WORLD MAPS OF N(z) 7 



map was then carefully checked by another an- 

 alyst to make certain all data points had been 

 properly considered. The contours were modi- 

 fied in some areas on the basis of other informa- 

 tion or considerations not accounted for in the 

 machine-analyzed radiosonde data. For in- 

 stance, supplementary surface data [Knoll, 

 1941; Serra, 1955; Bean and Cahoon, 1957; 

 UNESCO, 1958; Bean et al., 1960b; Air Min- 

 istry, 1961 ; Dodd, 1965] were considered in the 

 contouring of those parameters (D and W ) 

 dependent upon surface observations. Also, if 

 spurious "high" centers of the wet term (such 

 as the isolated values found at Tananarive, 

 Malagasy Republic) were produced at high ele- 

 vation stations by reduction-to-sea-level pro- 

 cedures, these values were smoothed to some 

 extent. It was also found that the wet scale 

 heights derived from mean A/-profile data for 

 stations at altitudes greater than 1 km tend to 

 give more unrealistic sea-level wet terms than 

 the average wet-term scale height of 3.0 km 

 suggested by Hann [List, 1958]. When this 

 "standard atmosphere" scale height was sub- 

 stituted for Hw, the maximum error of N(z) 

 values calculated from (3) or (4) for all sta- 

 tions above 1 km which are listed in table A-l 

 was 6.2 percent of the true 5-year mean value 

 at Tananarive in August ; the second largest 

 error was 5.5 percent at Nairobi in February. 



Another contouring check was made of all 

 modified contours ; a third analyst reviewed the 

 smoothing to be sure it was consistent with the 

 original plotted data as well as with the sup- 

 plementary information. 



To further check the contouring, calculated 

 N(z) values (using (3) and (4) with values 

 read from figs. A-l through A-25) were com- 

 pared with actual observed values at 32 repre- 

 sentative stations. The results of this check 

 (reported in detail in table 2 of sec. 7) em- 

 phasize that, although some error undoubtedly 

 results in N(z) values below 1 km due to con- 

 touring, it is not a problem for N(z) values at 

 3 km or above. 



3.4. Problem Areas of N(z) Maps 



The use of wet and dry scale heights in a bi- 

 exponential radio refractive index model has 

 proved to be a good indicator of climatic differ- 

 ences [Bean, 1961; Misme, 1964]. The dry 

 term, or atmospheric density component of re- 

 fractivity, decreases with height in a uniform 

 manner throughout the troposphere so that its 

 scale height is an accurate indicator of the 

 degree of density stratification, but the water- 

 vapor component (the wet term) is not so well- 

 behaved. Because the saturation vapor pres- 

 sure, e s , is itself an exponential function of 

 temperature (which generally decreases linearly 



with height), one of the best wet-term models 

 is probably an exponential curve [Reitan, 1963 ; 

 Dutton and Bean, 1965]. However, an exponen- 

 tial model of the wet term must be used with 

 discretion because humidity is extremely vari- 

 able, both vertically and horizontally (because 

 of its high dependence upon the temperatures 

 within the different air masses, as well as var- 

 ious terrain and land-water effects) . 



To show actual physical changes in H w , the 

 wet scale height, it would be desirable to pre- 

 sent contoured values based not only on a large 

 number of stations, but also on data representa- 

 tive of various times of day. Figures A-6, A-12, 

 A-18, and A-24 present the seasonal values of 

 H w , but worldwide maps of the diurnal vari- 

 ability of the wet scale heights are not yet avail- 

 able. 



There are three specific areas of the world 

 where the assumption of an exponential distri- 

 bution of the wet term is largely invalid and can 

 be used only with reservations. Two of the 

 areas have one thing in common — a low sea- 

 level wet term. At continental stations in high 

 latitudes where strong temperature inversions 

 persist during winter months, the wet term at 

 3 km may be as large as, or even larger than, 

 that at the surface (because of the increase of 

 water vapor "capacity" with temperature) , and 

 the result is a negative or a very large positive 

 value of the wet scale height, neither of which 

 is physically realistic. At any tropical desert 

 station where the sea level wet term is < 30 

 A7-units, deceivingly high wet scale heights also 

 may result. Fortunately, because of the small 

 contribution of the wet term in these cases, the 

 total A/-error remains small. The wet-term pro- 

 files at nine stations with low values of W were 

 examined, and the largest error at any height 

 was 3.7 percent of the true 5-year N(z) value 

 at Niamey (a desert station) in February (fig. 

 3) . In the arctic areas (represented by Barrow, 

 Alaska, in this same figure) the maximum error 

 never exceeded 1.5 percent of the total N (z) 

 value. 



The third area presents a more serious prob- 

 lem because it exists in a subtropical climate 

 (15°-35° north and south of the equator) where 

 the wet term contributes from 14 to 1/3 of the 

 total N. The sharp decrease of humidity and 

 increase of temperature which is found in at- 

 mospheric layers between the surface and 3 km 

 in the subsidence regions of semipermanent sub- 

 tropical highs destroys the exponential distri- 

 bution of the wet term. In fact, in regions such 

 as this, the exponential fit may be valid only at 

 two or three points. This can be noted in figure 

 4, where the mean wet term for May is graphed, 

 and the H w value from the least-squares fit from 

 0-3 km of log W versus height is indicated, for 



