Mark Eubank has been working on a long term (1-year) weather forecast for over 40 years. At one point he nearly gave up because he decided it would be impossible. Then he refocused on a 6-month forecast using the remote sensor technologies that have been developed over the past 20 years. I still think it is too complicated a system to adequately model over anything but 2-weeks for the time being. You can get a good forecast for one or two days, but anything over three is very flexible still.
When Jurassic Park came out, Chaos Theory came into vogue and people liked to talk about the butterfly flapping its wings in the amazon causing tornados in Oklahoma. I don't think people really understood the point of the conversation though. The idea behind chaos theory is that no matter how good the model is, eventually there will be divergence when you try to model very specific outcomes over a long period of time. Because we can't incorporate every input into a model, we have to make simplifications that result in sources of error. The butterfly didn't cause the tornado in Oklahoma, though it may have resulted in a slight deviation in the path that could not be predicted, resulting in a failure of the intended purpose of a modeled system.
Models also work off of past observations. When conditions in an area match up with past experiences, we tend to predict that the same thing will occur. For instance, when we have a heavy wind and the barometer drops, we generally expect a storm to be moving into the area and anticipate rainfall. Most weather models do the same thing, but from much greater distances, using much more complicated instruments, but it is all about observation. Sometimes, the weather system deviates significantly from past experiences though, as we saw this past winter when past experience caused NWS and most other forecasters to predict a below average winter. With this additional data, along with an attempt to determine what was different from the past that resulted in wet conditions, the models are being updated and improved. Over time, the models will get much better in their percent accuracy, but will never be 100%.
The best weather instrument I have ever seen is still the weather rock that hangs from my dad's cabin porch. It has a means of interpretation next to it:
- If the rock is warm, it is sunny.
- If the rock is wet, it is rainy.
- If the rock is white, it is snowy.
- If the rock is swaying, it is windy.
- If the rock is bouncing, it is an earthquake.