基于高分辨率再分析海流数据的三种涡旋识别方法在南海的评估

Assessment of three eddy detection methods based on high-resolution reanalysis of ocean current data in the South China Sea

本文基于南海11 a逐日高分辨率再分析海流数据,利用三种涡旋识别方法——缠绕角(Winding Angle,WA)法、速度矢量几何(VectorGeometry,VG)法和OW(OkuboWeiss)参数法对南海中尺度涡进行识别和追踪,对比了三种方法对中尺度涡的探测能力,并分析了在高分辨率再分析产品中的适用性。研究表明,产品空间分辨率的提高能够放大OW方法的W参数噪声,使该方法存在涡旋过量检测和涡旋分割现象,导致识别的涡旋数量偏多、半径偏小,涡旋平均成功探测率最低(76.2%);与OW方法相比,WA方法在一定程度上降低了涡旋过量探测率和漏判率,涡旋平均成功探测率提高至85.1%,但涡旋识别时间较长;与前两种方法相比,VG方法的涡旋平均成功探测率可达93.2%,综合评估参数(过量探测率×漏判率÷成功探测率)优势显著(4.5%),且计算高效。因此,在基于高分辨率再分析产品进行中尺度涡识别时,VG方法具有更加合理的涡旋探测结果和更强的适用性。

Based on 11-years of daily high-resolution reanalysis of current data,three eddy detection methods,i.e.,the Winding Angle(WA)method,the Vector Geometry(VG)method,and the Okubo Weiss(OW)parameter method,are used to detect and track mesoscale eddies in the South China Sea.The detection capabilities of the three methods for mesoscale eddies are compared,and their applicability in high-resolution reanalysis products is analyzed.It has shown that the improved spatial resolution of the product can amplify the W-parameter noise of the OW method,which makes the method characterized by eddy over-detection and eddy splitting phenomena,leading to a higher eddy number,a smaller eddy radius,and the lowest average successful detection rate of eddies(76.2%).Compared with the OW method,the WA method reduces the eddy's excessive detection rate and the missing detection rate to a certain degree,and the average successful detection rate of eddies improves to 85.1%,but the eddy identification takes longer.The VG method has an average eddy successful detection rate of up to 93.2%compared to the previous two methods,combined evaluation parameter(excessive detection rate×missing detection rate÷successful detection rate)has a significant advantage(4.5%),and the computation is efficient.Therefore,the VG method has more reasonable eddy detection results and better applicability when using it for mesoscale eddy identification based on high-resolution reanalysis products.