Staff Writer, Paul, Sarah Crespi discussed on Science Magazine Podcast
Welcome to the science podcast for February. Fifteenth two thousand nineteen I'm Sarah Crespi in this week's show. I'm talking to staff writer, Paul loosen about whether there's a hard limit on. Whether prediction is fifteen days as far as we can get and Meghan, Cantwell and Trig Faucheux discuss his paper on an Thomas bought. That's mapping final plankton off the coast of Norway. How far out can we predict the weather we've gotten pretty good at it? But is there a limit staff writer, Paul vision is here to discuss a study that says, yeah. About two weeks, that's the maximum for the mid latitudes. Anyway. Hi, paul. So the app on my phone gives me ten days of weather, and it's pretty good. I expect a little variation in those like ladder days. But is that the best that we can do right now is the best we can do right now the top and models from Europe, and the US really they max out at skillful predictions about ten days we've been getting better at this over the last few decades, what has been that rate of increase each decade. They've added one additional day of predictability, this is a really great success. But is that gonna stop we can't just keep adding one day of prediction a decade forever? Probably not it would be nice. It'd be great for job security for forecasters. Yeah. But this is something meteorologists have wondered about for fifty years, and they have often said maybe it's about two weeks. And this new study seems to say, yeah, it does seem to be about two weeks. How can a study prove a negative that we can't get better than you know, a certain limit. It can't to a certain extent. You know there. There's always gonna be ways you could improve this method. But yes, so they took the last version of the European model which contributed the best in the world and ran at about a hundred and twenty times, which is not really operationally expensive, these are supercomputer type models. So they ran out hundred twenty times a day change something about the initial conditions or what they were measuring or what they were excluding in these different sessions. Yeah. So a lot of the air that we see in miles could also drive from our uncertainty about what were observing today. So we don't have these perfect understanding of the weather and so trying to tease out. What is a problem do? Our imperfect knowledge versus the inherent chaos of the atmosphere has always been issue. But if you do this kind of big ensemble suite of models, Ryan them all together, you can artifice -cially make it seem like you have no error in the observations. You can narrow the range reduced to ten percent from the current uncertainty and that creates kind of artificial certainty. This is kind of presuming that or even better in the future. Measuring that we know more about how weather works and that we're better at inputting that data into these models. And when all those assumptions, we still get this two week limit. Yep. Implausibly better. We'll probably never get a good and still you comes to this two week limit that Edward Lorenzo who is a famous meteorologist mathematician. Father of chaos theory is fly Mr. butterfly effect. He hostile laid in a nineteen sixty-nine study that probably seem to be about two weeks law. People quibble about the model said may be didn't, you know, really? Like the atmosphere. But now these new models that do look a lot like that Mr. to the point of even having convection, the kind of cloud thunderstorm systems for me in them, which is hasn't been the case in the past still run into chaos after two weeks and no better than your me. Guess in base off climate records. Let's go back to the butterfly effect. Great moot. Natta great for people who don't know what this was about. I was kinda surprised that there was a paper kind of just thought it was like a thing that people say from scifi there was that too freshest butterfly as the crush history. This is the flap of butterflies eggs affecting weather. Yes. Yes..