32 



rates. Depths at 2,000 ft offshore varied somewhat throughout the storm 

 due to sediment movement; therefore, the depth determined by SBEACH 

 for each hour of the storm was used for analysis. The profile in the flume, 

 of course, remained constant. 



Task B 



All wave flume tests conducted for the SPN used a wave period of 

 15.9 sec. Because this value was a constant for all tests, it was not used 

 in the analysis. Gravitational acceleration was therefore the only term 

 available by which to nondimensionalize overtopping rate in time. All 

 other parameters required only a length scale, and either /or H were rea- 

 sonable candidates for the repeating variable. After trying both variables, 

 it was found that results were somewhat improved by using/. After many 

 variations and combinations of terms were tried, the dimensionless vari- 

 ables that provided the best fit to the data were arranged as follows: 



Q' = QKg*f)^'^ 



PI\ = b/f 



Pll = H/f 



PB = d/dlOOO 



PI4 = cot 



PIS = d/f 



Data collected in the physical model tests were converted to prototype 

 scale for the regression analysis. Input data are shown in Table 8. Note 

 that the last three lines in Table 8 give the input data for the three points 

 in Table 6 determined by regression analysis. 



Examination of the residuals from one of the regression models that 

 was tried indicated that higher-order terms were required (a residual is the 

 difference between Q' predicted by the regression model and measured 

 Q'). Second-order terms (squares of the PI variables) and higher were 

 therefore added to the analysis. 



Regression analysis was conducted on the dimensionless variable Q'. 

 Any negative overtopping rates predicted were set to zero, and results 

 were converted to predicted dimensional overtopping rates. Model selec- 

 tion was then based on the sum of squares of differences between ob- 

 served and predicted overtopping rates. 



SAS assumes a null hypothesis that the coefficient of a term in the 

 model is zero, then computes the probability that the null hypothesis is 

 true. Only terms in the model with a low probability of having zero 



Chapter 3 Research Tasks A, B, and C 



