Surface-water quality and flow Modeling Interest Group
Summary of Water Temperature Calibration for a CE-QUAL-W2 Model of
Shasta Lake, California
by Laurel Saito and John Bartholow
USGS, Biological Resources Division
4512 McMurry Avenue
Fort Collins, CO 80525-3400
Internet:
laurel_saito@usgs.gov,
john_bartholow@usgs.gov
Phone: (970) 226-9328, (970) 226-9319
Editor's note:
This article was drafted by the authors for a CE-QUAL-W2 Symposium held
in Portland, OR on 9/12/97. The data and interpretations provided on this
page are provisional. They may not have received final review and approval;
therefore, they are subject to change and are not to be cited until published.
Contents
Introduction
This report summarizes the calibration of the CE-QUAL-W2 model of Shasta Lake
for water temperature. Shasta Lake is located in northern California about 12
miles north of the town of Redding. It is the largest storage reservoir in
California, with a surface area of 11,940 hectares (29,500 acres), a maximum
depth of 157.6 meters (517 feet), a length of 56 kilometers (35 miles), and a
shoreline of 587 kilometers (365 miles). The reservoir is fed by the Pit,
McCloud, and Sacramento Rivers (BOR, 1994). Figure 1 shows the reservoir
along with the CE-QUAL-W2 model segmentation.
Figure 1. Shasta Lake CE-QUAL-W2 model segmentation and sampling
stations. (Larger map)
A new temperature control device (TCD) was recently installed at Shasta Lake
and is the focus of the modeling project. The TCD was installed because of
the reservoir's multipurpose operations and the recent focus on downstream
stream temperature management for salmon propagation. The reservoir has three
outlet options: the spillway, penstocks that release water through power
generating facilities, and release outlets at three levels. Prior to the
installation of the TCD, the penstocks were only capable of withdrawing water
from one elevation, and any releases through the low-level outlets bypassed
the penstocks. In order to maintain the downstream temperature requirements,
releases were often required through these outlets, reducing the amount of
hydropower that could be generated. The installation of the TCD allows the
passage of water from various elevations through the penstocks, which should
allow increased hydropower generation while still maintaining the downstream
temperature requirements (Hovecamp et al., 1996).
The CE-QUAL-W2 model of Shasta Lake is part of a collaborative project at
Shasta Lake that will assess the impacts of the TCD operations on
in-reservoir water quality and ecology. The model will be combined with fish
habitat and production models to perform this assessment. The intent of the
project is to provide an overall management strategy for the reservoir that
recognizes the linked ecosystems that are inherent to the reservoir system
(NBS, 1996).
Calibration Data for 1995
To calibrate the Shasta Lake modeled water temperatures, five general
categories of data were investigated: bathymetry, inflow distribution, inflow
temperatures, outflow distribution and meteorology.
Bathymetry
Measurements of bathymetric characteristics were made using USGS topography
maps that are based on a survey done in 1934 before the reservoir was filled.
Earlier bathymetry measurements were spot-checked and some minor revisions
were made. The main changes made during the bathymetry calibration were:
- Varying layer depths were used in place of constant (i.e., 3-meter)
layer depths. The total number of layers remained the same. The
configuration of the varying layer depths from the top of the
reservoir to the bottom is as follows: 21 - 1.5 m layers, 19 - 3.0 m
layers, and 11 - 6.0 m layers.
- The PQC switch was set to OFF because some of the runs with the
varying layer depths would not run with the PQC switch set to ON. The
PQC switch determines how the inflows are distributed amongst the
layers in the upstream reach of each branch. If the PQC switch is OFF,
the inflows are distributed evenly amongst all of the layers. If the
PQC switch is ON, the inflows are distributed to the laycr(s) whose
density most closely matches the inflow density.
- The method of calculating the bathymetry parameters for the lowest
layers of each segment was modified.
Inflow distribution
Although the model has five branches, historical inflow data are only
available for the four main branches. 1995 inflow data provided by the USGS
were used for the Pit, McCloud and Sacramento Rivers. Squaw Creek inflows
were calculated using a regression relationship with the McCloud River
inflows that was based on the historical data.
The BOR also provided computed inflow data at Shasta Dam. The four branches
are by far the largest rivers in the watershed above Shasta Dam and the USGS
gages are located several kilometers upstream of the dam. Thus, differences
between the sum of the USGS daily flows at the three gages and the BOR
computed inflows at Shasta Dam are primarily attributable to precipitation
and evaporation. In order to balance the inflows from the rivers with the BOR
computed inflows, the flow differences were distributed to the four branches
according to the average fraction of the total flow of the four branches that
each had carried over the historical period of record (i.e., January 1946 to
December 1963).
The main changes made to the inflow input files during the inflow distribution calibration include the following:
- Newly obtained USGS gaged inflows were used from October 1, 1995 to
December 1, 1995 for the Pit and McCloud Rivers.
- 1995 Squaw Creek inflows were calculated using a new regression
relationship with McCloud River gaged inflows. A minimum flow value of
15 cubic feet per second (cfs) was assumed.
- The fractions used to distribute flow discrepancies between the gaged
inflows and BOR computed inflows at Shasta Dam to Squaw Creek, the
McCloud River, the Pit River and the Sacramento River were revised.
Inflow temperatures
Water temperature data used in the model were obtained for the Pit, McCloud
and Sacramento Rivers from the California Data Exchange Center's (CDEC)
website. Gaps in the available data for these rivers were filled by using a
regression relationship between flow and water temperature. Some manual
adjustments were made to fit the regressed data to the gaps in the actual
data. Water temperatures for Squaw Creek were assumned to be the same as
temperatures for the McCloud River.
Most of the changes to the inflow temperature input files were due to the
changes in the inflow distributions mentioned previously because of the
regression relationship between inflows and inflow temperatures used to fill
the gaps in available temperature data. The regression relationship was not
changed during the calibration process.
Outflow distribution
For 1995 operations, the BOR provided daily operation records for three
options for releases from Shasta Dam: power, spill, and outlet. Power
releases are those that go through the penstocks for power generation. Spill
releases are those that go over the crest of the dam, and there were no such
releases in 1995. In 1995, the outlet releases went through bypass pipes
(i.e., they bypassed the penstocks) that are located at three levels. The
actual distribution of the outlet releases at each level is unknown for 1995,
although BOR personnel were able to estimate seasonal operating
priorities.
The changes made during the outflow distribution calibration involved the
incorporation of these priorities into the distribution of the outlet
releases amongst the three levels of bypass pipes.
Meteorology
The following meteorological data are needed for the meteorological input
file: air temperature, dewpoint temperature, wind speed, wind direction, and
cloud cover. There are no meteorological data available at Shasta Lake until
March 1996, when a meteorological station located at Shasta Dam began
collecting data. These data are available through a BOR computer in Idaho.
Prior to March 1996, the closest available meteorological data are from a
station located at the Redding airport approximately 15 kilometers to the
south of Shasta Dam and at about 200 meters less in elevation. These data are
available from the Western Regional Climate Center for a fee.
Because no meteorological data are available at Shasta Lake for 1995, all
data for the input file were developed in some manner from 1995
meteorological data at the Redding airport. The following changes were made
during the calibration process:
- The factor subtracted from the air temperatures at Redding to obtain
air temperatures at Shasta Lake was changed from 1.13 oC to
1.71 oC. The first factor was determined using the moist
adiabatic lapse rate for the change in air temperature with elevation,
while the second factor was determined using the dry adiabatic lapse
rate.
- The methodology for calculating the dewpoint temperatures remained the
same, but because the dewpoint temperature calculation depends on the
air temperature at Shasta Lake, the estimated dewpoint temperatures at
Shasta changed.
- Wind speeds at Shasta Lake were calculated using a multiple regression
equation with the Redding wind speeds. Previously, Redding wind speeds
were used without adjustment at Shasta Lake.
- A wind sheltering coefficient of 1.0 was used instead of 0.85. This
coefficient is used to account for the effect of surrounding terrain
on sheltering a waterbody from winds observed at a meteorological
station that may be located at a distance from the waterbody. A wind
sheltering coefficient of 1.0 indicates that the waterbody is not
sheltered from the measured wind speeds. Since the multiple regression
of the Redding wind speeds effectively locates the winds at Shasta
Lake, the use of a wind sheltering coefficient of 1.0 is reasonable.
No changes were made to the wind direction or cloud cover data in the
meteorological input file. Wind direction and cloud cover data at Redding
were used at Shasta Lake without adjustment.
Measured water temperature data
Temperature and other water quality parameters have been measured by the USGS
and BOR at several sampling stations. Table 1 summarizes the locations and
dates of USGS samplings in 1995 that were used to calibrate the model.
Table 1. Locations and dates of USGS water quality sampling in 1995.
(Larger table)
Results
Table 2 summarizes the runs that were made that represent the revised model
input files after each category of data was calibrated. Many other runs were
made that are not shown in the table to determine the effects of changing
various parameters on predicted water temperatures.
Table 2. Description of calibration changes in key model runs.
(Larger table)
The changes described in Table 2 are as follows:
-
New bathymetry
- Varying layer depths; PQC switch OFF; new method of calculating bottom
layers
- New inflow distribution
- New USGS flows for Pit and McCloud Rivers; new regression for Squaw
Creek flows with minimum flow of 15 cfs; new fractions to distribute
flow discrepancies
- New inflow temperatures
- New calculated inflow temperatures due to the new inflow distribution
where there are gaps in available inflow temperature data
- New outflow distribution
- New bypass release distribution as described by the USBR
- New meteorology
- New air temperatures due to change in reduction factor; new dewpoint
temperatures because of the new air temperatures; new wind speeds
calculated using multiple regression techniques; wind sheltering
coefficient changed to 1.0
Table 3 below shows the overall r-squared values, root mean squared error
(RMSE), and absolute mean error (ABSE) for each run when compared with
measured data. Generally, RMSE and ABSE values decreased and r-squared values
increased as additional categories of data were calibrated. The greatest
improvement in the model results occurred with the bathymetry and
meteorological modifications, while changes to the inflow distribution,
inflow temperatures, and outflow distribution had little effect on model
results.
Table 3. R-squared values, root mean squared error, and absolute mean
error for Oldbath, Newbath, Newflows, Newtemps,
Newout, and Newcalib.
(Larger table)
Figures 2 through 5 compare the modeled water temperatures from runs
Oldbath and Newcalib with measured water temperature data for
Segment 16 in the late spring, summer, early fall, and winter. Segment 16 is
located just downstream of the confluence of the McCloud and Pit Rivers and
is considered representative of the overall model temperature calibration
results.
Figure 2. Modeled and measured water temperatures for Segment 16 on
Julian day 130.
Figure 3. Modeled and measured water temperatures for Segment 16 on
Julian day 206.
Figure 4. Modeled and measured water temperatures for Segment 16 on
Julian day 264.
Figure 5. Modeled and measured water temperatures for Segment 16 on
Julian day 317.
Isotherms for runs Oldbath and Newcalib along with measured
water temperature isotherms are shown in Figures 6 and 7. Note that the plots
show that water levels predicted with the old bathymetry are noticeably
different from the actual water level data supplied by the USBR, but the runs
with the new bathymetry predict water levels that are essentially the same as
the actual data.
Figure 6. Comparison of Oldbath and measured isotherms for
Segment 16. (Larger
figure)
Figure 7. Comparison of Newcalib and measured isotherms for
Segment 16. (Larger
figure)
The large improvement in the model results due to the changes in the
bathymetry are largely due to the use of varying layer depths instead of
constant layer depths. While the total number of layers is unchanged, the
finer resolution of the layers at the top of the reservoir has improved the
estimates of the water temperature throughout the year. The tradeoff has been
in the computation time, with the new configuration having a run time that is
about 2.5 hours, while the constant-layer configuration ran in approximately
1.5 hours.
In addition to these general observations, other insights gained during the
calibration process include the following:
Bathymetry
- The effects of changing the widths of the segments were investigated.
While changing the widths by 25% resulted in significant changes in
the model-predicted temperatures, such an adjustment was not
considered reasonably representative of the reservoir's bathymetry.
Smaller changes in segment widths (i.e., by applying width factors
between 0.90 and 1.10) produced inconclusive results.
- Modeled water temperatures resulting from runs with the PQC switch set
to OFF were generally slightly higher than corresponding temperatures
resulting from runs with the PQC switch set to ON. Differences in
modeled temperatures were usually less than 0.5 oC.
Inflow temperatures
- A run was made with all inflow temperatures reduced by 1 oC
for all of 1995 (Lowtemps). Temperatures predicted by this run
were generally less than 1 oC lower than those predicted by
the run without the temperature reduction (Newtemps).
Interestingly, the predicted temperatures for both runs were
essentially the same in the upper layers of the reservoir on the days
compared (i.e., Julian days 172, 206, and 241), while the predicted
temperatures began to diverge by more than 0.5 oC at about
8-11 meters below the water surface. This observation indicates that
the temperatures in the surface layers are more influenced by
air-water interactions than by the inflow temperatures.
Outflow distribution
- As expected, the modeled temperatures change the most in Segment 19,
which is the segment closest to the dam (where the outflows occur).
Meteorology
- The model is sensitive to the wind sheltering coefficient. The
greatest effect of changing the wind sheltering coefficient occurs at
or near the water surface, with lower coefficients resulting in higher
estimated water temperatures at the surface. At deeper parts of the
temperature profile, the wind coefficient has very little effect.
- The model is sensitive to air temperature to a lesser degree than the
wind sheltering coefficient. The highest sensitivity occurs at or near
the water surface, but the effect is not consistent with changes in
air temperature. In other words, if the air temperature is lowered
consistently by a certain amount on all days, noticeable changes in
modeled water temperatures are not seen on all days. This result
indicates that changes in air and dewpoint temperature alone may not
significantly affect model results, However, the interaction of other
model components such as wind speed combined with air temperature may
result in greater effects on the water temperature profile.
- The model appears to be sensitive to wind speed, but a sensitivity
analysis was not done for this paraineter. The greatest improvement in
the model predictions during the meteorological calibration occurred
when the method of estimating the wind speed at Shasta was changed.
- The model is sensitive to cloud cover, with modeled water temperatures
decreasing as cloud cover increases. When comparing runs with constant
cloud cover throughout the year, modeled water temperatures were as
much as five degrees cooler with 100 percent cloud cover than with
zero percent cloud cover.
- The model is not sensitive to wind direction. Runs were made with
constant wind direction throughout the year (for example, with wind
always blowing from the north), and modeled water temperatures were
very similar regardless of which direction the wind was blowing
from.
- Because the model is fairly sensitive to the meteorology data, the use
of actual measured data at Shasta Lake for validation runs and other
runs utilizing data after March 1996 should result in very good water
temperature predictions.
References
Hovecamp, S., C. Sarsfield, and J. DeStaso III, 1996, Shasta Dam temperature
control device: ecological study design, U.S. Department of the Interior,
Bureau of Reclamation, Northern California Area Office, Shasta Lake, CA.
National Biological Service (NBS), 1996, Research strategic plan: Shasta
limnology and fish habitat (Work Unit #229, Study/Task #1), Midcontinent
Ecological Science Center, River Systems Management Section, Fort Collins,
CO.
U.S. Department of the Interior, Bureau of Reclamation (BOR), 1994, Fact
Sheet: Shasta Dam, Shasta Powerplant, Shasta Lake, Keswick Dam, Powerplant,
and Reservoir, Mid-Pacific Region, Sacramento, CA.
Back to the SMIG Features Page
Home |
Mailing List |
Features |
Conferences |
Classes |
Reading |
Model Archives |
Feedback
Stewart Rounds, SMIG coordinator
<sarounds@usgs.gov>
U.S. Geological Survey
http://smig.usgs.gov/SMIG/features_0997/shasta_w2.html
Last modified Wednesday, 17-Dec-2003 14:07:00 EST
Privacy Statement ·
Disclaimer ·
FOIA ·
Accessibility
Introduction