OK, so, weather models predicted that a big winter snow storm would hit the Denver area late next week. This made me rather nervous because I have to make a drive on one of those days, and I don't want to cross through any mountainous regions when it's snowing. But as luck would have it, somewhere, some time ago, a butterfly (the thing on the left) flapped its wings, and now the storm is gone!
ACTUALLY, before my friend, the idiot, calls me out on it, it is actually NOT TRUE that either a butterfly or a fly on butter could influence the weather on this scale (yes yes, I was not telling the truth, but rather trying to make people laugh the best that I know how...). He is absolutely correct on this. At the very small scale of wing flapping, there just isn't enough "umph" to affect the weather on larger scales. The little poofs of air produced would be quickly wiped out by larger scale processes.
But, the fact that weather models just a few days ago predicted this big snow storm and now they are predicting hardly anything happening at all points to the chaotic nature of weather. Are the models wrong? Not really, no. It's just a reiteration of my previous point that no model can account for every variable and every perturbation.
As we get closer to the actual date for the prediction, the models tend to get better at getting an accurate forecast. This is because the perturbations haven't had time to grow and drastically change the weather yet. These perturbations don't lead to different outcomes popping up immediately - instead, differences between predictions grow more slowly. So, if the weather models make a prediction for tomorrow, this prediction is more likely to be accurate than a prediction for a week from tomorrow.
So, to summarize, there are two bottom lines:
1) Background Dominated doesn't understand insects
2) Trust weather predictions for the near future to be more accurate than those for the far future.