342 REFRACTION AND REFRACTIVE INDEX MODELS 



how much of the observed correlation of the heterogeneous sample is 

 derived from correlation between "classes" of data (in the statistical 

 sense). For a more thorough treatment of the CRPL Standard Sample 

 and the associated regression coefficients for range error and elevation 

 angle error, the reader is referred to Bean and Thayer [19]. 



For the Standard Sample, the standard error of estimate is equal to 

 the standard prediction error within ±1 percent over the range of A^s from 

 200 to 470, and will be used interchangeably with the latter. 



8.3.4. Comparison With Independent Data 



Before turning to an examination of the experimental refraction data 

 and the degree of success realized in applying the theoretical prediction 

 model to those data, it seems appropriate to examine the accuracy of the 

 prediction model when ai^plied to some independent theoretical (i.e., ray- 

 traced) data. For this purpose, four check stations were selected which 

 were not only independent in the sense of not having been included in 

 the original 13-station Standard Sample, but were from locations widely 

 differing from the region of selection of the original sample. It was de- 

 cided to select one station representative of an arctic type climate, one 

 temperate, one tropical, and one from a "problem" climate area. 



Amundsen-Scott station at the South Pole (lat. 90°S) was chosen as the 

 arctic type ; this station was expected to present the most rigorous test of 

 the prediction model (as based on the Standard Sample) that could be 

 obtained anywhere in the world. In the first place the extreme arctic- 

 continental climate, with almost no water-vapor contribution to the re- 

 fractive index and the nearly incessant temperature inversion, is more 

 alien to the Standard Sample than any other type ; in the second place the 

 station elevation is 2800 m, which is 900 m in excess of the highest station 

 (Ely, Nev., 1908 m) included in the Standard Sample. These two effects 

 were expected to augment each other as regards refraction. 



Dakar, Senegal, on the western coast of Africa, was selected as a 

 "problem" climate station; an inverse relationship exists there between 

 N s and AN (the A'-gradient over the first kilometer above the surface). 

 A Congo basis station, Bangui, in what was French Equatorial Africa, 

 was selected as the tropical location, and Moscow, U.S.S.R., was selected 

 as the temperate location. 



In order to combine brevity with comprehensiveness, ray tracings were 

 done of the total refraction (bending at 70 km target height) at two 

 elevation angles, 20 mrad and 100 mrad, for six profiles from each loca- 

 tion. The six profiles were selected as representing roughly the range of 

 Ns in winter (February), summer (August), and spring-fall (May and 

 November), two profiles being selected from February, two from August, 

 and one from each of May and November. The 20-mrad elevation angle 



