1. Plug flow NOT realistic.
2. The mixing model is not as important as getting the general trends for cumulative impact. Better topography and velocity distributions should be included. Depth is a key to exposure and needs to be included more realistically into the proposed model -- possibly in form of available depth distributions for the river defined by the bathymetry.
3. Include seasonal patterns into dissipation parameter h, since it is highly sensitive to temperature.

1. Goal for model: What % of Volume is > 120%? Get that as close to correct as possible.
2. Mortality rates are affected by:
   • Fish length,
   • Water temperature, and
   • Topography (can fish escape or not?).
   • Another possible source is more project specific structural mortality, incorporating higher mortality rates at flip lips and ski slope like deflectors. Cramer's PIT-tag analysis on stilling basin affects looked at spillway mortality at LGS and found high mortality thought to be due to passage through the outside, non-deflector bays. Can also refer to IHR Deflector Report for information about this type of structural mortality.
3. Solutions? Possibly break up the reservoir so that particularly complicated parts of the reservoir are described with more resolution, i.e. including particularly important topography characteristics.

4.0 Unsteady Flow and Transport Model. Marshall Richmond presented the graphic output from the model he has developed for flow and gas transport on the Columbia and Snake Rivers. Currently it is a 1-D model, with Dworshak, Hell's Canyon Dam and Priest Rapids inputting gas into the system and output given up to the Snake-Columbia confluence. A 2-D near project hydrodynamic model is currently being developed with plans of adding a particle tracking simulation for 2-D mixing. Richmond also mentioned the use of these meteorological values: wind speed, temperature, relative humidity, and %-cloud cover but expressed concern over the lack of stations close enough to the regions of interest. Airport data is not a sufficient substitute for regions in the river canyon.

The goals for output of this model are: gas levels, velocities, and general hydrodynamics, particularly in the tailrace of the dams. Richmond also stated that he would coordinate with those on the CRiSP project to get the desired output in a usable form for their model. The variables of interest from this model for the CRiSP project including: gas levels, velocity, and river depth.

5.0 TDG Mass Transfer. Mike Schneider presented his analysis of gas/spill data. A linear fit of kcfs spilled and %-gas produced was shown at several dams. Day versus night spill plots showed the affects of different night and day spill patterns leading to different gas production curves. The variation around the fitted gas production curve was about 5%. The question was then raised as to the usefulness of predictive curves which are only accurate to within 5-10% , particularly considering the difference between 115% and 125% supersaturation. Carlson commented that this band of uncertainty may be too wide for this predictive curve to be used to evaluate alternative structural changes. A suggestion was made to use CRiSP in monte carlo mode to incorporate the uncertainty into survival analysis.

6.0 Biological Modeling. Larry Fidler discussed his progress on the development of new mortality models. He identified the following important parameters for mortality rates: fish species, fish length, exposure time, effective total gas pressure distributed over (x,y,z,t), water temperature, dissolved oxygen, and activity level. Fidler did an extensive literature search and review on Gas Bubble Trauma and then implemented neural nets to 1) do initial data screening of Gas Bubble Trauma data in the literature, and 2) Develop steady state mortality models. Fidler warned that when using mortality data from the literature, care is needed to not include mortality that was an artifact of the study design.

Time to mortality versus %-mortality was focussed on with the model parameters being: exposure time, species and PO2. Fidler commented that very little data for %-mortality exists for gas levels below 120% which makes it difficult to calibrate a model.