Land cover mapping using aerial and VHR satellite images for distributed hydrological modelling of periurban catchments: Application to the Yzeron catchment (Lyon, France)

Abstract

The rapid progression of urbanization in periurban areas affects the hydrological cycle of periurban rivers. To quantify these changes, distributed hydrological modelling tools able to simulate the hydrology of peri-urban catchments are being developed. Land cover information is one of the data sources used to define the model mesh and parameters. The land cover in periurban catchments is characterized by a very large het-erogeneity, where the vegetated and the artificial surfaces are finely overlapping. The study is conducted in the Yzeron catchment (150 km 2), close to the city of Lyon, France. We explore the potential of very high-resolution (VHR) optical images (0.50–2.50 m) for retrieving information useful for those distributed hydrological models at two scales. For detailed object-oriented models, applicable to catchments of a few km 2 , where hydrological units are based on the cadastral units, manual digitizing based on the 0.5 m resolution image, was found to be the most accurate to provide the required information. For larger catchments of about 100 km 2 , three semi-automated mapping procedures (pixel based and object-oriented classifications), applied to aerial images (BD-Ortho O IGN), and two satellite images (Quickbird and Spot 5) were compared. We showed that each image/processing provided some interesting and accurate information about some of the land cover classes. We proposed to combine them into a synthesis map, taking profit of the strength of each image/processing in identifying the land cover classes and their physical properties. This synthesis map was shown to be more accurate than each map separately. We illustrate the interest of the derived maps in terms of distributed hydrological modelling. The maps were used to propose a classification of the Yzeron sub-catchments in terms of dominant vegetation cover and imperviousness. We showed that according to the image processing and images characteristics, the calculated imperviousness rates were different. This can lead to significant differences in the hydrological response.