Regional Scale Simulation of Water Temperature and
Dissolved Gas Variations in the Columbia River Basin
By Marshall C. Richmond, William A. Perkins, and Yi-Ju Chien
Paciﬁc Northwest National Laboratory, Richland, Washington, USA
In this paper we present a numerical model used for simulating hydrodynamics, water
temperature, and total dissolved gas transport on a regional scale in the Columbia River
Basin. The model is one-dimensional, unsteady, and applicable to branched channel sys-
tems. The use of the model is presented for two example studies. The ﬁrst examines the
effect of impoundments on water temperatures. The second example discusses how the
model was used to perform a comparative analysis of different structural and operational
alternatives to reduce total dissolved gas concentrations. In both studies, frequency anal-
ysis was performed on the simulated values to calculate temperature and dissolved gas
exceedance levels (as compared to water quality standards) at critical locations.
The management of water resources in the Columbia River Basin requires attention to
many complex economic, sociological, and ecological issues. One of the central issues is
the impact of the Columbia and Snake River dams on the ecology of the river system. Of
particular concern are the impacts on water quality and on anadromous salmonid species,
several of which are listed as threatened or endangered.
One of the current operational strategies to beneﬁt migrating juvenile salmonids is fo-
cused on increasing spillway ﬂows at Columbia and Snake River mainstem dams with the
goal of reducing overall migration times and decreasing the number of migrants passing
through turbines and other non-spill routes (NMFS, 1995). Increased spillway discharges
to aid ﬁsh passage, to conform with ﬂood or reservoir management rules, or to meet vary-
ing power generation demands, also introduce supersaturated levels of dissolved gases
into the river. The supersaturated gas levels increase the potential for violation of water
quality standards (USEPA, 1985) and can cause ﬁsh to develop gas bubble trauma, which
can be fatal (Weitkamp and Katz (1980), Fidler (1988)). The evaluation of potential phys-
ical modiﬁcations to the dams or alternative operational strategies requires a quantitative
understanding of the linkages between dissolved gas production mechanisms, project
operations, dissolved gas transport, water quality criteria, and the exposure of ﬁsh to
potentially harmful levels of dissolved gases.
Thermal conditions are also signiﬁcant and one important question is what are the tem-
perature differences between the current, impounded conditions and a free-ﬂowing, unim-
pounded river? Water temperature is a key physical quantity that affects the time of smolt
emergence and predator dynamics, among many other members of the river ecosystem.
In this paper, a one-dimensional unsteady hydrodynamics and transport model is pre-
sented as a tool to perform comparative analyses of different river conﬁgurations (im-
pounded and unimpounded) and total dissolved gas (TDG) abatement alternatives for the
Lower Columbia and Snake River systems.
Hydrodynamics and environmental transport over large reaches of the mainstem Columbia
and Snake River have been simulated using two models: the one-dimensional (1D)
MASS1 model and the two-dimensional depth-averaged (2D) Modular Aquatic Simula-
tion System (MASS2) model (see Richmond et al. (1999). The geographic domain for
each model is shown in Figure 1. Because a 2D model simulation of an entire season (5
months) or multiple years can be computationally expensive (in terms of computer time) to
perform, the 1D MASS1 model has been employed to perform regional scale simulations
over long time frames.
025 50 75 100
Figure 1: Geographic domain for the 1D MASS1 (red line) and 2D MASS2 models (black
line) along the Columbia and Snake Rivers.
The analyses discussed in this paper were performed using the MASS1 code which is
a one-dimensional, unsteady hydrodynamic and water quality model for river systems.
MASS1 simulates cross-sectional averaged processes (one-dimensional) in branched
channel systems at discrete cross-sections. The model accepts time varying inﬂows of
water and constituents at points or as lateral inﬂow/outﬂow along a channel reach. Hy-
draulic structures such as dams, weirs, and culverts can be also included in a channel
The hydrodynamic equations are discretized using the Preissmann four-point implicit
ﬁnite-difference scheme and the resulting system of nonlinear algebraic equations are
solved using the double sweep method as described in Cunge et al. (1980). The various
transport equations are solved using the split-operator method. The advective part of the
system is solved using an explicit TVD (total variation diminishing) scheme presented by
Gupta et al. (1991). Explicit methods are also used for the diffusive (ﬁnite-volume) and
source term parts (Euler method) of the transport equation. A time sub-cycling scheme
is used to allow the hydrodynamics to run at the larger time steps allowed by the implicit
Preissmann scheme while using a smaller time step that satisﬁes the explicit stability
criteria (Courant number less than one). A detailed description of the mathematical for-
mulation of MASS1 is presented in Richmond et al. (2000).
Heat exchange processes at the water surface are modeled using standard parameter-
izations given in Edinger et al. (1974); meteorological conditions can be assigned on a
zonal basis using multiple weather stations. The model includes a general mass transport
equation solver that allows any number of constituents (up to computer memory limits) to
be simulated by specifying individual source/sink processes. The transport equation is
solved using an accurate transport scheme so that rapidly-varying concentration distribu-
tions can be simulated.
Gas exchange at the air-water surface of the river is a process that acts as either a source
or sink of dissolved gas. For supersaturated water, the exchange process acts to reduce
TDG in the river and move it toward equilibrium with the saturation concentration of air.
Rates of air-water gas exchange increase with wind speed as waves and turbulence are
produced. Degassing through air-water exchange can be a very important process in
reservoirs which are longer, shallow, or subject to more intense wind-waves. The air-
water surface transfer coefﬁcient in MASS1 is a function of wind speed is computed using
a curve ﬁt to empirical data presented in O’Connor (1982).
The MASS1 model has been applied to over 800 miles of river system that includes 15
hydroprojects and many tributary inﬂows. The model has been calibrated and validated
using water temperature and dissolved gas measurements from locations in the forebay
and tailrace of each mainstem dam. For the Lower Snake River, mean absolute errors
between simulated and measured water temperature in the tailraces (Lower Granite, Little
Goose, Lower Monumental, and Ice Harbor Dams) ranged from 0.3 to 1.26 degrees C for
the years 1996, 1997, and 2000. RMS error for simulated TDG for the Lower Snake and
Lower Columbia (McNary, John Day, The Dalles, and Bonneville Dams) ranged from 17
to 48 mmHg for the years 1996 and 1997. Note that 38 mmHg nominally corresponds
to 5 saturation.
Gas production relationships for dams on the Lower Columbia and Snake Rivers have
either been physically-based (mechanistic) or observational-based (regressions). Exam-
ples of attempts to formulate mechanistic production equations are provided in the works
of Roesner and Norton (1971) and more recently by Geldert et al. (1998). In these simula-
tions, the regression approach, based on ﬁeld measurements, of TDG has been adopted
The application of the one-dimensional MASS1 model is presented for two studies. The
ﬁrst study uses the model to compare simulated water temperatures for current impounded
conditions and unimpounded conditions for the Lower Snake River. The second example
uses the model to perform a comparative analysis of different operational and structural
dissolved gas abatement (reduction) scenarios for federal hydroprojects on the Lower
Columbia and Lower Snake Rivers.
In each study, probabilistic approach was adopted in performing and analyzing the simu-
lations rather than a “design-ﬂow” methodology. The probabilistic approach relies on com-
puting cumulative frequency distributions for various criteria using each model-simulated
time-series of dissolved gas concentrations. This approach provides information about
how frequently water quality criteria may be violated at a particular compliance point or
over a spatial area. Such an approach accounts for the unsteady nature of real river
conditions and shows how the TDG levels of the system respond to imposed operational,
hydrologic, and meteorological conditions.
Lower Snake River Water Temperatures
Using historical main stem inﬂows, tributary inﬂows, and meteorological conditions for a
35- year period spanning 1960 to 1995, MASS1 was used to simulate the water tempera-
ture with the dams in place (current impounded conditions) and supposing the dams were
removed (unimpounded conditions). Frequency analysis was performed on the results
(saved values of daily average and maximum temperatures) to compare simulated water
temperatures for impounded and unimpounded conditions. Figure 2 shows the median
value of the daily average temperature (over the35-year period) at Ice Harbor dam during
April through October. The unimpounded river warms up faster and has higher mid-
summer temperatures than the impounded river, but it cools down more quickly starting
in September. The fraction of time that daily average temperatures exceeded the water
quality criteria and the average value of that exceedance are shown in Figure 3. In both
cases the simulations indicate that the criteria are exceeded by about 1.4 to 1.7 degrees
C for 44-52 days along the Lower Snake river. Individual dams are often referred to by a
three-letter code for brevity. Table 1 lists the codes forseveral of the Snake and Columbia
These simulated trends are in general agreement with water temperature measurements
from the mid-Columbia River shown in Figure 4. In this reach of the river, the numerical
water temperature criteria is 18 degrees C. These measurements show that prior to the
completion of Grand Coulee Dam water temperatures at the Rock Island Dam (scroll
case) could exceed the criteria during the summer months. After the completion of Grand
Coulee Dam, water temperatures downstream at Rock Island Dam and at the Hanford
B reactor (about 10 miles downstream of Priest Rapids Dam) have somewhat reduced
peaks, but remain warmer into fall and early winter. Note that the Hanford B reactor data
are from a period prior to the completion of the Wanapum and Priest Rapids Dams.
Table 1: Codes used for selected Snake and Columbia river projects.
River Project River Mile Code
Columbia Bonneville BON
The Dalles TDA
John Day JDA
Priest Rapids PRD
Rock Island RIS
Rocky Reach RRH
Chief Joseph CHJ
Grand Coulee GCL
Snake Ice Harbor IHR
Lower Monumental LMN
Little Goose LGS
Lower Granite LWG
Hells Canyon HCD
Dissolved Gas Abatement Alternatives
The U. S. Army Corps of Engineers Dissolved Gas Abatement Study (DGAS) is examining
measures to reduce dissolved gas produced by the eight federal hydroelectric dams on
the Lower Columbia and Snake Rivers. These measures include a number of structural
and operational modiﬁcations to the dams intended to reduce dissolved gas concentra-
tions produced by spillway discharges and thus move toward meeting water quality criteria
and reducing potential mortality from gas bubble trauma. Implementation of these mea-
sures may also provide additional operational ﬂexibility to increase spillway discharges for
ﬁsh passage purposes. DGAS will use the relative performance of each alternative mea-
sure in reducing dissolved gas concentrations as one basis for comparing the various
The MASS1 model was run for the gas abatement scenarios over the April 1 through
September 1 season and the time-varying results were analyzed to assess the compara-
tive performance of the alternatives.
Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1
Median Mean Temperature for Day
Day of Year
Ice Harbor Dam (SRM 9.5)
Unimpounded River Scenario
Current Conditions Scenario
Figure 2: Comparison of the simulated median mean temperature at Ice Harbor Dam lo-
cation for current impounded and unimpounded conditions.
Four so-called “fast-track” scenarios were simulated. The general features of these sce-
1. This scenario is used to examine the possible beneﬁt of project spill pattern changes
only. Current project conﬁgurations were assumed, but the spill pattern was changed
at all projects so that spill was evenly distributed only over bays with deﬂectors. At
The Dalles, having no deﬂectored bays, a uniform pattern over all bays was as-
sumed. A uniform spill pattern tends to lower gas production at those projects where
the spill pattern is highly nonuniform. The high spill ﬂow through a few bays is spread
over many bays, thus lowering the gas production.
2. Starting with the baseline scenario, it was assumed deﬂectors would be installed in
any spill bays in which they were are not currently installed, except for The Dalles,
where current conditions were used. At each project with a full compliment of de-
ﬂectors, a uniform spill pattern was assumed (except at Lower Granite).
3. In addition to the modiﬁcations of scenario # 2, powerhouse/spillway ﬂow divider
(splitter) walls would be installed on those projects susceptible to powerhouse ﬂow
entrainment, and The Dalles dam was assumed to have a full complement of de-
4. In addition to the scenario #3 modiﬁcations, increasing the bottom elevation of the
tail races would be implemented at those projects which might beneﬁt most (Lower
Granite, Little Goose, Lower Monumental, McNary, and Bonneville).
Fraction of Time Daily Mean Exceeded Standard
Average Amount Daily Mean Exceeded Standard, °C
Figure 3: Fraction of time (year) water temperatures exceeded temperature criteria and
average value of that exceedance for current impounded conditions and unim-
0 5 10 15 20
Jan 1 Mar 1 May 1 Jul 1 Sep 1 Nov 1 Dec 31
Day of Year
Median Temperature for Day, C
Rock Island 1933−1941
Rock Island 1943−1960
International Border 1957−1997
Hanford B Reactor 1952−1959
Figure 4: Measured water temperatures in the Columbia River.
Hourly simulated TDG saturations from the MASS1 model for the fast-track scenarios
were statistically summarized in order to compare the effects of each scenario. Simulated
hourly saturation values were recorded at three important locations: the spillway average
(does not include any powerhouse TDG and associated mixing), the cross section aver-
age at the tailwater monitoring location, and the cross section average at the downstream
forebay monitoring site.
From these hourly values a cumulative frequency distribution (CFD) curve was developed,
showing the percentage of time any given value of TDG was exceeded during the sim-
ulated season.The simulated hourly TDG data was examined to determine the number
of days the waiver to the 110 saturation water quality standard was violated. Figure 5
compares them graphically. Several waiver criteria have been deﬁned by the States of
Washington and Oregon, as well as NMFS. Water quality waiver refers to the measure
computed based on the daily highest 12-hourly values in a single calendar day exceeding
120% at the tailwater and 115% saturation at the downstream forebay monitor locations,
respectively. This criteria closely corresponds to the Oregon and NMFS waiver deﬁnitions.
Scenarios 3 and 4 provided the most beneﬁt in the both the Columbia and Snake Rivers.
In the Columbia this is primarily due to the installation of deﬂectors and use of uniform
spill patterns. In the Snake River, the beneﬁts are primarily due to the addition of the
divider walls and raised tailraces. The Columbia derives more beneﬁt from Scenario 2
and from deﬂector installation at The Dalles in Scenario 3. The addition of deﬂectors
at Bonneville, in fast-track Scenarios 3 and 4 does signiﬁcantly reduce the number of
days the water quality waiver criteria is exceeded at the downstream monitoring site. The
addition of deﬂectors at The Dalles in Scenarios 3 and 4 results in signiﬁcant reductions
in the number days exceeding the water quality waiver.
Spillway Medium/High Flow Season (1996)
LWG LGS LMN IHR MCN JDA TDA BON
Days in Exceedence
Fast Track #1
Fast Track #2
Fast Track #3
Fast Track #4
Tailwater Fixed Monitor Location
LWG LGS LMN IHR MCN JDA TDA BON
Days in Exceedence
LGS LMN IHR MCN JDA TDA BON TID
Days in Exceedence
Figure 5: Total days exceeding the water quality waiver for the daily highest 12-hour av-
erage TDG percent saturaturation. Simulations for a medium-high ﬂow season.
Project codes are listed in Table 1
Application of the MASS1 model to simulate hydrodynamics, water temperature, and dis-
solved gas transport has produced a broad range of metrics that have been used to
compare the performance of different river system conﬁgurations and gas abatement al-
The Snake River long-term temperature simulations showed that the primary difference
between the current and unimpounded river scenarios is that the reservoirs decrease
the water temperature variability. The reservoirs also create a thermal inertia effect that
tends to keep water cooler later into the spring and warmer later into the fall compared
to the unimpounded river condition. Vertical average temperatures in mid-August tended
to be about 1 degree C warmer near the Ice Harbor Dam location for unimpounded river
conditions. In September, impounded conditions were about 1-2 degrees warmer than
unimpounded conditions. However, since the model is vertically averaged temperatures
in the upper part of the water column in the current conditions may be underestimated.
The simulations for the gas abatement study quantitatively show that as additional gas
abatement measures are implemented at a project, or series of projects, the number
of days exceeding the water quality standard decreases. Flow deﬂectors are an effective
gas abatment measure. Alternatives which reduce TDG concentrations at The Dalles and
Bonneville Dams will lead to proportional TDG reductions in the Columbia River estuary.
This indicates that a possible implementation schedule would prioritize these two projects
in order to realize the beneﬁts in the estuary. Reducing TDG in the estuary could be
signiﬁcant because more of the available habitat is less than 2 meters deep which means
there is less opportunity for ﬁsh and other aquatic species to avoid gas bubble disease by
being below the compensation depth. Improvements in estuary conditions beneﬁt salmon
as all smolt must pass through the estuary during an outmigration.
These applications also show the potential for using unsteady, one-dimensional ﬂow and
transport models for real-time in-season operations control to manage operations and
water quality forecasting in a large river system. For example, the 600 river mile Lower
Columbia and Snake River region was simulated (on a Pentium 4 processor) for a 5-
month period in about 2 hours at a time increment of 15 minutes. This would roughly
translate to about 6 minutes to forecast out a week. Thus it would be feasible to couple
the model with optimization routines to examine tradeoffs between spill, power, and other
The majority of this work was sponsored by the U. S. Army Corps of Engineers, Walla
Walla and Portland Districts under contract DACW68-96-D-002.
Cunge, J., F.M. Holly, J., and Verwey, A. (1980). Practical Aspects of Computational
Hydraulics. Pitman Advanced Publishing Program.
Edinger, J., Brady, D., and J.C.Geyer. (1974). Heat exchange and transport in the en-
vironment. Technical report, Electric Power Research Institute (EPRI), Palo Alto, CA.
Publication No. 74-049-003.
Fidler, L. (1988). Gas Buuble Trauma in Fish. PhD thesis, Department of Zoology, Uni-
versity of British Columbia, Vancouver, British Columbia.
Geldert, D., Gulliver, J., and Wilhelms, S. (1998). Modeling dissolved gas supersaturation
below spillway plunge pools. J. of Hydraulic Engineering, 124(5):513–521.
Gupta, A., Lake, L., Pope, G., Sephernoori, K., and King, M. (1991). High-Resolution
Monotonic Schemes for Reservoir Fluid Flow Simulation. In Situ., 15(3):289–317.
NMFS (1995). Endangered species act section 7 consultation, biological opinion: Reiniti-
ation of consultation regarding 1994-1998 operation of the federal columbia river power
system and juvenile transportation program in 1995 and future years. Technical report,
National Marine Fisheries Service, NMFS, Northwest Region, 7600 Sand Point Way
N.E., Seattle, Washington.
O’Connor, D. (1982). Wind effects on gas-liquid transfer coefﬁcients. J. Environmental
Richmond, M., Perkins, W., and Chien, Y. (2000). Numerical model analysis of system-
wide dissolved gas abatement alternatives. Final report, Battelle Paciﬁc Northwest
Division, P.O. Box 999, Richland, Washington, 99352. Prepared for the U.S. Army
Corps of Engineers, Walla Walla District under Contract DACW68-96-D-0002.
Richmond, M., Perkins, W., and Scheibe, T. (1999). Two-dimensional simulation of hy-
drodynamics, water quality, and ﬁsh exposure in the columbia/snake river system. In
Proceedings of WaterPower99.
Roesner, L. and Norton, W. (1971). A nitrongen gas model for the lower columbia river.
Final report, Water Resources Engineers, Inc. Prepared for the U.S. Army Corps of
Engineers, Portland District.
USACE (1999). Dissolved gas abatement study, phase ii, 60% draft. Draft technical
report, U.S. Army Corps of Engineers, Portland and Walla Walla Districts. Distributed
USEPA (1985). Quality criteria for water. Technical Report U.S. Environmental Protection
Agency Report 440/5-86-001, Washington, D.C.
Weitkamp, D. and Katz, M. (1980). A review of dissolved gas supersaturation literture.
Transactions of the American Fisheries Society, 109:659–702.
Marshall Richmond, Ph. D. is a Chief Engineer in the Hydrology Group at Paciﬁc North-
west National Laboratory. He specializes in the development and application of computa-
tional ﬂuid dynamics to engineered and environmental systems.
Bill Perkins is a Senior Science and Engineering Associate in the Hydrology Group at
Paciﬁc Northwest National Laboratory. His areas of expertise include numerical model
development and application, geographic information systems, and scientiﬁc visualization.
Yi-Ju Chien is a Research Scientist in the Applied Geology and Geochemistry Group at
Paciﬁc Northwest National Laboratory. She specializes in geostatistics and statistical data