EFAR dataset

An eddy-front association recognition (EFAR) dataset from 2006 to 2015 was constructed. The eddies were produced by an SSH-based approach, and SST fronts were generated by figuring up the SST gradient using the Sobel operator. Daily eddies and SST fronts were plotted concurrently in a figure. The minimum external matrix of the maximum geostrophic velocity boundary of each eddy was expanded by 1° to plot the AE fronts and CE fronts separately based on eddy polarity. Figs. A and B show sea surface temperature (SST) and sea level anomaly (SLA) data in the Kuroshio Extension (KE) region, respectively. The SST front is denoted by the maximum SST gradient (Fig. E). Three types of eddy-front categories are represented by different colors (Fig. C). Specifically, the eddy-front category is regarded as weak association when the eddy and front are close to each other but there is no crossover, and is shown in the green graph. The eddy-front category is regarded as medium association when the eddy and the front intersect, and is shown in the yellow graph. The eddy-front category is regarded as strong association when the eddy and front boundaries almost coincide, as shown in the red graph. The EFAR dataset was constructed to train the eddy-front association recognition network.