5 ms·

Not science? Better tell the computer scientists and the computational physicists... Science is, at its most basic, our attempt to understand the world. As suc

Not science? Better tell the computer scientists and the computational physicists... Science is, at its most basic, our attempt to understand the world. As such, science employs many different tools and approaches, including numerical modeling. In particular. numerical analysis is a very well understood field of mathematics that allows us to rigorously analyze complicated problems by using one of the most innovative tools ever devised: the computer. Of course, one can make an inaccurate model as easily as one can write down an inaccurate equation or make a mistake in a derivation. That doesn't mean that the tool isn't a part of science.

LargeWu 18y ago
I'm not saying models aren't tools of science. They are very important. However, the most important part of science is the part about independently recreating and verifying results. You simply cannot do that with long range climate models without waiting for that time to elapse and seeing if the model predicted what it was supposed to predict. I'm not saying the models are wrong, or that climate change doesn't exist. What I'm saying is that calling the predictions of these models "science" is a bit misleading. The underlying factors maybe be based on scientific knowledge, but the models themselves are based on statistics, not science, because they can't be empirically verified.
- joeyo 18y ago
  It sounds like you are arguing that science needs to be observational to be "real." I'd argue that science needs both observations and theory and that either half alone is incomplete. Think of the models as a manifestation of the theory. It would be nice if you could solve a set of equations for the atmosphere in closed form, wouldn't it? But you can't so you run a numerical simulation. Anyway, it's hardly surprising that so much effort has gone into theoretical climate science, since, as you observed, it's difficult running a controlled experiment with the atmosphere.
ironsam 18y ago
One of the most ubiquitous and profitable types of models would be short-term weather forecasting. These are regularly wrong by several degrees, misjudging precipitation, etc, when trying to predict just 12 hours in advance. There's absolutely no point in looking at forecasts greater than three days, they're barely a guess, throw all the "accurate" equations and you want at it. [Disclaimer: I'm from the midwest - capricious weather.]
- glymor 18y ago
  Weather and climate forecasting confront different problems. I can't predict the behaviour of a single gas module but I can predict the behaviour of the gas as a whole as the randomness averages out. Climate forecasting is a lot more complex a system but the daily variability, at least, averages out in the same way. (Another analogy some people use is that we can predict that winter is colder than summer even if some days in the season aren't etc. this isn't really an argument for modeling but I suppose it's easy to grasp.)

LargeWu 18y ago

I'm not saying models aren't tools of science. They are very important. However, the most important part of science is the part about independently recreating and verifying results. You simply cannot do that with long range climate models without waiting for that time to elapse and seeing if the model predicted what it was supposed to predict. I'm not saying the models are wrong, or that climate change doesn't exist. What I'm saying is that calling the predictions of these models "science" is a bit misleading. The underlying factors maybe be based on scientific knowledge, but the models themselves are based on statistics, not science, because they can't be empirically verified.

joeyo 18y ago
It sounds like you are arguing that science needs to be observational to be "real." I'd argue that science needs both observations and theory and that either half alone is incomplete. Think of the models as a manifestation of the theory. It would be nice if you could solve a set of equations for the atmosphere in closed form, wouldn't it? But you can't so you run a numerical simulation. Anyway, it's hardly surprising that so much effort has gone into theoretical climate science, since, as you observed, it's difficult running a controlled experiment with the atmosphere.

ironsam 18y ago

One of the most ubiquitous and profitable types of models would be short-term weather forecasting. These are regularly wrong by several degrees, misjudging precipitation, etc, when trying to predict just 12 hours in advance. There's absolutely no point in looking at forecasts greater than three days, they're barely a guess, throw all the "accurate" equations and you want at it. [Disclaimer: I'm from the midwest - capricious weather.]

glymor 18y ago
Weather and climate forecasting confront different problems. I can't predict the behaviour of a single gas module but I can predict the behaviour of the gas as a whole as the randomness averages out. Climate forecasting is a lot more complex a system but the daily variability, at least, averages out in the same way. (Another analogy some people use is that we can predict that winter is colder than summer even if some days in the season aren't etc. this isn't really an argument for modeling but I suppose it's easy to grasp.)