CREAF has provided ecological and remote sensing expertise to the ALFAwetlands project, applying techniques that improved the spatial knowledge of the wetlands in the project’s Living Labs. In Work Package 1, CREAF developed algorithms to map land cover in Living Labs using Sentinel-2 data and generated 30-year land cover change maps with Landsat data. In Work Package 3, CREAF applied machine learning models to track environmental changes, linking ALFAwetlands field data with satellite spectral and thermal observations.
Mapping land cover and land cover change
CREAF has been involved into the mapping of land covers in the ALFAwetlands’ Living Labs. Using machine learning, CREAF’s team created a high-resolution land-cover map at 10-meter spatial resolution, including 27 land cover types. The map was created based on Copernicus Sentinel-1 and -2 satellite data, two missions from the EU Copernicus programme. Furthermore, CREAF was involved into the creation of land cover change maps depicting the changes in the Living Labs that occurred from 1985 to 2023. The land cover change maps were created using Landsat imagery, with a coarser 30-meter spatial resolution.
The results show that peatlands and mineral wetlands, along with their management practices, can be accurately classified using Sentinel imagery and auxiliary data. GIS data on peat soils and land management are crucial for achieving the highest accuracy. In some areas, classification accuracy was low given the complexity of the land cover types within the Living Labs’ wetlands. Despite this, the method developed by CREAF offers a streamlined approach for land cover classification and change detection across large regions.
Please also read: Advancing wetland type and management mapping: Machine learning unlocks insights for peatlands and wetlands in Europe
Bridging field and satellite data insights for environmental monitoring
CREAF has also carried out several key tasks regarding the linking of field data, collected by other project partners, and linking them with satellite measurements. CREAF’s team extracted Landsat data corresponding to the spatial and temporal locations of field measurements, including field variables such as Gross Primary Productivity, GHG fluxes, soil temperature, and soil humidity. Then, CREAF trained a machine learning model based on the field data and satellite observations. Such model can predict environmental variables across the Living Labs and at a continental scale. These algorithms will enable the production of high-resolution, long-term Gross Primary Productivity (GPP) maps for the Living Labs based on Landsat data.
This post was prepared and graphs provided by CREAF team – ALFAwetlands partner.