Marine pollution is often invisible: potentially toxic elements (PTEs) move unseen through seawater, sediments, and living organisms, and the science that tracks them usually stays locked inside tables, indices, and statistical models that few outside the laboratory can read. Drawing on our multimedia study of eight PTEs (Cu, Pb, Zn, Cd, Cr, Hg, As, and Ni) in Sanmen Bay — the “Cradle of Chinese Mud Crabs” — this presentation explores how scientific visualization can act as a bridge between ocean science and art, turning abstract data into an interpretable visual language. We show how three families of images speak to different audiences: GIS pollution-hotspot maps that give contamination a place and a geography; a graphical abstract that compresses the biogeochemical journey of trace elements — from water to sediment to the shellfish and crabs of a fishing community — into a single readable scene; and, most strikingly, the topological grids of a self-organizing-map (SOM) neural network, whose colour fields and clustered patterns render the machine's “reading” of the bay as an almost painterly composition. Together these visuals translate quantitative findings — the marked enrichment of Cd and Hg as key ecological stressors, and four apportioned pollution sources — into forms that invite attention, curiosity, and dialogue. We argue that the graphical abstract and the SOM map are not merely decorative endpoints but genuine ArtScience objects: designed images that make the imperceptible perceptible and carry environmental knowledge back to the communities it concerns. The talk shares these visuals, reflects on the aesthetic choices behind them, and considers how scientist–artist collaboration could deepen the ocean literacy of coastal audiences.