I was just thinking about the Weather Channel and their *10 Day Weather* page. It got me thinking, how can we know what the weather will be in advance and how accurate is it?
In: Earth Science
Most weather is driven by very large and fairly predictable patterns in the atmosphere. They aren’t hard to see on a weather map and they usually move in basically the same ways. So even you, with some decent weather maps and a few weeks to study them, could probably make reasonably accurate forecasts most of the time.
It’s not perfect. Sometimes unusual patterns set up that break the usual rules. For example, Hurricane/Superstorm Sandy was the result of storm systems that usually move in a regular west-to-east pattern instead backing up east to west and picking up a hurricane in the process. So to make better predictions, we have sophisticated models that take what we know about the forces at play in the atmosphere and try to simulate them. Different models do this differently, and so how much the models do or don’t agree can tell you about how complex a situation you’re looking at. When models all agree, they’re usually right; when they all disagree you should be very uncertain about what will happen.
For example, here’s an excerpt from [a forecaster’s discussion](https://www.nhc.noaa.gov/archive/2020/al13/al132020.discus.016.shtml?) of their reasoning for a forecast path for Hurricane Laura last year (I’ve added some bracketed notes):
> Laura continues to move briskly west-northwestward or 285/18 kt [direction/speed]. The track forecast reasoning remains the same as the previous advisory. Laura should continue to move west-northwestward to the south of a deep-layer ridge [“big high pressure area” – hurricanes tend to move around them] that is forecast to build westward across Florida and the eastern Gulf of Mexico during the next day or two. The track guidance [models] has continued to edge southward for the portion of the forecast near Cuba, and the NHC [National Hurricane Center, the agency issuing this forecast] forecast has again been moved in
that direction. Laura should continue moving west-northwestward
over the southeastern Gulf on Tuesday, but a turn toward the
northwest is expected Tuesday night as the cyclone [i.e., Laura] nears the western portion of the ridge [the aforementioned big high pressure area]. A northwestward to north-northwestward motion should then continue around the western portion of the [high-pressure] ridge until the cyclone reaches the northwestern Gulf coast. The latest run of the ECMWF [model] has shifted significantly eastward, however its ensemble mean and many of the stronger ensemble members [ensemble = same model run a bunch of times w/diff data] remain farther west as a stronger cyclone is likely to be steered more westward by the deep-layer ridge [again, that high pressure blob]. The GFS, UKMET, and HWRF [other models] remain close to the previous NHC track, so little change was made to the official forecast was made after 48 hours.
So in this forecast, the big question occupying the forecaster’s mind is “how strong and exactly where will the ridge be?”. One model, the ECMWF, says it’ll be further east, which will let Laura go further east; the other models agree that it’ll be further west. As it happened, the other models were right, and Laura made landfall in western Louisiana almost exactly in line with the other models’ forecast.
We know how pressure, temperature, humidity and sunshine work to create weather, and we have very complicated simulations that rely on field data collected by stations and satellites.
An example:
There’s a beach. A sunny beach with cold water in front of it. When the sun shines on the sand, it gets hot very quickly and heats up the air above it, but the water stays cold. Warm air rises up, and the surrounding air fills the “hole” the rising air has left. This air comes from the cold water, and moving air is what we call wind. Now, depending on how sunny it is, we can calculate how much warmer the beach will get than the water, and then get the wind speed out of that. How do we know how sunny it is? Through satellite pictures that show clouds, and through sensors at the ground.
There’s lots and lots of complicated models and atmospheric mechanisms, but with those we can predict many many things. Since we have loads of data, experience and computing power, we can do large simulations that predict roughly up to two weeks. That roughness adds up though for each day, because one cloud or one degree can already make the difference, but most models are accurate up to three days.
Latest Answers