CRITICAL APPRAISAL OF RESULTS 107 



Approximating dn/dh in (4.7) by AN, we may determine that the radio 

 horizon distance of an antenna located 150 m above the earth would vary 

 from 48 km when A''^ = 200 to 59 km when Ns = 400. Yet another ap- 

 plication of the N s charts is to the exponential models of the decrease of 

 refractive index with height which have been proposed to date [9, 17]. 

 These models are completely specified by N s and may be used to account 

 for seasonal and geographic variations of such refraction effects as radar 

 range and elevation angle errors. 



4.2.6. Critical Appraisal of Results 



The world maps presented above were based upon data from 306 

 weather stations. This number of stations appears to be consistent with 

 the scale of map used. The map scale is so small, however, that only 

 large climatic differences can be expected to be discerned. For the 

 climate of any given area one should refer to detailed studies of N such 

 as those currently in preparation for the United States at the Central 

 Radio Propagation Laboratory. 



The accuracy of the present charts may be assessed from the charts of 

 maximum range, R, of monthly means as given by figures 4.12 and 4.13. 

 The standard deviation of the individual monthly means may be esti- 

 mated from [18] 0.43 R, where the coefficient 0.43 is appropriate for five 

 individual observations. Since, in general, R < 20 N units, then 

 0.43 i? < 9N units, although this standard deviation may be as large as 26 

 N units for the month of February in Australia and 17 N units in the 

 southwest of the United States during August, or in the African Sudan 

 during February. 



Further, the standard error of estimating a 5-year mean from five 

 individual monthly values is determined from 



0A3R 



_ ) 

 \/n 



where n for our case is 5 and thus the error of the 5-year mean would be 

 0.192 R. Remembering that R < 20 N units and assuming perfect skill 

 in drawing the contours, one would expect the standard error of estimate 

 to be less than 4 A^ units. This standard error can be as large as 12 A" 

 units in Australia where R = GO N units. 



The value of A^s at each of the 20 test stations of table 4.2 was estimated 

 from the A^o contours with an rms error oi 5 N units which is consistent 

 with the standard error of estimate obtained from the range charts. In 

 the large areas of sparse data, such as the oceans and Russia, this un- 

 certainty rises to about 10 N units and thus the contours in these regions 

 are dashed. 



