date: 2015-11-19
Online published: 2016-04-19
Copyright
The quality assessment of remote sensing land cover data is the premise and guarantee of using it reasonably, and it's helpful to improve remote sensing mapping methods. In this study, the 1:100000 land use data of Taoyuan county in 2010 was used as the reference data to validate four different resolution land cover data: RapidEye_5 m, FROM_GLC (30 m), GlobCover2009 (300 m) and MODIS_V005 (500 m). We evaluated the four different resolution land cover data from three aspects, including area relevance and consistency, spatial consistency, and window analysis based on the conversion of classification systems. The results show that: the overall accuracy of RapidEye_5 m data is the highest, MODIS_V005 and FROM_GLC are intermediate, and GlobCover2009 data is relatively lower. The land cover data with higher resolution have a certain superiority for classifications of residence, transportation land, water and other fine material, and the area relevance and the overall consistency of primary types is higher than secondary types. Producer accuracy and user accuracy of the four types of land cover data in crop land, woodland and water is better, while in the construction land and other unused land is lower. Moreover, the spatial consistency of FROM_GLC and MODIS_V005 data is poor in the shrub grassland. The spatial consistency of the four different resolution land cover data is better in the flat areas. Confusion occurred mainly among shrub grassland, woodland and crop land. With the increase of resolution for land cover data, more and more land cover types can be distinguished. The number of land cover types separated from RapidEye_5 m and FROM_GLC (30 m) land cover data focuses on the range of 7-16, in contrast, data with lower resolution focuses on the range of 1-5. Furthermore, data with higher resolution are better to distinguish the grand objects in hilly and mountainous areas.
XU Guangming , YANG Yaping , YANG Fei , JING Wenlong , CHANG Zhongbing . Quality Assessment of Multi-resolution Remote Sensing Land Cover Data: A Case Study in Taoyuan County of Hunan Province[J]. Journal of Geo-information Science, 2016 , 18(4) : 553 -563 . DOI: 10.3724/SP.J.1047.2016.00553
Fig. 1 Location of Taoyuan county图1 湖南省桃源县地理位置 |
Tab. 1 Basic information of land cover data表1 土地覆被数据基本信息 |
数据名称 | 空间分辨率/m | 数据时间/年 | 卫星/传感器 | 分类方法 |
---|---|---|---|---|
RapidEye_5 m | 5 | 2010 | RapidEye | 面向对象 |
FROM_GLC | 30 | 2009-2010 | Landsat/TM、ETM+ | 监督分类 |
GlobCover2009 | 300 | 2009 | ENVISAT/MERIS | 神经网络分类 |
MODIS_V005 | 500 | 2010 | TERRA/MODIS | 决策树 |
Tab. 2 Primary types and secondary types after conversion表2 分类系统转换后的一级类和二级类 |
一级类 | 二级类 |
---|---|
1乔木林地 | 11针叶林、12常绿阔叶林、13落叶阔叶林、14混交林 |
2灌木草地 | 21灌木林、22草地 |
3耕地 | 31农田 |
4水体 | 41水体 |
5建筑用地 | 51居民地、52工业用地、53交通用地 |
6其他未利用地 | 61裸地/裸岩 |
Tab. 3 Conversion of classification systems of land cover data表3 土地覆被分类系统转换 |
分类编码 | MODIS_V005 | GlobCover2009 | FROM_GLC | RapidEye_5 m |
---|---|---|---|---|
1乔木林地 | 常绿针叶林、常绿阔叶林、落叶针叶林、落叶阔叶林、混交林、树林草原 | 常绿阔叶/半落叶林、郁闭常绿针叶林、郁闭落叶阔叶林、混交林 | 阔叶林、针叶林、混交林 | 常绿阔叶林、落叶阔叶林、针叶林、混交林、竹林 |
2灌木草地 | 郁闭灌木林、稀疏灌木林、稀树草原、草原 | 植被(50%~70%)/耕地(20%~50%)镶嵌、灌木/林地(50%~70%)/草地(20%~50%)镶嵌、草地(50%~70%)/林地(20%~50%)/灌木镶嵌、灌木林、草地 | 天然草地 | 草地、茶园、柑橘、灌木林、其他经济林 |
3耕地 | 耕地、农田与自然植被镶嵌 | 水田、旱地、耕地(50%~70%)/植被(20%~50%)镶嵌 | 其他农地、裸农地 | 菜地、水田、旱地 |
4水体 | 水体、永久湿地 | 水体 | 水库/鱼塘、河流 | 水体、沼泽 |
5建筑用地 | 城镇 | 城镇 | 不透水层(高反照率、低反照率) | 居民地、道路、工矿用地 |
6其他未利用地 | - | - | 砾/岩石地 | 裸地、裸岩 |
注:— 表示在研究区没有分出此类型 |
Fig. 2 Comparison of classification effects between land cover data with different resolutions图2 不同分辨率的土地覆被数据产品分类效果对比 |
Fig. 3 Comparison of producer accuracy and user accuracy between different resolution land cover data图3 不同分辨率土地覆被数据生产者精度和用户精度对比 |
Tab. 4 Comparison of area consistency and relevance in the primary types表4 不同分辨率土地覆被数据一级类面积一致性和相关性比较 |
一级土地覆被类型 | 一级类面积一致性系数 | |||
---|---|---|---|---|
RapidEye_5 m | FROM_GLC | GlobCover 2009 | MODIS_V005 | |
乔木林地 | 0.71 | 0.97 | 0.60 | 0.11 |
灌木草地 | 0.88 | 0.01 | 0.94 | 0.004 |
耕地 | 0.74 | 0.22 | 0.36 | 0.66 |
水体 | 0.97 | 0.51 | 0.55 | 0.86 |
建筑用地 | 0.68 | 0.14 | 0.06 | 0.36 |
其他未利用地 | -3.98 | 0.99 | 0.00 | 0.00 |
一级类相关性系数 | 0.9468 | 0.7775 | 0.8272 | 0.7831 |
Tab. 5 Comparison of area consistency and relevance in the secondary types表5 不同分辨率土地覆被数据二级类面积一致性和相关性比较 |
二级土地覆被类型 | 二级类面积一致性系数 | |||
---|---|---|---|---|
RapidEye_5 m | FROM_GLC | GlobCover 2009 | MODIS_V005 | |
针叶林 | 0.65 | 0.09 | 0.10 | 0.55 |
常绿阔叶林 | -7.84 | -1.87 | 0.89 | 0.01 |
落叶阔叶林 | -14.09 | 0.00 | 0.67 | 0.03 |
混交林 | -3.82 | -35.11 | -33.55 | -172.89 |
灌木林 | 0.70 | 0.00 | 0.94 | 0.001 |
草地 | -9.31 | 0.34 | 0.74 | 0.41 |
农田 | 0.74 | 0.22 | 0.36 | 0.66 |
水体 | 0.97 | 0.51 | 0.55 | 0.86 |
居住地 | 0.62 | 0.14 | 0.06 | 0.36 |
工业用地 | 0.70 | 0.00 | 0.00 | 0.00 |
交通用地 | 0.86 | 0.00 | 0.00 | 0.00 |
裸地/裸岩 | -6.49 | 0.99 | 0.00 | 0.00 |
二级类相关性系数 | 0.862 | 0.3656 | 0.5974 | 0.2119 |
Tab. 6 The overall accuracy of each land cover data(%)表6 各土地覆被数据的总体精度(%) |
土地覆被数据 | 总体精度(OA) |
---|---|
RapidEye_5 m | 63.4 |
FROM_GLC | 52.2 |
GlobCover | 48.4 |
MODIS_V005 | 55.1 |
Fig. 4 Statistical chart of classification categories under different windows图4 不同大小窗口各土地覆被数据分类类型统计图 |
Fig. 5 Statistical chart of classification categories under different terrain conditions图5 不同地形条件下各土地覆被数据分类统计图 |
Fig. 6 Spatial distribution of consistency between the four types of land cover data and the reference data图6 4种土地覆被数据与参考数据一致性的空间分布 |
Fig. 7 Confusion analysis between the land cover data with different resolutions and the reference data图7 不同分辨率土地覆被数据与参考数据的空间混淆分析 |
The authors have declared that no competing interests exist.
[1] |
[
|
[2] |
[
|
[3] |
|
[4] |
|
[5] |
|
[6] |
|
[7] |
[
|
[8] |
|
[9] |
|
[10] |
|
[11] |
|
[12] |
|
[13] |
|
[14] |
[
|
[15] |
[
|
[16] |
[
|
[17] |
[
|
[18] |
[
|
[19] |
[
|
[20] |
[
|
/
〈 | 〉 |