Biological Opinions - Bureau of Reclamation

Evaluation of baseline hydrology involved analyses of data for UKL and the Sprague and
Williamson Rivers. Even though the proposed action was developed based on the 1981 through
2011 period of record (2012 data were not available when the proposed action was developed),
data from water year 2012 and years before 1981 were incorporated into the baseline hydrology
evaluation where applicable. Data sets used for hydrologic analysis included daily observed
flow data for water years 1921 through 2012 in the Sprague River, 1918 through 2012 in the
Williamson River, and 1981 through 2012 in UKL.
The daily data were reduced to median monthly values for seasonal time frames. Median flow
values for each season were calculated from the daily flow values for that season for each year
during the period analyzed. For example, the median for the October through February period
was calculated based on 151 daily flow values (data for February 29 were excluded). For the
March through June period, the median was calculated based on 122 daily flow values. For July
through September, the median was calculated based on 92 daily flow values, and for the water
year it was based on 365 daily flow values.
Trends in median seasonal and water year flow in the Sprague and Williamson rivers and net
inflow into UKL were evaluated by fitting a LOcally WEighted Scatterplot Smoothing
(LOWESS) curve (Helsel and Hirsch 2002) to flow data, and statistical testing for trend using the
Mann-Kendall trend test (Helsel and Hirsch 2002; Helsel et al. 2005). Trends were evaluated
based on the entire period of record for each water body: water years 1921 through 2012 for the
Sprague River, 1918 through 2012 for the Williamson River, and 1981 through 2012 for UKL.
UKL inflow data for water years 1961 through 1980 were not used because daily calculated net
inflows are not available.
LOWESS smoothing emphasizes the shape of the relationship between two sets of variables; and
in this case, the variables are flow volume and time. LOWESS smoothing provides a way to
evaluate changes in data without the constraint of a prior assumption of an equation that best
models the data.
The Mann-Kendall method is a nonparametric trend test that determines whether a statistically
significant upward or downward change in flow has occurred over the period of record.
Nonparametric tests are most appropriate where data are expected to be non-normally distributed
or where a specific distribution is unknown (Helsel and Hirsch 2002). The Mann-Kendall trend
test is superior to simple linear regression because the Mann-Kendall test was developed
specifically to determine if the median, or central value, changes over time (Helsel and Hirsch
2002). The effects of extreme values do not influence Mann-Kendall tests as substantially as
they influence simple linear regression. Flow data are strongly serially correlated, which is a
correlation between a value and previous values in the dataset. Although simple linear
regression and the Mann-Kendall test are biased when serial correlation is present, the use of
monthly medians reduces these effects substantially. In our analysis, the Mann-Kendall
equations test for a monotonic trend (the dependent Y variable changes in a consistent direction)
in the flow data over time (Helsel et al. 2005).
A significance level (alpha) of 0.10 was selected for assessing the Mann-Kendall trend test data.
The alpha does not depend on the data, but is a management decision regarding the level of
significance to be applied to the statistical test results. It is a subjective value used to evaluate
80