Large-extent Classification of Wetland and Management Types Using Satellite Remote Sensing and Auxiliary Data

There have been few studies examining how effectively land use / land cover (LULC) classes of wetland types and their management can be detected with remote sensing and auxiliary data. We conduct wetland-specific present-day LULC classification and long-term change detection in seven European study areas. We assess how accurately LULC types and their changes can be detected with satellite remote sensing and to what extent auxiliary data improves classification accuracy. We conduct 10 m spatial resolution 27-class LULC classifications using Sentinel-1 and -2 satellite imagery and auxiliary GIS data about soil, LULC, and topography. For the change detection, we distinguish 8 LULC classes between 1985 – 2023 using Landsat imagery at a 30-meter resolution. The classification accuracy for the present-day analysis varies between 52–93 %, with the omission of auxiliary datasets reducing accuracy by up to 41 %-points. The change detection classifications have 73–94 % agreement with the Corine Land Cover data. Our results indicate that wetland and management types can be classified using 10-meter imagery, but the use of good quality auxiliary data is required to improve the classification accuracies. Additionally, 30-meter LULC changes are detectable using Landsat imagery. The developed approach allows for quick and straightforward implementation of LULC classification and change detection over large areas.