PhD defence: Gyula Mate Kovács

Gyula Mate Kovács defends his thesis, 

Remote Sensing in Wetland Studies
The application of Earth Observation for detecting change, mapping carbon storage, and assessing land use impact in wetlands – A synopsis.

Supervisors:
Professor Rasmus Fensholt, IGN
Associate Professor Stéphanie Horion, IGN

Assessment Committee:
Professor Susan E. Page, SGGE, University of Leicester, UK
Professor Sebastian van der Linden, Department of Geography and Geology, University of Greifswald, DE
Associate Professor Alexander Prishchepov (chair), IGN

Summary:
Globally, despite their crucial role in providing essential ecosystem services, wetlands face degradation and a 16–23% historic reduction in area, resulting in diminished carbon storage and increased greenhouse gas emissions through land use. The scarcity of wetland datasets underscores the necessity of Earth Observation (EO) to bridge these gaps. This thesis showcases EO applications across spatial scales, from the Global South to Europe, demonstrating its effectiveness in delivering wetland inventories and trend monitoring. The aim is to provide domestic and continental policy-relevant information for wetland conservation, emphasizing preserving ecological services and the prevention of additional emissions from land use changes and practices.
In arid regions, the world's largest floodplain wetlands face threats from anthropogenic disturbances. In Africa, wetlands, which are vital for water, nutrients, and local populations, are endangered by economic motives. Applying a piece-wise regression model to the Inner Niger Delta wetlands, using 30-meter resolution Landsat data, reveals significant long-term changes, exceeding the MODIS accuracy. Moreover, creating reference datasets for change detection products becomes laborious in the absence of baseline wetland maps, underscoring the need for detailed inventories in future change detection efforts.
In Europe, as wetland loss is concentrated in industrialized regions, the European Union aims to restore 30% of degraded wetlands by 2030. Utilizing 10-meter satellite data, CORINE Land Cover annotations, and machine learning, our study mapped six wetland types across Europe with 94±0.5% accuracy in 2018. Despite carbon storage losses to human disturbances up to 26%, our findings reveal a substantial wetland carbon pool, estimated at 11.07 Gt to 49.09 Gt of CO2 equivalent, highlighting uneven restoration needs and opportunities across European countries.
In Denmark, nitrous oxide emissions linked to land use, especially in drained croplands, were quantified using a deep-learning approach on 3-meter PlanetScope imagery across Zealand. These depressions in cultivated fields, covering less than 1% of the total cropland area, released 80 times more nitrous oxide, accounting for 30 ± 1% of the nitrous oxide 8 budget post-fertilization. Urgent action is needed to assess and mitigate these hotspots in managed croplands. In conclusion, this thesis aims to contribute to wetland conservation's spatially explicit information needs. EO and machine learning provide a universal perspective, offering tailored tools to address wetland challenges. This synthesis provides a comprehensive overview of current knowledge and the state of wetland research, highlighting its interconnections with sustainable development, Earth Observation, machine learning, biogeochemistry, and environmental policy. It serves as a practical solution for the global preservation and sustainable management of wetlands, addressing gaps in our understanding and promoting effective wetland conservation and restoration practices.

A digital version of the PhD thesis can be obtained from the PhD secretary at phd@ign.ku.dk