Discrimination of herbaceous habitats using multitemporal Spot4 and Landsat8 images (Massif central – France)
Résumé
In the framework of the French national project CarHAB (Cartographie des HABitats) managed by the MEDDE*, the use of remote sensing data, combined with environmental data, aims to provide information about the physiognomic and environmental characteristics of vegetation. To this end, this study explores the potential of multitemporal imagery for improving the discrimination of herbaceous habitats.
Study area is the Forez plain and the Pilat mountains in the department of Loire. The altitude of the area ranges from 140 to 1432 m. Open vegetation is dominated by grasslands: 87 % of the agricultural area (UAA) of the department is occupied by permanent grasslands and forage crops allocated to the production of forage and/or grazing (Agreste Rhône Alpes, 2011). The CBNMC (Conservatoire Botanique National du Massif Central) is associated with CarHAB project. From 2012 to 2014, botanists have conducted inventories on the field concerning: crops, temporary grasslands and different permanent grasslands based on the floristic composition/ uses, trophic and hydromorphic gradients. Field surveys distributed in different environments cover about 1740 agricultural parcels.
Mapping herbaceous habitats involves several steps :
1/ the delimitation of the herbaceous patches: forests, water bodies and buid-up areas are masked using vectorial database (BD Forêt V.2®IGN and BD TOPO®IGN data),
2/ the distinction between seasonal cropping land and perennial cover
3/ and the physiognomic characterization of the herbaceous habitats.
We analyze the contribution of multitemporal HRS imagery (Spot4 and Landsat8 data, level 2A) between March 2013 and September 2014 (14 images). Annual crops mapping and more specially the extraction of bare soils are carried out by calculating the NDVI values at each date. Surfaces classified as bare soil at least once a year are identified as crops. The method is validated by means of the “Registre Parcellaire Graphique” and Terruti Lucas data.
If the temporal signatures of the crops are still distinct from those of the perennial vegetation thanks to the period of ploughing, the signatures of perennial grasslands, almost permanently vegetated, are very close to each other. On the basis of field data, we analyse their values by means of two indexes: NDVI and NDWI (Gao, 1996) on four test areas from the plain to the highlands of Pilat. The analysis of responses of the different types of grassland (distinguished by botanists on the field) of each area, brings out the specificities of vegetative cycles, agricultural practices but also the climatic conditions. For each grassland type in each area, it highlights the presence or absence of particular signature. This analysis also reveals differences in signatures within the same type of grassland (for example, hayfields with different mowing dates). The recognition of temporal signatures is very useful to better known the characteristics of vegetation types to map. Thanks to this prior step, we try to define protocols for the classification of the herbaceous habitats.
We thank CNES and Spot 4 Take Five experience – and - the French Ministry of Ecology, Sustainable Development and Energy (MEDDE).