,DOF.S 



sphere. This is due partly to their 

 spacing and frequency and partly to 

 errors inherent in the equipment. 



In addition, there are data voids 

 in the areas surrounding the United 

 States, such as the Gulf of Mexico, 

 the Atlantic waters adjacent to the 

 east coast, and portions of Mexico 

 and Canada. All of these contribute 

 to serious lateral boundary prob- 

 lems, the most pressing being the 

 Gulf of Mexico. Texas, Louisiana, 

 Mississippi, Alabama, Florida, and 

 Georgia are all high-incidence areas 

 for destructive tornadoes, and the 

 lack of any direct meteorological data 

 over the Gulf of Mexico has made 

 objective analysis and prediction dif- 

 ficult. 



To augment the conventional sur- 

 face and upper air networks, use has 

 been made of radar and satellite 

 photographs. The processing and 

 display of either method is still in 

 its infancy; considerable experimenta- 

 tion will be required to obtain con- 

 tinuous readout of radar- and satel- 

 lite-produced information. At present, 

 neither the radar nor satellite output 

 is woven into conventional analyses 

 in a systematic and objective manner. 



Forecast Methods — Present meth- 

 ods are largely subjective, drawing 

 heavily on case studies and the ex- 

 perience of the individual forecaster. 

 This is slowly being replaced by 

 objective, computer-oriented methods, 

 partly dynamical and partly statisti- 

 cal. (See Figure V-10) Considerable 

 improvement is needed for either 

 method. The most promising avenue 

 for dynamical methods concerns the 

 development of a fine-mesh primitive 

 equation model for multi-layers. Such 

 a model would be of limited value at 

 this time because of the data limita- 

 tions noted, but it will become in- 

 creasingly important as the average 

 spacing between stations is reduced. 

 The statistical approach involves a 

 search for predictors through the use 

 of multiple-screening regression tech- 

 niques. It has not been possible to 

 gather all of the possible predictors 



Figure V-10 — SEVERE WEATHER WARNING 



The table illustrates an experimental severe weather warning of a thunderstorm 

 cell moving from 256° at 23 knots. The warning gives the time of closest approach 

 to airports near the forecast path. It also gives the distance and the direction of 

 the echo from the airport. Finally, it estimates potential error of the forecast in 

 terms of the time period of closest approach. This warning was prepared auto- 

 matically by a computer using statistical properties of radar echoes such as those 

 measured in Figures V-8 and V-9. 



along with tornado occurrences, so 

 this approach will require further 

 work. 



Research and Development — Com- 

 paratively little research on forecasts 

 is being performed in this country. 

 In allied fields, considerable research 

 and development is under way on 

 hail suppression, doppler radar, 

 LIDAR (light detection and ranging), 

 and remote-sensing techniques. Im- 

 proved equipment and techniques will 

 have application to the warning prob- 

 lem. 



Modeling 



Several theories have been ad- 

 vanced to explain the Great Plains 

 tornado. These theories do not, how- 

 ever, explain the hurricane-induced 

 tornado, the western U.S. tornado, 

 or the waterspout. A great deal more 

 work is needed in modeling tornado 

 formation. 



Prediction Techniques 



The same problems apply to the 

 warning as to the forecast. A vast 

 majority of reported tornadoes do not 



come close enough to any of the 

 reporting stations to be detected, ei- 

 ther visually or by instruments. 



Radar Detection — The radar net- 

 work is being expanded throughout 

 the United States, using 10-centimeter 

 radar. This is effective to 125 nautical 

 miles in defining severe thunder- 

 storms capable of producing torna- 

 does, but even a highly skilled radar 

 operator cannot clearly identify a 

 tornado by radar or give a 15-minute 

 forecast that a certain cloud will 

 produce a tornado. Certain charac- 

 teristic shapes provide some informa- 

 tion on the probability of tornadoes, 

 but the pattern is not present for 

 every tornado. 



Instrument Detection — There are 

 no mechanical methods at this writ- 

 ing that can make an objective dis- 

 tinction between the pressure fall or 

 rise produced by a strong squall line 

 and that produced by a tornado. 

 Even if there were such a device, the 

 spacing required to insure its useful- 

 ness would be prohibitively expen- 

 sive. 



Volunteer Spotters — Most warn- 

 ing is based on a combination of 



147 



