An improved algorithm for detecting mesoscale ocean fronts from satellite observations: Detailed mapping of persistent fronts around the China Seas and their long-term trends

Global mapping and evolution of persistent fronts in Large Marine Ecosystems over the past 40 years

“…2a ). In addition to the numerical increase, the objective delineation method and improved front detection algorithm contribute to presenting more accurate locations and delineating finer-scale distributions compared to previously hand-drawn persistent fronts 29 , 30 . Persistent fronts are prevalent in most LMEs, with the exceptions being Central Arctic LME 64 where fronts cannot be found due to the sea-ice cover all year round (see the location and number of each LME in Supplementary Fig.…”

“…This algorithm is designed to identify a boundary that separates two water masses as a front, and consists of five main steps: preprocessing, histogram analysis, cohesion testing, locating frontal pixels, and combining multiple windows 29 , 73 . In the preprocessing stage, this improved algorithm adopts inverse distance weighting to create buffer zones between SST-available pixels and SST-unavailable pixels (such as land and cloud-contaminated regions) and uses a 3 × 3 pixel median filter to reduce the random noise and anomalous values in SST images.…”

“…The histogram analysis utilizes the ratio of between-cluster variance to within-cluster variance, using a threshold of 0.75 to determine if two distinct SST populations exist in each window. In comparison to a previous study 29 , it is worth noting that we introduced an additional treatment where all windows of the two populations with an average SST difference of <0.25 °C were excluded, aiming to reduce the detection of false fronts in waters with low SST gradients 75 . When both the entire window and the two SST populations exhibit cohesion coefficients exceeding the threshold of 0.92, the window passes the cohesion testing and identifies pixels on the edge of the warm population connected to the cold population as potential frontal pixels.…”

“…What is most remarkable is that numerous persistent fronts have been found within Large Marine Ecosystems (LMEs), and these fronts have further been manually delineated based on global frontal occurrence probability and named according to their geography 13 . Recently, advancements in front detection algorithms and high-resolution satellite observations have significantly enhanced our knowledge of persistent fronts, including the better detection of coastal fronts and the elimination of repeated detections 29 , along with the objective delineation method of persistent fronts 30 . These techniques have led to the identification of quite a few new persistent fronts and have corrected previously coarse hand-draw positions 29 , 30 .…”

“…Recently, advancements in front detection algorithms and high-resolution satellite observations have significantly enhanced our knowledge of persistent fronts, including the better detection of coastal fronts and the elimination of repeated detections 29 , along with the objective delineation method of persistent fronts 30 . These techniques have led to the identification of quite a few new persistent fronts and have corrected previously coarse hand-draw positions 29 , 30 . However, although existing research has presented global persistent front fields based on the relatively objective and automatic delineation algorithm 30 , a comprehensive global atlas with a specific focus on depicting each individual persistent front in LMEs, such as each separately extracted and named front in ref.…”