February 8, 2019 Slides
Participants:
Jim Peterson, Erin McCreless, Bruce Digennaro, Chris Kwan, Emanuel Rodriguez, Kevin Clark, Kenneth Kundargi, Scott Hamilton, Shawn Acuña, You Chen Chao, Mike Beakes, Corey Phillis, Denise Reed, Felipe La Luz, Duane Linander, Mike Urkov, Brad Cavallo, J.D. Wikert, Sheila Greene, Ben Geske, Chuck Hanson, Rod Wittler, Rene Henery, Rod Wittler, Maggie Christman, Mike Eakin, Shelley Ostrowski, Sally
Chinook model update
- New coded wire tag report came out last week, with better estimates of escapement of hatchery origin fish for 2013
- There are significant discrepancies from the previous estimates (see graphics in meeting presentation for details)
- We're recalibrating the models to these values. Calibration should be done next week, then we can rerun the models.
Delta smelt model update
Rose (2013) model was originally on daily time step but since CalSim and DSM2 are monthly, we modified model to monthly time step
Rose model used empirical food availability 1995-2006. Some of our candidate actions affect food availability so we are modeling it. Using data from Will Smith/Ken Rose and calculating monthly average food availability for 6 taxa and each life stage in a GLMM as a function of flow, temperature, secchi, and EC. We now have best fitting, taxon-specific models.
The R code, data, and parameters for the food models are at this link; feel free to play around with it and let us know if you have feedback
CalSim and DSM2 do not produce temperature outputs. Emanuel Rodriguez at FlowWest has modeled water temperatures in the Delta as a function of air temperatures at CDEC stations in a simple regression; this is consistent with the literature that suggests air temperature is the main driver of water temperatures. We considered adding flow as a covariate but sufficient data were not available.
All model inputs for the delta smelt model are complete and we've started the model calibration – calibrating to FMWT index
Focusing on time period 1995-2006, consistent with Rose model
Using a genetic algorithm (heuristic) – this is a simulation optimization method for estimating parameters or optimal values. This method has a tendency to find local minima/maxima, so we're running it multiple times with multiple seeds to find values the model converges on
When calibration is completed next week, we'll start modeling the scenarios
Before the next in-person meeting we'll convene a conference call for the Delta smelt subgroup to go over the results of both models. Tentatively planning this for Tuesday Feb. 19.
Comment from Corey: should we consider extending the model past 2006, since the system has changed over the last decade?
- Yes, this is worth considering. We'll try to pull data together for more recent years and extend the model if we can.
Comment from Scott: Differences in results between Scott's model and the Rose/Smith model might be due to the fact that Scott's model incorporates predation by silversides but the Rose/Smith model does not account for predation. Evidence suggests that silversides have a dramatic impact on eggs and larvae. In the final report, we'll just have to explain this source of discrepancies and the fact that the two models are different interpretations of system dynamics.
New business
- Ag revenue modeling should be done by the end of next week
- Timeline: The government shutdown put us about a month behind schedule but we're hoping to summarize all model outputs by the Feb. 22 workshop. After that, we will send out the draft final report for review. We'll have either one more in-person meeting or a conference call to go over the report.
- Jim will be talking to Bruce and Josh about the plan for continuing on to Phase 2.
- Comment from Rene: It would be good to get some feedback on Phase 1 – lessons learned, things to keep in mind for Phase 2. We could potentially have a small workshop focusing on this. General agreement that this is a good i