3 ms·

> Have you written about this at all? I'll be giving a (short, somewhat shallow) talk about this and some other things at this year's Akademy, about "Taking KD

by sho_hn 25d ago

> Have you written about this at all? I'll be giving a (short, somewhat shallow) talk about this and some other things at this year's Akademy, about "Taking KDE to the Pyramids". > Curious if you've played around with a MRMS-esque (Multi-Radar Multi-Sensor) approach to ingesting multiple passes on the same area to "get more resolution for free" I haven't! We do have some interesting bespoke highly-stacking radar hardware, and do somewhat unusual things like complex transilluminations, but I'm really more a nuts and bolts software engineer than the actual sensor expert :-)

monster_truck 25d ago
Oh I'm glad that's soon, I don't usually use a calendar but I've set a reminder to check. Ask chat about it! I have a pretty limited understanding of any of this shit but: I wasn't really able to find a satisfactory explanation for why we do not currently do this with doppler radar[1]. A great deal of CONUS has 3+ overlapping local/national radars from ideal angles & elevations, the smallest cell on the mosaic being ~1km felt "like bullshit/not very 2026". I don't care if the radars are nearly twice my age, they're still quite good. With Codex 5.2 (and some preprints wrt MRMS from countries that don't even really get tornadoes) I got to the point where this 5800X3D (and 128gb of ram) could attenuate and plot all of conus about as fast as new sweeps are published, but it makes my room very hot. Now that fall is coming and the models are so much better I will revisit it. Will email you if anything good happens. Here's an old mp4 in the meantime https://i.imgur.com/EZjM46W.mp4 https://i.imgur.com/EZjM46W.mp4 I'm certain if you threw Astra at it on high with a pile of slightly different scans it would do something useful. 1) This isn't something they would do half-cocked, I learned weather nerds are very serious about only presenting exactly what can be measured with known error bounds and detest anything resembling speculation

by sho_hn 25d ago

monster_truck 25d ago

Oh I'm glad that's soon, I don't usually use a calendar but I've set a reminder to check. Ask chat about it! I have a pretty limited understanding of any of this shit but: I wasn't really able to find a satisfactory explanation for why we do not currently do this with doppler radar[1]. A great deal of CONUS has 3+ overlapping local/national radars from ideal angles & elevations, the smallest cell on the mosaic being ~1km felt "like bullshit/not very 2026". I don't care if the radars are nearly twice my age, they're still quite good. With Codex 5.2 (and some preprints wrt MRMS from countries that don't even really get tornadoes) I got to the point where this 5800X3D (and 128gb of ram) could attenuate and plot all of conus about as fast as new sweeps are published, but it makes my room very hot. Now that fall is coming and the models are so much better I will revisit it. Will email you if anything good happens. Here's an old mp4 in the meantime https://i.imgur.com/EZjM46W.mp4 https://i.imgur.com/EZjM46W.mp4 I'm certain if you threw Astra at it on high with a pile of slightly different scans it would do something useful. 1) This isn't something they would do half-cocked, I learned weather nerds are very serious about only presenting exactly what can be measured with known error bounds and detest anything resembling speculation