Laserfiche WebLink
34 - 6 PITLICK AND CRESS: DOWNSTREAM CHANGES IN CHANNEL GEOMETRY <br />Table 3. Parameter Estimates for Regression Relations Between <br />In Transfonmed Values of Grain Size, Width, Depth, Shear Stress, <br />Shields Stress, and Distance' <br /> n a b SE,, r' F p <br />In surface D84 78 57.6 0.0015 0.00050 0.11 9.4 0.003 <br />In surface D5), 78 33.4 0.0015 0.00049 0.11 9.1 0.003 <br />In surface Die 78 18.4 0.0013 0.00041 0.12 10.8 0.002 <br />In subsurface Ds, 27 45.9 0.0021 0.00075 0.24 8.0 0.009 <br />In subsurface D5() 27 16.5 0.0023 0.00081 0.25 8.3 0.008 <br />In subsurface D16 27 0.8 0.0034 0.00202 0.10 2.9 0.10 <br />In width 132 167.0 -0.0011 0.00036 0.07 9.7 0.002 <br />In depth 132 7.8 -0.0035 0.00026 0.58 180 <0.001 <br />In shear stress 132 27.6 0.0014 0.00039 0.08 12.0 <0.001 <br />In Shields stress 132 0.048 -0.00005 0.00026 <0.01 0.03 0.858 <br />"Here a is the number of observations; a and b are parameters in the <br />relation In r = /n a + h.r. will ix measured in km; SE,, is the standard error of <br />the coefficient h: r' is the coerficient of detennination, F is the value of the <br />F distribution: andp is the significance probability. <br />Table 4 indicate that the trends in surface D50 are not <br />significantly different from zero (p > 0.05), except in the <br />segment between DeBeque Canyon and Westwater Canyon. <br />We would be surprised if this was not the case, given that <br />the segment is 100-km long, and includes a 40% increase in <br />discharge from the Gunnison River. The other segments fail <br />the test of significance because the standard error of the <br />regression SEb is of the same order as the coefficient b itself. <br />To determine if fining is indeed stronger within the DeBe- <br />que-Westwater segment, we then performed an analysis of <br />covariance (ANCOVA), using a binary predictor variable <br />for the separate segments (DeBeque-Westwater vs. the <br />remainder of the data). The results of the ANCOVA indicate <br />that there is a marginal difference in the slopes of the <br />separate regression relations for InD50 (p = 0.08), suggest- <br />ing that the rate of downstream fining within the DeBeque- <br />Westwater segment is not much different from the trend <br />defined by the remainder of the data. What is perhaps more <br />important to the analysis and the main hypothesis of this <br />study is the correlation between gain size and shear stress, <br />which we pursue later in the section on channel geometry. <br />fix] Trends in the percentiles of the subsurface sediment <br />(Figure 5) indicate that the bulk bed material of the <br />Colorado River also fines systematically downstream. <br />Regression analysis of these data indicates that the subsur- <br />face D,ua and Dso are modeled reasonably well by exponen- <br />tial relations that parallel each other (Table 3); the relation <br />for D16 is not as strong (p = 0.10), and the trend is slightly <br />steeper. Comparison of the fining coefficients b for surface <br />and subsurface D50 (Table 3) suggests that the bulk bed <br />material fines downstream at slightly higher rates than the <br />pavement. The difference reflects an increase in the pro- <br />portion of sand and granules in the bed, and implies that the <br />surface coarsens downstream relative to the subsurface. <br />However, a plot of the ratio of surface D50 to subsurface <br />D50 derived from paired samples at the same locations <br />(Figure 6) shows that this trend is not very strong (r2 = <br />0.22, p = 0.02). On average the surface D50 is about 1.5 <br />times the subsurface D50- <br />4.3. Channel Geometry <br />[I9] The downstream trends in bank-full channel width <br />and depth, derived from individual cross section measure- <br />ments, are plotted in Figure 7. These data are further <br />subdivided according to reach type (alluvial or quasi- <br />alluvial), and fit with separate regression equations. To <br />assess the influence of reach type on overall trends, we <br />compared the regression estimates of the intercept a and <br />slope b of the separate relations using an analysis of <br />covariance (ANCOVA), with distance as a covariate and <br />reach type as a factor. The purpose of these tests was to <br />establish whether changes in channel geometry are driven <br />more by one reach type than the other, and to examine <br />potential interactions between reach type and distance. <br />[20] The results of this analysis are summarized in Table <br />5, which lists the regression estimates of a and b for each <br />relation, standard measures of the strength and significance <br />of the relation, and a summary of the ANCOVA results with <br />different levels of complexity. A test of the regression <br />equations for In bank-full width indicates that there is no <br />significant difference in the slopes of the separate relations <br />(p = 0.44); therefore reach type does not appear to affect <br />downstream trends in bank-full width. Removing the inter- <br />action term (distance x reach type) results in a significant <br />difference in adjusted means (p < 0.001), thus bank-full <br />widths are consistently higher in alluvial reaches than in <br />quasi-alluvial reaches. Tests of the regression equations for <br />In bank-full depth likewise indicate no significant difference <br />in the slopes of the separate relations (p = 0.53); thus <br />downstream trends in bank-full depth do not appear to be <br />affected by reach type. Removing the interaction term <br />results in a significant difference in adjusted means (p = <br />0.01) suggesting that bank-full depths are, on average, <br />lower in alluvial reaches than in quasi-alluvial reaches. <br />[21] Previous studies of downstream changes in channel <br />geometry have shown that, when discharge is used as the <br />Table 4. Parameter ',Estimates and Measures of Statistical <br />Significance of Within-Segment Relations Between Individual In <br />Transformed Variables and Distance' <br /> n a b SE,, r2 F p <br /> rkm 365- 315 <br />surface D511 17 20.1 0.0029 0.0049 0.02 0.35 0.565 <br />width 31 7.8 0.0075 0.0029 0.19 6.84 0.014 <br />depth 31 14.7 -0.0051 0.0024 0.14 4.73 0.038 <br />tau 31 15.9 0.0031 0.0021 0.07 2.13 0.155 <br />tau*. 31 0.039 0.00075 0.0023 <0.01 0.11 0.743 <br /> rkm 300- 200 <br />surface D50 36 14.69 0.0048 0.0016 0.20 8.69 0.006 <br />width 57 107.03 0.0010 0.0016 0.01 0.39 0.536 <br />depth 57 9.16 -0.0044 0.0013 0.18 11.79 0.001 <br />tau 57 15.70 0.0034 0.0012 0.14 8.96 0.004 <br />tau* 57 0.054 -0.00066 0.0012 <0.01 0.29 0.592 <br /> rkm 180- 140 <br />surface D50 16 24.14 0.0027 0.0026 0.07 1.04 0.325 <br />width 24 86.20 0.0029 0.0048 0.02 0.37 0.550 <br />depth 24 17.07 -0.0082 0.0032 0.23 6.68 0.017 <br />tau 24 17.66 0.0026 0.0037 0.02 0.48 0.497 <br />tau* 24 0.045 -0.000067 0.0035 <0.01 <0.01 0.985 <br /> rkrrt 139- 105 <br />surface D50 9 22.94 0.0062 0.0216 0.01 0.08 0.782 <br />width 20 19.50 0.0164 0.0084 0.17 3.81 0.067 <br />depth 20 12.86 -0.0074 0.0057 0.08 1.66 0.213 <br />tau 20 0.20 0.0452 0.0104 0.51 18.86 0.0004 <br />tau* 20 0.015 0.0106 0.0056 0.17 3.61 0.074 <br />'The notation is the same as in Table 3