APPENDIX A 



MIXING SAMPLES FOR COMPOSITE ANALYSIS 



It was suggested in Section III that essentially the same composite 

 frequency-size distribution could he obtained either by averaging the 

 frequencies per size class for each sample from a set or by analyzing a 

 single mixture of all samples. The two approaches are identical from a 

 mathematical standpoint, but can similar laboratory results be expected 

 using the two approaches? The following experiment was conducted to look 

 into this practical problem. 



Eight test samples were prepared by blending predetermined weights of 

 1/4-phi-sized sands. Each test sample was normally distributed with mean 

 and sorting values like those typically observed across a profile, on an 

 east coast beach, and at elevation spacings of 4 feet (1.2 meters) from 

 +4 feet MSL (test sampie 1) to -24 feet (-7.3 meters) MSL (test sample 8). 

 Six random splits were then obtained from each test sample using Jones 

 Sample Splitter. Each test split contained approximately the same amount 

 of sand as determined by visual comparison of their relative volumes. 

 This semiaccurate procedure was selected as being fairly consistent with 

 what could be expected with normal sample preparation procedures. Three 

 analysis procedures were then followed. 



First, two sets of samples containing one split of each test sample 

 were analyzed at CERC using settling techniques. The size data from each 

 set were averaged and their calculated composite distributions are the two 

 samples labeled "No mixing" in Table A-l. Secondly, splits from test sam- 

 pleg 1 to 4 (onshore and nearshore) and samples 5 to 8 (offshore) were 

 physically mixed to create two samples representative of the test set. 

 This procedure was followed twice and the samples were analyzed; these cal- 

 culated composites are shown as the "Mixture 2" in Table A-l. The third 

 procedure was to physically mix all eight splits and then analyze only one 

 sample from the mixture. Again, the procedure was repeated twice (Mixture 

 1, Table A-l). Finally, the column labeled "Test composite" shows the 

 composite-size distribution for the test samples as computed from the weight 

 proportions used to initially make up the samples. 



Comparison of the calculated composites with the test composite shows 

 fairly high variation in actual percentages of sediment observed in the 

 various size fractions. These variations probably are caused more by the 

 technique used to obtain the "equal" splits than by analysis error. Still, 

 there is little variation of the phi mean and phi sorting values from the 

 test composite parameters and these results do not seem to favor a partic- 

 ular analysis procedure. 



Recommendations resulting from this small experiment are: (a) Mixed 

 samples can be used to obtain accurate composite data; (b) care should be 

 taken to obtain random splits of appropriate size from samples and to thor- 

 oughly homogenize the mixtures; (c) it is probably best to combine samples 

 into related mixtures such as onshore, nearshore, and offshore samples, or 



43 



