7930 - Laserfiche WebLink

Home Browse Search

As discussed earlier, annual sediment yields from total sediment rating equations were found to average less than the sediment dam yields at the 90 percent confi- dence level but not at the 95 percent level. Total sedi- ment yield was also estimated by determining annual suspended sediment and annual bedload from the respec- tive suspended sediment and bedload rating equations and adding the results. These estimates were somewhat less than those from the total sediment rating equations (fig. 9) and tested significantly less than the sediment dam estimates at the 95 percent confidence level. This data set, however, was smaller (n=17) than the earlier data set (n=29) because there were fewer years when both the suspended sediment and bedload equations were significant. 1o0 • a E SO ' u A+ w o 60 A F a O4 ?l Z O 40 ?. • ' W W * ? TOTAL SEDIMENT O Z ? • SUM OF SUSPENDED F„ 20 IL AND BEDLOAD SEDIMENT a = o 20 40 60 80 100 120 140 SEDIMENT YIELD FROM SEDIMENT DAMS, mg/YR Figure 9-Comparison of sediment yield from sediment dams with sediment yield from total sediment rating equations and the sum of yields from suspended and bedload sediment rating equations. Data from Silver Creek watersheds. CONCLUSIONS Sediment rating equations similar to equation 1 that are based on small sample sizes are not in themselves reliable for documenting watershed disturbances in streams dominated by bedload sediment in the Idaho batholith. A large proportion (33 percent) of the rating equations developed during this evaluation were not statistically significant, and an additional 24 percent, while significant, were not "useful" because streamflow explained less than 60 percent of the variance in sedi- ment. Year-to-year watershed response evaluations were extremely difficult. The large natural variance of sedi- ment data resulted in rating equations with wide confi- dence bands. This made it hard to show statistical sig- nificance of suspected shifts in rating equations. Rating equations for a given stream changed from year to year. Sometimes the change could be attributed to activities in the watershed, but other times there was no apparent reason for the shift. It may have resulted from the sam- pling scheme or purely from sample variance. In snowmelt-dominated and bedload-dominated sys- tems, time integration of sediment yield from sample date to sample date can provide accurate estimates of annual sediment yields with as few as 15 or 20 well- taken samples. This requires that samples adequately describe the average conditions for which they apply. Misleading estimates of sediment yield will be obtained if sampling does not properly reflect streamflow and sediment transport distributions. Sampling schedules should be well thought out and flexible so that atypical runoff events can also be sampled. Thus, accurate sedi- ment yield estimates can be expected using time integra- tion. For the Silver Creek data, 74 percent of the esti- mates of sediment yield by time integration were within 50 percent of the sediment dam yields. The use of pumping samplers to generate more sam- ples does not necessarily increase the chances of having better predictive rating equations. Of the rating equa- tions developed for pumped samples in Silver Creek, 62 percent were significant, and many of the significant equations had limited usefulness for predictions due to the small range in streamflow for which they applied. Pumping samplers with intakes positioned as those in Silver Creek do not adequately provide total sediment load samples. When placed close to the bottom of the stream they will sample some bedload but only enough to overestimate suspended sediment load. Separate suspended and bedload sediment rating curves estimate smaller sediment yields than do the total sediment rating equations from the same data. Total suspended sediment equations appear to estimate too much sediment for low annual yield years, and increasingly too little sediment for greater sediment years. 11