Use of bi-Seasonal Landsat-8 Imagery for Mapping Marshland Plant Community Combinations at the Regional Scale

Abstract

Coastal marshlands may provide ecosystem services but their vegetation and related services may be impacted by environmental changes. Habitat mapping is a key step to monitor the spatio-temporal dynamics of vegetation and detect on-going changes. However, it is still a challenge to produce reliable vegetation maps at the regional scale. This study aims to evaluate the potential of new Landsat-8 imageries (acquired in September and December 2013) for mapping fine-grained plant communities in coastal marshlands. Field-based vegetation maps were collected for 270 km of marshlands along the French Atlantic coast. In order to be identifiable on the satellite image, fine-grained vegetation units were aggregated into fewer plant community combinations. The classification accuracy was assessed by comparison with field-based vegetation data and compared between the supervised methods used, including Minimum Distance, Mahalanobis, Maximum Likelihood, Random Forest and Support Vector Machine. The best result was obtained with the Maximum Likelihood classifier and by combining the two Landsat-8 images (85.9 % accuracy overall). Three main habitat types dominated the coastal Atlantic marshlands: croplands, Trifolio maritimae-Oenantheto silaifoliae geosigmetum and Puccinellio maritimae-Arthrocnemeto fruticosi geosigmetum. The reliability of the vegetation map produced will provide a good basis for monitoring the conservation status of the various habitats.