Performance of Clustering for Upwelling Area Detection Combining Ocean Reanalysis Data and Bathymetric Data Sets

Upwelling areas have been shown to be centers of high biological activity and fisheries production. Upwelling plays a critical role in marine ecosystems and impacts marine resource production, making the detection of its occurrence vital. Previous attempts have largely focused on detecting upwelling areas based on Sea Surface Temperature (SST) and Chlorophyll-a (Chl-a), with some studies also utilizing Sea Surface Height (SSH). However, there has been a lack of comprehensive research into the detection of upwelling areas considering bathymetry, which is believed to influence the occurrence of upwelling.In this study, we add bathymetry to the traditionally used SST, Chl-a, and SSH data sets and employ the fuzzy c-means clustering algorithm (FCM), which has been used in previous studies to identify upwelling areas, to classify regions into upwelling and non-upwelling areas. Our results indicate that using only SST and Chl-a lead to misclassification of some offshore areas as upwelling regions. By including SSH and bathymetry in the input data, the misclassification of these areas was reduced. This suggests that adding indicators with relatively little daily variation, like SSH and bathymetry, may allow for more stable classification in areas where SST and Chl-a alone are more prone to false detection.Furthermore, we used the V
up
index, employed in previous studies to validate classification results of upwelling and non-upwelling areas. We confirmed that our classification method gave comparable performance to previous studies.