Operational ocean forecasting systems such as the CMEMS Global Ocean Physics Analysis and Forecast produce sea water velocity fields at 1/4° horizontal resolution (~25 km). While these products underpin a wide range of applications, their coarse spatial scale limits their utility in contexts where fine-scale circulation features — mesoscale eddies, fronts, and boundary currents — are dynamically important. Statistical downscaling driven by deep learning offers a pathway to derive high-resolution ocean fields from operational model outputs, but systematically evaluating and communicating the added value of such approaches requires accessible and intuitive tooling.
We present OceanLens, a web-based application designed to interactively visualize and compare ocean current forecasts at the native 1/4° resolution with AI-downscaled equivalents at 1/36° (~3 km). The application is built on Next.js and TypeScript, uses MapLibre GL JS for cartographic rendering, and reads multidimensional oceanographic data directly from Zarr archives, enabling efficient access to large datasets without server-side preprocessing. Ocean speed fields (√(u² + v²)) are rendered in real time via a WebGL-based layer using configurable colormaps, while a canvas-based side-by-side comparison slider allows visual inspection of structural differences between the base and downscaled products over user-defined geographic domains. A multi-day temporal timeline supports navigation through the forecast horizon.
OceanLens serves a dual purpose: as a diagnostic tool for AI downscaling researchers to rapidly assess inferred spatial structures, and as a communication platform for demonstrating the operational relevance of high-resolution ocean forecasts to end users and decision-makers.