ABROLOGY OF TROPICAL STORMS 
of the graph where the zonal motion approached zero. 
The mean deviation was about 1 degree of latitude 
per day. 
Steering Principle. It is no accident that the study of 
motion of disturbances has stood in the foreground of 
literature on the tropics. Contrary to other aspects of 
meteorology, this aspect of forecasting is more de- 
veloped for the tropics than for middle latitudes. The 
reason for this can be seen readily when we examine 
what Petterssen [24] has termed the ‘‘area of uncer- 
tainty.’”’? Most tropical storms are small and have a tight 
circulation compared to extratropical cyclones. An error 
of 1-2 degrees of latitude in forecasting the position of 
a middle-latitude storm affects the forecast verification 
but little. In the tropics, however, an entire storm is 
often contained within this distance. An error of 1-2 
degrees will make all the difference at any given station 
between wind of hurricane force and no wind at all. 
The steering concept at first was used in a directional 
sense only. Hurricane tracks appeared to parallel the 
sea-level isobars over the Atlantic to a large extent. 
Tt is only since the late 1930’s that some forecasters 
attempted to obtain the speed of displacement from 
the upper wind field. Difficulties arose at once in choice 
of steering level since speeds varied with height. This 
difficulty also applied to direction whenever the wind 
direction changed with height, especially near the sub- 
tropical ridgelme. Simpson [38] pointed out that there 
has been a tendency to shift the level thought repre- 
sentative of steering higher and higher as the upper- 
wind observations were extended upward with the ad- 
vent of rawins. The difficulties of making a proper 
choice of steering level have led to doubts on the part 
of many forecasters as to the usefulness of the method 
in principle. 
In the writer’s opinion, the difficulties result mainly 
from application of steering techniques to situations in 
which they should not be used. In a strict sense, steer- 
ing refers to the forces that guide a very small disturb- 
ance in a broad zonal current that is steady and usually 
considered uniform along all space axes. The disturb- 
ance also does not change shape or intensity with time. 
These restrictions never are fulfilled in reality. Two 
questions come up: Under what actual conditions can 
the simple theory be applied without much loss of fore- 
cast accuracy? and To what extent can theory itself 
modify the restrictions listed above? Not much has been 
tried along the latter line. But the problem is quite 
urgent. It is easy to see that formal solutions of the 
equations of motion will not be possible if many re- 
strictions are dropped. But a numerical integration 
procedure may carry quite far, provided some good 
information exists as to what quantities to put into a 
computing machine. In this respect, studies of the 
kind attempted by Riehl and Burgner [31] may provide 
the necessary background. They can determine which 
theoretical restrictions are immaterial for the forecast 
and which impose limitations. It is almost certain, for 
example, that variations along the vertical axis will 
turn out to be important. This is amply indicated by 
the literature of the last ten years cited above. Probably 
911 
the concept of “‘steering level” must be replaced by 
that of “steering layer.” Tropical storms extend through 
almost the entire troposphere. It is quite reasonable to 
expect that some integral function of conditions at all 
heights will serve much better than conditions at one 
level alone. 
Recurvature. The motion of tropical storms from the 
tropics into the middle latitudes can take place in 
several ways. Sometimes the forecast is very easy, as 
when upper winds are nearly uniform through a deep 
layer and gradually turn from east through south to 
southwest as we follow them around the western edge 
of a subtropical high pressure cell. For such cases, 
methods like those developed by Yeh [44] should be 
fully applicable. Important difficulties arise whenever 
winds are weak and/or when the vertical wind-shear is 
large. In such situations we can no longer speak of 
steering. 
The literature on the subject, referred to above, gives 
empirical forecast rules for several types of tropical 
storms. No one, however, has claimed to have devel- 
oped a perfect qualitative forecast method. The occa- 
sional failures of all systems stand out. Quantitative 
calculations of recurvature from empirical data have 
not been attempted or have been unsuccessful. Theory 
has bypassed the problem. Thus the state of knowledge 
with respect to the many situations when recurvature 
hangs in the balance is unsatisfactory. 
It is not easy to make suggestions for future work 
that can readily be carried out. The trouble stems from 
two sources. One of the missing elements in the recurva- 
ture forecasts is a knowledge of the internal forces 
within a storm. These cannot be measured at all at 
present. The other difficulty lies in the fact that the net 
external force operating on storms in recurvature posi- 
tions is a small resultant of large opposing forces; this 
is also indicated by the slow motion of storms at re- 
curvature points. It is practically impossible to evaluate 
the net force from north or south acting on a storm even 
at one level, to say nothing of an integral from sea level 
to the tropopause. The station network necessary to 
calculate such resultant forces simply does not exist 
anywhere. 
Thus we must again demand more data if physical 
analysis of recurvature is to make much headway. 
Unfortunately, it is probable that the great density of 
stations really necessary for this problem will never be 
realized on a routine basis because of cost. For practical 
purposes, future recurvature work should follow statisti- 
cal rather than physical approaches. Mitchell [21], for 
example, has already shown the importance of the rate 
of motion of extratropical troughs. Riehl and Shafer 
[83] have emphasized the vertical structure of the sub- 
tropical high-pressure belt. Together with the paper by 
Klem and Winston [19] this literature yields a great 
number of starting points for statistical correlation. If 
future efforts are applied in this direction, it may be 
possible to bypass the obstacle furnished by the balance 
of forces at the recurvature point. 
Large Storms. The problem of forecasting the motion 
of large storms occurs mainly in the Pacific Ocean in 
