Accuracy assessment of multi-source land use products in the loess hilly and gully region

State Key Laboratory of Urban and Regional Ecology,Research Center for Eco-Environmental Sciences,Chinese Academy of Sciences

Clc Number:


Fund Project:

  • Article
  • |
  • Figures
  • |
  • Metrics
  • |
  • Reference
  • |
  • Related
  • |
  • Cited by
  • |
  • Materials
  • |
  • Comments

    Evaluating the accuracy of multi-source land use products is essential for obtaining reliable surface information, supporting land planning and management, and promoting regional ecological protection and high-quality development. However, the capability of existing land use products to accurately depict the surface cover characteristics in complex terrains or fragmented habitats, such as the loess plateau hilly and gully regions, remains uncertain. Taking the Guanchuan River Basin, a secondary tributary of the Yellow River, as an example, this study assessed the accuracy of six land use products with spatial resolutions of 10 m and 30 m (WorldCover 10 m, ESRI 10m, GLC_FCS30-2020, GlobleLand30, CNLUCC, and CLCD) in the loess hilly and gully region using high-precision GCLUCC land use data. This GCLUCC data, with an overall accuracy exceeding 95%, was derived from GF-2 (0.8m), DEM (5m), and 6400 field sampling points, employing the object-oriented method and manual visual interpretation. The evaluation results showed that (i) In terms of classification characteristics, most products could extract main land categories, yet significant differences existed in extraction efficiency and accuracy, especially regarding the spatial distribution of terraced fields, forest lands, and construction lands; (ii) For area consistency, there were significant discrepancies in land category areas between various products and GCLUCC. For instance, the grassland area in some products was more than twice that of GCLUCC, while forest and water areas accounted for only 0.13%-12.11% and 1.03%-5.86% of GCLUCC, respectively; (iii) In terms of overall accuracy, GlobleLand30 and WorldCover 10m demonstrated relatively higher accuracy, reaching 58.21% and 50.19%, respectively. The accuracy of CLCD and CNLUCC was comparatively lower, with notable classification confusion between forests and grasslands, terraced fields and grasslands, and construction lands and terraced fields; (iv) Significant spatial discrepancies existed between various products and the actual ground surface, particularly in accurately classifying forests, shrubs, bare land, and water bodies. In conclusion, current land use products still face notable challenges in precisely describing surface cover characteristics in the loess plateau hilly and gully regions. Future product development should place greater emphasis on topographical and geographical features and strengthen the recognition of specific land use types like forests, shrubs, bare land, and water bodies to enhance data accuracy and decision-making reliability.

    Cited by
Get Citation
Article Metrics
  • Abstract:
  • PDF:
  • HTML:
  • Cited by:
  • Received:October 31,2023
  • Revised:December 25,2023
  • Adopted:December 27,2023
  • Online: April 29,2024
  • Published: