Sammendrag
Oceanographic fronts between different water masses are important for marine animals and birds [1] [2], but are also a challenge for ocean modelling and acoustic propagation modelling. Fronts have conventionally been found by detecting sudden horizontal changes in surface salinity, temperature and even chlorophyll [3] [4]. However, such methods fail to take into account the full depth dependence of the oceanographic field. Also, when using only a single state variable, some types of fronts may remain undetected.
Hjelmervik et al. [5] suggest that climatological oceanographic regions may be classified by using Empirical Orthogonal Functions (EOF) [6] and clustering [7]. They showed that EOFs may be used to compress the information contained in depth dependent salinity and temperature profiles into a few parameters.
In this study we apply Sobel edge detection [8] on EOF coefficients in order to detect oceanographic fronts. In theory, fronts present at any depth may then be detected using edge detection techniques. Furthermore, the method introduces a formalised approach that includes different state variables in order to detect fronts. The method is demonstrated on modelled temperature and salinity data extracted from two different dynamical ocean models; one for the North Atlantic and one for the Nordic Sea.
Vis fullstendig beskrivelse