26 SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. I39 



it, the results and comments representing this project. Even with 32 

 stations, the United States is so vast in area and so varied in con- 

 trasting conditions that with the fullest use of my results no adequate 

 country-wide coverage of the expected precipitation to 1967 can be 

 made. As stated above, confidence in the forecasts must depend 

 largely on the fidelity with which the first half of the forecast, 1950- 

 1958, inclusive, fits the observed record. 



Table 10 presents in parallel columns for all 32 stations the monthly 

 percentage departures of forecasts and observed records, 1950-1958, 

 from the normals given in table 9. 



That readers may see from a graphical standpoint to what degree 

 the forecasts represent the events, I present figure 10. It gives the 

 march of forecasts and events from 1950 to 1958 for Cincinnati, one 

 of the best, and Denver, a less favorable station. 



Table 9a. — Percentage departures (O-F) 1950-1958, from table 



City Abilene Augusta Bismarck Charleston Cincinnati 



% A —12 — 17 —6 —11 +4 



City Detroit Eastport Helena Independence Little Rock 



% A —4 +23 —11 —17 +4 



City Natural Bridge Peoria Sacramento Salisbury Salt Lake 



% A —7 —6 —4 —5 —7 



City San Bernardino Santa Fe St. Louis St. Paul Thomasville 



% A +10 —17 —8 —11 —9 



Figure 10 shows for a more favorable and a less favorable station 

 a graphic view of data taken from table 10. 



A glance at figure 10 shows for both cities an obvious similarity of 

 the features of the forecasts and of the events for the majority of 

 months covered. There are, to be sure, differences in amplitude of 

 features observed and forecasted. In many cases the forecast, built 

 on average conditions of about 1,000 previous months, hits the fea- 

 tures found in the observed record from 1950 to 1958 on the exact 

 months. But in the better station, as well as in the worse, there occur 

 relative displacements of features common to both forecast and event. 

 These displacements are rarely as great as 5 months for any station, 

 but may extend through durations sometimes as great as several years 

 before returning to agreement. 



Displacements of features. — Several years ago I published the ac- 

 count of a forecast for 104 years of St. Louis precipitation, including 

 a comparison with the observed records. I quote from my discussion * 



* Text continued on page 44. 



