Numerical Weather Prediction 2D: Grid Resolution & Forecast Skill
Interactive 2D numerical weather model: a real advection-diffusion solver runs side-by-side truth and forecast grids on canvas, showing live how grid resolution, timestep (CFL number), diffusion and initial-condition error each degrade forecast accuracy. Drag to pan, scroll to zoom.
This 2D canvas sibling solves the identical real 2D advection–diffusion equation for a pressure/vorticity field carried by a jet-stream wind, drawn side by side at two resolutions: a fixed fine-grid "truth" run (left panel) and a user-controlled coarser "forecast" run (right panel) — the same tradeoff every numerical weather prediction center makes between grid resolution and compute cost. Drag the resolution slider to coarsen the forecast grid and watch it lose fine structure and drift from the truth field beside it; push the CFL safety factor above 1 to trigger genuine numerical instability; and dial in a small initial-condition error to see how a chaotic flow amplifies tiny observation errors into large forecast divergence, exactly the mechanism that motivates ensemble forecasting. Live readouts track the CFL number, elapsed forecast time and RMSE forecast error against the truth run. Drag the canvas to pan and scroll or pinch to zoom into either grid.
A 2D canvas advection-diffusion solver runs a fixed fine-resolution 'truth' grid beside a user-controlled coarser 'forecast' grid side by side, with real drag-to-pan and scroll-to-zoom, showing live how grid resolution, timestep (CFL number), diffusion and initial-condition error each degrade forecast accuracy against a chaotic jet-stream flow.
2D · HTML5 Canvas 2D · 60 FPS target · runs fully client-side, no install