Models and Forecasts: Sandy Edition

A few days ago, I blogged about the impossibility of predicting chaotic systems.  One characteristics of chaotic systems is that the behavior of those systems over time can be said to be sensitively dependent on initial conditions.  That means that small, seemingly insignificant differences in the inputs to the system may result in major differences in the behaviors of the system over time.  The current forecast models for hurricane Sandy are a great example of this.

Sandy originated in the western Gulf of Mexico and is currently crossing over Jamaica, moving essentially due north toward Haiti and then the Bahamas.  For the short-term forecast, nearly every forecast agrees, however, beginning Monday morning, there is a remarkable divergence in the forecast paths.  Below are the model tracks from the National Weather Service Environmental Modeling Center at the National Center for Environmental Prediction.

There is a remarkable divergence of paths on this map, with a spread of nearly 120 degrees in the direction of motion of the storm, leading to predictions ranging from a landfall on the Delmarva Peninsula through landfalls all the way up the rest of the Northeast coast, to models showing Sandy turning harmlessly out toward the central Atlantic Ocean. One of the most widely respected models, the GFS model, shows Sandy  looping east and then back west to come ashore in southern Maine, cross the White Mountains and dissipate over the Adirondack Mountains of northern New York.  In addition to track uncertainty, there is considerable variation in the forecast central pressure, which has implication for winds, surge, and rainfall and therefore for potential hazards.  MyFoxHurricane has some nice barometric pressure map visualizations of the European and GFS models.


Here is what the National Hurricane Center is saying about the long-range potential for Sandy:

THE INITIAL MOTION IS 015/11. THE SHORT-TERM TRACK FORECAST
REASONING REMAINS UNCHANGED FOR THE FIRST COUPLE OF DAYS...AS SANDY
SHOULD MOVE GENERALLY NORTHWARD INTO A BREAK IN THE SUBTROPICAL
RIDGE OVER THE BAHAMAS AS IT INTERACTS WITH AN UPPER-LEVEL LOW TO
ITS WEST. LATE IN THE PERIOD...THERE REMAINS CONSIDERABLE SPREAD IN
THE GUIDANCE DUE TO DIFFERENCES IN HOW MUCH SANDY INTERACTS WITH
ANOTHER TROUGH MOVING INTO THE NORTHEASTERN UNITED STATES. THE
ECMWF HAS BEEN CONSISTENT IN SHOWING MORE INTERACTION AND A TRACK
FARTHER WEST...WHILE THE GFS SHOWS SANDY TURNING EAST AND MISSING
THE TROUGH. OVERALL MORE OF THE GUIDANCE HAS BEEN TRENDING TOWARD
SHOWING MORE INTERACTION WITH THE EASTERN U.S. TROUGH LATE IN THE
PERIOD. GIVEN THIS TREND...THE NHC FORECAST HAS BEEN ADJUSTED TO THE
LEFT AT DAYS 4 AND 5...AND LIES ROUGHLY BETWEEN THE ECMWF AND THE
TVCA MULTI-MODEL CONSENSUS. NEEDLESS TO SAY...THE TRACK FORECAST
UNCERTAINTY REMAINS HIGH AT THE END OF THE PERIOD.

In other words, "We don't know, either."  Early next week, the storm will have moved far enough north that the uncertainty should have diminished somewhat.  It could be an interesting Halloween, stay tuned.