3 ms·

>If you go look at the climate science literature, you'll find that all different kinds of models and explanations have been proposed and studied. Those alterna

>If you go look at the climate science literature, you'll find that all different kinds of models and explanations have been proposed and studied. Those alternatives have been looked at in detail by people in the field, just like any other study. None of them have been found to better match the data. Because the system is chaotic and we can't predict it beyond a week. Maybe a day at best. Once you can predict hurricanes, volcanoes, and tropical storms, then let me know. You can over-fit any data. I guess ML still has to learn this basic concept from statistics 101. Anyways, I appreciate the discussion, thank you. I'm a bit irate this week. You can harness chaos in space because things are really far apart. It doesn't work on Earth, or our stock market would be linear.

solinent 6y ago
To summarize. The existing models just scratch the surface.
onion2k 6y ago
Because the system is chaotic and we can't predict it beyond a week. Maybe a day at best. Once you can predict hurricanes, volcanoes, and tropical storms, then let me know. You're confusing weather and climate. We can't model the weather very well. It's hard. We can model the climate reasonably well (our models fit historical data really well), but the models turn out to be bad at predicting the future because we underestimate how much of an impact we're having (the article is a good example).
- solinent 6y ago
  I'm not a reductionist so I don't conflate the weather and climate, but I'd consider hurricanes, volcanoes, tropical storms, and fluid flows in the ocean to be part of climate, not weather. Hurricanes have a huge impact on animal populations, volcanoes have a huge impact on atmosphere and temperature of the core, and tropical storms are still big enough that the sum total of them is not negligible at the climate scale. Chaotic systems sometimes don't have higher level descriptions--sufficiently chaotic systems can't be predicted at any level, and you can actually measure this. Look into the mathematical theory of chaos.

solinent 6y ago

To summarize. The existing models just scratch the surface.

onion2k 6y ago

Because the system is chaotic and we can't predict it beyond a week. Maybe a day at best. Once you can predict hurricanes, volcanoes, and tropical storms, then let me know. You're confusing weather and climate. We can't model the weather very well. It's hard. We can model the climate reasonably well (our models fit historical data really well), but the models turn out to be bad at predicting the future because we underestimate how much of an impact we're having (the article is a good example).

solinent 6y ago
I'm not a reductionist so I don't conflate the weather and climate, but I'd consider hurricanes, volcanoes, tropical storms, and fluid flows in the ocean to be part of climate, not weather. Hurricanes have a huge impact on animal populations, volcanoes have a huge impact on atmosphere and temperature of the core, and tropical storms are still big enough that the sum total of them is not negligible at the climate scale. Chaotic systems sometimes don't have higher level descriptions--sufficiently chaotic systems can't be predicted at any level, and you can actually measure this. Look into the mathematical theory of chaos.