Monitoring ecosystems in the Amazonian basin with ERS, ENVISAT and ALOS SAR data: Case studies in French Guyana and the Brazilian Jau basin
Pierre-Louis Frison(1), Cedric Lardeux(1), Habib Mahi(2), Emmanuel Fritsch(3), Nadia Do Nascimento(4) and Jean-Paul Rudant(1)
(1) Université Paris-Est, Institut Francilien des Sciences Appliquees, 77454 Marne la Vallee Cedex 2, France
(2) Agence Spatiale Algérienne, 1 avenue de la palestine, BP 13 ARZEW, Algeria
(3) IMPMC, UMR 7590, et IPGP, Case 115, 75252 Paris cedex 05, France
(4) Insituto de Geociencias e Ciencias Exatas, DEPLAN/IGCE/UNESP, 13500–230 Rio Claro, Brazil
This paper presents 2 study cases assessing the potential of ASAR and PALSAR data for the monitoring of ecosystems in the Amazonian basin.
Past studies have shown the potential of Multi frequency and polarimetric data (fully or partial) for land surface classification. In particular the use of the SVM (Support Vector Machine) algorithm has shown significant improvement for the supervised classifications with respect to the classical method based on the a priori Wishart distribution observed by the radar data. The main advantage of the SVM algorithm is its ability to take into account different heterogeneous polarimetric indices obtained from multi frequency radar data, such as the elements of the coherency matrix, the degree of coherence or the H / a / alpha parameters obtained from the Cloude and Pottier decomposition. Very good results were obtained for tropical vegetation cartography over a French Polynesian Island, allowing the discrimination between different forest species.
Here the same method is applied over 2 test sites in the Amazon basin: the Jau basin, located in Brazil, and a site near Kourou, in French Guyana.
The Jau basin is characterized by the presence of hydromorphic Podzols which are associated with low activity clay soils. They are a major concern for the upper Amazon basin according to the superficies affected by such processes. The cartography of the Jau basin with SAR data will allow for a better understanding of the different formation processes. PALSAR data acquired in fully polarimetric and Dual-Pol modes as well as one ERS SAR acquisition are available.
The site located in the French Guyana is concerned with the monitoring of the vegetation. It is a coastal plain surrounded by continental forest. Land use consists in mangrove forest, marshes, dry and moist savannas, and agricultural fields, evolving due to sedimentation and erosion processes as well as anthropic pressure. In particular, deforestation is particularly watched on. Acquisitions by PALSAR as well as ASAR and ERS SAR sensors allow to investigate the contribution of full/partial polarimetry and L and C band contribution for land use monitoring in this environment.