Research On Automatic Recognition Algorithm Of Mesoscale Vortex Based On Satellite Altimeter

Mesoscale eddies are one of the mesoscale ocean phenomenon and exist almost everywhere in the world ocean. For decades, following the rapid development of satellite altimeter, many scholars have used altimeter data which is large covering, quasi-synchronous and continuously observed to research the statistical characteristics and propagation characteristics of mesoscale eddies in specific or the world ocean. So it is important to automatically identify the mesosclae eddies rapidly and accurately in large volumes of altimeter data, which is the foundation of quantitatively analyzing the statistical properties of mesoscale eddies.In this paper, three automatic identification algorithms(WA methods, OW methods and threshold-free contours method), which are the most widely used to identify mesoscale eddies, are improved respectively to enhancing their accuracy and computational efficiency. Then, through the comparison among three improving algorithms, their main advantages and disadvantages are detailed respectively. Finally, using the most accurate algorithm(WA algorithm) and CTD data, we identify and track the anti-cyclonic eddy off southeast Vietnam during summer and autumn 2013, analyzing the evolution character, hydrological character and internal structure of it. The main work of this paper includes:1) We first introduce some dealing skills of WA algorithm which is crucial for this approach and then propose a new “clustering” algorithm which can group the streamlines together that belong to the same vortex. Comparing with the old one, the new clustering algorithm is more accurate, more reliable, without artificial threshold and has lower computational complexity. At the same time, we propose a new “eddy-edge recognition” algorithm to detect vortex edge. The combination of the two algorithms we proposed here enable WA method to recognize the multi-core structures from altimeter data, which is unrealizable for unimproved WA method.2) For OW algorithm, there always exist two common problems: first, the algorithm is susceptible to noise from altimeter data; second, eddies with opposing polarities sometimes occur within a single closed contour of W. To solve these problems and enhance the accuracy, we attempt to improve OW algorithm with different means. Especially for the latter problem, we propose two different “separation” algorithms(the algorithm based on relative vorticity and the algorithm based on simple connected region) to separate eddies with opposing polarities and then analyze the characteristics for each of them.3) We adopt threshold-free contours method, which has been implemented only by SSH data, by using SLA data and confirm its validity. On this basis, to improve this approach that could not recognize the eddy core and multi-core structures, we propose a new “matching” algorithm that can match the core-edge to the corresponding eddy-edge accurately. The new matching algorithm can recognize not only the number of eddy cores which belong to the same vortex but also their intensity and specific location.4) To evaluate the efficiency and accuracy of identification algorithms, we invite five oceanographic experts to identify mesoscale eddies visually from 12 available SLA maps chosen randomly. We calculate the success of detection rate(SDR), the excess of detection rate(EDR), the default of detection rate(DDR), the SDR to EDR ratio and the area overlap ratio for every algorithm. Because the expert maps are the basis for assessing the accuracy of identification algorithms, we also estimate the reliability of the result of expert visual judgement.5) Based on the comparison of the three automatic algorithms, we adopt the most accurate one(WA algorithm) to identify and track the anti-cyclonic eddy off southeast Vietnam during summer and autumn 2013. At the same time, combing the identification algorithm with CTD data, we analyze the evolution character, hydrological character and internal structure of the warm eddy.