EGU23-5352, updated on 22 Feb 2023
EGU General Assembly 2023
© Author(s) 2023. This work is distributed under
the Creative Commons Attribution 4.0 License.

Mapping vegetation cover on rewetted fen peatlands using hyperspectral spaceborne images from DESIS and PRISMA

Arasumani Muthusamy, Fabian Thiel, Vu Dong Pham, Christina Hellmann, and Sebastian van der Linden
Arasumani Muthusamy et al.
  • Institute of Geography and Geology, University of Greifswald, Partner in the Greifswald Mire Centre, Friedrich-Ludwig-Jahn-Straße 16, 17489 Greifswald, Germany

Undisturbed peatlands constitute relevant carbon sinks. However, drained or degraded peatlands cause carbon emissions and more than 90 % of the peatlands in Mecklenburg-Western Pomerania, Germany, have been drained. In order to achieve the goals for climate change mitigation, several initiatives have been taken to restore drained and degraded peatlands by rewetting. Previously, drained peatlands were used for agricultural and grassland production, which led to significant carbon emissions. The Paludiculture scheme was introduced to allow sustainable and climate-friendly agricultural production under permanently wet conditions, which allow carbon sequestration of the soil. Phragmites australis (Reed) and Typha spp. (Cattail) are key plants for paludicluture in rewetted areas of north-east Germany.

It is necessary to regularly monitor the plant communities in rewetted areas, as this can be an indicator of rewetting success. Distinguishing peatland vegetation communities require high spectral resolution images, whereas multispectral images may show comparable spectral signatures in different peatland vegetation types. Recently launched hyperspectral sensors offer new possibilities regarding accurate vegetation monitoring in rewetted peatlands. In this study, we investigated multi-date hyperspectral images from DESIS, mounted on the International Space Station (ISS), and satellite-based PRISMA for peatland vegetation mapping. In addition to the increased spectral information, both sensors allow multiple observations per year, which was not the case for airborne hyperspectral data. The 30-m spatial resolution of both sensors, however, brings along multiple peatland vegetation communities within single pixels; hence we used a sub-pixel classification strategy using a Regression-based unmixing approach with synthetic training data. Our analyses focussed on the influence on map accuracy by i) the hyperspectral information and ii) the observation dates. 

We observed that combining April and June PRISMA (MAE = 16.4%) and April and June DESIS (MAE = 17.3%) datasets produced the best results for mapping the peatland vegetation fractions. Analysis of single dates showed that June data leads to slightly better results than April. We found that PRISMA images produced slightly better results than DESIS, which may be caused by the shortwave infrared information missing in the DESIS data. In contrast, DESIS has only visible and near-infrared bands (400-1000 nm) despite having a higher spectral resolution (2.55 nm) than PRISMA (10 nm). 

In conclusion, the hyperspectral information, especially from the short wave infrared > 1 µm, together with the multi-date observation could be shown to contribute to sub-pixel mapping accuracy. In the future, PRISMA and DESIS images can be coupled with EnMAP and the forthcoming SBG and CHIME missions to further improve the space-borne monitoring of rewetted peatlands.

How to cite: Muthusamy, A., Thiel, F., Pham, V. D., Hellmann, C., and van der Linden, S.: Mapping vegetation cover on rewetted fen peatlands using hyperspectral spaceborne images from DESIS and PRISMA, EGU General Assembly 2023, Vienna, Austria, 24–28 Apr 2023, EGU23-5352,, 2023.