Premium
Mapping grassland plant communities using a fuzzy approach to address floristic and spectral uncertainty
Author(s) -
Rapinel Sébastien,
Rossignol Nicolas,
HubertMoy Laurence,
Bouzillé JanBernard,
Bonis Anne
Publication year - 2018
Publication title -
applied vegetation science
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.096
H-Index - 64
eISSN - 1654-109X
pISSN - 1402-2001
DOI - 10.1111/avsc.12396
Subject(s) - vegetation (pathology) , grassland , remote sensing , geography , plant community , multispectral image , fuzzy logic , cartography , environmental science , computer science , ecology , ecological succession , artificial intelligence , medicine , pathology , biology
Abstract Aims The mapping and monitoring of natural vegetation is a challenging but important objective for environmental management. Although remote sensing has been used to map plant communities for several years, the maps produced are not sufficiently accurate to meet management requirements. This can be explained by the cumulative effects of floristic and spectral uncertainty. The objective of this study was to accurately map grassland plant communities using a comprehensive fuzzy approach in order to address floristic and spectral uncertainty. Location Sub‐brackish wet grasslands, Marais Poitevin, France. Methods We first created a compromise typology—floristically and spectrally consistent—to perform fuzzy noise clustering on a joint PCA matrix derived from vegetation relevés and remote sensing data. This typology had two levels, which corresponded to spectral signatures and plant communities, respectively. Second, we mapped grassland plant communities to predict the fuzzy model from the remote sensing data. We applied this approach using (1) a very high spatial resolution multispectral satellite image and a Li DAR ‐derived Digital Terrain Model acquired on a 73 km 2 wet grassland site and (2) more than 200 relevés collected in the field. Results The results show that (1) the compromise typology yields significantly higher mapping accuracy than classic phytosociological typology (62% and 26%, respectively); (2) compared to a crisp approach, the fuzzy approach improves mapping accuracy by 17 percentage points and (3) a single plant community can be defined by several (1–4) distinct spectral signatures. Conclusions The comprehensive fuzzy procedure successfully mapped herbaceous plant communities at the ecosystem scale using inexpensive remote sensing data. Floristic and spectral uncertainty was considered in a fuzzy approach, resulting in the mapping of nine herbaceous plant communities with acceptable accuracy. As the natural habitats were characterized at the plant community level, correspondence with functional properties of the species or with ecosystem services can be easily inferred. These encouraging results open up new ways to meet the requirements for monitoring the conservation status of natural habitats in the EU Habitats Directive.