DRY AND WET TERM SCALE HEIGHTS 319 



of the more dense and stratified continental air that customarily flows 

 offshore during the winter months. The same pattern is repeated on the 

 summertime map along the west coast but is less i^ronounced on the east 

 coast due to the combination of more uniform heating and also onshore 

 advection of maritime air |)roduced by the circulation pattern of the 

 Bermuda high-pressure area. The high value of H d = 10.5 km observed 

 in the southwest during the summer appears to be due to the intense heat- 

 ing with resultant convective mixing to great heights so common in that 

 desert area. A somewhat opposite pattern is evidenced by the //„, maps. 

 For example, the coastal areas generally have the lowest values and thus 

 reflect the characteristic strong humidity stratification of maritime air. 

 The smaller humidity gradients of the inland regions produce somewhat 

 larger scale heights for that area. The summer Hy, map is quite sur- 

 prising in that very little variation is shown, perhaps indicating uniform 

 vertical convection of the available moisture at all locations throughout 

 the country. The strong convection indicated in the southwest on the 

 summer H d map is again reflected by the high value of //,„ = 3.0 km for 

 that same area. 



It is quite evident from figure 8.9 that within the troi)osphere, /i < 10 

 km, the bi-exponential model has a lower rms error for the common, 

 near-zero angles of departure used in tropospheric propagation of radio 

 waves. Both models yield about a 12 percent error in determining 

 T for dt) — Q and h = W km. At ^o = 100 mrad, however, the percentage 

 error decreases to 4 percent for the bi-exponential model and 7 percent for 

 the exponential reference atmosphere. The rather marked errors of the 

 single exponential model at 10 km simply reflect that that model is 

 deliberately fitted to the average A'^ structure over the first few kilometers 

 with the result that this model systematically departs from the average 

 atmospheres in the vicinity of the tropo pause. This is particularly 

 apparent at the higher values of ^o where the integral tends to become a 

 function of the limits of integration. That is, using the theorem of the 

 mean for integrals. 



cot d dn ^ -cot da dn ^ [N^ - N] 10-" cot ^o (8.9) 



under the assumption that cot 6 may be replaced by cot ^o over the interval 

 of integration. For do < 20 mrad this assumption introduces less than 

 a 10 percent error for the interval < /i < 10 km. It is ai)parent from 

 (8.9) that the error in j^redicting r for large do is then simj)ly a matter of 

 how closely the model approaches the true value of n in the atmos}ihere. 

 At ^0 near zero, however, the integral for r is very heavily weighted to- 

 wards the effect of n gradients near the earth's surface [1]. Since the 

 values of do commonly used in tropospheric propagation are as near as 



