Oceanic front sample points from five estuaries

This dataset contains 1,055 ground-truth sample points used for training and evaluating oceanic front identification models in estuarine environments. The samples were collected from five estuaries with contrasting hydrodynamic regimes: the Yangtze River Estuary, Pearl River Estuary, Qiantang River Estuary, Mississippi River Estuary, and Colorado River Estuary. Each sample is labelled as either “oceanic front” (positive) or “non-front uniform water body” (negative). Positive samples correspond to locations with sharp optical transitions in Sentinel-2 true-color, NDWI, and red-band images, verified by two independent researchers. Negative samples cover uniform water areas with different turbidity levels (high-turbidity, low-turbidity, and clear offshore waters). Samples affected by cloud shadows, sun glint, or land-boundary mixed pixels have been excluded.