On the Use of Fully Polarimetric Radarsat-2 Time-Series Datasets for Wetland Monitoring
Pottier, Eric1; Hubert-Moy, Laurence2
1IETR - UMR CNRS 6164, University of Rennes 1, FRANCE; 2COSTEL - UMR CNRS 6554 LETG, University Haute Bretagne, FRANCE

The project entitled Evaluation of RADARSAT-2 quad-pol data for functional assessment of wetlands (Id6842), developed in the framework of the CSA-ESA SOAR-EU (Science and operational applications research for Europe) aims to contribute to the application development in demonstrating the exploitation of fully polarimetric time-series datasets for the functional assessment of wetlands. The objective of this paper is to address the issue of evaluating fully polarimetric RADARSAT-2 time-series datasets to determine the water cycle dynamics, in order to delineate precisely potential, effective and efficient wetlands.

Remotely sensed data are currently used to identify, delineate and characterize wetlands. Optical data provide interesting information on land-use and land cover but are limited to cloud-free periods. For these reasons it is not possible to precisely inventory wetland vegetation and agricultural practices, as well as water cycles and water levels in these areas with optical data. For this purpose, spaceborne fully polarimetric and time-series datasets show great potential for mapping wetlands with a sufficient level of precision. The objective of this paper is to evaluate fully polarimetric RADARSAT-2 time-series datasets to determine the water cycle dynamics, in order to delineate precisely potential, effective and efficient wetlands. To that end, the development and validation of a supervised PolSAR segmentation including multi-temporal analysis of wetland evolution and polarimetric decomposition methods are presented. The proposed methodology is based on the segmentation of a polarimetric descriptor, the Shannon Entropy, which has been shown to be a very sensitive parameter to the temporal variability of flooded areas. The results have been validated using ground truth measurements in the field and a LiDAR image. They show that it is possible to produce detailed water feature maps useful for delineating water tables as well as water-saturated areas, and for monitoring water area dynamics. These products provide useful information to identify and delineate wetlands in order to support conservation and management in these ecosystems across large areas.

The investigated area is the site of Pleine-FougA''res, located in Britanny near the Mont St Michel and is referenced in the LTER-Europe (lter-europe.net) and the ILTER networks. Different fully-polarimetric RADARSAT-2 datasets have been acquired in ascending orbit (18:02 - GMT) and in Fine Quad Polarisation mode (SLC products), thus presenting a nominal swath width of 25km with a spatial resolution of 5.6m (range) x 7.8m (azimuth). FQ23 beam is used that corresponds to a radar incidence angle range from 41.9A° (near range) to 43.3A° (far range). This project is set from February 2010 to February 2011 (15 images) over the study site. The lower valley of the Couesnon River presents five marshes that are regularly flooded. Our studying area focuses on two of them. These two marshes, that cover an area of around 4km2, are entirely made up of grazing land. Up today, 15 RADARSAT-2 quad-polarimetric SAR images have been acquired, with a repeat time of 24 days, in the FQ23 beam and in the Fine Quad Polarisation mode. A LiDAR image of the study area has been acquired on April 5th, 2009, to conduct a very precise spatial analysis and interpretation of the obtained results.

Information on the land cover is a paramount importance for monitoring and management of the environment on a local, regional and global scale. In natural areas, land cover presents complicated structures and highly complex scattering responses, due to various scattering contributions, dielectric and shape properties and volumetric structures. The complementarities of polarimetric high-resolution observations and time-series datasets will be necessary to provide enough information for general land classification, characterization and mapping. The proposed research activities that are presented in this paper, aim at developing and validating a general supervised and/or unsupervised PolSAR segmentation methodology, including multi-temporal analysis of land cover evolution, and polarimetric decomposition methods. The RADARSAT-2 quad-polarimetric SAR image processing has then been conducted using the PolSARpro v5.0 software (http://earth.esa.int/polsarpro). Different polarimetric decomposition methods have been investigated in order to extract one polarimetric descriptor that could be very sensitive to the temporal variability of the marsh flooded areas. The Shannon Entropy (SE) has shown this sensitivity. It can then be seen that this polarimetric descriptor is a very sensitive parameter to the temporal variability of the marsh flooded area. The original concept proposed in this methodology is based on the segmentation of the SE image in order to detect and extract the limits of the open water area and its evolution during the time period under study. Every 24 days, corresponding to the RADARSAT-2 revisiting period, ground-truth campaigns have been organized in order to measure the distance between a reference fix point and the limit of the water table. We shall show that there exists a prefect correspondence between the results of the segmentation and the limit of the water table detected during the different ground-truth campaigns.

Studies have shown that the length of the contact between wetlands and the dry land play an important role in the denitrification process, and that the area between effective and potential wetlands can be considered as negotiation areas for restoration purposes. For these reasons, a precise determination of the limits of effective and potential wetlands appears as an important stake from an environmental point of view. The combination of radar and optical data is needed to provide a better assessment of wetlands functionalities, almost for hydrologic processes, but also for ecological and biogeochemical processes. The use of fully polarimetric time-series datasets (RADARSAT-2) and the development of novel remote sensing techniques fusioning radar and optical data is an important aspect for future research activities. This will permit to improve our knowledge on wetlands functional processes. The information derived from remotely sensed data will thus be integrated in hydrological and ecological models to predict water fluxes and biodiversity levels. Furthermore, the common work between remote sensing and practices analysis in wetlands opens the possibility to detect evidences for changes in agricultural practices in these areas.