Normally I'd be concerned, even at the best of times, about letting LLMs do science things because they're complete black boxes that nobody, even their creators, know how they actually work. Or what to do if they stop working right. But... this is meteorology we're talking about. It's such an inexact science already that the only practical difference from actual occult fortune telling is that it's generally more reliable. When the weather forecast is wrong, the people behind it already couldn't tell you why. I say go for it. And when it inevitably goes haywire and starts spitting out garbage all the time, that's just called "back to square one."