You may be wondering why we meteorologists can be so confident when there are so many different possible paths. How can we get so many different projections, and how can we be sure that the storm won't go over our heads?
Looking at the possibilities for Hurricane Irene, you may be confused by the lines going all over the place:
A bit confusing, no? You may be wondering why we meteorologists can be so confident when there are so many different possible paths. One turns left, and goes right over D.C., continuing north through Pennsylvania and up to Canada. On the other extreme, one doesn't even touch North Carolina. How can we get so many different projections, and how can we be sure that the storm won't go over our heads?
First, let us change gears for just a second: You're driving around the Beltway (you’re brave) and you want to speed up a bit, so you press the car’s accelerator. The car’s speed increases, and you hold the pedal steady to maintain a constant speed. Your car, its accelration, speed and braking are all linear systems. A small initial force on the accerator pedal produces a very preditcable corresponding (linear) change in the car’s speed.
Now image you are drving (this time, not on the Beltway) and you press the accerator. This time, your car’s speed begins to increase. You ease off the accelerator, but the car continues to speed up. Suddenly, it slows down. Then, a few secords later it speeds up again. Your foot is now off the accerator, but the car continues to change speeds on its own — accelerating, slowing down but its overall speeds increase. A small touch of the gas pedal produces all sorts of weird changes in the car’s speed. This is a non-linear system — thankfully, this only exists in this imaginary example — and soon you have no idea what this car will do next. It has become unpredictable and chaotic.
You’ve probably heard of the "butterfly effect,” the famous tongue-in-cheek term coined by Dr. Edward Lorenz, a famous meteorologist, to describe the non-linear, chaotic behavior of our weather. The weather is indeed an incredbily complex system, with many different physical processes — winds, evaoration, friction, radiation etc. — influencing the future state of the atmosphere. Small differences on very small scales or initial conditions — such as the “breeze” from the flap of a butterfly’s wings — can eventually grow (remember our “nonlinear car” example above). Suddenly, these small things turn into big factors, especially when discussing the future or predicted state of the atmosphere. Lorenz concluded that detailed long range weather forecasts were not possible.
We now can measure the atmosphere in greater detail than possible 40 years ago when Lorenz decribed the chatoic nature of the weather. But even making local initial measurements of the atmosphere — miles — will only extend our ability to make detailed weather forecasts for days.
But if small changes in our initial weather or initial measurements all still produced similar — if not the same — weather forecast results, we would have more confidnece in those forecasts. We could say that for some weather and weather patterns, little changes or errors in our measurement of wind, temperture, humidity and so on might not matter that much. The atmosphere and our weather is not as chaotic on some scales as others.
The core of the forecasts for Hurricane Irene is numerical weather prediction, or NWP. The physical laws — the basic science of weather — are known. We actually solve a number of fundamental equations of physics that govern the weather in order to predict the future of weather. Detailed measurements from satellites, the ground, radar, aircraft, ocean buoys and ships provide the “inital” data, and supercomputers use all this data to solve the equations and give us meteorologists numerical outputs which we use — along with our won knowledge, experience and observations — to produce the forecast you hear, read or see.
If we purposly make small changes to the inital data that we use to generate our numerical (computer model) forecasts or make very small changes in the equations we use, we can get a sense of how chatoic the system is based on our range of results. We can also know how confident we are about the forecast.
Let's look again at the above image of Hurricane Irene's possible paths. Do the lines look a bit like spaghetti to you? We meteorologists call them spaghetti diagrams (hence the title of this post).
These models of all possible paths are becoming a more powerful tool to forecasters. In addition to providing us information about the “spread" (in other words, the uncertainity) of weather variables, the mean (or average) of the ensemble gives us useful information to campare with single models. Ensembles are now used in not only hurricane track forecasts,but winter precipitation forecasts, as well as day to day and long range forecasts.
The first thing we do is eliminate the outliers. They're highly improbable (but we won't say impossible). Harrisburg, you're safe. Outer Banks, you're not. By looking at the "clumps" of spaghetti, and the more-likely tracks, we can forecast the path with a lot more certainty.
In the future, ensemble forecasts will be used to help us better forecast very short term weather, such as summer thunderstorm squall lines, and even help us give probability estimates of different events because of global changes. The stastical and probabilistic information we can gather from ensemble models is also quite helpful to anyone who has to make weather-related decisions — from power companies to drivers — who hope thier vehicles continue to operate in a linear manner.
Spagetti, fun to eat and fun to have as a weather forecasting tool.