多源土地利用/覆被产品在黄土高原应用和精度评估

为准确掌握黄土高原土地利用/覆被的时空格局，基于6种不同来源的土地利用/覆盖数据产品，MCD12Q1,ESA-CCI-LC,GlobeLand30,GLASS,CAS-LUCC和ChinaCover,开展了黄土高原地区退耕还林(草)实施以来土地利用/地表覆被时空格局和转类特征的研究，对不同来源数据产品的精度进行了验证分析。结果表明：(1)分类特征和变化趋势上，MCD12Q1和GLASS产品与其他产品的差异最大，其次为ESA-CCI-LC产品，GlobeLand30,CAS-LUCC和ChinaCover产品的差异相对较小。(2)除MCD12Q1产品外，不同来源土地利用/覆被产品各类型面积随时间变化趋势基本一致。(3)空间转类上，GLASS产品与其他产品的差异最大，解译类型明显错误；其次为MCD12Q1产品，转林类和转草类也存在明显错误，退耕还林(草)特征表现不明显；ESA-CCI-LC和GlobeLand30在黄土丘陵沟壑区退耕还林(草)的转类特征表现不足；CAS-LUCC和ChinaCover退耕还林(草)特征明显。(4)抽样精度上，GlobeLand30,CAS-LUCC,Chin...

Application and Accuracy Evaluation of Multi-source Land Use/Cover Products on the Loess Plateau

To accurately capture changes in spatial and temporal patterns of land use/cover on the Loess Plateau, based on six different sources of land use/cover data products, MCD12Q1, ESA-CCI-LC, GlobeLand30, GLASS, CAS-LUCC and ChinaCover, a study on the spatial and temporal patterns and transhape characteristics of land use/surface cover in the Loess Plateau region since the implementation of fallowing to forest(grass)was carried out, and the accuracy of data products from different sources was verified and analyzed. The results showed that:(1)in terms of classification characteristics and variation trends, MCD12Q1 and GLASS products were the most different from other products, followed by ESA-CCI-LC products, GlobeLand30, CAS-LUCC and ChinaCover products.(2)except for MCD12Q1 product, the trend of area change over time was generally consistent for all types of land use/cover products from different sources.(3)in terms of spatial classification, the GLASS product was the most different from other products, and the interpretation type was obviously wrong; the second was MCD12Q1 products; there were also obvious errors in the conversion of forest and grass, and the features of fallow forest(grass)were not obvious; ESA-CCI-LC and GlobeLand30 showed insufficient characteristics of conversion from farmland to forest(grassland)in the loess hilly-gully region; CAS-LUCC and ChinaCover had obvious characteristics of returning farmland to forest(grassland).(4)In terms of sampling accuracy, GlobeLand30, CAS-LUCC and ChinaCover products had the highest overall accuracy, which were 86.61%(K=0.84), 91.68%(K=0.90)and 89.45%(K=0.87), respectively. followed by ESA-CCI-LC(73.83%, K=0.69); the accuracies of MCD12Q1 and GLASS products was the lowest, which were 59.43%(K=0.51)and 36.31%(K=0.24), respectively.(5)the fusion of land use/cover multi-source data products was beneficial to improve the accuracy of the products; compared with the pre-fusion products GlobeLand30, CAS-LUCC and ChinaCover, and the total accuracy of the fusion products could be increased by 2.25%～7.30%. In conclusion, the accuracy of CAS-LUCC product has the highest accuracy, and the changes in the conversion of farmland to forest(grass)from 2000 to 2010 are more consistent with the changes in the vegetation pattern of the Loess Plateau.