Consistency and Accuracy Evaluation Analysis of Four Cropland Cover Products: A Case Study of Indochina Peninsula

Abstract

Information on the area and spatial distribution of cropland resources is very important for agricultural monitoring and sustainable development. The existing evaluation and analysis of cropland resources mainly focus on the first-class cropland types, and the spatial resolution of cropland cover products studied is low (<30 m), which cannot meet the needs of food security assessment and sustainable analysis of cropland resources. Therefore, in this paper, four published cropland cover products with a spatial resolution of 30 m (GFSAD30SEACE, GLC_FCS30-2015, FROM-GLC2015, and SERVIR MEKONG) are evaluated for consistency and accuracy through area validation, comparison of spatial pattern distribution, and fieldwork data. The results showed that (1) the area and spatial pattern of the four first-class cropland cover types were consistent, especially the GFSAD30SEACE product. However, the area and spatial consistency among different products decreased significantly for the finer type of secondary cropland cover, and the area percentage of rice field was only 3.03%. (2) The spatial inconsistency areas of cultivated land for the four products have different distribution rules under different topographic factors, mainly distributed in the range of elevation (<300 m), slope (2°-15°), and topographic relief (0-10 m). (3) The GFSAD30SEACE product had the highest overall accuracy of 88.33% for the primary type of cropland cover. The overall accuracy was lower for all secondary types of cropland cover, with the SERVIR MEKONG product having the highest overall accuracy (71.64%) and the other products having an overall accuracy of less than 62%. Therefore, future land cover mapping of cropland needs to focus on technical solutions for the classification of fine cropland types.