When Good Models Go Bad

When Good Models Go Bad
On 29 December 2000, forecasters from Washington to Boston were watching with concern a powerful upper-air disturbance moving southeastward from south-central Canada into the Midwest. All the forecast models were predicting rapid cyclogenesis and subsequent explosive deepening somewhere on or near the U.S. Mid-Atlantic coast during the night of the 29th and morning of the 30th. With cold air in place over the Northeastern U.S., it was clear that a major snowstorm was about to occur, though it was not clear where. All the models agreed there would be a sharp back edge to the precipitation swath and that there would be rain to the east of the cyclone center, which might or might not track across Massachusetts. The location of cyclogenesis and the subsequent track and orientation of the storm would result in significant differences in where the impacts would be in the Washington to Boston urban corridor. Thus, the forecast problems of the day were:

How far south and how close to the East Coast would the initial cyclogenesis take place?

How rapidly would the storm deepen once it developed, and what path would it take?

How much precipitation would wrap around the cyclone as it developed, and how would the precipitation bands be oriented?

How much moisture would be available to wrap into the storm after a previous southern-stream wave had swept out most of the moisture?