9521

• UCREFRP Catalog Number: 9521
• Author: Liebermann, T. D., D. K. Mueller, J. E. Kircher and A. F. Choquette.
• Title: Characteristics and Trends of Streamflow and Dissolved Solids in the Upper Colorado River Basin, Arizona, Colorado, New Mexico, Utah, and Wyoming.
• USFW Year: 1989.
• USFW - Doc Type: Denver.
• Copyright Material: NO

where
In (C) = estimated natural logarithm of dissolved-solids
concentration;
In(Q)=natural logarithm of streamflow; and
bo and bl = regression coefficients.
Detransformation of the regression model enabled estima-
tion of dissolved-solids concentration in the original units:
C=exp(60)Q6, exp(1/262) (4)
where
exp(11262) =the bias-correction factor (Miller, 1984), and
62 =the estimated residual variance from calibra-
tion of the model (eq 1).
The residuals, or flow-adjusted concentrations, were com-
puted as the difference between the observed dissolved-solids
concentrations and the corresponding estimates (C) from
equation 4. Flow-adjusted concentrations were computed for
both the monthly and annual time series for all sites.
Nonparametric Trend Analysis
Trend analyses were made to test for significant
changes in streamflow and water quality during the period
of record. Nonparametric (rank or distribution-free) analyses
were used rather than parametric (least-squares regression)
analyses, because water-quality data commonly do not meet
the assumptions of parametric analyses (Hirsch and others,
1982). Hirsch and others reported that, even when all the
normality assumptions were met, the nonparametric Kendall
test was almost as powerful as parametric analyses based on
least-squares regression. They also reported that when skew-
ness or seasonality were introduced, the seasonal Kendall
test was better than regression, and that the effects of serial
correlation were no worse on the seasonal Kendall test than
on the regression test.
Nonparametric analyses were applied to the monthly
and annual time series of streamflow, dissolved-solids
concentration, dissolved-solids load, and flow-adjusted con-
centration for all sites having 10 or more years of dissolved-
solids records. Determination of trend significance was based
on the following criteria: p < 0.01, highly significant; 0.01
< p < 0.05, significant; 0.05 < p < 0. 10, marginally signifi-
cant; p > 0. 10, not significant; where p is the attained, two-
sided significance level for the test.
The seasonal Kendall test (Crawford and others, 1983)
was used to identify monotonic trends. The result of this test
is analogous to the slope of the least-squares regression line
(parametric) when the independent variable is measured in
years. Analyses of annual data yielded the median annual
change during the period of record, with an associated
significance level for the change. Analyses of monthly data
yielded the median annual change for each month of the year.
For example, dissolved-solids concentration may have
decreased by 3 mg/L per year during January, may have in-
creased by 1 mg/L per year during June, and may not have
changed significantly during September. Testing for trends
in both the annual and monthly time series was necessary
because the significance of trends in the two time series may
be different. For example, a site may have no significant
trend in annual streamflow, but may have an increasing trend
in monthly streamflow during January and a decreasing trend
in monthly streamflow during June.
The Mann-Whitney-Wilcoxon rank-sum test (Crawford
and others, 1983) was used to identify changes in the annual
and monthly data caused by an intervention in the watershed.
An intervention is some definable change that has an effect
on streamflow or water quality. Interventions identified in
the study area included construction of reservoirs, implemen-
tation of salinity-control projects, and initiation of transbasin
diversions. The data for a specific site were divided into two
periods: preintervention and postintervention. The Mann-
Whitney-Wilcoxon test analyzes the significance of the dif-
ference between the median values of the two periods and
is analogous to the Student's nest for the significance of the
difference between two means. The statistic used to estimate
the change in the median between the two periods is called
the step trend. It is the median of the differences calculated
from all possible combinations of the values for the two
periods. The step trend is a unbiased estimate of the change
in median and is a less variable estimator than the simple
difference between the medians of the two periods.
Thirteen sites were tested for step trends caused by an
upstream intervention. Results of the analyses for annual data
are summarized in table 4. Each site had a substantial period
of record before and after the intervention. Construction of
a large reservoir or transbasin diversion system was the major
intervention at 12 sites, and 1 site was evaluated for interven-
tion due to a salinity-control project. Other salinity-control
projects in the Upper Colorado River Basin were too recent
to enable adequate intervention analysis. In addition to the
step-trend analysis, the monotonic-trend analysis was done
for the preintervention period and the postintervention period
for each of these sites.
CHARACTERISTICS AND TRENDS OF
STREAMFLOW AND DISSOLVED SOLIDS
The study area was divided into three major regions:
the region drained by the Colorado River and tributaries
upstream from the confluence with the Green River, the
region drained by the Green River and its tributaries, and
the region from the confluence of the Green and Colorado
Rivers to Lee Ferry, Ariz., including the San Juan River and
tributaries and all tributaries to the Colorado between the
18 Characteristics and Trends of Streamflow and Dissolved Solids in the Colorado River Basin