https://blog.ringsaturn.me/en/tags/woof/Recent content in WOOF on ringsaturnHugo -- 0.167.0enringsaturnSat, 10 Oct 2026 00:00:00 +0900https://blog.ringsaturn.me/en/posts/recast-gpu-woof/?utm_medium=rssSat, 10 Oct 2026 00:00:00 +0900https://blog.ringsaturn.me/en/posts/recast-gpu-woof/<p>Today I ran a 12-hour simulation around Mount Fuji at 500 m resolution in the <a href="https://woof.recastsystems.com">WOOF</a> web app from <a href="https://www.recastsystems.com">Recast Systems</a>, with boundary conditions from GFS 0.25°. I think the result is excellent: it ran on an RTX PRO 6000 (on-demand instance), took 11 minutes 40 seconds and cost 1.34 USD in total. Video demo below:</p> <script> (function(){ function wantsHD(){ var c = navigator.connection || {}; if (c.saveData) return false; if (/(^|-)2g$|^3g$/.test(c.effectiveType || '')) return false; if (window.matchMedia && matchMedia('(prefers-reduced-data: reduce)').matches) return false; if (window.matchMedia && matchMedia('(pointer: coarse)').matches) return false; var dpr = window.devicePixelRatio || 1; var edge = Math.max(screen.width || 0, screen.height || 0) * dpr; return edge >= 2560; } function upgrade(){ if (!wantsHD()) return; document.querySelectorAll('video[data-hd]').forEach(function(v){ if (v.dataset.hdApplied) return; v.dataset.hdApplied = '1'; var s = v.querySelector('source'); if (!s) return; s.src = v.dataset.hd; v.load(); }); } if (document.readyState === 'loading') { document.addEventListener('DOMContentLoaded', upgrade); } else { upgrade(); } })(); </script> <figure class="align-center video-figure"> <video data-hd="https://blog-assets.ringsaturn.me/pic/2026/2026-10-10-recast-gpu-woof/recast-area-tsk-2026-10-10-1600gmt%2B9.mp4" controls playsinline preload="metadata" style="max-width: 100%; height: auto;"> <source src="https://blog-assets.ringsaturn.me/pic/2026/2026-10-10-recast-gpu-woof/recast-area-tsk-2026-10-10-1600gmt%2B9-mobile.mp4" type="video/mp4"> <a href="https://blog-assets.ringsaturn.me/pic/2026/2026-10-10-recast-gpu-woof/recast-area-tsk-2026-10-10-1600gmt%2B9-mobile.mp4">https://blog-assets.ringsaturn.me/pic/2026/2026-10-10-recast-gpu-woof/recast-area-tsk-2026-10-10-1600gmt%2B9-mobile.mp4</a> </video> </figure> <p>This surprised me. It suggests that in the future, with higher-resolution real-time observations (satellite remote sensing in particular), rapid rolling corrections to the assimilation output of large-scale models, followed by a model like WOOF, a port of WRF-ARW to the GPU, could bring a large improvement to local weather forecasting and hazard warning, especially for severe convective weather.</p>