De Boissieu F. auteur aut IRD Menkes Christophe auteur aut IRD Dupouy Cécile auteur aut IRD Rodier Martine auteur aut IRD Bonnet Sophie auteur aut IRD Mangeas Morgan auteur aut IRD Frouin R. auteur aut IRD text bookSection eng text/pdf born digital access In recent years great progress has been made in global mapping of phytoplankton from space. Two main trends have emerged, the recognition of phytoplankton functional types (PFT) based on reflectance normalized to chlorophyll-a concentration, and the recognition of phytoplankton size class (PSC) based on the relationship between cell size and chlorophyll-a concentration. However, PFTs and PSCs are not decorrelated, and one approach can complement the other in a recognition task. In this paper, we explore the recognition of several dominant PFTs by combining reflectance anomalies, chlorophyll-a concentration and other environmental parameters, such as sea surface temperature and wind speed. Remote sensing pixels are labeled thanks to coincident in-situ pigment data from GeP&CO, NOMAD and MAREDAT datasets, covering various oceanographic environments. The recognition is made with a supervised Support Vector Machine classifier trained on the labeled pixels. This algorithm enables a non-linear separation of the classes in the input space and is especially adapted for small training datasets as available here. Moreover, it provides a class probability estimate, allowing one to enhance the robustness of the classification results through the choice of a minimum probability threshold. A greedy feature selection associated to a 10-fold cross-validation procedure is applied to select the most discriminative input features and evaluate the classification performance. The best classifiers are finally applied on daily remote sensing datasets (SeaWIFS, MODISA) and the resulting dominant PFT maps are compared with other studies. Several conclusions are drawn: (1) the feature selection highlights the weight of temperature, chlorophyll-a and wind speed variables in phytoplankton recognition; (2) the classifiers show good results and dominant PFT maps in agreement with phytoplankton distribution knowledge; (3) classification on MODISA data seems to perform better than on SeaWIFS data, (4) the probability threshold screens correctly the areas of smallest confidence such as the interclass regions. specialized TELEDETECTION SPATIALE DONNEES SATELLITE PHYTOPLANCTON CLASSIFICATION ALGORITHME COULEUR DE L'OCEAN 126TELAPP05 034BIOVEG02 Frouin R.J. ed. edt IRD Pan D. ed. edt IRD Murakami H. ed. edt IRD 9261 92611R [14 en ligne] Bellingham WA SPIE 2014 SPIE Asia-Pacific Remote Sensing, Pékin (CHN), 2014/10/13-16 9261 https://www.documentation.ird.fr/hor/fdi:010063840 10.1117/12.2083730 [F B010063840] https://www.documentation.ird.fr/hor/fdi:010063840 https://www.documentation.ird.fr/intranet/publi/depot/2015-07-23/010063840.pdf Accès réservé (Intranet de l'IRD) IRD - Base Horizon / Pleins textes 2015-07-02 2018-02-12 fdi:010063840 fre