PART IV — DYNAMICS OF THE ATMOSPHERE-OCEAN SYSTEM 



other modern methods of communi- 

 cation and to improvements in or- 

 ganization and management. 



Evaluation of Forecast 

 Performance 



The "bread and butter" products 

 of the meteorologists, however, are 

 the hour-to-hour and day-to-day 

 forecasts of rain, snow, and tem- 

 perature for the general public and 

 of airfield and low-level flying 

 weather. These have not fared so 

 well. Reliable tigures to demonstrate 

 improvement of accuracy over the 

 past few decades are not available 

 for scientific judgment of the per- 

 formance of the meteorological com- 

 munity. Statistics do exist for a 

 large number of forecast targets 

 (cities, airports) for a limited time 

 period and for a few selected targets 

 for two or three decades. However, 

 this sparse data sample (which may 

 be quite misleading) and subjective 

 evaluations over the years suggest 

 that, in an over-all sense, short- 

 period forecasts have demonstrated 

 little improvement for several de- 

 cades. 



Routine Forecasts of Temperature 

 and Precipitation — Between 1942 

 and 1965, for example, the Chicago 

 office of the National Weather Service 

 (NWS) showed a steady improve- 

 ment in their combined weather and 

 temperature forecasts of about .33 

 percent per year. Large temperature- 

 forecast errors (10° F. or more) made 

 by the Salt Lake City office decreased 

 from one such error every 6 days 

 to one each 14 days. (Statistics for 

 the over-all temperature-prediction 

 capability of this office are not avail- 

 able.) A study of 260 NWS stations 

 discerned no noticeable change in 

 the ability to forecast rain "today" 

 during the first half of the 1960's, 

 but did note an increase of about 

 3 percent in the number of accurate 

 predictions of rain "tonight" and 

 "tomorrow." Scattered data such as 

 these suggest that the accuracy of 

 routine, classical forecasts of tem- 



perature and precipitation has in- 

 creased — but only very slowly. 



Hurricane and Typhoon Posi- 

 tions — Forecasts for special types 

 of weather events in some, if not 

 most, cases have fared better. For 

 example, from 1955 to 1965, the 

 NWS's 24-hour forecasts of hurricane 

 positions improved from an average 

 error of about 125 nautical miles to 

 one of about 110 nautical miles. 

 With regard to similar forecasts for 

 typhoons in the western Pacific, made 

 jointly by the Air Force and Navy 

 weather services, errors diminished 

 from nearly 170 nautical miles in 

 the mid-1950's to about 110 in 1969. 



Winds — Forecasts of winds for 

 high-flying aircraft are in the "new" 

 and specialized area. Between the 

 early and late 1960's, wind-prediction 

 errors at 20,000 feet dropped from 

 over 15 to under 11 knots. With 

 regard to similar forecasts for low- 

 flying aircraft (e.g., at 5,000 feet) 

 which, although part of a specialized 

 activity, can hardly be classed as ex- 

 otic, the reduction was about one- 

 half that for the higher level. 



Visibility and Cloud Cover — The 

 predictions of airfield ceiling and 

 visibility made by the Air Weather 

 Service (U.S. Air Force) are repre- 

 sentative of those made by other 

 services. Their statistics for the pe- 

 riod January 1968 through January 

 1970, compiled from the records of 

 200-odd airfields, show a small im- 

 provement, not necessarily repre- 

 sentative of performance improve- 

 ments of previous years. The accuracy 

 increased between 3 and 4 percent 

 for forecasts with time ranges of 3, 

 6, 12, and 24 hours. By 1970, the 

 forecasts were better than persistence 

 (no change from "time of observa- 

 tion") by nearly 4 percent at 3 hours 

 and nearly 8 percent at 24 hours. 

 Statistics for predictions of low-level, 

 in-flight clouds and weather are not 

 available, but are likely to be about 

 the same as those for airfield con- 

 ditions. 



Verification systems used for the 

 kinds of forecasts discussed above 

 necessarily vary considerably. Opin- 

 ions of atmospheric scientists regard- 

 ing the representativeness of the 

 data, and the value of the methods, 

 also differ widely. On the whole, 

 however, it can be said that the 

 status of forecasting is about the 

 same for cities, airfields, and low- 

 level flying — on the average not 

 bad, on occasion seriously deficient, 

 and improving very slowly. 



Factors Responsible for 

 Improvements in Forecasting 



The Norwegian Theory — The Nor- 

 wegian air-mass and frontal theory, 

 developed around 1920, began to 

 influence meteorological research and 

 application on a large scale by the 

 late l Q 30's. It represented a scien- 

 tific and conceptual revolution that 

 substantially improved the capabili- 

 ties of the atmospheric scientist. The 

 Norwegian theory was largely sub- 

 jective, and its application relied on 

 the individual skill and imagination 

 of trained and experienced practi- 

 tioners. A large part of the theory 

 was concerned with the distribution 

 and intensity of rain, surface tem- 

 perature and wind changes, and 

 cloudiness — elements that directly 

 influence man in his daily activities. 



The Rossby Theory — A second 

 revolution in concept was initiated 

 by Rossby in the 1940's. In con- 

 trast to the Norwegian theory, Ross- 

 by's approach emphasized the im- 

 portance of upper-level wind and 

 temperature patterns, whose influ- 

 ences on sensible weather were broad 

 and ill-defined. The theory was ob- 

 jective and lent itself to mathematical 

 calculation. Almost at the outset, 

 after refinements by a number of 

 atmospheric scientists, Rossby's basic 

 theory began to produce usable prog- 

 noses of upper-level wind and tem- 

 perature fields. Further refinements 

 produced relatively large-scale fields 

 of vertical air-movement from which 

 it has become possible to predict 



102 



