CRiSP.1 models downstream migration and survival of juvenile salmon.
The Columbia River Salmon Passage model (CRiSP.1) tracks the downstream migration and survival of migratory fish through the tributaries and dams of the Columbia and Snake Rivers to the estuary. CRiSP.1 describes in detail the movement and survival of individual stocks of natural and hatchery-spawned juvenile salmonids and steelhead through hundreds of miles of river and the major dams. Constructed from basic principles of fish ecology and river operation, CRiSP.1 provides a synthesis of current knowledge on how the major hydroelectric system in the country interacts with one of its major fisheries. Biologists, managers and others interested in the river system can use this interactive tool to evaluate the effects of river operations on smolt survival.
There are five major uses of the model:
There are two modes in which CRiSP.1 can be executed. Scenario Mode illustrates the interactions of model variables. Monte Carlo Mode, which is stochastic, provides measures of variability and uncertainty in predicting passage survival. Using Monte Carlo Mode, estimates of probability distributions for survival and travel time can be determined for any stock between any two points in the river system.
CRiSP.1 has advanced programming features including:
The model is designed to run on Windows 95/NT operating systems or Sun SPARC workstations running Solaris2.X operating system. CRiSP.1 has been tested in both environments.
CRiSP.1 was developed at Columbia Basin Research, School of Fisheries, University of Washington, under a contract from Environment, Fish and Wildlife, Bonneville Power Administration (BPA).
The model is being applied to the Columbia River. As a result, the files describing the river and release points of fish into the river are an essential part of the model. A list of references pertaining to model theory and calibration can be found in the Theory and Calibration Manual.
CRiSP.1 is a mechanistic based stochastic model simulating fish survival.
CRiSP.1 is a mechanistic based stochastic model that treats passage and survival of pre-selected salmon substocks through the Snake and Columbia rivers and their tributaries. Factors that affect survival of hatchery and wild juvenile stocks include: daily flow, river temperature, predator activity and density, total dissolved gas supersaturation, and river operations such as spill, fish transportation and bypass systems.
Daily river flow is generated from headwater flows calibrated to historical data or from period averaged flows at dams provided from system hyrdomodels HYSSR and HYDROSIM, which are operated by the Army Corps of Engineers and Bonneville Power Administration. Diel variations of spill and daily water budget operations are modeled. Flow is converted to velocity using geometries and pool elevations of the reaches.
Dam passage is based on fish guidance efficiency, passage mortalities, and diel passage behavior. These factors are modeled on a species and dam specific basis.
Fish migration rate is treated in terms of water velocity and fish behavior that includes fish choice of river habitat over the day.
Reservoir mortality is a function of fish travel time and the resulting exposure to predation and gas bubble disease. Predation in turn depends on water temperature, predator activity and density. Gas bubble disease depends on total dissolved gas supersaturation levels, fish depth, and duration of exposure.
Total dissolved gas supersaturation is modeled in terms of spillway geometry, river flow and temperature, and fraction of flow spilled.
CRiSP.1 has two modes of operation: scenario and Monte Carlo.
The model tracks multiple fish releases and assigns passage and survival parameters to individual stocks. A model run covers migration of all stocks over a year. Input parameters describing fish behavior and system operations are specified for each year. Menu driven graphical input tools allow selection of specific combinations of parameters.
Survival and migration of fish in CRiSP.1 can be modeled in two modes: Scenario Mode and Monte Carlo Mode.
Scenario Mode runs one year with as many releases as desired. It can be used to gain insight on the effects of changes in system parameters on the survival and migration of fish during a single water year. In this mode, natural unregulated inflows are specified. These flow into the mainstem rivers and storage reservoirs at headwaters. System operation and fish biological parameters can be varied stochastically according to user specifications.
Monte Carlo Mode runs CRiSP.1 with as many releases as desired for one or more combinations of water year and system operations. The main variable changed in each run is river flow. Flows are specified at the project (dam), not headwaters. In each run, a different flow regime and other model parameters are used. Fish survival is determined for each run and the distribution of survivals from all runs provides an estimate of the probability distribution of survival under the specified conditions. Flow is generated from runs of the hydroregulation models maintained by the Army Corps of Engineers (HYSSR) and the BPA (HYDROSIM). The hydroregulation models use historical water data and a projection of electrical demand to simulate system flows, which are designated flow.archive files. These files give period-averaged flows at operating projects which are modulated by CRiSP.1 to represent daily flows. CRiSP.1 uses the modulated flows along with yearly input data files describing the system operations and fish biological parameters to produce histograms of survival.
A graphical user interface allows viewing and modification of variables and model outputs through an interactive map of the river that includes headwaters storage reservoirs, run of the river dams, reaches, tributaries, and release sites.
An analysis tool allows graphical and statistical comparison of travel time and survival probabilities for any number of alternative runs river locations.
A report generation tool makes tables of model inputs and outputs for inclusion in documents. Graphical model output can be printed or captured with a standard graphics program for inclusion in a document.
Help gives context-sensitive help on model use and the underlying model theory.
CRiSP.1 is calibrated against independent data sets.
Calibration is a major effort in simulation modeling. Since CRiSP.1 describes mechanisms of fish survival and migration it can be calibrated from a variety of independent data. Currently calibration efforts have focused on spring and fall chinook and steelhead.
Fish travel time depends on flow and fish behavior. Flow is input from hydromodels (HYSSR or HYDROSIM) or from historical records of natural runoff. A flow/velocity relationship is defined by reservoir elevation and geometry and was calibrated with data supplied by the U.S. Army Corps of Engineers and the U.S. Geological Survey.
Period averaged hydro-model flows are modulated into daily flows. The modulators use a fixed weekly variation and a daily stochastic variation. Spectrograms of modeled and observed flows have been compared for several recent years.
Fish migration behavior has been calibrated, via a minimization of sum of squares, to PIT tag travel time data and to NMFS survival estimates. This is performed iteratively; alternating between calibration of travel time and of survival. All other model data and parameters are chosen or calibrated first so that the model best fits the travel time and survival data sets.
Reservoir mortality has been calibrated with a variety of independent studies conducted over two decades. These include measures of predation from the stomach contents of predators, measures of survival of PIT tagged fish which are recaptured at downstream dams, and from differences in the numbers of returning adults that, as juveniles, migrated through different segments of the river.
Dam passage has been calibrated using studies on fish guidance efficiency and passage mortality through turbines and bypass systems. While the studies have been conducted over two decades, the recent PIT tag studies have significantly improved the calibration of dam passage factors.
Total dissolved gas supersaturation calibrations were based on the work published by U.S. Army Corps of Engineers, Waterways Experiment Station (WES) for the majority of the eight Lower Snake and Lower Columbia dams. Some of WES's calibrations were not used as a result of structural modifications to the dam or more available data that suggested a different dynamic. Recent studies on total dissolved gas supersaturation have been included in the calibration.
Laboratory studies on the affect of total dissolved gas supersaturation on fish mortality have been incorporated as a dose-response curve. A threshold total dissolved gas level for mortality is included.
Agencies in the Columbia River Basin have access to CRiSP.
The model has been distributed to state, federal and public agencies and tribes in Idaho, Oregon and Washington. Since early 1998, a Windows95/NT version of the model has been made available to the public from the Columbia Basin Research website.
The model is being used to evaluate hydrosystem operations and salmon recovery plans as part of PATH (Plan for Analyzing and Testing Hypotheses) and providing Inseason Migratory Forecasts.
The CRiSP model is programmed with a robust and flexible structure that allows for easy future accommodation of new theory with a minimum of reprogramming effort.
Our understanding of salmon ecology indicates that smolt migration and survival depend on fish behavior, physiology and morphology and fish interactions with their forage base, and predators.
Future work will focus on defining these interactions with emphasis on hatchery practices disease control programs, bypass systems, transportation programs, and predator population control.
Inclusion of fish condition elements requires an integrated effort of theory development, laboratory experiments, and field testing. Many of these efforts are underway. Others are yet to be designed.
Successful integration of this research in a timely manner will require a coordinated effort of biologists, modelers and statisticians. Models must be developed on first principals describing how fish behavior, physiology, and morphology affect predator-prey interactions. Experiments must be specifically designed to calibrate the model parameters. The experiments must also be designed along sound statistical principles.
Model calibration needs to proceed with top down and bottom up approaches. From the top, field studies are required to identify survival rates over river segments. From the bottom, experiments on individual mechanisms of survival are required to calibrate model parameters. Survival predicted from the model calibrated with experiments should agree with survival estimated from field studies. The amount of disagreement in the two approaches provides a measure of uncertainty in our analysis.
CRiSP1 is one component of a planned series of models to define salmon in their ecosystem and the impact of human activities on that ecosystem. A harvest model, CRiSP Harvest, that accepts input from the juvenile passage model, is also available. An adult upstream migration model is under development.