Laserfiche WebLink
<br />Discussion of Regression Results <br /> <br />Although initial multiple regressions considered all 13 basin characteristics. <br />final analyses included only those independent variables that were statistically <br />significant at the 5-percent level and that were not highly correlated with another <br />variable. The most significant variables in these relations were AE' E, Ss' and <br />SC' Because the regression equations containing E ,SS ' and Sc in addition <br />to AE had average standard errors of estimate and correlation coefficients essen- <br />tially the same as the equations containing only AE, only the following relations <br />between each selected flood characteristic and effective drainage area are <br />reported: <br /> <br />, <br /> <br />Q10= 525AEO.68 <br />Q2S= 92OA.EO.73 <br />Qso =1 ,280AE 0.77 <br />Q10o=1,680AEO.81 <br /> <br />(5 =46.8, R=0.81) <br />e <br />(5 =42.8, R=0.85) <br />e <br />(5 =39.8, R=0.88) <br />e <br /> <br />(Se=37.1, R=0.90) <br /> <br />(8) <br />(9) <br />(10) <br />(11) <br /> <br />To determine if these relations might be improved, the Qlooand Q10 residuals <br />(observed value minus calculated value) for each of the 15 stations were studied <br />to detect trends associated with their areal distribution or with the magnitude of <br />the effective drainage area. Residuals were found to be generally negative for <br />those stations in the southern one-half (Las Animas and Baca Counties) and <br />positive in the remainder of the study area (fig. 1). Thus, regression equations <br />8 to 11 may have a tendency to over-predict in the south and under-predict in <br />the north. In addition, the relation between the residual and the effective drain- <br />age area indicates nonlinearity between observed and calculated values. Because <br />of the limited sample size (n=15) and the lost degrees of freedom in statistical <br />analysis, the data could not be regionalized based on both location and effective <br />drainage area. The effect of drainage area was found to be most significant and, <br />therefore, the data were regionalized for two ranges of basin size (1) 0.50 to <br />2.99 mP, and (2) 3.00 to 15.0 mi2. Selection of these two particular ranges was <br />arbitrary to allow for adequate sample sizes of 6 and 9, respectively, and the <br />regressions were done graphically to prevent a discontinuity at 3.0 mP. <br />Following are the final equations: <br /> <br />20 <br />