What is the Delta STARS Model?

STARS = Survival, Travel Time, and Routing Simulation

The Delta STARS Model is an individual-based simulation model that predicts survival, travel time, and routing of juvenile salmon migrating through the Sacramento–San Joaquin River Delta The model’s structure and paramaters are based on a recent analysis (Perry et al. 2018) that relates individual survival, travel time, and routing of late-fall Chinook salmon to daily Sacramento River flows at Freeport and Delta Cross Channel operations.

What does the model simulate?

The STARS model simulates travel time, routing, and survival of individuals in a daily cohort as they migrate through eight unique reaches of the Delta. A daily cohort is defined as all fish that enter the Delta on a given day at Freeport. Because travel time, routing, and survival depends on river flow when an individual enters a given reach, overall survival of a daily cohort depends on the entire time series of daily flows during their migration time through the Delta. For example, two cohorts may enter the Delta at the same discharge, but their overall survival will differ if one cohort enters during an ascending hydrograph and the other enters when it descending.

This Shiny App summarizes the output of each daily cohort for the current water year to display 1) overall survival and median travel time through the Delta between Freeport and Chipps Island, 2) route-specific survival and median travel time, and 3) the proportion of fish using each unique migration route. In addition to median travel time, users may also examine travel time distributions of daily cohorts or animate how travel time distributions change over time.

How should model output be interpreted?

It is important to note that the STARS model is based on a set of relationships fitted to hatchery-origin late-fall Chinook salmon that migrated through the Delta between late November and mid-March over a five-year period (2007 - 2011). Therefore, model output should be thought of as a “historical expectation”. That is, the model provides predictions about survival based on what we know about a particular race that migrated through the Delta during a particular time of year under the environmental, operational, and physical characteristics of the Delta that occurred between 2007 and 2011.

Output from the model can be useful when used in conjunction with real time telemetry data to help understand what we might expect to occur based on what we have seen in the past. When real-time data and in-season observations deviate from our “historical expectation”, such deviations provide an important opportunity to learn about system dynamics.

More details on simulation methods…

The following pseudocode describes how travel time, survival, and routing are simulated given a daily time series of discharge and Delta Cross Channel gate operations:

1. Select parameter set i from the joint posterior parameter distribution (see Perry et al. 2018 for details).
2. Initiate the simulation with 1,000 fish at Freeport on day t.
3. Calculate survival in reach 1 given discharge on day t and parameter set i.
4. Draw individual travel times through reach 1 from a log-normal distribution where the mean of the distribution depends on parameter set i and discharge on day t. This yields a distribution of arrival times at the junction of Sutter and Steamboat Slough with the Sacramento River.
5. Draw the route taken by each fish from a Bernoulli distribution where the probability of entering Sutter and Steamboat Slough is a function of discharge on the day each fish arrives at the junction.
6. Calculate the survival probability of each individual for the next reach downstream (Sacramento River or Sutter and Steamboat Sloughs) given the discharge on the day each fish entered the next reach.
7. Draw travel times for each individual for the next downstream reach given the discharge on the day each fish entered the reach.
8. For fish remaining in the Sacramento River, draw the route taken by fish at the junction of the Sacramento River with the Delta Cross Channel (DCC) and Georgiana Slough from a multiple Bernoulli distribution where the probability of entering each route depends on the position of the DCC gates and discharge on the day each fish arrived at the junction.
9. Repeat steps 6 and 7 for all remaining reaches.
10. Repeat steps 2-9 for all days in the simulation.
11. Repeat 1-10 for all iterations of the joint posterior distribution.

This simulation yields a posterior distribution of reach-specific survival probabilities, reach-specific travel times, and routing histories for a cohort of 1,000 individuals entering the Delta at Freeport on each day of the daily time series.

Other noteworthy items:

The STARS model now incorporates California and Nevada River Forecast Center flow forecasts for five days from the present. The CNRFC was kind enough to provide us with a time series of historical forecasts from which we built a model to estimate forecast uncertainty. For each parameter set, i, we now draw a new flow prediction for each forecasts river day. In this manner we propagate flow forecast uncertainty through the STARS model.

Questions or Comments?

This Shiny App was developed by the USGS, Western Fisheries Research Center, Quantitative Fisheries Ecology Section. We plan to continually evolve the Shiny App to improve the user’s experience. Please send questions, comments, and bug reports to the Quantitative Fisheries Ecology Section.

This Shiny App is supported by Columbia Basin Research, School of Aquatic and Fisheries Sciences, University of Washington. Questions, comments, and requests related to this implementation can be sent to: Columbia Basin Research

Route-specific travel times

About route-specific travel time distributions:

The graph on the right shows expected travel time distributions between Freeport and Chipps Island for a cohort of fish entering the Delta on the given date shown by the vertical line in the graph above.

Tips:

- Use slider bar to view travel time distributions for different Delta entry dates.

- Click 'play' button (located below slider on right) to animate travel time distributions over time. The slider can be moved to desired position before pressing play.

Perry et al. 2018, Flow-mediated effects on traveltime, routing, and survival of juvenile Chinook salmon in a spatially complex, tidally forced river delta. Can. J. Fish. Aquat. Sci. 75: 1886-1901 (2018) dx.doi.org/10.1139/cjfas-2017-0310

Hance et al. 2021, From drought to deluge: spatiotemporal variation in migration routing, survival, travel time and floodplain use of an endangered migratory fish. Can. J. Fish. Aquat. Sci. 00: 1-19 (0000) dx.doi.org/10.1139/cjfas-2021-0042

What is the Delta STARS Model?

STARS = Survival, Travel Time, and Routing Simulation

The Delta STARS Model is an individual-based simulation model that predicts survival, travel time, and routing of juvenile salmon migrating through the Sacramento–San Joaquin River Delta The original model’s structure and paramaters are based on a recent analysis (Perry et al. 2018) that relates individual survival, travel time, and routing of late-fall Chinook salmon to daily Sacramento River flows at Freeport and Delta Cross Channel operations. The model was updated in 2022 to predict the survival, travel time and routing of winter-run Chinook salmon smolts based on the analysis of Hance et al. 2021. The updated winter-run model related survival, travel time and routing to additional covariates and includes routing and survival through Yolo Bypass.

What does the winter-run STARS model simulate?

The winter-run STARS model simulates travel time, routing, and survival of individuals in a daily cohort as they migrate through the Delta. A daily cohort is defined as all fish that enter the Delta on a given day at Knights Landing. Because travel time, routing, and survival depends on river flow when an individual enters a given reach, overall survival of a daily cohort depends on the entire time series of daily flows during their migration time through the Delta. For example, two cohorts may enter the Delta at the same discharge, but their overall survival will differ if one cohort enters during an ascending hydrograph and the other enters when it descending.

This Shiny App summarizes the output of each daily cohort for the current water year to display 1) overall survival and median travel time through the Delta between Knights Landing and Chipps Island, 2) route-specific survival and median travel time, and 3) the proportion of fish using each unique migration route.

How should model output be interpreted?

It is important to note that the STARS model is based on a set of relationships fitted to hatchery-origin winter-run Chinook salmon released near Redding (river kilometer 551) in February or March over a five-year period (2014 - 2018). While the migration timing varied between years and among individual, some of these fish were present in the Delta through late-April. The model output should be thought of as a “historical expectation”. That is, the model provides predictions about survival based on what we know about a particular race (winter-run) and lifestage (smolts) that migrated through the Delta during a particular time of year under the environmental, operational, and physical characteristics of the Delta that occurred between 2014 and 2018. For example, the model is less applicable to winter-run passing Knights Landing earlier in the year at a smaller sizes, and is not representative of changes such as the Yolo Bypass Salmonid Habitat Restoration and Fish Passage Project which is likely to change the probability of fish entering Yolo Bypass.

Output from the model can be useful when used in conjunction with real time telemetry data to help understand what we might expect to occur based on what we have seen in the past. When real-time data and in-season observations deviate from our “historical expectation”, such deviations provide an important opportunity to learn about system dynamics.

More details on simulation methods…

The simulation method for the winter-run STARS model differs from the methods used for the late-fall STARS model which reflects difference in how the models were constructed. In practice, for each day t at Knights Landing the winter-run model calculates the complete set of survival probabilities, arrival distributions, and routing probabilities for each downstream junction and reach and integrates over all possible combinations using the mathematical formulas described in Hance et al. 2021. Covariates includes Fremont Weir stage, discharge at Freeport, maximum daily water temperature at Freeport, discharge at Rio Vista and discharge and maximum daily water temperature in Yolo Bypass. Fish can only enter Yolo Bypass when the Sacramento River stage at Fremont Weir exceeds 32 feet. The model uses observed covariates and forecasted flows (courtesy of the California and Nevada River Forecast Center) and temperatures for five days from the present. Uncertainty in future flow and temperatures are propogated through the model. This simulation yields a posterior distribution of overall, route- and reach-specific survival probabilities, travel times, and routing probabilites for winter-run smolts entering the Delta at Knight Landing on each day of the daily time series.

Questions or Comments?

This Shiny App was developed by the USGS, Western Fisheries Research Center, Quantitative Fisheries Ecology Section. We plan to continually evolve the Shiny App to improve the user’s experience. Please send questions, comments, and bug reports to the Quantitative Fisheries Ecology Section.

This Shiny App is supported by Columbia Basin Research, School of Aquatic and Fisheries Sciences, University of Washington. Questions, comments, and requests related to this implementation can be sent to: Columbia Basin Research

Perry et al. 2018, Flow-mediated effects on traveltime, routing, and survival of juvenile Chinook salmon in a spatially complex, tidally forced river delta. Can. J. Fish. Aquat. Sci. 75: 1886-1901 (2018) dx.doi.org/10.1139/cjfas-2017-0310

Hance et al. 2021, From drought to deluge: spatiotemporal variation in migration routing, survival, travel time and floodplain use of an endangered migratory fish. Can. J. Fish. Aquat. Sci. 00: 1-19 (0000) dx.doi.org/10.1139/cjfas-2021-0042