Paula J. White
School of Fisheries
University of Washington
Senior Project (FISH 495)
Dr. Jim Anderson, Faculty Advisor
August 16, 1996
Introduction | Materials and Methods | Results | Discussion | Acknowledgments | References
Pacific salmon are one of the most valuable commercial species in fisheries today, ranking second in quantity and value in 1983 (Allen & Hassler, 1986). However, high fishing pressure and habitat loss and degradation has severely reduced the abundance of many stocks. Some stocks of Chinook (Oncorhynchus tshawytscha) in the Sacramento and San Joaquin River systems have declined dramatically since the construction of terminal dams. The current California stocks of Chinook are heavily supplemented by hatchery fish that are released as fry or fingerlings (Allen & Hassler, 1986).
Chinook, and other Pacific salmon are anadromous, meaning they spawn in fresh water, migrate to salt water and return again to fresh water to spawn. After spawning the adults die. The young may rear in freshwater anywhere from 4 months to over a year and then migrate seaward to spend up to the next 3-4 years of their life at sea. They will then return to their natal stream to spawn. This interesting, and multifaceted life cycle makes the management of this genus (Oncorhynchus) much more complex.
The Sacramento River system is unique in that nearly all stocks of native Chinook are ocean-type; approximately 82-90% (Groot & Margolis, 1991). After a relatively short freshwater residence, the juveniles migrate to sea as sub-yearlings. Adults arrive anywhere from mid-summer to winter, and fry may begin their seaward migration as young as 4 months. Time of year, water temperature, stream flow and fish size all influence the time and speed of the downstream migration.
Fisheries managers have turned part of their focus from managing adults to managing juveniles in the effort to maintain a healthy abundance and a commercial fishery. Computer modeling has become increasingly useful in estimating the passage and survival of the migrating fry. Daily estimates can be made regarding the percent survival of the populations and their approximate position in the river by altering influential factors such as temperature and flow, and by altering the physical parameters of the river such as depth and the addition of stream blocking structures such as dams. Although modeling is not a precise tool such as actual data collection and analysis, a well constructed model can give a fairly accurate estimate and allows for future projection.
The computer model "Columbia River Salmon Passage" (CRiSP) was developed at the School of Fisheries, University of Washington. Created to model juvenile salmon passage and survival in the Columbia River, it pays special attention to physical obstacles, in particular, dams. It provides fishery and river managers with an interactive tool to evaluate effects of individual operation strategies on salmon recovery efforts. The model is stochastic, providing measures of variability and uncertainty in its predictions. It operates on Sun SPARCstations (UNIX computers).
The model can be easily reconfigured to any other river system by altering the description and data files. The description file contains information describing the shape of the river system, depth, dam location and specifics, release sites and maximum and minimum flow. The data file includes information on predation coefficients, temperature, fish guidance efficiency, fish behavior and nitrogen supersaturation. These files may be created for any river system and CRiSP will make predictions for that system.
In this project, CRiSP was reconfigured to the Sacramento River and its major tributaries. The objective was to create preliminary data and description files of the Sacramento system which would be read and modeled by CRiSP.
Materials and Methods
Data collection for the building of the description and data files was done by searching for previously existing data. No actual data was collected during this project.
The construction of the river description file which includes the map, maximum and minimum flow, release sites and dam location was done using a number of sources.
GIS information was found on US GeoData Hydrography and Transportation CD- ROM. Using a built in program on the disc, hydrology layer data (1:100,000) was extracted for the Sacramento area and fed into ArcInfo. From ArcInfo a piece of code was written which allowed streams, double banked rivers, and reservoirs to be displayed. Another piece of code allowed the individual selection of rivers and streams relevant to the project. Computer generated arcs and double banked rivers were refined by hand to follow the most direct path of the river. This allowed the production of a single set of latitude and longitude (latlon) data to describe the river course. This refined data was converted to the CRiSP format of "degrees, minutes, seconds" and placed into a text editor.
Flow information was found on the World Wide Web. California Data Exchange Center (http://cyclone.water.ca.gov/) allows users to select specific stations by county and to extract historical daily flow data for that station. A number of stations were selected along the course of the Sacramento as well as for Cottonwood Creek, Battle Creek, Butte Creek, Feather River, Yuba River, Bear River, American River and the San Joaquin River. Data was selected for approximately the past 30 years, depending upon availability, and searched for maximum and minimum values. These values were built into the description file. The daily values were entered in the CRiSP data file.
Approximate release sites were obtained from the U.S. Fish and Wildlife Service (US-FWS), Stockton, CA. Sites were shown on a map, then a latitude and longitude (latlon) point was approximated by finding the corresponding river arc and general location upon the CRiSP description map. Numbers and dates of releases were also supplied by US-FWS. Group survival and a survival index were determined by the agency and will be valuable in determining the sensitivity of the model by comparing the actual data to the computer generated estimate.
Dam information was sketchy at best. Terminal dams were treated as headwaters in the description file and only one dam, Red Bluff Diversionary Dam, was placed on the map. Additional water removal sites such as Anderson-Colusa Irrigation District (ACID) and Glenn-Colusa Irrigation District (GCID) where fish screens are employed were not built into the description file due to incomplete data and the nature of the structure. Water loss is reflected in the data file.
Depth information was unavailable and was estimated under the assumption the Sacramento and its tributaries were comparable to the Columbia River system.
The data file for the Sacramento holds information such as loss, water temperature, juvenile release numbers, predation coefficients, water velocity, fish migration behavior, dam passage (based upon fish guidance efficiency) and nitrogen supersaturation levels.
Temperature data was found in the publication "USGS Water Resources Data." The station used for obtaining flow data was the preferred choice for obtaining temperature data. Temperature profiles for each station selected was recorded in the data file. The period of record for most stations was narrow and often over 10 years old. The profile of a single year was often the best data available. However, it provides an overview of the yearly patterns.
Data for the other parameters reflected by the data file were primarily approximations from the existing Columbia River file. Again, the assumption the two systems were comparable allowed for the Columbia data to be applied to the Sacramento.
The models sensitivity was evaluated by releasing fish at designated release sites in the model and observing the predicted survival at the final reach (Bay). Factors such as ôreach predator coefficientö and ôreach predator densityö were altered to observed the effects these parameters had on successful passage. They ranged from 7 to 39 and 200- 1000 respectively. For speed calibration the values were set to 39 and 1000. Travel times were adjusted by changing the values of maximum and minimum speed and flow effect on travel rates. The survival predicted by the model was compared to the actual survivals and travel times reported by US-FWS to gauge how accurate the model was when run with the data and description files for the Sacramento River system.
The CRiSP model was run using the Sacramento river description and data files. From these files, the model generated a map and a profile of the system. The model map (Figure 1), created by the description file, shows the layout of the river system as well as Red Bluff Diversionary Dam and the numerous release sites. It describes the reaches, and therefore the shape of the river, by a string of latlon points and describes the dam by listing its physical characteristics (Figure 2).
Figure 2 Sacramento Description File Example
The model's sensitivity when using the Sacramento files appeared, at first, to be greatly impaired when contrasted with the relatively good sensitivity it has when using the Columbia files. Passage was estimated at every release site to be at least 70% and up to 98%. Survival estimates could be lowered slightly by altering the predator density from 200 up to 1000 (default is 200) and the predator coefficient for Chinook 0 from a mean value of 7 to 39 (default is 7). However, the model predicted at most a 10% drop in survival when these factors were changed, resulting in percent passage estimates still well into the 70's and 80's. Actual survival values reported by the US-FWS range from approximately 5% to 64% with most falling between 35% and 45%. The models predictions are far greater than the known values.
A more accurate prediction by the model occurred when travel rates were calibrated for each individual release. Once the speed had been adjusted to allow the fish to arrive within the actual time period reported by US-FWS, the survival predictions made by the model were closer to the actual values Table 1 shows how the predicted survivals changed as the travel time was adjusted. The predator behavior remained the same while travel behavior was altered to fit the model to the actual data. The adjusted values used to fit the model are shown in the table.
|Site||Release jday||collection dates||TT julian day||Survival|
|Battle Creek 116/103 adj||15||25||.84||0|
|Red Bluff adj||.08||.95||.13||0|
|Yuba City adj||5||7.06||.23||0|
|Coleman Hatchery 104 adj||9.84||14.68||.58||0|
|Coleman Hatchery 04 adj||6.19||8.49||.12||0|
Computer modeling has the potential to be easily employed in many biological and managerial applications. As this project has shown, computer models, in particular CRiSP, can be easily reconfigured with appropriate data to fit many different environments. However science, especially computer modeling, is only as good as the data available.
The nature of this project, in that the data was borrowed rather than collected specifically for the purpose of modeling, causes the output to be much more general than would be desired. While running CRiSP with the Sacramento files the results may be directly influenced by the quality and quantity of the data available. Data regarding width, depth, fish behavior, predator behavior, temperature and river obstacles such as dams was incomplete, scarce or not available. Often it was out of date. The data used in this project was collected from a number of agencies, studies, and databases and was therefore presented in different formats, units or terms. The most significant problem encountered was having no data at all.
Even with the lack of ideal data, computer modeling can still be a valuable tool. The accuracy may be, however, compromised. To determine how realistically the model makes a prediction, its results must be compared to actual studies. How fry/smolt passage and survival is effected by temperature, flow, time of release and fish size has been studied in other river systems. This information may be applied to the Sacramento under the assumption the two systems being compared are similar. Comparison of the model's output to scientific study results allows for sensitivity evaluation.
Temperature can have a significant influence on the timing of smolt runs. A threshold water temperature or a pattern of variation for a prolonged period may initiate the downstream migration. Evidence suggests a strong correlation between daytime migratory activity and water temperature (Greenstreet, 1992). Although many juveniles migrate at higher numbers at night, a temperature cue may be their initial prompt to begin seaward migration.
Temperature is also known to be a highly significant factor in determining mortality rates. There are optimum temperatures for survival and growth in which mortality is minimized. However as temperatures reach minimum and maximum threshold values, stress levels elevate and mortality is increased. Beyond the threshold temperatures mortality is high and can have a significant impact on abundance.
The feeding behavior of the predator is also influenced by temperature. Metabolism increases with rising temperature, therefore the predator is capable of consuming more prey. Temperature has other physiological effects which may influence the amount of prey consumed as well as the density of the predator itself. Increased temperatures in the Sacramento may not only be having a significant factor on the survival and passage of salmon, but also on the feeding and survival behavior of the predator.
It should be noted that the temperatures used in the data files were often profiles of a single year. Although they provide a general pattern of temperature throughout the seasons, the year selected may have been biased and not representative of the normal temperatures. Again this is a case of incomplete data which may lead to bias and decreased sensitivity in the model.
Water flow and net river discharge have been shown to be highly influential in the rates at which young salmon migrate. At times of high water flow the rate at which fish passed a fish ladder increased dramatically and was no longer correlated to temperature (Greenstreet, 1992). Increased flow seems to flush the migrants downstream, increasing their rate of passage. Survival of smolts passing through the Sacramento-San Joaquin River delta is highly correlated with the discharge of the Sacramento River (Groot and Margolis, 1991), presumably due to less time to interact with obstacles and potential threats along the course of the migration when the juveniles are being carried downstream by high flows.
Time of release is another influence upon passage time and survival. As seasons progress the physical characteristics of the river as well as temperature and flow undergo change. Time of release is therefore a secondary influence, based upon the more dramatic affect of changing environmental conditions as time progresses. Food availability at different times of year also will influence the survival and overall health of the migrating juveniles. Lack of food in itself may be a cue to begin seaward migration. Predator abundance and feeding rates and efficiency may also have some influence upon the time at which juveniles migrate, the predator putting a selective pressure upon the time, and rate, of migration.
The size of the juveniles during migration certainly has an effect upon survival and passage. The ocean-type populations of the Sacramento River and its tributaries migrate as sub-yearlings. These juveniles migrate as fry, often with yolk still visible. Size influences the position in the river the fish choose to migrate within. Larger fish appear to prefer the middle, swifter moving water, where smaller fish, presumably fry, choose the slower moving water along the banks (Groot and Margolis, 1991). Smaller fish, therefore, migrate at a slower rate than larger fish and are exposed to the dangers of the river environment for a longer period of time.
Sensitivity analysis allows the model to be compared to current knowledge and theory. However, if the model is lacking in its physical parameters, its prediction and output will not be reflective of reality. For the Sacramento data file, much of the data was estimated based upon the assumption the Columbia River and the Sacramento River were comparable. Estimates for depth, predator coefficients, reservoir mortality, fish behavior, dam passage and nitrogen supersaturation were all made from data currently employed in the Columbia River data file. If the rivers are indeed comparable this should not influence the sensitivity of the model, however if the assumption is incorrect it could potentially have serious consequences regarding the accuracy of the output.
The Sacramento River system may not be as similar to the Columbia River system as might be hoped. It would probably be safe to assume that water temperatures are higher in the Sacramento and its tributaries than in the Columbia. The concerns of migrating juveniles between the two rivers also differ. The Columbia River is heavily dammed for generation of hydroelectric power. The Sacramento, however, only has one prominent dam which is passable. Another difference is in the amount of water loss to irrigation. Although the Columbia is used as a water source for agricultural irrigation, the Sacramento and its tributaries are heavily taxed for irrigation, affecting the flow and depth of the river. The raised temperature and altered flows may make the Sacramento different enough from the Columbia as to invalidate the assumption that the two systems are comparable.
The Sacramento system is most certainly different in some aspects from the Columbia. Fish behavior and predator behavior are both highly variable, even within a single system. The Sacramento is already unique in that most of the migrating juveniles are sub-yearling, so it is possible these migrants are also exhibiting different behavioral characteristics. Predators in the Sacramento are most likely exhibiting different behavioral patterns than in the Columbia as well. The primary predator in the Sacramento is a different species than in the Columbia. This species may be capable of eating more or less, which would therefore affect the rate of predation. Metabolism and the rate at which prey are consumed as temperature changes is modeled by CRiSP, however the parameters determining the predation must be correct for the model to accurately predict the rate at increased temperatures. The calibration of the model to the Sacramento River files focused primarily on these possible differences.
The predator coefficients and predator density were increased to observe how sensitive the model was to such changes. As the two parameters were set to higher values, the survival estimates dropped accordingly, however were not enough to bring the predicted passage near the actual passage. It could be that the parameters were not increased to the true values, or perhaps there is another factor causing the high rates of survival other than predation. Predator behavior and metabolism is most likely influenced by temperature. The elevated temperatures in the Sacramento, in comparison to the Columbia, may be affecting predator behavior, therefore making the chosen parameter values invalid. Further study and data retrieval are necessary to determine the significance of predation on survival estimates in the Sacramento system.
Another possible factor which influences survival rates is travel time. The behavior of the fish, the size, and the water flow are all factored into the travel rate for migrating juveniles. The longer the fish is in the river environment, the higher chance it has of becoming prey or experiencing mortality due to physical stress. Slower moving fish are therefore exposed to risk for a longer period of time, thus increasing the mortality rate. Perhaps because these fish are migrating as sub-yearlings and are quite small they are exhibiting different swimming speeds and are influenced by the river flow in a way different from migrating salmon in the Columbia. The swimming and travel behavior was altered for each individual release in attempt to better calibrate the model. It appears that swimming activity and travel time, in conjunction with predator behavior adjustments, allows the model to predict survival rates closer to the actual data.
It is ironic that the Sacramento River and its tributaries are home to some threatened and endangered stocks of Chinook, yet there is relatively little published about the system in comparison to the Columbia. The uniqueness of having nearly all sub- yearling migrants appears to dominate the literature regarding the Sacramento, rather than the factors of the system which influence salmon survival and abundance. To better understand the system, and to generate more reliable data for use in computer modeling, future studies need to include, or focus upon, those factors such as temperature and flow which heavily influence juvenile migration and salmon survival/abundance.
The assumption that similar ecosystems function in the same way is the premise for efficient computer modeling. If this assumption were invalid, computer modeling would not be as useful because a new model would have to be built for every individual ecosystem. Instead it is a more logical approach to assume that ecosystems are, for the most part, similar and only when substantial evidence is found to the contrary should a computer model be abandoned or overhauled for that particular system.
The applications of computer models are endless. In the case of juvenile salmon migration, CRiSP and other models are currently being used to determine the best way to pass the migrants downstream. Hypothesis regarding the best way to increase survival and passage of fish can be "tested" by the model before large amounts of money and time are invested in field research. Models do not answer the question of what is best. However, they can offer suggestions and offer preliminary analysis of potential practices. For maximum efficiency and learning potential, computer models should be a major part of all future planning, construction, and operating procedures involving salmon conservation and recovery efforts.
Dr. Jim Anderson for allowing me the opportunity to work with his staff and to learn about the world of computer modeling.
Josh Hayes for providing invaluable guidance and encouragement throughout this project
Nick Beer for spending excessive amounts of time helping to extract the latlons and construct the final map, and for teaching me how to stay afloat in the UNIX world
Susannah Iltis for assisting in the construction of the web page
Dr. Frieda Taub for exposing me to computer modeling and CRiSP
Allen, M. A., and T.J. Hassler. 1986. Species profiles: life histories and environmental requirements of coastal fishes and invertebrates (Pacific Southwest)ùChinook salmon. U.S. Fish Wildl. Serv. Biol. Rep. 82(11.49). U.S. Army Corps of Engineers, TR EL-82-4. 26 pp.
Greenstreet, S. P. R. 1992. Migration of hatchery reared juvenile Atlantic salmon, Salmo salar L., smolts down a release ladder. 1. Environmental effects on migratory activity. J. Fish. Biol. 40:655-666
Groot, C. and L. Margolis (eds.) 1991. Pacific Salmon Life Histories. UBC Press, Vancouver. pp. 314-393.
Nicholas, J.W. and D.G. Hankin. 1988. Chinook salmon populations in Oregon coastal river basins: Description of life histories and assessment of recent trends in run strengths. Oregon Dept. of Fish. And Wildl. R&D.