Thursday, October 31, 2013

Spaghetti (without meatballs)

When we look at the weather forecast on our handy dandy weather app and it says "50% chance of rain today," how was that forecast made? And why in the world would the meteorologist say there is a 50% chance of rain." It might sound like a ridiculous forecast, but that forecast actually means something very important. Sometimes we take the forecast for granted, but the process is quite complicated and quite meaningful.

Weather models take terabytes of data and computer power to calculate solutions to complex equations that describe, approximately, the physical interaction in the atmosphere. The initial weather data and the equations are the basic components  needed to start a weather model and make a forecast.

Since we can't know the exact condition of the atmosphere, we try to make some guesses. One method for making forecasts is to run a model several times with slightly different initial conditions. A collection of different models is called an ensemble forecast. Because the atmosphere is a chaotic system, each solution can be quite different. When we plot the solutions from each model run we get what we call "Spaghetti Plots." You can see why in the figures below--they look like strings of spaghetti noodles! Each line in the figure below represents a different solution for each of the model runs. This first figure shows the current conditions of each model before any forecast is made.
All these models generally agree with one another, except for some discrepancy along the 582 height line off the coast of the Baja peninsula and out in the middle of the Atlantic ocean. When we look at solutions forward in time, however, the errors in the models begin to grow--the spaghetti starts to spread out! Below is the forecast 24 hours after the model started. Each model tends to show the same features, but there is uncertainty in the exact location of the features. Keep in mind that this map shows the entire United States, so a little perturbation between models would be the difference between a storm hitting Payson instead of Logan. 
These plots don't show precipitation, but precipitation ensemble forecasts are similar. A 50% chance of rain means half of the models show a possibility of rain while the other half show no rain. 

The longer these models are run the larger the errors grow. Four days after the model is initialized each model run puts the troughs and ridges in different locations.

Keep running the model to 16 days after the forecast is initialized and just about anything is possible. It's just a big bowl of Spaghetti!

No comments:

Post a Comment