|
<br />~xpected signs for the coefficients, However, the models for several sites
<br />~ lid not have the expected sign for net diversion or net reservoir release, At
<br />~ these sites, the net diversion and net reservoir-release terms were not
<br />tv representing the physical occurrence of diversions and reservoir releases in
<br />tv the basin, but may have been accounting for monthly variations in dissolved-
<br />solids discharge caused by some other factor.
<br />
<br />The R2 values, listed in table la, ranged from 0.83 to 0.97, indicating a
<br />strong correlation between actual and estimated dissolved-solids discharge at
<br />all 16 sites. The smallest R2 values were for site 15, San Juan River near
<br />Bluff, Utah; site 11, White River near Watson, Utah; and site 10, Duchesne
<br />River near Randlett, Utah. The standard errors ranged from 11 to 32 percent
<br />of the actual mean monthly dissolved-solids discharges. The sites having the
<br />larger dissolved-solids discharge generally had the smaller standard errors.
<br />The largest standard errors occurred for site 13, San Rafael River near Green
<br />River, Utah, and site 15, San Juan River near Bluff, Utah. Statistical
<br />estimates of mean annual natural dissolved-solids discharge generally were
<br />comparable to the mass-balance estimates. At the farthest downstream site,
<br />site 16, Colorado River at Lees Ferry, Ariz., the difference was less than 0,1
<br />percent. Differences greater than 10 percent occurred at only four sites:
<br />site 15, San Juan River near Bluff, Utah (15 percent); site 6, Green River
<br />below Fontenelle Reservoir, Wyo. (18 percent); site 10, Duchesne River near
<br />Randlett, Utah (31 percent); and site 13, San Rafael River near Green River,
<br />Utah (91 percent).
<br />
<br />
<br />'~ -l
<br />
<br />';E
<br />
<br />aiL,. ';;:<'~"">'_ ,""
<br />
<br />
<br />With the exception of site 15, San Juan River near Bluff, Utah, the sites
<br />having the poorest model fit also had the smallest dissolved-solids discharge.
<br />The mass-balance estimates of natural dissolved-solids discharge for site 6,
<br />the Green River below Fontenelle Reservoir, Wyo.; site la, Duchesne River near
<br />Randlett, Utah; site 11, White River near Watson, Utah; and site 13, San
<br />Rafael River near Green River, Utah, were all less than 200,000 tons per year
<br />and, in total, accounted for less than 11 percent of the estimate for site 16,
<br />Colorado River at Lees Ferry, Ariz. The sites of most concern were site la,
<br />Duchesne River near Randlett, Utah, and site 13, San Rafael River near Green
<br />River, Utah, because the differences between the statistical estimates and
<br />mass-balance estimates were so large.
<br />
<br />The entire period of record for site 10, Duchesne River near Randlett,
<br />Utah, was affected by upstream regulation and transbasin exports from
<br />Strawberry Reservoir. However, the net diversion and net reservoir-release
<br />variables were not significant in the model for this site, This unexpected
<br />result could have been caused by lack of variability in the historical data,
<br />which is common at regulated sites. Because the effect of the reservoir is
<br />not explicitly considered in the model, estimates of predevelopment conditions
<br />could be inaccurate. Tarns and others (1965) also reported that some of the
<br />return flows from irrigation in the Duchesne River basin may be bypassing
<br />site 10 and entering the Green River directly. Therefore, a linear com-
<br />bination of consumptive use terms from preceding months may not be the best
<br />estimator of return flows at the site,
<br />
<br />The San Rafael River basin also has a number of diversions that were not
<br />included in the model for site 13, San Rafael River near Green River, Utah.
<br />
<br />33
<br />
|