Privacy Policy Disclaimer
  Advanced SearchBrowse




Journal Article

Improving the characterization of global aquatic land cover types using multi-source earth observation data


Xu,  Panpan
External Organizations;

Tsendbazar,  Nandin-Erdene
External Organizations;


Herold,  Martin
0 Pre-GFZ, Departments, GFZ Publication Database, Deutsches GeoForschungsZentrum;

Clevers,  Jan G. P. W.
External Organizations;

Li,  Linlin
External Organizations;

External Ressource
No external resources are shared
Fulltext (public)
There are no public fulltexts stored in GFZpublic
Supplementary Material (public)
There is no public supplementary material available

Xu, P., Tsendbazar, N.-E., Herold, M., Clevers, J. G. P. W., Li, L. (2022): Improving the characterization of global aquatic land cover types using multi-source earth observation data. - Remote Sensing of Environment, 278, 113103.

Cite as: https://gfzpublic.gfz-potsdam.de/pubman/item/item_5012035
The sustainable management of aquatic resources requires spatially explicit information on the water and vegetation presence of aquatic ecosystems. Previous Global Aquatic Land Cover (GALC) mapping has been focused on water bodies while lacking information on vegetation, and aquatic types have always been characterized by low accuracies in global land cover products, calling for specific attention to improve GALC mapping. The availability of a wealth of open Earth Observation (EO) data on cloud-computing platforms provides opportunities to map aquatic land cover globally. This study aims to evaluate the potential of multi-source freely available EO data, including optical, Synthetic Aperture Radar (SAR), and various ancillary datasets, for improving the characterization of aquatic land cover comprising both water and vegetation types on a global scale. Using different combinations of features derived from these data, the classification performance of five land cover classes (i.e., trees, shrubs, herbaceous cover, bare/sparsely vegetated lands, and water bodies) in aquatic areas was cross-validated. Results showed that Sentinel-2 data alone achieved similarly good overall accuracy as those combining multi-source data. However, the single-sensor Sentinel-2 data cannot discriminate highly mixed and spectrally similar types, such as shrubs, trees, and herbaceous vegetation. Integrating SAR features from the ALOS/PALSAR mosaic and Sentinel-1 data with optical features provided by Sentinel-2 data could help address this limitation to some extent. Although with a lower spatial and temporal resolution, the ALOS/PALSAR mosaic had a stronger impact on GALC classification than Sentinel-1 data when they were synergistically used. Features provided by ancillary datasets did not lead to significant improvement in the overall GALC classification. At class-level, topographic and soil features helped to reduce the commission error of shrubs, and climate variables were useful to improve the characterization of bare aquatic lands. The Global Ecosystem Dynamics Investigation (GEDI) forest canopy height dataset helped to characterize trees but also resulted in a decrease in accuracies of shrubs. By assessing multi-source earth observation data, this research represents an important step forward in the global mapping of comprehensive aquatic land cover types at high spatial resolution (i.e., 10 m).