Discriminating stocks of striped red mullet (Mullus surmuletus) in the Northwest European seas using three automatic shape classification methods

Stock identification is of primary importance for population structure assessment of economically important species. This study investigates stocks of striped red mullet using three automatic methods of stock identification based on otolith shape and growth marks. Otolith shape is known to be a promising approach for stock identification but interpreting patterns of variance is a difficult problem. In this study, images in reflected and transmitted light were acquired from 800 otoliths sampled in the Northwest European seas from South Bay of Biscay to North Sea. The growth marks are pointed out manually by an expert. The external shape of otoliths was automatically extracted by computer vision process and then three automatic classification methods were compared, two classical state-of-the-art methods based on Fourier descriptors and principal component analysis (PCA), and a recently proposed method based on shape Geodesics. From a methodological point of view, results show that the shape geodesic approach significantly outperforms other classical methods. From a biological point of view, this study shows that the population of striped red mullet in Northwest European seas can be divided in three geographical zones: the Bay of Biscay, a mixing zone composed of the Celtic Sea and the Western English Channel and a northern zone composed of the Eastern English Channel and the North Sea (67% of correct classification rate using both shape and growth pattern information). Moreover, it shows that for a given zone, two subsets of the same year have a lower variability in shape than two subsets from two consecutive years.