Automatic identification of oceanic eddies in infrared satellite images Armando Manuel Fernandes a,n,1 , Susana Nascimento a,b , Dmitri Boutov c a CENTRIA—Centre for Artificial Intelligence, Universidade Nova Lisboa, Quinta da Torre, 2829-516 Caparica, Portugal b Departamento de Informa ´tica, Universidade Nova Lisboa, Quinta da Torre, 2829-516 Caparica, Portugal c Instituto de Oceanografia, Faculdade de Ciˆ encias, Universidade de Lisboa, Campo Grande, 1749-016 Lisbon, Portugal article info Article history: Received 30 July 2009 Received in revised form 19 October 2010 Accepted 16 December 2010 Available online 2 February 2011 Keywords: Infrared Satellite images Ocean Eddies Size abstract Oceanic eddies have a large impact on climate and human activities; consequently, it is worthwhile to characterise them. One of their main features is size; however, it is a difficult task to obtain user- independent estimates of this feature from brightness temperature maps for eddies near the Iberian Peninsula. The reason is that the current methods in the scientific literature are unable to handle the variability in the shape and size of these eddies as well as the weak temperature gradients associated with them, especially those found off Iberia or those methods employ user-defined values that influence the estimate of the eddies’ sizes. Our new method solves these problems using orientation fields and clustering methods. Its outcome is an ellipse that characterizes the size of the eddies with good precision. & 2011 Elsevier Ltd. All rights reserved. 1. Introduction Oceanic eddies, which are mesoscale water structures exhibit- ing rotating flow patterns (Paillet, 1999), have an impact on the global climate (Garabato et al., 2007), on air–sea fluxes of heat and gases (Vecchi et al., 2004), on acoustic wave propagation (Jian et al., 2009), on surfactant slick spread (Schuler et al., 2004), on plankton dynamics (Smith et al., 1996), and even on oil rig tow, owing to water drag (Horizon Marine Inc., 2009). Therefore, the study of this phenomenon is very important. Recently, it has been shown that satellite images, namely brightness temperature maps, are a good tool for studying mesoscale eddies due to their spatial and time resolution (Oliveira et al., 2000). Satellites from the National Oceanographic and Atmospheric Agency (NOAA) equipped with an Advanced Very High Resolution Radiometer (AVHRR) are able to provide brightness temperature maps from approximately the same geographic zone every 6 h, when the cloud coverage allows it. These brightness temperature maps contain important information regarding the mesoscale phenom- ena and are known to be less noisy than sea surface temperature maps (Oliveira et al., 2000; Torres et al., 2003). For this reason, we use brightness temperature maps in the present study. The quantitative characterization of mesoscale eddies by human analysts should be avoided because it is not reproducible owing to the subjective interpretation of the images. Consequently, the aim of this paper is to present a new method that allows us to determine algorithmically the size of eddies. The method is able to calculate an optimal value for the radius of a circle that surrounds the eddy. Consequently, the value for this radius is adjusted to each eddy, which is important in our case because eddies have a large range of sizes. When we tried to use the same initial estimate of the radius value for all eddies, setting the value equal to the eddies’ maximum expected radius, the final radius values calculated for eddies of different sizes tended to be similar and to have values close to the one we set initially. This finding explains why we cannot use methods published in the scientific literature where the user chooses values for the maximum sizes of the eddies. Previous works where this value is not set by a user are available (Thonet et al., 1995; Yang et al., 2001; Chaudhuri et al., 2004), but they cannot be employed in our case because they cannot handle the weak temperature gradients and the structure variability of the eddies off the Iberian Peninsula. Our method defines the optimum radius of the eddies and finds its core by applying clustering algorithms in an innovative way. Then it determines an ellipse that accurately fits the eddies’ borders, by fitting the points within the optimum radius whose flow orientations are closest to those of an elliptical flow. Various works in the scientific literature are capable of determining an outline for eddies, but they are incapable of finding the optimal Contents lists available at ScienceDirect journal homepage: www.elsevier.com/locate/cageo Computers & Geosciences 0098-3004/$ - see front matter & 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.cageo.2010.12.007 n Corresponding author. Permanent address: Rua Bar ~ ao de Sabrosa, 134, 3 1C, 1900-094 Lisbon, Portugal. E-mail addresses: arm.fernandes@gmail.com (A. Manuel Fernandes), snt@di.fct.unl.pt (S. Nascimento), dboutov@fc.ul.pt (D. Boutov). 1 Current address: CITAB—Center for the Research and Technology of Agro-Environmental and Biological Sciences, Universidade de Tra ´ s-os-Montes e Alto Douro, Quinta de Prados, 5000-911 Vila Real, Portugal. Computers & Geosciences 37 (2011) 1783–1792