HURRICANES GENTRY AND SIMPSON 315 



pending upon the amount of data available. Professor Riehl and 

 Commander Sanborn of Project AROWA made a series of forecasts 

 during the 1955 season in which they found that the average error for 

 47 forecasts made 24 hours in advance was 63 nautical miles. These 

 results are somewhat better than those obtained using the AROWA 

 technique at the various hurricane forecast offices. 



Numerical prediction. — During 1955 and 1956 extensive tests were 

 made using numerical prediction techniques to forecast movement of 

 hurricanes, and these forecasts have competed in quality with those 

 made by other methods. Forecasts made by the numerical prediction 

 technique in current use are also based on a form of "steering." This 

 technique has the advantage over many of the others, however, in that 

 it is purely objective. As improved and more realistic mathematical 

 models of the wind circulation become available for use in numerical 

 prediction, there should be continued improvement in forecasts made 

 by this method. 



SPECIAL PROBLEMS IN FORECASTING HURRICANE MOVEMENT 



So long as the hurricane continues to move along the same path at 

 approximately the same rate of speed, either the persistence of the past 

 movement or the statistical method of forecasting usually gives good 

 results. Unfortunately, in the areas near the United States coast, 

 hurricanes frequently change their direction of movement and will 

 often accelerate or decelerate in their rate of forward movement. 

 Hurricane Carol of 1954 and hurricane lone of 1955 are two storms 

 that illustrate one of the problems continually facing the forecaster. 

 While hurricane Carol was located south of Cape Hatteras, N. C, it 

 had a net forward speed of less than 3 m.p.h. over a period of 60 

 hours. Then, within a period of a few hours it had accelerated until 

 it moved at a rate of about 40 m.p.h. Hurricane lone, by contrast, 

 moved toward the coast of North Carolina at a rate of 15 to 20 m.p.h. 

 until the time it crossed the coastline. Instead of accelerating, as 

 hurricane Carol had done, it slowed down, made several loops in its 

 course, and had a net forward speed for several hours of only 2 to 3 

 m.p.h. Whereas hurricane Carol had moved in the direction be- 

 tween north and north-northeast at about 40 m.p.h. from the vicinity 

 of Cape Hatteras, N. C, to Long Island, N. Y., hurricane lone turned 

 suddenly toward the east and east-northeast as it left the coast of 

 North Carolina just a few miles south of Norfolk, Va., and moved out 

 into the Atlantic ocean. 



Figures 7 and 8 are surface maps of hurricanes Carol and lone when 

 they were near Cape Hatteras, N. C. An inspection of these maps 

 will reveal few clues as to the widely diff erent types of movement that 

 were to follow. It was pointed out earlier that we have learned that 

 the movement of hurricanes is largely controlled by the flow of air 



