distributions. This is an important property of composites often mis- 

 understood. These distributions must encompass the entire range of sizes 

 encountered, and their sorting is not simply the average sorting of the 

 samples. The graphical composite mean is about equal to the averaged sam- 

 ple mean (1.41 versus 1.40 for the composite), but the graphical composite 

 sorting is much greater than the averaged sample sorting (0.49 averaged 

 versus 1.50 for the composite). Figure 4 also reflects this relationship 

 where the sorting of each individual sample is less than for the overall 

 and sampled composites. 



In general, a composite of 20 or more beach samples will tend to plot 

 as a relatively straight line on log probability paper. The degree of 

 log normality can be quickly checked by comparing the fit of the plotted 

 curve with a straight line drawn through the 16th and 84th percentile 

 intersections. Often, a composite curve will tend to straighten out when 

 more samples per profile or more profiles are added, and when a smaller 

 phi interval is used during analysis. It is best to use a constant phi, 

 ■interval; for most purposes, an interval of one-half phi is usually ade- 

 quate (Table 3) . 



An alternative method for obtaining a computed composite is presented 

 by Krumbein (1957). He assumes that the samples are nearly equally sorted 

 and vary in phi mean grain size with coarse and fine extremes labeled Mj, 

 and M a , respectively. In this approach, the composite phi mean, M c , 

 and variance, o^, are computed as follows: 



CMfc - Ma) 2 



? 2 



o u a 12 



(7) 



where a^ is the average squared sample sorting. For the data in Table 

 6, o| is 0.25, M c is 1.43, and a a is 1.18. This approach might be followed 

 in cases where the original size percentage data are unavailable or where a 

 beach has been lost by erosion and the "native beach" data consist of mean 

 sizes and sortings that seem most probable, considering the processes pre- 

 sent. However, a composite distribution obtained by averaging is preferable 

 to a computed composite. The averaging approach is less affected by extreme 

 values and with its use, no assumption is made regarding the equal sorting 

 of the samples. 



The same techniques are used for calculating the composite-size dis- 

 tribution for both borrow site and native beach sediments. The sampling 

 plan should be tailored to the characteristics of each borrow area but the 

 concept of the composite remains the same. The quality of any beach-fill 

 calculations is at best, only as good as the native beach and borrow site 

 composites. 



Finally, in concept, the averaging of individual sample size frequencies 

 is equivalent to mixing parts of the actual samples and then doing a single- 



27 



