a true condition or may be due to lack of adequate data samples. To 

 eliminate the variation and to obtain a clearer picture of trends, means 

 of the probabilities for all months were computed (table 20). The results 

 indicate that probability of persistence is highest for all thermal struc- 

 ture parameters when comparing means of areas during the same time periods. 

 The highest probability is found by comparing monthly means of adjacent 

 1-degree quadrangles. With exception of surface temperature, all param- 

 eters have a mean probability greater than 90 percent that no significant 

 difference exists (95^ level of significance). This indicates that vari- 

 ability between adjacent 1-degree quadrangles is partially masked by the 

 use of monthly means. 



The next highest mean probability resulted from comparison of 5-day 

 mean adjacent 10-minute quadrangles. However, this probability was only 

 slightly higher than that for the comparison of 5-day means of alternate 

 10-minute quadrangles. Thus, the variability within a 10-minute quad- 

 rangle over a 5-day period is nearly the same in an area roughly 20 miles 

 away as in one 10 miles away. In general, all thermal structure parameters 

 varied more from one 5-day period to the next in one area than they do 

 between 10-minute quadrangles dviring the same 5-day time period. The 

 lowest mean probability of all parameters was recorded for 10-minute quad- 

 rangle, alternate 5-day means. It was the only recorded pararoet^^r with 

 mean probability below 50 percent. 



Parameters exhibiting greatest spatial persistence are the gradients 

 of the first and maximum gradients and the thickness of the maximum gra- 

 dient. Parameters exhibiting least spatial persistence are the gradient 

 of the mean gradient, the layer depth, and the temperature of the upper 

 bound of the maximum gradient. As mentioned previously, the magnitude of 

 the mean gradient is related to the BT tjrpe ('4-50 or 900 feet) and is not 

 a reliable statistic. 



The only parameter exhibiting a high degree of temporal persistence 

 is the thickness of the first gradient. The gradients of the first and 

 maximum gradients apparently persist from one 5-day period to the next 

 but vary from year to year. Surface temperature and the temperature of 

 the upper bound of the maximum gradient show the greatest temporal vari- 

 ability. 



One objective of this study is determination of area size for which 

 a point temperature structure forecast is applicable within statistical 

 limits. A direct solution of this problem necessitates comparison of 

 individual observations. Since locations of individual observations may 

 be in error by as much as 5 miles, such comparisons may be invalid. It 

 is hoped that errors may be minimized through processing of large amounts 

 of data. Thus, the quantity of data available may indicate the analysis 

 potential or limitations. 



If the spatisil variation in the thermal structure occtirring at any 

 given instant within a 10-minute quadrangle is assumed to be equal to or 

 less than the variation in a 5-day period within the same area, table 26 

 can be used to give an estimate of the spatial variation. For example, 

 variation in surface temperature during March within a 10-minute quad- 

 rangle is ( + 3 X .37) or + 1.11°F for 100 percent probability j and 



