WEATHER TORLCASTING 



perature, and moisture — at spatial 

 intervals of less than a kilometer and 

 nearly continuously in time (cf., pre- 

 sent spacing of about 400 km. and 

 time intervals of 12 hours). Eco- 

 nomically, this is a prohibitive re- 

 quirement, quite apart from its prac- 

 tical feasibility in terms of the 

 instrumentation and observing sys- 

 tems currently envisaged. 



Despite the present hopelessness 

 of straightforward applications of 

 physical-numerical methods to the 

 prediction of small-scale phenomena, 

 there is no doubt that opportuni- 

 ties exist for improved forecasting 

 through properly directed research 

 efforts. New developments in instru- 

 mentation and measuring systems — 

 doppler and acoustic radars and the 

 geostationary satellites, to mention 

 the most promising — utilized in con- 

 junction with special observing pro- 

 grams planned for the future, offer 

 great opportunities for advancing un- 

 derstanding of severe storms. From 

 this understanding, improved tech- 

 niques are bound to emerge. For in- 

 stance, it has been found that, unlike 

 the typical thunderstorm, very large 

 thunderstorms tend to move to the 

 right and slower than the steering 

 current. A better physical under- 

 standing of the cause of this behavior 

 would undoubtedly lead to superior 

 forecast techniques. 



Hurricanes — Hurricane prediction 

 has improved steadily during the past 

 decade or two. The improvement has 

 been brought about by the use of 

 aerial reconnaissance, radar, and, 

 more recently, meteorological satel- 

 lites to detect and track the hurricanes 

 and by the development of better 

 techniques for predicting their move- 

 ment. Skill is still largely lacking in 

 forecasting their development, but 

 fortunately they form sufficiently 

 slowly and usually far enough away 

 from land areas that the development 

 problem is seldom critical. 



In the past, extrapolation and steer- 

 ing methods have been the mainstays 

 in predicting hurricane movement. 



Currently, the most accurate method 

 is a statistical one that uses past 

 weather records to derive regression 

 equations relating future movement 

 to previous movement and to various 

 measures of the large-scale atmos- 

 pheric structure in the region sur- 

 rounding the hurricane. With this 

 method, hurricane positions can be 

 predicted 24 hours in advance with 

 an average error of about 100 nauti- 

 cal miles. While this figure leaves 

 considerable room for improvement, 

 there can be no doubt about the 

 enormous value of current forecasts 

 in terms of lives saved and property 

 damage reduced. 



Further refinement of the statistical 

 method and better observations of 

 the broad-scale features of the hurri- 

 cane environment could lead to some 

 improvement in hurricane prediction, 

 but it seems likely that the statistical 

 method has already approached its 

 limits of accuracy. Development of 

 numerical prediction methods would 

 seem to hold the key to further prog- 

 ress in this area. Methods of numeri- 

 cal prediction have already been tried 

 which forecast the large-scale steering 

 flow in the vicinity of hurricanes and 

 thereby allow better use of the steer- 

 ing principle. These methods have 

 met with some degree of success, 

 yielding errors comparable to, or 

 slightly larger than, the statistical 

 method. 



More significant and promising for 

 the future has been the development 

 in recent years of theoretical models 

 which, starting from assumed initial 

 conditions, are able to simulate many 

 important features of hurricanes. 

 These models have reached the stage 

 where they could be tested routinely 

 in the atmosphere if the proper initial 

 data — i.e., observations of wind, 

 temperature, and humidity at suffi- 

 ciently close intervals to resolve the 

 atmospheric structure in and near the 

 hurricane — were available. The in- 

 terval required is 100 kilometers or 

 less, well beyond present observa- 

 tional capability. However, it is con- 

 ceivable that geostationary satellites 



with visual and infrared sensoi 

 eluding sounders, could go a long 

 way toward providing the type of 

 information needed for carrying out 

 physical-numerical prediction of hur- 

 ricane formation, movement, and in- 

 tensity. 



Despite these promising theoretical 

 and observational developments, it 

 would be premature to enter on a 

 crash program of hurricane predic- 

 tion. Emphasis now must be put on 

 improving the physical basis of hur- 

 ricane models and on developing the 

 full potential of the geostationary 

 satellite as an observing platform. 

 Tropical field experiments planned as 

 part of the Global Atmospheric Re- 

 search Program (GARP) will assist 

 theoretical studies of hurricanes by 

 providing data suitable for investi- 

 gating the nature of the interaction 

 of mesoscale convective phenomena 

 with the larger-scale flow patterns of 

 the tropics. 



Synoptic Systems — During the 

 past decade or two, thanks to the in- 

 troduction of high-speed computers 

 and the development of numerical 

 weather prediction, remarkable prog- 

 ress has been made in predicting the 

 genesis and movement of high and 

 low pressure systems and tropos- 

 pheric circulation features in general. 

 Prognostic weather maps prepared by 

 computer now surpass the efforts of 

 even the most skilled and experienced 

 forecasters. 



Despite these successes, short-range 

 forecasts of specific weather elements 

 often leave much to be desired. In 

 part, the shortcomings are due to 

 small-scale phenomena which, as ex- 

 plained earlier, are not predictable, 

 except in a statistical sense, more than 

 a few hours in advance. But, in con- 

 siderable measure, they can also be 

 attributed to deficiencies or limita- 

 tions in the numerical prediction 

 models. The models are most suc- 

 cessful in predicting pressure and 

 wind fields; they are less successful 

 in predicting cloud and precipitation 

 amounts and patterns and in answer- 



95 



