Today I ran a 12-hour simulation around Mount Fuji at 500 m resolution in the WOOF web app from Recast Systems, 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:
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.
From my own experience, summer convection in Beijing often develops over the northwestern mountains in the afternoon and then moves quickly toward the city. Whether it comes down from the mountains, and whether it weakens or intensifies after it does, are both highly uncertain. It often weakens as it reaches the edge of the urban area, with thunder but no rain inside the city, and re-intensifies after crossing it. There are also cases where it sweeps across the city with almost no weakening.
Tokyo is similar: besides triggering over the northwestern mountains, sea-breeze convergence over the plain is also a common trigger.