Production Potential
The 1982 ODFW Coho Salmon Management Plan identified production goals for wild
coastal coho. Because of a number of factors, including unfavorable marine survival,
these production levels have never been realized. Much new information is now
available about the factors affecting production of coho salmon. For example, extensive
habitat inventory data are now available, a model has been developed to estimate coho
salmon smolt capacity from habitat data, spawner numbers are estimated using
statistically valid methods, and we have a better understanding of the effects of variability
in climate on salmon production. New understanding of the interactions between
freshwater and marine survival of coho salmon are of particular interest to the
development of realistic production goals for wild fish.
Research has demonstrated that the quality of freshwater habitat (particularly over-winter
habitat) has a direct influence on freshwater survival rate. Habitat and population
modeling has demonstrated that to be equally productive, salmon inhabiting a stream with
poor quality habitat will require a higher rate of marine survival than salmon inhabiting a
stream with good quality habitat. As a result of these interactions, marine survival plays a
dominant role in determining the productivity and sustainability of coho salmon
populations.
The modeling predicts that extended periods of low marine survival cause extirpation of
coho salmon from all but the best freshwater habitats. In fact, this is exactly what we
observe today. A prolonged period of poor marine survival has occurred for coho off
Oregon since the late 1970s. Random sampling of coho spawner abundance indicates
that very few stream reaches have large spawner populations, and that most stream
reaches have few or no spawning coho salmon.
Thus, the concept of a single production goal has become obsolete. The concept of
production potential is more appropriate. Production potential is the estimated number of
adult salmon that might be expected from a population under a particular set of natural
environmental circumstances. When estimating production potential, both the quality of
the freshwater habitat and the probable levels of marine survival must be considered.
Production potential and range of coho salmon abundance within a basin would be
expected to expand and contract as marine survival increases and decreases.
The estimates of production potential presented in this chapter were developed based on
actual measurements of habitat in individual stream reaches made during the period 1990-95 and two assumed levels of marine survival: 3% and 5% (see Table 1). Therefore, two
tiers of freshwater habitat would be capable of supporting coho production,
corresponding
Table 1. Estimated Production Potential of Current Habitat for Coho Salmon in Oregon Coastal ESUs.
| Basin | Marine Survival |
Production potential |
Spawners needed |
1990 | 1991 | Estimated 1992 |
Spawners 1993 |
1994 | 1995 | Preliminary 1996 |
|
| Oregon Coastal ESU | |||||||||||
| Nehalem | 5% | 59,100 | 31,700 | 1,600 | 4,000 | 1,300 | 2,300 | 2,400 | 1,600 | 1,100 | |
| 3% | 24,000 | 17,500 | |||||||||
| Tillamook | 5% | 8,300 | 5,700 | 300 | 3,000 | 300 | 900 | 900 | 300 | 700 | |
| 3% | 2,400 | 2,000 | |||||||||
| Nestucca | 5% | 10,500 | 6,400 | 200 | 700 | 700 | 400 | 300 | 1,800 | 500 | |
| 3% | 2,400 | 1,800 | |||||||||
| Siletz | 5% | 13,100 | 7,400 | 400 | 1,000 | 2,400 | 400 | 1,200 | 600 | 800 | |
| 3% | 5,500 | 4,300 | |||||||||
| Yaquina | 5% | 21,700 | 11,800 | 400 | 400 | 600 | 500 | 2,400 | 5,700 | 4,600 | |
| 3% | 9,100 | 7,100 | |||||||||
| Alsea | 5% | 42,600 | 21,500 | 1,200 | 1,600 | 7,000 | 1,100 | 1,300 | 700 | 1,600 | |
| 3% | 20,200 | 15,100 | |||||||||
| Siuslaw | 5% | 69,000 | 39,200 | 2,700 | 3,700 | 3,400 | 4,400 | 3,000 | 6,100 | 8,800 | |
| 3% | 28,500 | 22,800 | |||||||||
| Coastal Lakes | 5% | 20,000 | 4,400 | 7,300 | 2,000 | 10,100 | 5,800 | 11,200 | 13,500 | ||
| 3% | 12,000 | 6,700 | |||||||||
| Umpqua | 5% | 106,200 | 62,200 | 3,700 | 3,600 | 2,200 | 9,300 | 4,500 | 11,000 | 14,400 | |
| 3% | 38,400 | 29,400 | |||||||||
| Coos | 5% | 25,100 | 14,600 | 2,300 | 3,800 | 15,600 | 15,300 | 14,600 | 10,400 | 12,100 | |
| 3% | 8,900 | 7,200 | |||||||||
| Coquille | 5% | 28,600 | 18,900 | 2,700 | 5,600 | 2,100 | 7,400 | 5,000 | 2,100 | 16,200 | |
| 3% | 7,700 | 5,400 | |||||||||
| Direct Ocean | 5% | 26,400 | 16,100 | 1,100 | 1,600 | 2,000 | 2,300 | 2,200 | 900 | 4,000 | |
| Tributaries | 3% | 9,500 | 7,300 |
||||||||
| Total ESU | 5% | 430,600 | 235,100 | 20,900 | 36,300 | 39,700 | 54,400 | 43,700 | 52,400 | 78,300 | |
| 3% | 168,600 | 126,600 |
|||||||||
| Transboundary ESU | |||||||||||
| Rogue | 5% | 28,900 | 14,200 | 2,800 | 800 | 1,900 | 200 | 5,300 | 4,200 | 5,800 | |
| 3% | 6,800 | 5,400 | |||||||||
to the two levels of marine survival. Estimates of production potential were also made
for 10% marine survival, however these levels are not considered attainable and are
therefore not included in Table 1. All estimates of production potential were derived with
the assumption of having fully seeded freshwater habitat, and should be viewed as
potentially achievable levels of production based on current habitat condition. For the
Transboundary ESU that included southern Oregon and Northern California, estimated
production potential was calculated for the Rogue Basin only. Production potential for
coho salmon is thought to be very small in other Oregon streams in this ESU.
Because estimates of potential production are based on modeling of freshwater habitat
capacity, which relies heavily on winter habitat conditions, these estimates may be
optimistic in some cases - especially for areas where high summer water temperatures
may occur such as the Umpqua and Rogue basins. Temperature may be a more severe
constraint than winter habitat on populations in some streams in these basins, and limit
production below the maximum levels estimated (see Table 1). Consequently, current
estimates of potential production should be viewed as giving general guidance.
Undoubtedly, this guidance will be revised in the future as population models are
improved and more habitat data are collected.
To assess the status of a population relative to its potential, it is necessary to consider its
history of relative marine survival. Potential production levels vary as marine survival
changes. Thus, because marine survival for the last 2 decades has been poor, attaining
the production potential of the higher levels of marine survival will occur only after
achieving adequate spawner abundance in the poor habitat that currently has few, if any,
spawners. Achieving adequate spawner abundance in these poorer habitats may require
that several generations experience improved marine survival.
For current habitat conditions, the modeling predicts that wild coho salmon production
could range from about 168,000 in about 800 miles of habitat to about 430,000 in about
2,100 miles of habitat in the Oregon Coastal ESU. Spawner needs are in the range of
126,000 to 235,000. Similarly, production potential for the Rogue Basin ranges from
about 7,000 to about 29,000 with spawner needs of about 5,000 to 14,000.
Recent Population Trends
Since 1990, coho salmon spawner populations in the Oregon Coastal ESU have been
estimated using statistically designed, stratified random surveys. From 1950 to 1990,
populations were monitored using standard survey sites. Whereas the standard surveys
provided an index of abundance from year-to-year, the new methods provide actual
population estimates. In the Rogue Basin, populations estimates are made from ratios of
unmarked fish to marked hatchery fish collected in a seining operation at Huntley Park in
the lower river.
The populations estimated for each major coastal basin since 1990 are listed in Table 1. Abundance of spawners in the Coos and Coquille basins have been relatively strong since 1992, the first year of substantial harvest reduction on the south coast. In 1995, and
especially in 1996, spawning populations in the coastal lakes, and the Umpqua, Siuslaw, and Yaquina basins have also seen substantial increases in abundance. The preliminary estimates for 1996 show a significant increase in total abundance for all but the northern third of the Oregon Coastal ESU. The stronger spawner returns in the southern two-thirds of the ESU do not appear to be due to better habitat quality. For example, based on our habitat modeling, the Coos and Coquille basins have habitat of similar quality to that in the Nehalem and Nestucca basins, respectively, yet had a 15 to 20-fold greater density of spawners in 1996.
Total production of the coastal populations can be estimated by dividing escapement
estimates by 1 minus the exploitation rate. Figures 2 and 3 show these estimates for the
Oregon Coast ESU and for the Rogue Basin, respectively. In 5 of the last 7 years coastal
basins, as a whole, have been producing coho salmon at about one-half of their estimated
potential, given the poor marine survival conditions. The primary reason for this is lack
of adequate spawners. Figure 2 clearly demonstrates the effects of high exploitation rates
during periods of poor marine survival. In hindsight, exploitation rates of 50% to 70%
experienced by some coastal coho stocks as late as the early 1990s were clearly too high
given the poor marine survival conditions experienced by the fish. As harvest has
decreased, spawner abundance has increased. Since 1990, there has been a fairly steady
increase in spawner abundance despite a generally flat trend in total production. Spawner
to spawner ratios have ranged from 1.2 to 2.6 during the past 4 years. Estimated spawner
abundance has increased by about fourfold in two generations (1990-1996).
In the Rogue basin, the population has been much more variable (Figure 3). However,
marine survival of hatchery fish has greatly improved since 1994. Abundance of wild
spawners has also increased dramatically.
Prospects for the Future
The improved survival of coho salmon in the Rogue Basin and in the south and mid-coast
basins in recent years are hopeful signs. We know that climate is cyclic and strongly
influences marine survival. We have been in a poor survival cycle since 1977. The
improved marine survival of coho from the Rogue Basin since 1994, and apparent
improved survival in mid-coast basin in 1996 may be precursors of better survival in the
near future. Climatologists predict a return to a wetter climate similar to that experienced
in the 1960s, a period of good survival conditions for Oregon coho salmon. So what
trends in production are populations likely to experience in the future?
One possible answer has come from a simple model of projected populations that would result from the proposed OCRSI Harvest Strategy. If we consider two assumptions - that marine survival will remain poor (average 3%) or that marine survival will improve (average 5%) - we can develop a range of possible outcomes. Starting with the average spawner abundance of 50,000 fish for 1993-95, the model predicts an expected rebuilding trajectory for spawner abundance in the Oregon Coast ESU at 3-year intervals (Figure 4) while following the proposed harvest strategy. The typical cycle of Oregon coho salmon populations is 3 years from spawner to spawner. Thus the values for 1998 in Figure 4
would represent populations in 1996-98. Based on the model, the spawner population would be expected to be between 56,000 and 94,000 after one generation and between
75,000 and 216,000 after four generations. The estimated returns for 1996 fall near the
upper boundary of this predicted range after one generation.
A second answer to the question of where populations are likely to go in the future comes
from a more sophisticated habitat-based, life cycle model. This model is based on the
reach-level habitat data used to estimate production potential, and incorporates a range of
probable variation in survival at each life stage. The model also includes factors for
straying of spawners, multiple spawning periods, sex ratio, redd failure, and loss of
genetic fitness at low population size. Populations were modeled for 10 generations
using average marine survival rates and for 33 generations using a cyclic pattern of
marine survival (a detailed description of the model and results are presented in ODFW
Attachment 1).
We modeled populations in the Yaquina, Coos, and Tillamook basins. These basins have
the best, an intermediate level, and poorest habitat in the Oregon Coast ESU, respectively.
Results suggest that future population abundance will be heavily influenced by marine
survival and by exploitation rate when marine survival is low. Only the habitats with
high productivity remained viable when marine survival was low. Therefore, distribution
and abundance of fish was a function of long- and short-term variability in marine
survival and long term patterns of habitat quality. Within a reach, populations were
resilient unless numbers dropped to a level where demographic risk factors became more
important than density dependent population dynamics. Persistence of populations in a
basin during periods of poor marine survival depended on the highest quality reaches.
The model predicts that there is a high probability of persistence of coho populations in
all major basins of the Oregon Coast ESU for the next century if habitat condition
remains as it is today (we have not yet modeled the Rogue Basin). However, we have
estimated that quality of habitat in the streams currently considered to be coho salmon
habitat has likely declined by about 40% coastwide over the last century (See ODFW
Attachment 1). We therefore modeled declines in habitat quality ranging from 10% to
60%.
Based on these analyses, the model predicts that there would be a substantial increase in the risk of extinction (population of < 50 spawners occurring at any time) in basins with poor quality habitat, such as the Tillamook if habitat quality over the next century declines by 30-60%.. This would probably also be the case in the Nestucca, and Coquille basins based on our evaluation of habitat quality (see Figure 1). Similar declines in the quality of habitat in the remaining major basins in the Oregon Coast ESU would not result in an appreciable decrease in the probability of persistence of coho salmon populations in those basins. However, decreased habitat quality would result in smaller populations.
Summary
ODFW ATTACHMENT 1
POPULATION DYNAMICS OF OREGON COASTAL COHO SALMON:
APPLICATION OF A HABITAT-BASED LIFE CYCLE MODEL
Thomas E. Nickelson and Peter W. Lawson
Oregon Department of Fish and Wildlife
Abstract: To evaluate the 100 year extinction risk for Oregon coastal natural (OCN) coho salmon, a habitat based life cycle model was developed. Individual stream reaches (ca. 1 km) were characterized by estimated maximum smolt density using habitat survey data covering 16 to 67 percent of basins. Smolt output was a function of spawners, egg to parr survival, and overwinter survival. After natural mortality and harvest in the ocean, spawners returned to their natal reach. At low stock size, spawners could fail to reproduce due to random demographic events of straying, return timing, sex ratio, and redd failure. Accumulation of deleterious alleles was modeled at low abundance. Only the higher productivity reaches remained viable with low marine survival. Therefore, distribution and abundance of fish was a function of long- and short-term variability in marine survival and long term patterns of habitat quality. Within a reach, populations were resilient unless numbers dropped to a level where demographic risk factors became more important than density dependent population dynamics. Persistence of populations in a basin during periods of poor marine survival depended on the highest quality reaches.
Introduction
Population dynamics of Oregon coastal natural (OCN) coho salmon have been
investigated with stock-recruitment (e.g. Ricker 1975) functions, usually applied
to large areas of the coast as a single stock (Beidler, et al. 1980, Overholtz
1994), or to individual streams or stream sections (Overholtz 1994). When
applied to the stock aggregate, this approach has the advantage of describing
the general behavior of the stock, but fails to describe stock dynamics at low
abundance, cannot distinguish between freshwater and marine influences on
survival, and uses only a small portion of available data. Production functions for
single stream sections have little generality. This paper develops an alternative
approach to understanding the dynamics of OCN coho salmon using fine scale
freshwater habitat data as the basis for modeling freshwater production at the
scale of individual river basins. Modeling production at a fine spatial scale allows
us to incorporate metapopulation dynamics such as straying and depensatory
demographic effects such as variable sex ratios and run timing which become
important at low spawning escapements. Density-dependent survival occurs at
the reach level, while more general effects, such as marine survival, affect whole
populations.
The freshwater production model was used as the basis for simulations of OCN
coho population patterns over time. We incorporated stochastic variability at
many stages to represent the variability inherent in natural processes, and
experimental measurement error in parameter estimation. This allowed us to
estimate probability distributions of likely outcomes given specified starting
conditions, which then enabled us to identify likely population sizes and
extinction probabilities (Goodman, in press). Extinction probabilities are
commonly expressed as likelihood of extinction in 100 years, with an acceptable
risk level of 5 percent (Thompson 1991). Because this 5 percent represents the
left tail of a probability distribution it is sensitive to the structure of the model and
the parameters used to describe variability. To estimate extinction probabilities
rigorously would require incorporating "everything that is known and everything
that is not know about the dynamics of the population" (Goodman, in press). To
the extent we have not achieved this ambitious goal, our results in this area must
be viewed as exploratory.
The model was used to explore OCN coho salmon population dynamics at the basin scale. We used three basins of varying habitat quality to represent the range of conditions on the coast; Tillamook (poor), Coos (moderate) and Yaquina (good). We tested the sensitivity of populations in these three basins to varying levels of marine survival and exploitation rates over 10 generations. We simulated the effects of a range of starting population sizes, including the 1995 actual spawner escapements, on median population size and probability of extinction in 33 generations. We also modeled the effects of changes in habitat quality on persistence and population size after 33 generations.
Methods
A simulation model was developed that has both production and forward
simulation aspects. The production aspect addresses differences in habitat
quality and will subsequently be referred to as the habitat quality component.
The basis for this component is that the quality of freshwater habitat, which
varies both within and among watersheds, determines the number of coho
salmon smolts that a stream produces as well as the efficiency with which those
smolts are produced (i.e. survival rate). Production is estimated for individual
stream reaches within a basin, based on habitat quality data from the basin.
Estimates of smolt capacities and average survival rates at densities associated
with maximum smolt production were derived for ten large Oregon coastal
basins. These estimates were made for individual stream reaches (lengths of
stream between changes in gradient or valley and channel form) within each
basin using data in the Oregon Department of Fish and Wildlife (ODFW) Aquatic
Inventory Database (Moore et al. 1995) and data from the Siuslaw National
Forest (Bob Metzger, Siuslaw National Forest, Corvallis, OR, personal
communication June 1996), and represent sampling rates ranging from 16% to
64% of the available coho salmon habitat in each basin.
The temporal component of the model mimics the life cycle of coho salmon and simulates population fluctuation and random dispersal over generations by incorporating density driven compensation and depensation, short-term stochastic variation in survival, long-term climatic cycles, reduced genetic fitness because of small population size, and straying of spawners from their natal spawning areas.
The Habitat Quality Component of the Model
Estimates of Smolt Production Capacity
Estimates of smolt production capacity were derived for individual stream
reaches in two ways, depending on the level of inventory data available.
For stream reaches where winter habitat data were available, the latest version
of the habitat limiting factors model (HLFM Version 5.0) originally described by
Nickelson et al. (1992a), was used to estimate smolt potential. This model
estimates potential population abundance for the spawning, spring rearing,
summer rearing, and winter rearing life stages of coho salmon by multiplying
habitat-specific densities based on data from Nickelson et al. (1992b) by areas of
individual habitat types derived from stream inventory data collected during
summer and winter. It then estimates potential smolts by applying survival rates
from each of these life stages to the smolt stage (Table 1). The estimates of
potential coho salmon smolt capacity generated by this model have been shown
to be closely related to actual smolt production when summer habitat is fully
seeded with juveniles (approximately 1.5-2.0 parr/m2 of pool)(Figure 1).
Research has found that suitable winter rearing habitat typically is in least supply
in Oregon coastal streams compared with the other four types of habitat and
thus limits smolt production (Table 1; Nickelson et al. 1992a, 1992b). Thus we
can use the HLFM and data from inventories of winter habitat to estimate the
smolt capacity of a reach of stream.
Because stream habitat typically is surveyed only during summer, most stream
reaches lack data on winter habitat. Therefore, a multiple regression model was
used to relate summer habitat to winter habitat and estimate smolt potential for
these stream reaches. This model was developed from data for 74 stream
reaches where both summer and winter habitat surveys have been conducted,
and predicts smolt potential (as estimated by HLFM) from stream reach
characteristics determined during summer stream habitat surveys. To account
for differences in stream size, smolt potential was expressed as a density based
on reach area derived from active channel width. Some variables were
transformed to linearize the function or to normalize and equalize the variance.
The regression model shown below explained 80% of the variation in the
dependent variable (Table 2).
[1] C = (0.4000 - 0.0682logew - 0.0332g + 0.1030b + 0.2020p)2 ,
where C is the predicted potential smolt density for the reach expressed as
smolts/m2, w is the active channel width of the reach, g is the gradient of the
reach, b is the number of beaver dams per km in the reach, and p is the arc sine
square root transformation of the percent of pool in the reach. To test the
predictive power of this regression, the regression was estimated separately for
five randomly picked subsets consisting of 75% of the data and then used to
predict the remaining data in each case. The result was that smolt capacities
predicted by the multiple regression model were significantly correlated with
smolt capacities estimated using the HLFM (p<0.001; r = 0.874). To account for
uncertainty at the upper end of this relationship, where few values occurred,
maximum potential smolt density was capped at 1.15/m2 (the density expected if
the entire reach were made up of the best quality habitat).
Maximum smolt capacity (M) for each reach, expressed as a total number of smolts, was calculated by multiplying C by the total area of the reach (length multiplied by active channel width). The number of adults expected to be produced by these smolts was estimated by multiplying by marine survival, which for the purpose of this model was defined as the period of downstream smolt migration from the natal stream, ocean residence, and upstream adult migration back to the natal stream.
Over-Winter Survival
Observations of over-winter survival in a several streams was positively
correlated with potential smolt density (C) as estimated by HLFM. This
relationship is key to the influence of habitat quality on coho salmon population
dynamics. It is based on observed over-winter survival estimated for 5 streams
(four of which have been studied for 7 years) and the potential smolt capacity for
the streams estimated from winter inventory data using the HLFM (Figure 2).
This relationship yields the following equation:
[2] Sow = 0.1461logeC+ 0.5244,
where Sow is over-winter survival. The relationship explains 70% of the observed
variation in over-winter survival (Table 2). Thus, C is not only an estimate of
potential smolt density, but it is also an index of habitat quality that is related to
juvenile survival. Because this equation produces survival rates < 0 when C <
0.03 for a reach, all such reaches were assigned a survival rate of 2.5%, the
lowest value observed.
Egg deposition needed to produce maximum smolts
The egg deposition needed to produce maximum smolts (Dm) is synonymous
with the concept of full seeding of the habitat, and was calculated from:
[3] Dm = M / Ssmolt .,
where Ssmolt is egg-to-smolt survival rate which was calculated for each reach by multiplying over-winter survival rate by egg-to-summer parr survival rate. To estimate Dm we assumed a constant egg-to-summer parr survival of 7% for all reaches. This value was the approximate survival rate at the point of maximum parr production (full seeding) on a Ricker stock-recruitment curve based on data for three Oregon coastal streams from Moring and Lantz (1975).
Assumptions
Implicit to the habitat quality component of the model are the assumptions that winter habitat is the primary bottleneck to smolt production in each stream reach, and that survival from egg deposition to summer parr is 7% for all reaches when at full seeding. These assumptions are necessary because we have inadequate information upon which to base a more detailed analysis that would account for all the factors that influence survival. For example, some stream reaches may experience high water temperatures that exclude coho salmon during summer but then provide rearing habitat when waters cool in the winter. Depending on their location relative to the possibility of immigration of juveniles from other areas for over-wintering, these reaches may be limited by summer habitat. If we had adequate water temperature data, these reaches could be identified and adjustments could be made to the analysis. Similarly, sedimentation, and excess scouring can reduce egg survival. If information about these factors and their impact on survival were available for each reach, egg-to-parr survival could be appropriately adjusted. In lieu of such data we are forced to make the above assumptions.
The Forward Simulation Component of the Model
The elements that comprise the forward simulation component of the model
follow the life history stages of coho salmon (Figure 3). Coho salmon in Oregon
coastal streams typically spawn from early November through mid-January.
Juveniles emerge from the gravel in spring and typically spend a summer and
winter in freshwater before migrating to the ocean in their second spring. A very
small percent of juveniles (<5%) spend an additional winter in freshwater,
migrating to the ocean in their third spring (Moring and Lantz 1975). Precocious
males, called jacks, return to freshwater at the end of one summer in the ocean
as age 2 spawners. They comprise about 20% of each run (Moring and Lantz
1975), although this is variable depending on interannual variation in marine
survival, which is usually determined for a cohort during their first few weeks in
the ocean (Pearcy 1992). Adult coho return to freshwater after their second
summer in the ocean as age 3 spawners. Because of the predominance of age
3 adults in Oregon coho salmon populations, they are considered to have 3
brood cycles. For example, adults spawning in 1990, 1991, and 1992 will
primarily contribute offspring to adults spawning in 1993, 1994, and 1995,
respectively. Details of the modeling at each life stage are described below:
Spawners
Spawners were the starting point for the simulations and the ending point for
each generation. For the purpose of the model spawners included only age-3
adults. For simplification, jacks were not included in the calculations. Similarly,
because age-4 adults are very rare they were also excluded from the model.
The absence of these two age classes from the modeled populations could
possibly represent a slight underestimation of the productive potential of the
modeled populations.
The model incorporated a 5% within-basin straying rate to the population.
Labelle (1992) found that straying of wild adult coho salmon among Vancouver
Island tributaries to the Strait of Georgia ranged from 0 to 7.8%, averaged 4.2%
one year and 0% a second year, and averaged 2.1% overall. The value we used
for within basin straying was roughly double Labelle's among basin rate. The
straying rate was applied in the form of two components: 1) fish leaving a reach
at a random rate with a binomial probability distribution having p = 0.05, and; 2)
fish that have left a reach selecting a new reach at random with equal probability
for all reaches. The effect was to redistribute 5% of the spawners each
generation. Many strays were unproductive because they arrived in a reach with
poor habitat or arrived alone - two fish present at the same time, including one
male and one female, were required for spawning in this model.
Because wild coho in a given Oregon basin might spawn over a period of 2-3 months (Cooney and Jacobs 1995), fish spawning early cannot interact with fish spawning late. This is usually not a problem when populations are large; spawners should have little problem finding mates. However, when spawner populations are very small and some fish are present in a stream early and others late, finding a mate could become problematic. Spawners not finding mates is a depensatory effect of small spawner number. To simulate the effects of this depensation, time of spawning was split into two periods: early and late. If the number of spawners was >200, the spawners were divided evenly between the time periods. If the total number of spawners was <200, the number of spawners in the first period was generated from a binomial distribution having p = 0.5 and n = the total number of spawners and the number of spawners in the second period was derived by subtraction. This increased the probability that spawners would not be successful because they spawned at different times, thus increasing the likelihood of not finding a mate. Not including jacks in the model, makes this portion of the model conservative (i.e. increases the likelihood that the model will project a small population)
Eggs
The number of female spawners was calculated in two ways depending on the
total number of spawners. If the total number of spawners was >20 in a
particular time period, the sex ratio is assumed to be 1:1. If the number of
spawners was <20, the number of females was generated from a binomial
distribution having p = 0.5 and n = the number of spawners in the time period.
This adds an additional depensatory effect of small spawner number.
Egg deposition (D) was calculated as 2500 eggs per female (Moring and Lantz
1975, ODFW unpublished data for 1990-95) unless all spawners in a time period
were females, in which case, egg deposition for that time period was 0 (again the
model is conservative, as the inclusion of jacks would decrease the probability of
this happening). Egg deposition from the two time periods was summed.
Koski (1966) estimated no fry emerge from about 15% of coho salmon redds, the
likely result of gravel scouring. Thus we reduced egg deposition to account for
this mortality. This was done by estimating the number of successful redds in
each reach by adjusting the number of female spawners. When female
spawners was >200, the number of successful females was 85% of the total.
When the number of females was <200, the number of females was generated
from a binomial distribution having p = 0.85, and n = number of females.
Summer parr
The number of summer parr was calculated by multiplying egg deposition by
egg-to-parr survival rate (Sparr), which was estimated from a density dependent
function based on the relative level of seeding (P), where:
[4] P = D / Dm .
Relative seeding level was used as the independent variable in this relationship
because each reach had a different productive capacity. Thus, a given number
of eggs would represent a different level of seeding in each reach and therefore
a different point on the density dependent curve. Using the seeding level
provides standardization across reaches. The relationship between seeding
level and egg-to-parr survival rate(Table 2), based on data from Moring and
Lantz (1975), is shown in Figure 4 and yields the following equation:
[5] Sparr = 0.064P-0.743 eE,
where E is an error term derived by multiplying the standard deviation of the residuals from the relationship by a value chosen randomly from a normal distribution with mean 0 and standard deviation of 1. Because this fitted curve results in survival rates >100% when seeding level is < 2.5%, egg-to-parr survival rate was capped at 40%, just above the highest observed in the data set. The log normal form of the error term also has a tendency to produce unrealistically high survival rates all along the curve. To curb this tendency, the maximum random value chosen from the normal distribution was 1.167. This resulted in limiting the upper limit of variability to be very slightly above that actually observed in the data set (Figure 5A) providing a measure of conservatism to the model. Minimum survival rates were not affected. A new random error value was calculated each generation.
Also at this point in the life history, we added a factor to account for the genetic
effects of small spawner population size. When effective population size (Ne) is
small, generally on the order of 100 individuals or less, genetic fitness is reduced
because deleterious mutations accumulate due to random genetic drift (Lynch, in
press) whereas when Ne is relatively large (1 000 individuals) reduction in fitness
is generally not a problem. This reduction in fitness is in the range of 1.5% per
generation at very low Ne and is cumulative (Lynch, in press). Lynch further has
estimated that for salmonids, a conservative estimate of Ne is approximately 20%
of the actual number of spawners. Because there is genetic interaction among
successive broods of coho salmon, through mixing of age-2 jacks, age-3 adults,
and age-4 adults [estimated to be about 3% for Oregon streams resulting from
age 2 smolts (Moring and Lantz 1975)], Ne can be calculated as 20% of 3Ni,
where Ni is the number of spawners in a basin in generation i. We can model
reduction in fitness (f) as a reduction in survival, and describe the portion
attributable to any given generation by assuming: 1) f = 0 when Ne >1 000; 2) f =
0.001 when Ne =100; 3) f = 0.015 when Ne = 5, and; 4) the change in f is linear
between Ne = 5 and Ne = 100 and between Ne = 100 and Ne = 1 000. Thus for
any given generation i the reduction in fitness attributable to the spawner
population size that year is:
[6] fi = 0 when 3Ni > 5 000,
[7] fi = 1.11 * 10-3 - 2.22 * 10-7(3Ni) when 500 < 3Ni < 5 000,
[8] fi = 1.57 * 10-2 - 2.95 * 10-5(3Ni) when 25 < 3Ni < 500, and
[9] fi = 0.015 when 3Ni < 25.
The cumulative effect through time of deleterious mutations (g) can then be
expressed as:
[10] gi = (1 - f1 ) (1 - f2) (1 - f3 ) (1 - fi)
and in the model was multiplied by the egg-to-parr survival rate to effect a reduction in survival. As long as Ne remained at least 1 000, the value of g was 1.0. Maximum reduction in reproductive success occurred if all generations were below 3N < 25 (Equation 9). In this case gi = (1-0.015)n where n = the number of generations. For n = 10 the minimum gi = 0.859. For n = 33 the minimum gi = 0.607. These extreme values were rarely realized in the simulations.
Smolts
The number of smolts was calculated by multiplying summer parr by over-winter survival rate. The value for over-winter survival rate for each reach was derived by adding an error term to the value of Sow. (Equation 2) The error term for a given generation was calculated as the standard deviation of the observed residuals from Equation 2 multiplied by a value chosen randomly from a normal distribution with mean 0 and standard deviation of 1. This error term also has a tendency to produce unrealistically high survival rates. To curb this tendency, the maximum random value chosen from the normal distribution was 1.117. This confined the variability in maximum survival rates to the range of those observed in the data (Figure 5B) and added further conservatism to the model. Low survival values were not curbed, except that, any values < 0 were set at 2.5%, the lowest value observed.
Adults
The number of adults in the next generation was calculated by multiplying the
number of smolts by a marine survival rate. Unfortunately there are no direct
measures of marine survival available for wild coho salmon from Oregon.
However, Nickelson (1986) using an indirect approach, estimated that survival
rates for hatchery and wild coho salmon in Oregon were similar during periods of
favorable ocean conditions, but that wild smolts survived at roughly twice the rate
of hatchery smolts during periods of unfavorable ocean conditions. Data from
Washington (Seiler 1989) and British Columbia (Cross et al. 1991) also suggest
that marine survival of wild smolts is about double that of hatchery smolts during
a period of unfavorable ocean conditions. Marine survival of coho salmon smolts
from Oregon coastal hatcheries north of Cape Blanco have averaged 1.5% for
brood years 1982-91 (Lewis 1995). Assuming the above, this would represent
3% survival of wild smolts during this period.
Because separate simulations were run over two different time intervals (10
generations and 33 generations), marine survival was treated in two ways. For
simulations of 10 generation duration, the marine survival rate for a given
generation was the average rate set at initialization of the simulation (1.5%, 3%,
or 5%) plus an error term. The error term was derived by multiplying the
standard deviation of the average survival rate (approximated as the square root
of the average survival rate) by a randomly chosen standard normal deviate from
the mean of marine survival for hatchery coho salmon for brood years 1958-1992. The resulting distribution of errors was approximately log-normal.
Minimum marine survival allowed in the model was 0.4% Figure 5C depicts an
example of the distribution of marine survivals used by the model. This
approach was used because 10 generations was about as long a period as we
might expect between climatic regime shifts within which we might expect some
average survival with reasonable variation. For example regime shifts occurred
in the mid-1920s, mid-1940, and 1977, periods of 20 to 30 years (Francis and
Mantua, In Press).
For the long-term simulations, it was necessary to take into account the cyclic
nature of climate (i.e. the regime shifts) and the marine environment (Beamish
and Bouillon 1993; Hsieh et al. 1995). To accomplish this, we used the Aleutian
Low Pressure Index (ALPI) (Beamish and Bouillon 1993) as a template for the
pattern of long-term climatic variability. This annual index represents the
intensity of the low pressure system over the northern North Pacific Ocean
during winter and spring (December - May) for the years 1900-1989. Beamish
and Bouillon (1993) noted a strong positive correlation between ALPI and
salmon production in the Gulf of Alaska and, typically, production of coho salmon
in Alaska and Oregon have been inversely correlated (Nickelson and Lichatowich
1984). Thus, when ALPI is low, survival of coho salmon in Oregon has been
generally high, and when ALPI is high, survival has been generally low. For
modeling purposes, the long term cyclic pattern of ALPI was approximated by a
step function developed by dividing the smoothed trend (9y running average) by
a constant, and converting to an integer (Figure 6). Average marine survival
rates of 10%, 7.75%, 5.5%, 3.25%, and 1% were assigned to the resulting steps
of 0 through 4, respectively. Since the database runs for only 89 years, it was
doubled by appending the first year to the last year.
Because we are currently experiencing low survival conditions, simulations began with a randomly chosen year j having a step value of 3 or 4. For each subsequent generation of the simulation, the model proceeded through the ALPI cycle using the value of year j+3, j+6...j+99 (because of the 3y cycle of coho salmon) using the average survival rate each year that was assigned to the current step. A stochastic error term was then added in as was done for the 10 generation simulations.
Spawners
The number of spawners in a reach in the subsequent generation was calculated by multiplying the number of smolts times marine survival times 1 minus the fishery exploitation rate. Fishery exploitation rates were either 1) held constant for 10 generations or 2) varied with marine survival for 33 generations (See Simulations). The number of spawners in a basin was calculated by summing across reaches.
Depicting Habitat Quality
One product of the habitat quality component of the model is the depiction of
relative habitat quality and the ability to compare habitat quality among reaches,
streams, and basins. Two parameters are useful descriptors of habitat quality: 1)
smolts produced per adult spawner when maximum smolt production is
achieved, and; 2) the proportion of the habitat within a basin where the
population would, on average, replace itself if marine survival were some given
rate.
To calculate smolts per adult we first must calculate the number of adults
needed to produce the maximum number of smolts (Am). Two assumptions are
necessary: 1) fecundity is assumed to be 2 500 eggs per female (Moring and
Lantz 1975), and; 2) sex ratio is assumed to be 1:1. The value is then derived
from:
[7] Am = (Dm / 2 500) * 2. and smolts per adult equals M / Am..
The proportion of the habitat within a basin where the population would replace
itself if marine survival were some particular value, is derived by summing the
length of reaches that meet the following criteria:
[8] M * Smar > Am
where Smar is marine survival rate and M is maximum smolt capacity (See Estimates of Smolt Production Capacity), and then dividing by the total length of the basin sampled. We defined good quality habitat as those reaches that could sustain spawning populations at 3% marine survival. Lower quality reaches required higher marine survival rates to sustain spawning populations.
Forward Simulations
Monte Carlo trials of 1 000 iterations were conducted for individual river basins,
recording the coho salmon population size each generation for 10 generations or
33 generations for each iteration. The median population, probability of
population decline, and probability of extinction for a single population cycle were
calculated from the results from each trial. Because of uncertainty at low
population sizes, extinction was defined to occur in a given iteration if a
population size < 50 occurred at any time during the 10 or 33 generations
modeled, regardless of final population size. In addition, from the 33 generation
runs the minimum population and the minimum number of stream reaches
populated were recorded for each iteration. Three coastal basins were chosen
for these trials The basins represented high (Yaquina Basin), medium (Coos
Bay Basin), and low (Tillamook Bay Basin) levels of habitat quality based on the
results of the habitat component of the model.
The Tillamook Bay basin (Tillamook basin), located at 45o 30' N latitude, is
comprised of 5 major rivers and 249 miles of coho salmon habitat primarily in
second to fifth streams. The basin covers about 10 600 km2, much of which
burned in the late 1930s and 1940s. The Yaquina basin, located at 44o 36' N
latitude, is a small basin (about 4 600 km2) and has 109 miles of coho salmon
habitat. The Coos Bay basin (Coos basin), located at 43o 24' N latitude, covers
about 11 300 km2 and contains 208 miles of coho salmon habitat. All three
basins have large estuaries and the watersheds have been logged extensively
since the turn of the century.
Each basin was defined from the set of reaches surveyed for habitat quality and
represented 36%, 57%, and 25% of the coho salmon habitat in the Tillamook,
Yaquina, and Coos basins, respectively. For each reach, three reach level
parameters were provided from the habitat component of the model: 1)
maximum smolt capacity (M); 2) average over-winter survival rate at maximum
smolt capacity (Sow), and; 3) egg deposition needed to produce maximum smolts
(Dm). In addition, a starting spawner population number was specified for each
reach.
The distribution of the starting population across reaches was dependent upon
the quality of habitat in each reach because the capacity of a given reach to
support coho salmon varied with habitat quality. This distribution was
determined by using a spreadsheet form of the model with the stochasticity
removed. An iterative solver function was used to solve for the marine survival
that would result in the desired final population after 30 generations (a period
long enough for an equilibrium population size to be established in each reach).
This method produced distributions of spawners based on 1995 population
levels that were not significantly different (Wilcoxon Signed Rank Test; p > 0.5)
from the distribution of actual counts based on spawning surveys (ODFW
unpublished data) conducted in the three basins (Figure 7).
Reaches surveyed for habitat quality represented 36%, 57%, and 25% of available coho salmon habitat in the Tillamook, Yaquina, and Coos basins, respectively. To simulate effects of low spawner densities and straying we needed a representation of all reaches in each basin. We assumed that the distribution of habitat qualities in the sampled reaches represented all reaches in that basin. The total number of reaches present in each basin was calculated from the sampling fraction, and reaches were chosen from the sample randomly, with replacement, up to the total number of reaches. The reach population was bootstrapped for each iteration of the model (Efron and Tibshirani 1986; Efron 1987). As a result, uncertainty arising from sampling variability in the habitat data was incorporated in the range of modeled results.
10 Generation Trials
The 10 generation trials were used to examine the effects of marine survival and
harvest rate on the probability of persistence of coastal Oregon coho salmon in
each basin. Three levels of average marine survival (1.5%, 3%, and 5%) were
used as input parameters. Although a set survival rate was used as an input
parameter, the stochasticity built into the model caused marine survival to vary
around the average input value each generation as previously described.
To examine the effects of harvest on the probability of persistence of coastal Oregon coho salmon in each basin, exploitation rate was varied as an input parameter. Exploitation rates used were 0%, 10%, 20%, 30%, and 40%. This range was used to describe the relationship between harvest rate and measures of persistence at each of the marine survival levels. Initial populations for these trials were set at 1 000 spawners.
33 Generation Trials
The 33 generation trials were used to examine the long-term risk of extinction of
Oregon coastal coho salmon populations. As discussed above, marine survival
followed a cyclic pattern in these trials. Harvest rates were coupled with marine
survival to approximate the harvest strategy proposed for coho salmon in the
Oregon Coastal Salmonid Restoration Initiative (State of Oregon 1996). For
marine survivals of 1%, 3.25%, 5.5%, 7.75%, and 10% exploitation rates were
10%, 15%, 20%, 30%, and 35%, respectively.
To examine the effect of initial population size of coho salmon in a basin on the
probability of extinction, the initial population size was varied. Starting
populations of approximately 50, 100, 200, 300, 400, 600, 1 000, and 1 500 were
modeled for each of the three basins. In addition, the estimated population in
each basin in 1995 (ODFW unpublished data) was used as a starting point.
These populations were 275 for the Tillamook Basin, 5 671 for the Yaquina
Basin, and 10 400 for the Coos basin.
The trials described so far are based on current habitat remaining constant for the next century. It is unrealistic to expect this to be the case. However, it is also uncertain what the trajectory of habitat change over the next century will be. Habitat change was modeled as an exponential function that resulted in the specified change (DH) in habitat quality (H) over the time period of the run. Trials were run for changes in H of +10%, -10%, -20%, -30%, -40% and -60% over the next century using the 1995 estimated populations as the initial population for each basin. Each generation, the habitat quality (smolts/m2) in each reach was increased or decreased by multiplying potential smolt density (C) by e(a) where a = ln(DH/(# of generations-1)) .
RESULTS AND DISCUSSION
Quality of Habitat in Oregon Coastal Streams
The analysis indicates that the majority of coho salmon habitat in most coastal
basins is poor quality. Coast wide, about 20% of the coho salmon habitat is of
sufficient quality that spawners would at least replace themselves if marine
survival was 3% and exploitation rate was 0. However, in the Oregon coastal
basins between the Columbia River and Cape Blanco this equates to about 800
miles of habitat where coho salmon should sustain themselves when marine
survival is poor, as it has been for the past decade. The proportion of this
quality habitat varies by basin (Figure 8), ranging from 3.5% in the Rogue River
basin (which is south of Cape Blanco) to 42.5% in the Yaquina River basin.
These estimates of relative habitat quality appear to be realistic. We found that,
with the exception of the Coos and Coquille River basins, there was a very good
correlation (R = 0.92) between estimated habitat quality for a basin, and the
1990-95 mean coho salmon spawners per mile for the basin (ODFW
unpublished data; extremes removed)(Figure 9). The Coos and Coquille basins
are the two most southerly basins where spawner survey data are available.
These basins have experienced much higher spawner numbers in recent years
than the northern basins, most likely the result of lower exploitation rates and
better marine survival conditions (ODFW 1995).
Of particular interest is the question of how the quality of habitat has changed over the past century. It has been estimated that under natural disturbance regimes in Oregon coastal basins (i.e. before anthropogenic influence) about 60% of watersheds were productive for anadromous salmonids at any point in time (Benda 1994; Reeves et al. 1995). Reeves et al. (1995) describe a cyclic pattern of change that streams undergo over periods of about 300 years. In this analysis, productive was defined as habitat of a quality similar to Franklin Creek, a stream in the Umpqua Basin that they studied (Gordon Reeves, USDA Forest Service, Corvallis, OR, personal communication). The habitat data that we have for Franklin Creek predicts that coho salmon should at least replace themselves when marine survival is 4%. We have estimated that about 38% of the current coho salmon habitat in Oregon coastal basins north of Cape Blanco meets this definition of productive. If we assume that at the beginning of this century 60% of the habitat was productive, the quality of the current habitat represents a 37% decline over a period of approximately 100 years. It should be noted that this only applies to habitat that today is considered to be coho salmon habitat and does not include the total loss of habitat along the lower mainstems of many coastal rivers such as the coniferous marshes of the Coquille River (Benner 1992). Beechie et al. (1995) estimated that the productive potential of winter habitat in tributaries to the Skagit River, Washington had declined by 23% from historical levels, whereas the productive potential of habitats associated with the mainstem (including sloughs and side channels) had declined by 40%.
Forward Simulation Model Behavior
Quantitative results from this model depend on our estimates of a variety of
parameters and processes. These include habitat carrying capacity, survival
rates at various life-history stages, the shapes of density dependent survival
functions, straying rates, and the structure of variability in egg-to-parr, over-winter, and marine survival. Inaccuracies in our estimates of these factors affect
the numerical predictions of the model. However, all of these parameters are
based on data from studies of coho salmon. During the course of model
development, outputs were compared with known values and our understanding
of the behavior of these systems, so that we are confident that the numerical
outputs are in the correct range.
For example, smolt production values generated by the HLFM generally fall
within the range actually observed in field studies (Skeesick 1970; Moring and
Lantz 1975; Kadowaki et al. 1995) and distributions of spawners produced by the
model were similar to those actually observed (Figure 7). More importantly, the
relative distribution of change in population size from one generation to the next
in the simulation results was not significantly different (Wilcoxon sign rank test;
p>0.4) from that actually observed in wild coho salmon populations in Oregon
coastal basins over the period of 1990-95 (Figure 10). If anything, the model
tended to produce a greater percentage of declining populations than actually
observed, another indication of the conservative nature of the model.
The ten generation simulations are useful to test the sensitivity of the modeled
populations to a range of input conditions. From these simulations the effects of
marine survival rate, exploitation rate, and differences in habitat quality on
starting populations of 1 000 spawners in each basin can be observed.
The pattern of effects of marine survival and exploitation rate were similar for all three basins, differing only in magnitude (Figures 11-13). Changes in median population per mile after 10 generations, probability of population decline, and probability of extinction were all much greater when marine survival changed than when exploitation rate changed. Effects of decreased marine survival or increased exploitation rate were greatest in the Tillamook basin, where habitat quality is poorest, least in the Yaquina basin, where habitat quality is best, and intermediate in the Coos basin, where habitat quality is intermediate (See Figure 8). Only the habitats with high productivity remained viable when marine survival was low. Therefore, distribution and abundance of fish within a basin was a function of marine survival and the pattern of habitat quality. Within a reach, populations were resilient unless numbers dropped to a level where demographic risk factors became more important than density dependent population dynamics. Persistence of populations in a basin under conditions of poor marine survival depended on the highest quality reaches.
Sustainability of Oregon Coastal Coho Salmon
The 33 generation simulations are useful to examine the sustainability of Oregon
coastal coho salmon. We examined beginning populations that ranged from 50
to 1 500 spawners in each of the three basins. Figure 14 presents the results of
these simulations. In each basin, starting populations of 150 or more resulted in
similar ending populations after 33 generations. Also in each basin, the risk of
extinction (< 50 spawners at any time) increased for starting populations less
than 300-400. At starting populations of 50 and 100, probability of extinction was
inversely related to habitat quality. Probability of extinction was greater in the
Yaquina Basin than in either the Coos Basin or the Tillamook Basin because the
small starting populations were spread thinly across greater numbers of reaches
of good quality habitat. Thus, depensatory effects of small population size
resulted in a greater occurrence of extinction in individual reaches in the Yaquina
Basin. Median population after 33 generations in the Yaquina basin was 0 when
the starting population was 50.
The above results assume that current habitat quality would be maintained for the next 100 years. We also examined the effects of changes in habitat quality ranging from a 10% increase to a 60% decrease over the next century on the median ending population and probability of extinction based on a starting population equivalent to the 1995 level in each basin (Figure 15). Based on these analyses, the model predicts that there would be a substantial increase in the risk of extinction in basins with poor quality habitat, such as the Tillamook if habitat quality over the next century declines by 30-60%. Based on our evaluation of habitat quality (See Figure 8), this would probably apply to the Nestucca, Coquille, and Rogue basins as well. Similar declines in the quality of habitat in the remaining major coastal basins would have a much lesser effect on the sustainability of coho salmon populations in those basins. However, decreased habitat quality would result in substantial decreases in population size.
Implications
Based on results of the model, the population in most major coastal Oregon basins 100 years in the future will be independent of the current population size. Exceptions may be basins such as the Tillamook, where populations have dropped below a few hundred fish in some years. Trends in marine survival and habitat quality are much more influential. Future population abundance will be heavily influenced by marine survival and by exploitation rate when marine survival is low. Results from the model indicate that populations of Oregon coastal coho salmon have not lost their resiliency. This is consistent with the observed patterns of change in abundance (Figure 10), with some populations increasing by factors of 4 to 9 in a single generation. On the other hand, populations in basins with poor habitat may lose resiliency in the future if habitat quality continues to decline at the same rate as it has for the last century.
Where Do We Go From Here?
The model described in this manuscript is a work in progress. We continue to respond to reviews of the model and make appropriate refinements. Our next step is to include the following elements in the model.
Literature Cited
Beamish, R.J., and D.R. Bouillon. 1993. Pacific salmon production trends in
relation to climate. Can. J. Fish. Aquat. Sci. 50:1002-1016.
Beechie, T., E. Beamer, and L. Wasserman. 1994. Estimating coho salmon
rearing habitat and smolt production losses in a large river basin, and
implications for habitat restoration. N. Am. Jl. Fish. Manage. 14:797-811.
Beidler, W.M., T.E. Nickelson, and A.M. McGie. 1980. Escapement goals for
coho salmon in coastal Oregon streams. . Oreg. Dep. Fish and Wildl.,
Fish Div. Info. Rep. 80-10, Portland.
Benda, L.E. 1994. Stochastic geomorphology in a humid mountain landscape.
Doctoral dissertation, University of Washington, Seattle.
Benner, P.A. 1992. Historical reconstruction of the Coquille River and
surrounding landscape, Sections 3.2, 3.3 In The action plan for Oregon
coastal watersheds, estuaries, and ocean waters. Near Coastal Waters
National Pilot Project, Environmental Protection Agency, 1988-1991.
Portland, OR: Conducted by the Oregon Department of Environmental
Quality.
Cooney C.X., and S.E. Jacobs. 1995. Oregon coastal salmon spawning
surveys, 1993. Oreg. Dep. Fish and Wildl., Fish Div. Info. Rep. 95-3,
Portland.
Cross, C.L., L. Lapi, and E.A. Perry. 1991. Production of chinook and coho
salmon from British Columbia hatcheries, 1971-89. Can. Tech. Rep. Fish.
Aquat. Sci. 1816.
Efron, B. 1987. Better bootstrap confidence intervals. J. Amer. Stat. Assoc.
82:171-185.
Efron, B. and R. Tibshirani. 1986. bootstrap methods for standard errors,
confidence intervals, and other measures of statistical accuracy. Stat. Sci.
1 (1):54-77.
Francis, R.C. and N.J. Mantua. In Press. Climatic influences on salmon
populations in the Northeast Pacific. Workshop on Assessing Extinction
Risk for West Coast Salmonids, Seattle, WA , November 13-15, 1996.
Goodman, D. In Press. Population viability analysis: the risk assessment
paradigm. Workshop on Assessing Extinction Risk for West Coast
Salmonids, Seattle, WA , November 13-15, 1996.
Hsieh, W.W., D.M. Ware, and R.E. Thomson. 1995. Wind-induced upwelling
along the west coast of North America, 1899-1988. Can. J. Fish. Aquat.
Sci. 52:325-334.
Kadowaki, R., J. Irvine, B. Holtby, N. Schubert, K. Simpson, R. Bailey, and C.
Cross. 1995. Assessment of Strait of Georgia coho salmon stocks
(including the Fraser River). Dep. Fish. & Oceans, Draft PSARC working
Paper S94-9, Nanaimo, B.C.
Koski, K. V. 1966. The survival of coho salmon (Oncorhynchus kisutch) from
egg deposition to emergence in three Oregon coastal streams. Master's
thesis, Oregon State University, Corvallis.
Labelle, M. 1992. Straying patterns of coho salmon (Oncorhynchus kisutch)
stocks from southeast Vancouver Island, British Columbia. Can. J. Fish.
Aquat. Sci. 49: 1843-1855.
Lewis, M.A. 1995. Stock assessment of anadromous salmonids. Oreg. Dep.
Fish and Wildl., Fish Div. Fish Res. Proj. AFC-136, Ann. Prog. Rep.,
Portland.
Lynch, M. In Press. The genetic risks of extinction for pacific salmonids.
Workshop on Assessing Extinction Risk for West Coast Salmonids,
Seattle, WA , November 13-15, 1996.
Moore, K., K. Jones, and J. Dambacher. 1995. Methods for stream habitat
surveys, Version 5.1. Oreg. Dep. Fish and Wildl., R & D. Sec., Corvallis.
Moring, J.R., and R.L. Lantz. 1975. The Alsea watershed study: Effects of
logging on the aquatic resources of three headwater stream of the Alsea
River, Oregon. Part I - Biological studies. Oreg. Dep. Fish and Wildl., Fish
Div., Fish. Res. Rep. No. 9, Portland.
Nickelson, T.E. 1986. Influences of upwelling, ocean temperature, and smolt
abundance on marine survival of coho salmon (Oncorhynchus kisutch) in
the Oregon Production Area. Can. J. Fish. Aquat. Sci. 43:527-535.
Nickelson, T.E. and J.A. Lichatowich. 1984. The influence of the marine
environment on the interannual variation in coho salmon abundance: An
overview, p. 24-36. In W.G. Pearcy, [ed.] The influence of ocean conditions
on the production of salmonids in the North Pacific, A Workshop. Oregon
State University Sea Grant College Program, Corvallis, OR.
Nickelson, T.E., M.F. Solazzi, S.L. Johnson, and J.D. Rodgers. 1992a. An
approach to determining stream carrying capacity and limiting habitat for
coho salmon (Oncorhynchus kisutch), p. 251-260. In L. Berg and P.W.
Delaney, [eds.] Proceedings of the coho workshop, Nanaimo, B.C., May
26-28, 1992.
Nickelson, T.E., J.D. Rodgers, S.L. Johnson, and M.F. Solazzi. 1992b.
Seasonal changes in habitat use by juvenile coho salmon (Oncorhynchus
kisutch) in Oregon coastal streams. Can. J. Fish. Aquat. Sci. 49:783-789.
Oregon Department of Fish and Wildlife (ODFW). 1995. Oregon coho salmon
biological status assessment and staff conclusion for listing under the
Oregon endandered species act. Oreg. Dep. Fish and Wildl., Fish Div.,
Portland.
Overholtz, W.J. 1994. OCN coho: assessment, prognosis, and strategies for
stock rebuilding. Manuscript prepared for Oregon Trout, Inc. 207pp.
Pearcy, W.G. 1992. Ocean ecology of North Pacific salmonids. University of
Washington Press, Seattle.
Reeves, G.H., L.E. Benda, K.M. Burnett, P.A. Bisson, and J.R. Sedell. 1995. A
disturbance-based ecosystem approach to maintaining and restoring
freshwater habitats of evolutionarily significant units of anadromous
salmonids in the Pacific Northwest. Am. Fish. Soc. Sym. 17:334-349.
Ricker, W.E. 1975. Computation and interpretation of biological statistics of fish
populations. Bulletin of the Fisheries Research Board of Canada. Bulletin
191.
Seiler, D. 1989. Differential survival of Grays Harbor basin anadromous
salmonids: water quality implications, p. 123-135. In C.D. Levings, L.B.
Holtby, and M.A. Henderson [eds.] Proceedings of the national workshop
on effects of habitat alteration on salmonid stocks. Can. Spec. Pub. Fish.
Aquat. Sci. 105.
Skeesick, D.G. 1970. The fall immigration of juvenile coho salmon into a small
tributary. Res. Rep. Fish Comm. Oregon 2(1):1-6.
Thompson, G.G. 1991. Determining minimum viable populations under the
Endangered Species Act. NOAA Technical Memorandum NMFS F/NWC-198.
Table 1. Example of application of the coho salmon limiting factors model
(HLFM Version 5.0).
Stream: East Fork Lobster Creek
Stream inventories conducted in summer 1990 and winter 1990-91
Stream Length 3.8 km
| Season | Seasonal capacity | Life stage | Potential Smolts (Capacity*Survival) |
| Spawning | 1 330 000 | eggs | 266 000 |
| Spring | 32 400 | fry | 9 800 |
| Summer | 13 800 | parr | 6 900 |
| Winter | 4 500 | presmolts | 4 100 Limiting habitat and Smolt capacity |
| Stream area (m2) by habitat from inventories |
Seasonsl capacity by habitat (Area*Density) |
|||||
| Habitat type | Summer | Winter | Spawning | Spring | Summer | Winter |
| Cascades | 39 | 296 | - | 0 | - | |
| Rapids | 4 398 | 10 307 | 6 200 | 600 | 100 | |
| Riffles | 1 847 | 6 223 | 7 500 | 200 | 100 | |
| Glides | 2 966 | 1 911 | 3 500 | 2 300 | 200 | |
| Trench pools | 62 | - | - | 100 | - | |
| Plunge pools | 667 | 1 167 | 1 000 | 1 000 | 300 | |
| Lateral scour pools | 4 436 | 5 526 | 7 100 | 7 600 | 1 900 | |
| Mid-channel scour pools | - | - | - | - | - | |
| Dammed pools | 168 | 1 048 | 2 700 | 300 | 600 | |
| Alcoves | - | - | - | - | - | |
| Beaver ponds | 671 | 558 | 1 400 | 1 200 | 1 000 | |
| Backwater pools | 442 | 529 | 3 000 | 500 | 300 | |
| Spawning Gravel | 1 596 | 1 330 000 | ||||
| Total Calacity | 1 330 000 | 32 400 | 13 800 | |||
| Habitat type | Spring | Summer | Winter | |||||
| Cascades | 0.0 | 0.2 | 0.0 | Density independent survival rates | ||||
| Rapids | 0.6 | .01 | 0.01 | Egg to smolt | 0.2 | |||
| Riffles | 1.2 | .01 | 0.01 | Spring fry to smolt | 0.3 | |||
| Glides | 1.8 | .08 | 0.1 | Summer parr to smolt | 0.5 | |||
| Trench pools | 1.0 | 1.8 | 0.2 | Winter presmolt to smolt | 0.9 | |||
| Plunge pools | 0.8 | 1.5 | 0.3 | |||||
| Lateral scour pools | 1.3 | 1.7 | 0.4 | |||||
| Mid-channel scour pools | 1.3 | 1.7 | 0.4 | |||||
| Dammed pools | 2.6 | 1.8 | 0.6 | |||||
| Alcoves | 2.8 | 0.9 | 1.8 | |||||
| Beaver ponds | 2.6 | 1.8 | 1.8 | |||||
| Backwater pools | 5.8 | 1.2 | 0.6 | |||||
| Spawning Gravel | 2 500 eggs/redd / 3m2/redd = 833 eggs/m2 | |||||||
Table 2 ANOVA tables for regressions used in the model.
Multiple regression to predict habitat smolt capacity
| df | SS | MS | F | p | |
| Regression | 4 | 1.421 | 0.355 | 75.124 | <0.001 |
| Residual | 69 | 0.346 | 0.005 | ||
| Total | 73 | 1.747 | 0.024 |
Regression of overwinter survival on smolt capacity (Equation 2 and Figure 2)
| df | SS | MS | F | p | |
| Regression | 1 | 0.244 | 0.244 | 57.030 | <0.001 |
| Residual | 24 | 0.103 | 0.004 | ||
| Total | 25 | 0.347 | 0.014 |
Regression of egg-to-parr survival rate on percent ot full seeding (Equation 5 and Figure 4)
| df | SS | MS | F | p | |
| Regression | 1 | 9.984 | 9.894 | 52.606 | <0.001 |
| Residual | 25 | 4.702 | 0.188 | ||
| Total | 26 | 14.595 | 0.561 |
Figure 1. Performance of the coho salmon habitat limiting factors model (HLFM Version 5.0) in 7 study streams in terms of the relationship between the percent of the smolt capacity predicted by HLFM that was actually observed, and the density of juveniles present the previous summer.
Figure 2. Relationship between observed over-winter survival of coho salmon and potential smolt capacity as estimated by the HLFM for 5 study streams.
Figure 3. Flowchart showing the elements of the forward simulation component
of the model.
Figure 4. The relationship between egg-to-parr survival rate and percent of full
seeding that is the basis for the egg-to-parr survival parameter in the model.
Figure 5. Examples of frequency distributions of survival rates used in the simulation model: (A) egg-to-parr survival at four proportions of full seeding, (B) over-winter survival at four levels of habitat quality expressed as smolts/m2, and (C) marine survival rate when three average levels are used at initiation of the simulation.
Figure. 6. The pattern of the Aleutian low pressure index (dashed line) and the
corresponding step value (solid line) used in the simulation model to mimic
climatic variability and determine average marine survival rate and annual
exploitation rate.
Figure 7. Predicted and observed distributions of spawners into classes of fish
per mile in the 1995 run in the (A) Tillamook, (B) Yaquina, and (C) Coos basins.
Figure 8. The proportion of coho salmon habitat Oregon coastal basins where
coho salmon spawners will, at least, replace themselves if marine survival was
3% and exploitation rate was 0. NH = Nehalem; TB = Tillamook Bay; NS =
Nestucca; SL = Siletz; YQ = Yaquina; AL = Alsea; SI = Siuslaw; UM = Umpqua;
CB = Coos Bay; CQ = Coquille; RG = Rogue.
Figure 9. Relationship between 1990-95 mean coho spawners per mile
(extremes removed) in 11 coastal Oregon basins and habitat quality expressed
as the proportion of coho salmon habitat in each basin where coho salmon
spawners will, at least, replace themselves if marine survival was 3% and
exploitation rate was 0.
Figure 10. Frequency distribution of proportional change in population size from
one generation to the next based on results from the simulation model (3 basins;
99 000 observations) and on observed spawner abundance in 11 coastal Oregon
basins between 1990 and 1995.
Figure 11. Results of 10 generation model simulations for the Tillamook Basin comparing different levels of marine survival and exploitation rate: (A) median ending population per mile of habitat; (B) probability of population decline over the 10 generations, and; (C) probability of extinction.
Figure 12. Results of 10 generation model simulations for the Yaquina Basin comparing different levels of marine survival and exploitation rate: (A) median ending population per mile of habitat; (B) probability of population decline over the 10 generations, and; (C) probability of extinction.
Figure 13. Results of 10 generation model simulations for the Coos Basin comparing different levels of marine survival and exploitation rate: (A) median ending population per mile of habitat; (B) probability of population decline over the 10 generations, and; (C) probability of extinction.
Figure 14. Median population size and probability of extinction predicted for
model simulations of 33 generations with different levels of starting population for
the (A) Tillamook, (B) Yaquina, and (C) Coos basins.
Figure 15. Median population size and probability of extinction predicted for model simulations of 33 generations with different levels of change in freshwater habitat quality for the (A) Tillamook, (B) Yaquina, and (C) Coos basins.
Created March 4, 1997
Web Page Construction: Janet Demaris (503) 378-3397 x 234