@ State Water ResnurgeB&MLBuard @
`
`Division of Water Righis
`1001 1 Steeet, 14® Flooe # Sacramenio, Californie 95814 # 916.341,.5300
`l]-fl]‘fl PO, Box 2000 + Sacramenlo, California $58L7-2000 =
`M'fifl. mgh.l:l. 'Fn:xilil-fi.HL.!-l-l:l:I # waw walerrights ey % et - Md&w
`, T
`Magalie R. Salas, Secretary ':fir_:‘, T ?{#'.;';'-.‘
`Federal Energy Regulatory Commission fi% o
`BRB First Street, N. E. ?fi.-t =
`Washington, DC 20426 2 -
`Dear Ms. Salas:
`
`CONCERNS WITH WATER QUALITY MODEL, KLAMATH HYDROELECTRIC
`PROJECT, FERC #2082
`
`PacifiCorp contracted with Water Course Engineering to develop a water quality model to
`evaluate the impacts of the Klamath Hydroelectric Project (Project) on water temperature and
`water quality. During model development PacifiCorp agreed to develop a number of with-out-
`project (WOP) model runs that would provide insight on the Project’s contribution to water
`quality standards violations. The WOP runs for all water quality parameters have not been
`submitted as promised. At a meeting on November 23, 2004, Mike Deas with Watercourse
`Engineering stated that the model runs would be completed by December of 2004, and the
`documentation would be completed by January 2005. In Additional Information Request (AIR)
`W0Q-3, the Federal Energy Repulatory Commission (FERC) asked for input and output files for
`all of the modeling runs. This information was submitted in April 2005, on a set of 26 CDs. The
`files on the CDs were disorganized and difficult or impossible to use. While the FERC asked for
`input and output files, model code may be needed in addition to the files in order to duplicate
`runs. Additional information will be required to determine whether the model is “appropriate’
`and if it can accurately predict different scenarios,
`
`The Bureau of Land Management and the Karuk Tribe hired Scott Wells with Portland State
`University, to perform a review of the model. Dr. Well's identified significant issues with the
`model (report dated May 3, 2004), At the request of stakeholders, PacifiCorp hired Dr. Wells to
`work with Watercourse Engineering to implement the model review comments. In their response
`to ATR GN-2, PacifiCorp stated, “PacifiCorp’s responses have been reviewed by Dr. Wells and
`his feedback has been included. As modeling is still being completed, the interaction with Dr.
`Wells will continue. Therefore, the material in Appendix A is considered as draft
`documentation”. It is not clear when a final model calibration report, incorporating peer review
`comments by Dr. Wells, will be submitted.
`
`The response to GN-2 provided some validation information on certain water quality parameters.
`The ahility of the model to predict some parameters is poor. PacifiCorp stated in the report that
`“Discussion of model performance and results will be forthcoming in the final model
`
`documentation™. It is unknown to us when PacifiCorp will submit the final model
`documentation.
`
`California Environmental Protection Agency
`™ Recycled Paper
`
`
`
`
`
`
`
`
`
`Magalie R. Salas, Secretary i AUG 15 2005
`
`The U.S. Environmental Protection Agency, Oregon Department of Environmental Quality, and
`the California North Coast Regional Water Quality Control Board are currently developing a
`Total Maximum Daily Load (TMDL) for the Klamath River. PacifiCorp agreed to submit an
`executable version of the model to Tetra Tech (contractor) for use in development of the TMDL
`for the Klamath River. An executable version of the model was submitted, but without the WOP
`runs included. Tetra Tech has reviewed the model and has identified changes that need to be
`made so the model meets TMDL development requirements (enclosed). State Water Resources
`Control Board (State Water Board) staff are concerned that the current model performance is not
`adequate for use in evaluating Project alternatives. State Water Board staff encourage the FERC
`to require submission of a model by PacifiCorp that is calibrated, validated, and peer reviewed by
`Dr. Wells.
`
`The State Water Board will need a water quality model for the environmental review and water
`quality certification processes. If State Water Board staff determine the model submitted by
`PacifiCorp is not adequate, it will either be modified, or a new model will be developed. State
`Water Board staff would be pleased to work with FERC staff and other stakeholders to resolve
`
`problems with the model. Please contact me at (916) 341-5341 if you have any questions.
`Sincerely,
`w27
`Staff Environmental Scientist
`Enclosure
`cc: Klamath Service List
`Mark Filippini
`U.8. EPA Region 10
`
`1200 Sixth Avenue
`Seattle, WA 98101
`
`
`
`
`
`
`
`
`TETRATECH, INC.
`
`July 13, 2005
`
`U.S. EPA, Region 10
`1200 Sixth Avenue
`Seattle, WA 98101
`
`Re: Tetra Tech Klamath River Model Development
`
`Tetra Tech reviewed the Revised Klamath River Water Quality Model provided
`by PacifiCorp through Dr. Mike Deas from Watercourse Engineering on April 8, 2005.
`Since that date, we have been developing our own version of the Klamath River Model to
`meet TMDL development requirements. In doing so, we have begun making
`modifications to the Revised Klamath River Water Quality Model. The purpose of this
`letter is to identify the changes that we have begun and will continue to make.
`
`If you have any questions, please contact me at 703-385-6000 x155.
`Sincerely,
`
`fs/
`
`Andrew Parker
`
`Director, Water Resources Modeling and Assessment Group
`Tetra Tech, Inc.
`
`Fairfax, VA
`
`
`
`
`
`
`
`
`1. CBOD/Organic Matter Unification
`
`It is gur opinion that BOD should not be modeled in addition to organic matter. As such,
`we have eliminated the BOD compartment in the modeling system for both the riverine
`and impoundment sections.
`
`Additionally, in the current CE-QUAL-W2 models for the impoundments, particalate
`organic matter is not included in the tributary and distributed boundary condition files,
`This may result in an underestimation of particulate organic matter into the system.
`Therefore, for the major tnibutaries that are highly productive, such as the Lost River
`Driversion Channel, particulate organic maiter loading was represented based on data and
`appropriate assumptions. Concentration boundary condition files were thus modified.
`
`Z. Upper Lake Ewauna/Keno Reservoir Model Fine-tuning
`
`Designation of the two dominant boundary conditions into Upper Lake Ewauna/Keno
`Reservoir (Link River and Lost River Diversion Channel) was limiting the predictability
`of the model (particularly with respect to the nutrient budget, phytoplankton fate and
`transport, and orgamic matter). This was largely due to data limitations. Therefore, we
`applied artificial neural networks (ANN) to help improve the boundary conditions and
`resulting lake model prediction. This proposed approach involved training a NN model
`to represent the source-response relationship between the water quality at Miller Island
`and the concentration at Link River and LRDC boundary, and then using the Miller
`Island data to inversely derive the boundary condition at Link River and LRDC.
`
`Application of NN to improve the calibration in Lake Ewauna involved an iterative
`process. First, the existing Lake Ewauna model was run with different sets of boundary
`conditions at Link River and LRDC. The simulated water quality at Miller Island and the
`flows and water quality ar Link River and LRDC were then used to develop a series of
`NN models, The NN network structure for this study was a three-layer feed-forward
`structure with one input layer, one hidden layer, and one output layer. Three nodes were
`included in the input layer, each representing the simulated water quality at Miller Island,
`the flow at Link River, and the flow at LRDC. Two nodes were included in the output
`layer, each representing the boundary condition concentration at Link River and LRDC.
`The NN were trained using the data obtained from the Lake Ewauna hydrodynamic and
`water quality model. After the model was trained, the observed water quality data at
`Miller Island were plugged into the WN models to obtain an estimation of the boundary
`condition at Link River and LRDC, The above process was repeated for several
`iterations. For each iteration, if the simulated water quality at Miller Island was not
`satisfactory, the new training data obtained from the previous iterations, as well as
`additional training data obtained through applying 2 jittering method were used to train
`the new generation of NN model.
`
`The performance of the Lake Ewauna model in reproducing the observed water quality at
`Miller Island has been improved significantly, It should be noted that this approach only
`
`
`
`
`
`
`
`
`works for improving the model] performance for the upper section of the Lake since the
`impact from the major boundary conditions were not significantly dampened.
`
`3 Lower Lake Ewauna Algae Calibration Improvement
`
`Phytoplankton biomass in Lake Ewauna shows significant variability from upstream to
`downstream, i.e., from Miller Island (upstream) to Hwy56. Observed levels reduce
`dramatically. This trend, however, is not mimicked by the model. The simulated algae
`biomass (based on the model provided) is similar at both locations, This model behavior
`can be explained by the large upstream inflow that causes water to flow quickly from
`upstream to downstream. The algae biomass is transported with this quick moving water
`from upstream to downstream in a relatively short time, causing similar concentrations.
`
`It appears that with the existing lanetic structure in the model, it is impossible to
`reproduce this type of spatial distribution of algae biomass. Dr. Deas noticed that
`sometimes during the summer period the entire Lake Ewauna water column becomes
`hypoxic, and even anoxic. He communicated this observation to many lake and algae
`researchers around the world, and the information he collected led him to believe that the
`summer hypoxia/anoxia was related to the spatial variability in algae biomass in Lake
`Ewauna. Available data showed no other explanation for the observed phenomenon.
`
`Algae need oxygen to respire. Thus, when oxygen levels become low or depleted, algae
`metabolism 15 expected to be impacted. Growth is likely to be slowed down and
`death/excretion is likely to increase, Based on discussions with Dr. Deas, Tetra Tech has
`begun implementing a code modification to the existing CE-QUAL-W2 model to account
`for the dependence of algae metabolism on dissolved oxveen concenirations, as well as
`the length of time algae are exposed to hypoxic/anoxic conditions.
`
`4. Half-saturation for Algae Growth
`
`Currently, algae concentrations are relatively low in the RMA11 models (i.e., low growth
`rates). Tetra Tech plans to increase the half-saturation value for light inhibition of algae
`growth to decrease algae concentrations.
`
`5. Quantify Reaeration in the Copeco Dam to Iron Gate Reservoir Headwaters
`Reach
`
`In the current model for the lron Gate segment, predicted DO profiles do not correlate
`well with observations. Anoxic conditions occur in the metalimnion although DO is
`higher in the hypolimnion. Tetra Tech identified that this could be caused by neglecting
`dam reaeration from Copco Dam in the current model. Further analysis of the dam
`reaeration at Copco Dam, as well as its implications on the DO profile in Iron Gate
`reservoir, will be conducted in an attempt to improve model performance.
`
`
`
`
`
`
`
`
`6. Fine-tune Calibration for Copco Reservoir and Iron Gate Reservoir
`
`The Copco Reservoir and Iron Gate Reservoir model calibrations will be updated due to
`the following reasons: a) the updates for the Lake Ewauna model will change the
`boundary conditions for these downstream impoundments; b) in the current models, high
`settling velocities (5.0 meter/day) are used for particulate organic matter.
`
`7. Development of Estuarine Model
`
`Because the Klamath River impairments extend into the estuary, Tetra Tech has begun
`developing an estuarine model for linkage to the Klamath River Model provided. The
`modeling framework is the 3-D Environmental Fluid Dynamics Code (EFDC), an EPA-
`endorsed and widely applied model (particularly for TMDL development). EFDC allows
`for representing the complex geometry of the Klamath Estuary with a boundary-fitted
`curvilinear grid. The model is capable of simulating important physical processes and
`features, such as the eirculation patiern near the funnel-shaped mouth, islands, etc.
`
`
`
`
`
`
`
`
`