The Man And The ButterflyAdded : Friday at 10:15 Many of you will have noticed that when we have an event approaching, for example Easter or Christmas Day, we issue discussions based on the latest numerical weather models. At the same time, we try to explain that when these events are two weeks away then you should never take numerical weather output as displayed because at that range the accuracy is virtually impossible.
If we take a look at the chart for Easter Saturday issued today then we will eat our hats if that is the same chart which appears on Good Friday :-
The reason for this is that the further ahead you go in numerical models the larger the risk of errors. You, me and the cat know this as the butterfly effect which was discovered purely by accident by a chap called Edward Lorenz in the early 1960's.
Edward Lorenz was a meteorology professor at MIT (Massachusetts Institute of Technology) and one day was running a computer program to simulate weather patterns. While the program was running he popped away to make a coffee (nice to know meteorologists still have the same habits) but when he returned he noticed something strange.
The model was based on twelve variables such as temperature, wind, pressure etc and could show the variables changing over time. On this day however, Lorenz had been repeating a model run he had previously made, but rounded off one variable from .506127 to .506. This tiny alteration completely changed the output of the model the further ahead the model ran.
Lorenz published a paper on this in 1963 which had the catchy title of "Deterministic Nonperiodic Flow" which sadly never really made it big at the time and it wasn't until he published a paper years later called "Does the Flap of a Butterfly's Wings in Brazil Set Off a Tornado in Texas?" which brought chaos theory to the general masses.
By 1965 Lorenz had discovered the cause of this issue.... advection. Quite simply the uneven movement of heat, moisture and other properties in the atmosphere. Tiny errors in starting conditions, from the imprecise location of a single cloud to a model locating the centre of a hurricane 100 feet too far North would mean the error becomes accentuated over time as the model ran further and further ahead.
These days, we use ensemble forecasting to slightly adjust the starting conditions of the model run in order to take into account these errors.
It's why after T+372 hours the model with it's different starting conditions produce a range of output like this :-
×View All GFS ► PostStamp Ensembles Images
That's a lot of butterflies!
So, when you next have a cup of coffee today, raise your mug to Edward Lorenz and the chaos theory.
METEOROLOGIST : MARSH |