摘要
以最近发布的3种30 m分辨率的全球土地覆盖类型产品(GlobeLand30(2020)、GLC—FCS30(2020)和FROM—GLC30(2017))为研究对象,通过野外调查获取的福建省8170个土地覆盖类型样本,基于混淆矩阵、误差组分(数量、交换、转移)和强度分析结合的方法评价3种产品的精度,并分析3种产品的精度随地形(高程、坡度、坡向)的变化特征.研究结果表明:GlobeLand30、GLC-FCS30、FROM—GLC30在福建省的总体精度分别为70.13%、77.01%、82.02%.3种产品的耕地、林地、水体和人造地表分类精度相对较高(≥60%);裸地、草地、灌木精度相对较低(≤31%).GlobeLand30的数量误差组分最大(21.02%),GLC—FCS30的交换误差最大(13.46%),FROM—GLC30的转移误差最大(8.89%);3种产品的耕地和裸地的数量误差组分强度高于整体的数量误差组分强度;灌木、水体和人造地表的交换误差组分强度高于整体的交换误差组分强度;水体的转移误差组分强度高于整体的转移误差组分强度.3种产品的耕地和林地的精度随DEM和坡度的变化比较明显,林地的精度随着DEM和坡度增加而增加,耕地随着坡度的增加而降低.
Three recently released 30 m resolution Global Land Cover(GLC)datasets(GlobeLand30(2020),GLC—FCS30(2020)and FROM—GLC30(2017))were assessed by 8170 surveyed samples in Fujian Province.The accuracy of the three products was evaluated by calculating the confusion matrix,error components(Quantity,Exchange,Shift),and error components intensity.The variation of the accuracy changed with topography(elevation,slope,and aspect)was analyzed.The results showed that:The overall mapping accuracies of GlobeLand30,GLC-FCS30,and FROM—GLC30 in Fujian Province were 70.13%,77.01%,and 82.02%,respectively.The accuracies of Cultivated,Forest,Water and Artificial were relatively high(≥60%).The accuracies of Barren,Grassland and Shrub were relatively low(≤31%).GlobeLand30,GLC—FCS30,and FROM—GLC30 had the highest errors in Quantity(21.02%),Exchange(13.46%)and Shift(8.89%),respectively.The Quantity error intensity of the Cultivated and Barren for three GLC maps was higher than their Quantity overall error intensity.The Exchange error intensity of the Shrub,Water and Artificial for three GLC products was higher than their Exchange overall error intensity.The Shift error intensity of the Water for three GLC maps was higher than their Shift overall error intensity.The accuracy of Forest increased with the increase of DEM and slope,and the Cultivated decreased with the increase of slope.
作者
付东光
李登秋
陆灯盛
陈耀亮
FU Dongguang;LI Dengqiu;LU Dengsheng;CHEN Yaoliang(Institute of Geography,Cultivation Base of State Key Laboratory for Subtropical Mountain,Fujian Normal University,Fuzhou 350117,China)
出处
《福建师范大学学报（自然科学版）》
CAS
2023年第2期122-136,共15页
Journal of Fujian Normal University：Natural Science Edition
基金
国家自然科学基金资助项目(41701490)
福建省自然科学基金资助项目(2022J01640)
福建省科技厅公益类科研院所专项(2021R1002008)。
作者简介
通信作者:李登秋(1985-),男,副教授,研究方向为连续土地利用与覆被分类、时间序列变化检测和陆地生态系统碳循环,lidengqiu@fjnu.edu.cn。
