for the intercept. The confidence intervals for the expected value of 

 Yq given a certain x^, y^ = 0.84x^ - Q.06, are: 



[y^ + (0.268) y0.034 + (x^ - 1. 85)2/4. 543] ^^^^ 



\Yq + (0.222) y0.034 + (x^ - 1. 85)2/4. 543]gQO^ 



The visual fit (Fig. 14) is a good approximation to the linear regression 

 curve for RSA means between 1 and 2 phi (0.5 and 0.25 millimeter) which 

 includes the values of the most frequently occurring RSA means in this 

 study. 



One possible explanation for the RSA giving a consistently coarser 

 mean determination than the dry sieve analysis is the fluid mechanics 

 effect of other particles of the sample on the fall velocity of an indi- 

 vidual particle in the sample. Cook (1969, p. 781) suggested that smaller 

 particles falling in a grain-size mixture are entrained by the larger 

 particles, and therefore they may experience an increase of velocity by 

 as much as 15 percent over their normal fall velocities. This would cause 

 a sample to settle at a faster rate than predicted by the hydraulic char- 

 acteristics of the individual grain. The RSA would then interpret the 

 increased velocity as an increase in the concentration of larger particles, 

 thus giving a mean which would be coarser than that determined by dry 

 sieve methods. This explanation is effectively the same as that of Sanford 

 (1970) who noted that a vertical turbidity current develops in the RSA 

 settling tube, resulting in a sedimentation diameter which is consistently 

 coarser than the equivalent sieve diameter. For coarse particles the 

 effect may be just the opposite. The linear regression curve in Figure 



14 and the studies by CERC (C. Judge, geologist, personal communication, 

 November 1976) indicate that for coarser samples the RSA mean may be finer 

 than the dry sieve mean. In addition, turbulence may affect the fall 

 velocity of particles under some conditions. 



There has been increased interest in this problem by workers in fluid 

 mechanics. Further study should lead to a better interpretation of the 

 sizes obtained from fall velocity measurements in equipment such as the 

 RSA. 



6. McMaster's Procedure . 



There were some differences between McMaster's (1954) analysis proce- 

 dure and CERC's analysis procedure; however, the effects of these differ- 

 ences have been minimized where possible in the following ways. The most 

 important difference was in sampling method. At each sample site, McMaster 

 collected three 6-inch (15 centimeters) core samples from positions spaced 



15 feet (4.5 meters) apart along the high tide line. The top layer of 

 sand was scraped off and the three samples were combined in one container 

 and subsequently considered as one sample. Since CERC's samples are grab 

 samples, the surface layer is included and the samples did not penetrate 



33 



