| 摘 要： |
Accurate information on cropland area is essential for agricultural development and planning. Here, we evaluated the performance of six land cover products (CLCD, MCD12Q1, CCI-LC, GLC-FCS, CNLUCC, and CACD) in identifying cropland extent in a large irrigation district of China from 1995 to 2020 based on visual interpretation samples. The results indicate that CACD performs best with an average kappa coefficient (KC) of 0.89, followed by CLCD (KC = 0.87), GLC-FCS (KC = 0.75), CNLUCC (KC = 0.65), MCD12Q1 (KC = 0.49), and CCI-LC (KC = 0.29). Additionally, the cropland area provided by statistical yearbooks is significantly lower than that identified by CACD, with an average underestimation of -34%. We also find that these land cover products exhibit poor consistency in identifying cropland. The average percentage of grids labeled as completely consistent-where all six products identify those grids as cropland-is only 16.0% across the entire irrigation district, highlighting the uncertainty of existing land cover products in identifying cropland areas. |
在线留言
联系我们
所长信箱