lny = [ln(a) + b ln(x) ] ± t a/2 S y 



1 + N rin(x') ■ 

 N S, 



ln(x) 



(17) 



where y = dependent variable equal to sand flux 



a, b = power law coefficients presented in Table 6 



x = independent variable V - V* , where V and V* are 

 as defined previously 



t a/2 = t-statistic for 95 percent confidence interval 



x = average x , where x is as defined above 

 S xx = N E x 2 - (2 x) 2 , for x = 1 to N 



94. The 95 percent confidence interval is used in the next section to 

 determine whether certain test conditions were significantly affecting rates 

 of sand transport in the tank. 



Independent variables influencing sand transport 



95. Two interesting facets that emerged during the tests were explored 

 within time constraints of the project. It was noted that with elapsed time 

 sand in the test section became coarser due to sorting by the flow. Sieve 

 analysis indicated a median grain size of 0.30 mm after approximately 24 runs, 

 compared to 0.23 mm for the originally placed material (Figure 31). The 

 degree to which this sorting decreased quantities collected in the basin was 

 evaluated for one 2.5-min basin test and two 7.5-min basin tests. Between two 

 and eight shovelfuls of the winnowed (coarser) sand in the test section were 

 removed and replaced by the original, finer material. The measured 2.5-min 

 "new sand" flux was 3 percent greater than the flux predicted by a threshold 

 power equation fit to the data set (Figure 32), and the two 7.5-min "new sand" 

 fluxes were 17 and 19 percent less than equation-predicted fluxes (Figure 33). 

 However, these deviations from the equation-predicted fluxes were within the 

 95 percent confidence limit and, therefore, in the range of experiment 

 variability. The C nozzle measured 32 percent higher fluxes than predicted 

 after three shovelfuls of original sand replaced the winnowed -out material 

 (Figure 34). However, this deviation from the equation-predicted flux was 



66 



