Viable Salmon Populations and the Recovery of Evolutionarily Significant Units

U.S. DEPARTMENT OF COMMERCE

National Oceanic and Atmospheric Administration

National Marine Fisheries Service

NOAA Technical Memorandum NMFS-NWFSC-42

Viable Salmonid Populations

and the Recovery

of Evolutionarily Significant Units

June 2000

NOAA Technical Memorandum

NMFS Series

The Northwest Fisheries Science Center of the National

Marine Fisheries Service, NOAA, uses the NOAA

Technical Memorandum NMFS series to issue informal

scientific and technical publications when complete

formal review and editorial processing are not

appropriate or feasible due to time constraints.

Documents published in this series may be referenced in

the scientific and technical literature.

The NMFS-NWFSC Technical Memorandum series of

the Northwest Fisheries Science Center continues the

NMFS-F/NWC series established in 1970 by the North-

west & Alaska Fisheries Science Center, which has since

been split into the Northwest Fisheries Science Center

and the Alaska Fisheries Science Center. The

NMFS-AFSC Technical Memorandum series is now

being used by the Alaska Fisheries Science Center.

Reference throughout this document to trade

names does not imply endorsement by the National

Marine Fisheries Service, NOAA.

This document should be cited as follows:

McElhany, P., M.H. Ruckelshaus, M.J. Ford, T.C.

Wainwright, and E.P. Bjorkstedt. 2000. Viable

salmonid populations and the recovery of

evolutionarily significant units. U.S. Dept. Commer.,

NOAA Tech. Memo. NMFS-NWFSC-42,156 p.

NOAA Technical Memorandum NMFS-NWFSC-42

Viable Salmonid Populations

and the Recovery

of Evolutionarily Significant Units

Paul McElhany, Mary H. Rucklelshaus, Michael J. Ford, Thomas C.

Wainwright, and Eric P. Bjorkstedt*

Northwest Fisheries Science Center

Conservation Biology Division

2725 Montlake Boulevard East

Seattle, Washington 98112-2097

*Southwest Fisheries Science Center

Santa Cruz/Tiburon Laboratory, Salmon Analysis Branch

3150 Paradise Drive

Tiburon, California 94920-1211

June 2000

U.S. DEPARTMENT OF COMMERCE

William M. Daley, Secretary

National Oceanic and Atmospheric Administration

D. James Baker, Administrator

National Marine Fisheries Service

Penelope D. Dalton, Assistant Administrator for Fisheries

Most NOAA Technical Memorandums NMFS-NWFSC are

available on-line at the Northwest Fisheries Science Center

web site (http://www.nwfsc.noaa.gov)

Copies are also available from:

National Technical Information Service

5285 Port Royal Road

Springfield, VA 22161

phone orders (1-800-553-6487)

e-mail orders (orders @ntis.fedworld.gov)

iii

Table of Contents

List of Figures vii

List of Tables ix

List of Boxes xi

EXECUTIVE SUMMARY xiii

DEFINING A VIABLE SALMONID POPULATION 1

Introduction......................................................................................................................... 1

Purpose and Scope .................................................................................................. 1

Definitions............................................................................................................... 2

Short-term Risk Evaluations ............................................................................................... 4

Population concepts............................................................................................................. 4

General definitions .................................................................................................. 4

Definition of a population that NMFS will use in applying the VSP concept........ 5

Distinction between Population Definition and Tools for Estimation.................... 6

Structure below and above population level........................................................... 6

Borderline situations in defining populations ......................................................... 7

Relationship of the Population Definition to the ESU Definition........................... 7

Population Definition and Artificial Propagation................................................... 9

PARAMETERS FOR EVALUATING POPULATIONS 11

Introduction to Parameters................................................................................................ 11

Population parameters........................................................................................... 11

Guidelines for each population parameter............................................................ 11

Population Size.................................................................................................................. 12

Population Growth Rate and Related Parameters............................................................. 13

Spatial Structure................................................................................................................ 18

Diversity............................................................................................................................ 19

Integrating the parameters and determining population status.......................................... 23

ESU VIABILITY 25

Introduction....................................................................................................................... 25

Number and Distribution of Populations in a Recovered ESU......................................... 25

Populations not meeting VSP guidelines.......................................................................... 27

IMPLEMENTING THE VSP GUIDELINES 29

Introduction....................................................................................................................... 29

Practical Application......................................................................................................... 29

Uncertainty, Precaution, and Adaptive Management........................................................ 29

Interim Application........................................................................................................... 30

Examples........................................................................................................................... 31

APPENDIX 33

Applying VSP in the Regulatory Arena............................................................................ 33

Listing Criteria ...................................................................................................... 33

Recovery................................................................................................................ 34

Jeopardy................................................................................................................ 35

Relationship of VSP to Other Concepts............................................................................ 35

Relationship to Minimum Viable Population Concepts........................................ 35

Relationship to Quantitative Population Viability Analysis................................. 36

Relationship to Properly Functioning Conditions................................................. 36

Relationship to the Sustainable Fisheries Act (SFA) and Maximum Sustainable

Yield (MSY).......................................................................................................... 37

Relationship to Other Conservation Assessment Approaches.............................. 38

Identifying Populations..................................................................................................... 38

Introduction to identifying populations................................................................. 38

Types of Information Used in Identifying Populations......................................... 38

Evidence for independent populations ...................................................... 39

Indicators of population structure ............................................................. 41

Geographic and habitat indicators................................................. 44

Demographic indicators ................................................................ 46

Genetic indicators.......................................................................... 49

Identifying populations—combining the evidence................................... 51

Existing Approaches to Identifying Salmon Populations or Groups .................... 51

Population Size.................................................................................................................. 53

Introduction....................................................................................................................... 53

Density effects........................................................................................... 54

Environmental variation............................................................................ 57

Genetic processes...................................................................................... 58

Demographic stochasticity........................................................................ 61

Ecological feedback .................................................................................. 62

Assessment Methods............................................................................................. 63

Guidelines.............................................................................................................. 64

Population Growth Rate and Related Parameters............................................................. 64

Introduction........................................................................................................... 64

Why population growth rate is important ................................................. 67

Why intrinsic productivity and density dependence are important........... 68

Why stage-specific productivity is important ........................................... 69

Why ancillary data relevant to productivity are important ....................... 73

Estimating Population Growth Rate and Related Parameters............................... 77

Estimating population growth rate and changes in other parameters........ 78

Estimating population growth rate and detecting trends........................... 79

Detecting other pattern in time series: autocorrelation, interventions, and

epochs........................................................................................................ 80

Estimating intrinsic productivity and detecting density dependence........ 81

Critical assumptions .................................................................................. 81

Bias and methods for correcting it ............................................................ 82

Analyses that incorporate stage-specific dynamics................................... 83

Analyses of populations that include naturally spawning hatchery fish... 84

Spatial Structure................................................................................................................ 90

Introduction........................................................................................................... 90

General Spatial Patterns........................................................................................ 91

Why Spatial Structure is Important....................................................................... 91

Metapopulation theory.............................................................................. 91

Source-sink dynamics ............................................................................... 94

Importance of patch spacing...................................................................... 96

Fragmented habitats .................................................................................. 97

Assessing Spatial Structure................................................................................... 97

Straying..................................................................................................... 97

Habitat dynamics....................................................................................... 98

VSP Guidelines: Spatial Structure ...................................................................... 101

Diversity.......................................................................................................................... 101

Types of Diversity............................................................................................... 101

Why Diversity is Important................................................................................. 102

Factors that Affect Diversity............................................................................... 102

Risks to Diversity................................................................................................ 105

Selection.................................................................................................. 108

Straying and gene flow............................................................................ 114

VSP Guidelines: Diversity.................................................................................. 122

Viable ESUs.................................................................................................................... 124

Catastrophes........................................................................................................ 124

Long-term Demographic and Evolutionary Processes........................................ 125

ESU Viability Guidelines.................................................................................... 126

LITERATURE CITED 127

vii

List of Figures

Figure 1. This figure illustrates why subpopulations, populations, and ESUs are likely to have a

biological basis................................................................................................................................ 8

Figure 2. Theoretical types of spatially structured populations.................................................... 20

Figure A1a. Distributions of dispersal distances of eelgrass (Zostera marina) based on pollen

and seed dispersal.......................................................................................................................... 42

Figure A1b. Frequency distributions of pollen, seed and combined pollen and seed dispersal

estimated by identifying and mapping seedlings and parents....................................................... 43

Figure A2. The figure shows why demographic independence is theoretically a useful factor in

designating populations................................................................................................................. 47

Figure A3. Typical shape of parent-offspring (or stock-recruit) curves for populations with no

density-dependence (“Linear”), compensation only, and compensation plus depensation. ......... 55

Figure A4. Estimates of spawner:spawner production for Oregon coastal coho salmon

(Oncorhynchus kisutch) in three GCGs. ....................................................................................... 70

Figure A5. Estimates of pre-harvest-recruit:spawner production for Oregon coastal coho salmon

(Oncorhynchus kisutch) in three GCGs. ....................................................................................... 71

Figure A6. Estimates of exploitation rates for three GCGs (North-Mid Coast, Umpqua, Mid-

South Coast) of Oregon coastal coho salmon (Oncorhynchus kisutch) and estimated total

exploitation rate for coho salmon in the Oregon Production Index Area based on analyses of

coded wire tag recoveries.............................................................................................................. 72

Figure A7. Changes in parameters of a composite Beverton-Holt model for smolt yield as a

function of prespawner abundance for four putative life-history variants of spring chinook

salmon (Oncorhynchus tshawytscha) in the Grande Ronde River. .............................................. 74

Figure A8. Mean weights of coho salmon from rivers on the outer coast of Washington State.

Rivers are listed in order from south to north. .............................................................................. 75

Figure A9. Mean weights of coho salmon from rivers in Puget Sound. Rivers are listed in order

from north to south........................................................................................................................ 76

Figure A10. Estimated spawner escapement for Wenatchee River summer steelhead (O. mykiss),

and Natural Return Ratio (NRR) calculated for a range of values describing the relative

reproductive successes of hatchery- and natural-origin spawners................................................ 87

Figure A11. Examples of Ricker-type spawner-recruit models that include the influence of

naturally spawning hatchery fish in different ways, fitted to data for Wenatchee River summer

steelhead (Oncorhynchus mykiss) for 1984-1994......................................................................... 88

Figure A12. Theoretical types of spatially structured populations. .............................................. 92

Figure A13. Productivity estimated as spawners per spawner by index reach for coho in the

Snohomish River, WA. ................................................................................................................. 95

Figure A14. Map of probable spring chinook spawning areas in the Grande Ronde basin.......... 99

Figure A11-3a. Percent dead and infected with C. Shasta after 86 days of exposure................ 107

viii

Figure A12-1a.Mean spawn timing of Trask Hatchery coho...................................................... 109

Figure A12-1b.Average daily streamflow for the Nestucca, Siletz, Yaquina, Alsea, and Siuslaw

river basins, November through April......................................................................................... 110

Figure A12-2a. Discriminant scores of morphological variation between wild and hatchery

female coho salmon..................................................................................................................... 111

Figure A12-3a. The proportion of smolts that were age 1+ by year of migration, and the

observed numbers of yearling and two-year-old smolts by year of migration............................ 113

Figure A13-1a. Approximate relationship between G

and N

at equilibrium. ....................... 116

Figure A13-2a. Frequency of the A allele in population 1 with m

= 0.2 or m

= 0.05............. 118

Figure A14-2a. Time series of allele frequency data at six gene loci for fall chinook salmon from

the Snake and upper Columbia Rivers........................................................................................ 120

Figure A14-3a. Days to death by ceratomyxosis from initial exposure of rainbow trout to

Ceratomyxa shasta, 1989 and 1990............................................................................................. 121

List of Tables

Table A1. Estimates of straying (the percentage of marked fish returning to a location other than

that in which it was marked) for Pacific salmonids...................................................................... 40

Table A11-2a. (Data from Raleigh 1971, Table 1.).................................................................... 106

List of Boxes

Box A1. Assigning “resident” and anadromous salmonids to populations. ................................ 45

Box A2. Viable Population Size Guidelines................................................................................. 65

Box A3. Critical Population Size Guidelines................................................................................ 66

Box A4. Incorporating stage-specific productivity data in evaluations of abundance and

productivity trends: harvest estimates for coho salmon (Oncorhynchus kisutch) from coastal

Oregon........................................................................................................................................... 70

Box A5. Population dynamics in stage-structured populations and the fate of different life

histories in a population. ............................................................................................................... 74

Box A6. Example of ancillary data relevant to population viability: trends in size of coho

salmon from different regions....................................................................................................... 75

Box A7. Estimating productivity in populations that include naturally spawning hatchery fish. 85

Box A8. Population growth rate and related parameters guidelines............................................ 89

Box A9. Spatial Structure Guidelines........................................................................................ 100

Box A10. Examples of Diversity. Below are three brief examples illustrating trait diversity

within and among populations of chinook salmon...................................................................... 103

Box A11. Examples of adaptive diversity.................................................................................. 106

Box A12. Human caused selection. ............................................................................................ 109

Box A13. Models of genetic variation among populations........................................................ 116

Box A14. Examples of loss of diversity or adaptation due to human-caused gene flow alteration.

..................................................................................................................................................... 119

Box A15. Diversity guidelines................................................................................................... 123

Box A16. ESU viability guidelines............................................................................................ 126

xiii

EXECUTIVE SUMMARY

This document introduces the viable salmonid population (VSP) concept, identifies VSP

attributes, and provides guidance for determining the conservation status of populations and

larger-scale groupings of Pacific salmonids. The concepts outlined here are intended to serve as

the basis for a general approach to performing salmonid conservation assessments. As a specific

application, the VSP approach is intended help in the establishment of Endangered Species Act

(ESA) delisting goals. This will aid in the formulation of recovery plans and can serve as interim

guidance until such plans are completed.

The approach of the VSP concept and this document is to define a viable population,

describe techniques for determining population boundaries, identify parameters useful in

evaluating population viability and then set guidelines for assessing population viability status

with regard to each of the parameters. Finally guidelines are provided on how to relate

individual population viability to the viability of the Evolutionarily Significant Unit (ESU) as

whole. The document is based primarily on a review and synthesis of the conservation biology

and salmonid literature. A large portion of the document is an appendix devoted to describing

the technical rationale behind the population definition and viability guidelines.

We define a viable salmonid population as an independent population of any Pacific

salmonid (genus Oncorhynchus) that has a negligible risk of extinction due to threats from

demographic variation, local environmental variation, and genetic diversity changes over a 100-

year time frame. We define an independent population as any collection of one or more local

breeding units whose population dynamics or extinction risk over a 100-year time period are not

substantially altered by exchanges of individuals with other populations. In other words, if one

independent population were to go extinct, it would not have much impact on the 100-year

extinction risk experienced by other independent populations. Independent populations are

likely to be smaller than a whole ESU.

Population identification is the first step for a VSP analysis. The best method for

identifying independent populations uses direct observations of trends in abundance or

productivity from groups of fish with known inter-group stray rates. However, such data are

rarely available, and proxy evidence must be used to identify population boundaries. Such

evidence could include geographic and habitat indicators, demographic indicators and genetic

indicators (both neutral molecular markers and quantitative traits). The availability and

usefulness of each of these indicators will vary by ESU.

Four parameters form the key to evaluating population viability status. They are

abundance, population growth rate, population spatial structure, and diversity. The NMFS

focuses on these parameters for three reasons. First, they are reasonable predictors of extinction

risk (viability). Second, they reflect general processes that are important to all populations of all

species. Third, the parameters are measurable. To facilitate evaluation of populations, we

provide a collection of viability guidelines based on our interpretation of currently available data

and literature. As with all scientific endeavors, these guidelines can be modified as new data,

more rigorous analysis and clearer interpretations are generated.

xiv

Abundance is recognized as an important parameter because, all else being equal, small

populations are at greater risk of extinction than large populations, primarily because several

processes that affect population dynamics operate differently in small populations than they do in

large populations. These processes are deterministic density effects, environmental variation,

genetic processes, demographic stochasticity, ecological feedback, and catastrophes. Guidelines

relating minimum abundance to each of these processes are provided at both the “viable” and

“critical” level, where a critical level implies a high risk of extinction over a short time period.

Population growth rate (i.e., productivity over the entire life cycle) and factors that affect

population growth rate provide information on how well a population is “performing” in the

habitats it occupies during the life cycle. Estimates of population growth rate that indicate a

population is consistently failing to replace itself are an indicator of increased extinction risk.

Although our overall focus is on population growth rate over the entire life cycle, estimates of

stage-specific productivity—particularly productivity during freshwater life-history stages—are

also important to comprehensive evaluation of population viability. Other measures of

population productivity, such as intrinsic productivity and the intensity of density-dependence

may provide important information for assessing a population’s viability. The guidelines for

population growth rate are closely linked with those for abundance.

When evaluating population viability, it is important to take within-population spatial

structure needs into account for two main reasons: 1) Because there is a time lag between

changes in spatial structure and species-level effects, overall extinction risk at the 100-year time

scale may be affected in ways not readily apparent from short-term observations of abundance

and productivity, and 2) population structure affects evolutionary processes and may therefore

alter a population’s ability to respond to environmental change. Spatially structured populations

in which “subpopulations” occupy “patches” connected by some low to moderate stray rates are

often generically referred to as “metapopulations.” A metapopulation’s spatial structure depends

fundamentally on habitat quality, spatial configuration, and dynamics as well as the dispersal

characteristics of individuals in the population. Pacific salmonids are generally recognized as

having metapopulation structure and the guidelines for spatial structure describe general rules of

thumb regarding metapopulation persistence.

Several salmonid traits exhibit considerable diversity within and among populations, and

this variation has important effects on population viability. In a spatially and temporally varying

environment, there are three general reasons why diversity is important for species and

population viability. First, diversity allows a species to use a wider array of environments than

they could without it. Second, diversity protects a species against short-term spatial and

temporal changes in the environment. Third, genetic diversity provides the raw material for

surviving long-term environmental change. In order to conserve the adaptive diversity of

salmonid populations, it is essential to 1) conserve the environment to which they are adapted,

2) allow natural process of regeneration and disturbance to occur, and 3) limit or remove human-

caused selection or straying that weakens the adaptive fit between a salmonid population and its

environment or limits a population's ability to respond to natural selection.

The ESA is not concerned with the viability of populations per se, but rather with the

extinction risk faced by an entire ESU. A key question is how many and which populations are

necessary for a sustainable ESU. Three factors need to be considered when relating VSPs to

viable ESUs: 1) catastrophic events, 2) long-term demographic processes, and 3) long-term

evolutionary potential. We provide a number of guidelines related to these factors with an

emphasis on risks from catastrophic events.

The guidelines presented here are intentionally general so they can be applied equally

across the wide spectrum of life-history diversity, habitat conditions, and metapopulation

structures represented by Pacific salmon. It is left to Technical Recovery Teams and other

efforts to develop ESU-specific quantitative delisting criteria based on the principles outlined in

VSP. A main concern in translating the guidelines into specific criteria will be the degree of

uncertainty in much of the relevant information. Because of this uncertainty, management

applications of VSP should employ both a precautionary approach and adaptive management.

The precautionary approach suggests that VSP evaluations should error on the side of protecting

the resource and adaptive management suggests that management activities should be used as a

means of collecting more data to improve the quality of a VSP evaluation.

DEFINING A VIABLE SALMONID POPULATION

Introduction

This document introduces the viable salmonid population (VSP) concept, identifies VSP

attributes, and provides guidance for determining the conservation status of populations and

larger-scale groupings of Pacific salmonids. The concepts outlined here are intended to serve as

the basis for a general approach to performing salmonid conservation assessments. Pacific

salmonid risk evaluations can occur at small, local scales or over larger geographic regions—

depending on the salmon management entities involved and the purpose of the risk assessment.

In this document, we focus on conservation assessments of salmonid populations and

Evolutionarily Significant Units (ESUs) because there is an immediate need for such evaluations

under the Endangered Species Act (ESA)—a concern that the National Marine Fisheries Service

(NMFS) must address. This document is divided into two main sections: 1) an initial discussion

of the general concepts underlying the notion of a VSP, and 2) a detailed appendix where we

provide technical details to support population identification, population parameter guidelines,

and specific examples of how the guidelines pertain to salmonids.

We have confidence in the conceptual foundations underlying both the notion of a VSP

and what critical elements should be evaluated when determining viability at the population and

ESU scales. However, the approach to applying the VSP concept itself is still in the

development stage and is likely to change with experience. We expect that the means of

identifying population boundaries and establishing guidelines for population parameters will

continue to be refined as further empirical data and modeling efforts are brought to bear on these

important issues.

Purpose and Scope

The National Marine Fisheries Service (NMFS) is responsible for evaluating the status of

certain salmonids and other marine species under the Endangered Species Act (ESA).

For

species listed under the ESA, NMFS must determine whether particular management actions are

likely to appreciably reduce the species' likelihood of survival and recovery in the wild. NMFS

must also guide other entities in fulfilling listed species' needs and in taking actions necessary to

recover them to self-sustaining levels. The purpose of this document is to provide an explicit

framework for identifying attributes of viable salmonid populations so that parties may assess the

effects of management and conservation actions and ensure that their actions promote the listed

species' survival and recovery. The VSP concept and the criteria presented in this document are

NMFS shares ESA jurisdiction with the U.S. Fish and Wildlife Service (FWS) and generally retains ESA authority

over species that spend a majority of their lives in the marine environment, including anadromous Pacific salmonids

(FWS and NMFS 1974). In some cases, NMFS may possess ESA authority over salmonid species that spend all or

most of their life histories in freshwater as well. Consequently, the concepts contained in this document are

intended to apply broadly to all Pacific salmonid species under NMFS’ ESA jurisdiction.

intended both to help formulate recovery plans and to serve as interim guidance until such plans

are completed.

When making listing decisions regarding Pacific salmonids (members of the genus

Oncorhynchus), it is NMFS’ policy to list ESUs as “distinct population segments” under the Act.

However, there is wide recognition among NMFS, other agencies, and independent scientists for

the need to undertake conservation actions at scales smaller than the ESU (Waples 1991c, NMFS

1991, WDF et al. 1993, Kostow 1995, Allendorf et al. 1997). The population is at an appropriate

level for examining many extinction processes. As a consequence, the viability analyses

discussed in this document are applied primarily at the scale of what are called independent

populations, which will almost always be smaller than the scale of an ESU (see the following

section “Definitions”). We define population performance measures in terms of four key

parameters: abundance, population growth rate, spatial structure, and diversity. We then relate

performance and risks at the population scale to risks affecting the persistence of entire ESUs.

The VSP concept consists primarily of two components: 1) Principles for identifying

population substructure in Pacific salmonid ESUs, and 2) general principles for establishing

biological guidelines to evaluate the conservation status of these populations and, therefore, of

entire ESUs. The diversity of salmonid species and populations makes it impossible to set

narrow quantitative guidelines that will fit all populations in all situations. The concepts and

guidelines outlined in this document are therefore fairly general in nature. More specific

guidelines can only be determined through detailed analyses of case-specific information on

particular regions and particular species. As of Spring 2000, the concepts outlined in this VSP

document have been applied to salmonid conservation planning in the upper Columbia River

geographic region (Ford et al. 1999a). Populations have been identified for listed spring-run

chinook salmon and steelhead in the upper Columbia River ESUs. Viability targets at the

population and ESU levels have been established for both species in the Quantitative Analytical

Report (QAR) (Ford et al. 1999a). The “QAR” document was the result of the efforts of a multi-

agency team of scientists convened to provide an evaluation of the effects of the Columbia River

hydrosystem on ESA-listed Upper Columbia River spring chinook salmon and steelhead. This

VSP document provides a conceptual overview of important factors to consider in evaluating the

viability of salmonid populations and ESUs. The QAR document offers concrete examples of

how the general concepts outlined in VSP might be applied. As further applications of VSP

concepts are completed, we expect that more quantitative and general viability guidelines will

emerge.

Definitions

A viable salmonid population (VSP)

is an independent population of any Pacific

salmonid (genus Oncorhynchus) that has a negligible risk of extinction due to threats from

demographic variation (random or directional), local environmental variation, and genetic

diversity changes (random or directional) over a 100-year time frame. Other processes

Note that some early drafts of this document used the term “properly functioning population” or “PFP” in place of

VSP. We believe the term “viable population” more accurately reflects the authors’ intent, which is to describe the

population attributes necessary to ensure long-term species survival in the wild.

contributing to extinction risk (catastrophes and large-scale environmental variation) are also

important considerations, but by their nature they need to be assessed at the larger temporal and

spatial scales represented by ESUs or other entire collections of populations.

The crux of the population definition used here is what is meant by “independent.” An

independent population is any collection of one or more local breeding units whose population

dynamics or extinction risk over a 100-year time period is not substantially altered by exchanges

of individuals with other populations. In other words, if one independent population were to go

extinct, it would not have much impact on the 100-year extinction risk experienced by other

independent populations. Independent populations are likely to be smaller than a whole ESU

and they are likely to inhabit geographic ranges on the scale of entire river basins or major sub-

basins. The rationale underlying these definitions will be discussed further in “Population

Concepts” (p. 4).

While it is ultimately an arbitrary decision, the 100-year time scale was chosen to

represent a “long” time horizon for evaluating extinction risk. It is necessary to evaluate

extinction risk at a long time scale for several reasons. First, many recovery actions (such as

habitat restoration) are likely to affect population status over the long term. Second, many

genetic processes important to population function (such as the loss of genetic diversity or

accumulation of deleterious mutations) occur over decades or centuries and current actions can

affect these processes for a long time to come. Third, at least some environmental cycles occur

over decadal (or longer) time scales (e.g., oceanic cycles—Beamish and Bouillon 1993, Mantua

et al. 1997, Hare et al. 1999). Thus, in order to evaluate a population's status it is important to

look far enough into the future to be able to accommodate large-scale environmental oscillations

and trends.

Note that choosing a time scale of 100 years does not mean that we believe it is possible

to predict with great precision a population’s status that far into the future. Nonetheless, we can

describe those population attributes necessary for a species' long-term persistence. (This is

discussed in more detail in Part 2.) Although our time frame for evaluating population viability

is 100 years, we recognize and expect that many management actions and their subsequent

monitoring will occur over much shorter time scales, and some evolutionary and large-scale

demographic processes that can affect ESU viability will occur over much greater time periods.

One hundred years was chosen as a reasonable compromise: it is long enough to encompass

many long-term processes, but short enough to feasibly model or evaluate. It is worth noting that

quantitative and qualitative conservation assessments for other species have often used a 100-

year time frame in their extinction risk evaluations (Morris et al. 1999).

Although a population is the appropriate unit of study for many biological processes, it

may also be appropriate to evaluate management actions that affect units at smaller or larger

spatial and temporal scales. For example, ocean harvest plans may affect multiple-populations,

while a habitat restoration plan may only affect a small portion of a single population’s habitat.

The VSP concept does not preclude establishment of goals at these different scales. However,

management actions ultimately need to be related to population and ESU viability.

Short-term Risk Evaluations

In addition to evaluating population viability over long time periods, it is often important

to analyze short-term risks relating to population or species persistence. In particular, a number

of management decisions made at local, state, and federal levels are based on whether an action

will have a significant effect on salmonid population viability over short time spans (e.g., 10 or

fewer years). For example, in its decision on the 1995 Hydropower Biological Opinion, NMFS

established critical abundance thresholds below which the short-term survival of a population is

believed to be in considerable doubt. In another instance, federal, state, and tribal entities had to

determine the abundance levels at which a population is at such a high risk of extinction that a

captive broodstock program is needed in order to rebuild it (NMFS 1995b—Snake River Salmon

Recovery Plan). In most cases, a “critical” population status implies a high risk of extinction

over a short time period. In situations where such critical thresholds need to be established, the

same population parameters used in determining whether a salmonid population is viable should

be considered. In other words, evaluating whether a population is “critical” should involve

assessing its abundance, population growth rate, population structure, and diversity. Clearly, the

values of the four parameters in a critical population would be lower or less functional than those

in a viable population. In “Population Size” (p. 12) we describe guidelines for using abundance

to evaluate critical population status.

Population Concepts

General Definitions

In common biological usage, a population is broadly defined as a group of organisms.

For example, the Third Edition of the American Heritage Dictionary defines the ecological usage

of a population as “all the organisms that constitute a specific group or occur in a specified

habitat.” Other common definitions include “any specified reproducing group of individuals”

(Chambers Science and Technology Dictionary) and “any group of organisms of the same

species living in a specific area” (Academic Press Dictionary of Science and Technology). A

common definition of a population from ecology and population biology textbooks may be

summarized as “a group of organisms of the same species that occupy the same geographic area

during the same time” (e.g., McNaughton and Wolf 1973, Ehrlich and Roughgarden 1987).

Thus, the definition of a population is clearly broad enough to be tailored to specific

applications. For example, theoretical population genetic models often make use of a panmictic

population, defined as a group of individuals that randomly interbreed every generation (e.g.,

Crow and Kimura 1970). In an evolutionary context, a population is “a group of organisms,

usually a group of sexual organisms that interbreed and share a gene pool” (Ridley 1996). In

other situations, it may be useful to define populations much more broadly, up to and including

entire species (e.g., Ehrlich and Roughgarden 1987).

Definition of a Population NMFS Will Use in Applying the VSP Concept

In the VSP context, NMFS defines an independent population much along the lines of

Ricker's (1972) definition of a “stock.” That is, “an independent population is a group of fish of

the same species that spawns in a particular lake or stream (or portion thereof) at a particular

season and which, to a substantial degree, does not interbreed with fish from any other group

spawning in a different place or in the same place at a different season.” For our purposes, not

interbreeding to a “substantial degree” means that two groups are considered to be independent

populations if they are isolated to such an extent that exchanges of individuals among the

populations do not substantially affect the population dynamics or extinction risk of the

independent populations over a 100-year time frame. The exact level of reproductive isolation

that is required for a population to have substantially independent dynamics is not well

understood, but some theoretical work suggests that substantial independence will occur when

the proportion of a population that consists of migrants is less than about 10% (Hastings, 1993).

Thus independent populations are units for which it is biologically meaningful to examine

extinction risks that derive from intrinsic factors such as demographic, genetic, or local

environmental stochasticity.

The degree to which a group of fish has population dynamics that are independent from

another group's depends in part on the relative numbers of fish in the two groups. Ten migrants

into a group of 1,000 fish would have a much smaller demographic impact than 10 migrants into

a group of 10 fish. Practically speaking, applying our definition of a population will involve an

assumption about the degree of independence individual fish groups experienced under historical

or “natural” conditions (i.e., before the recent or severe declines that have been observed in many

populations). It is necessary to consider historical conditions to ensure that a population

designation is not contingent on relative conservation status among groups of fish. In some

cases, it may be determined that environmental conditions are so altered that either it is

impossible to evaluate an ESU's pre-decline population structure or the population structure of

the recovered ESU would be substantially different from what it was historically. In these cases,

it may be necessary to identify both the current population structure and what the population

structure is expected to be after recovery is achieved.

For species like pink and coho salmon, for which the age structure is relatively fixed

(e.g., pink salmon mature at 2 years and coho salmon often mature at 3 years), cohorts within a

breeding group could technically belong to separate populations as we have defined them.

Whether cohorts within a breeding group are treated as separate populations depends on the

degree of inter-cohort straying. In cases where there is less than 10% “migration” between

cohorts (as could occur when fish come back a year earlier or later than their normal pattern), the

cohorts should be treated as separate populations. In practice, because these “temporally

isolated” populations occupy essentially the same habitats in space, viability assessments at the

population and ESU level should take into account the highly correlated environmental

conditions such populations’ experience.

The Washington Department of Fish and Wildlife, tribal groups (WDF et al. 1993) and

the Oregon Department of Fish and Wildlife (OAR 635-07-501(38)) use population definitions

that require some level of reproductive isolation among populations. This focus on demographic

independence is consistent with the manner in which the population concept is often applied in

fisheries analysis. As discussed in “Population growth rate and related parameters” (p. 13),

estimating spawner/recruitment relationships is a common analytical tool in fisheries biology.

To apply these estimates, particularly where density-dependent reproduction is involved, it must

be assumed that populations are reproductively isolated. Indeed, inadvertently pooling groups of

fish from different independent populations is a major source of error in estimating

spawner/recruit relationships (Hilborn and Walters 1992, Ray and Hastings 1996). Whether

explicitly stated or not, most analyses using spawner/recruit relationships assume a population

(or “stock”) definition similar to the one used in this document.

Distinction between Population Definition and Tools for Estimation

In the Appendix “Identifying populations” (p. 38), we describe several ways to estimate

dispersal rates and population boundaries. These include performing mark-recapture studies,

exploring correlations in population fluctuations, assessing patterns of phenotypic variation, and

using molecular genetic markers to track individuals or to estimate similarity among groups of

fish. It is important to emphasize that these techniques are simply tools for estimating population

boundaries; they are not part of the population definition itself. For example, genetic marker

patterns may show the degree to which groups of fish are reproductively isolated. Our

population definition does not in any way stipulate how to interpret those patterns. As a case in

point, simply because one group of fish has a statistically detectable set of allele frequency

differences from another group, it does not necessarily mean that each group represents an

independent population.

Geographic characteristics are another tool that may be used to help identify populations

and their boundaries. Spatial distributions of spawning groups—and whole salmonid

populations—are constrained by geographic features such as basin and sub-basin structure. The

physical locations of suitable habitat within a basin and the fishes' dispersal capabilities combine

to determine, in part, the area over which a population is distributed. Nonetheless, it is important

to note that populations cannot be defined based on geography, rather they are defined based on

biological processes, (i.e., reproductive isolation and demographic independence). Thus biology

may cause a population’s geographic boundaries to be smaller or larger than a single basin or

sub-basin. Given seven species and many life-history variants, the geographic expanse that

different populations occupy is likely to vary substantially. An example of how one might use

such data to identify populations is provided by Ford et al. (2000).

Structure Below and Above Population Level

A population, as defined in this document, is described as a group of fish that is

reproductively isolated “to a substantial degree.” However, as a criterion for defining groups of

fish, the degree of reproductive isolation is a relative measure that may vary continuously from

pairs of fish to the isolation separating species. The “population” defined here is not, therefore,

the only biologically logical grouping that may be constructed. Within a single population, for

example, individual groups of fish are often reproductively isolated to some degree from other

groups but not sufficiently isolated to be considered independent by the criteria adopted here.

These groups of fish are termed “subpopulations.” (“Spatial Structure,” p. 18, describes

subpopulations and spatial structure.)

There may be structure above the level of a population as well as below it. This is

explicitly recognized in the ESU designations: an ESU may contain multiple populations

connected by some small degree of migration. Thus organisms can be grouped in a hierarchic

system wherein we define the levels of individual, subpopulation, population, ESU and, finally,

species. Other hierarchic systems made up of more or fewer levels could be constructed.

Though reproductive isolation forms a continuum, it is not a smooth continuum, and there exists

a biological basis for designating a hierarchy of subpopulations, populations, and ESUs (Figure

1).

Borderline Situations in Defining Populations

Because we are attempting to define discrete population boundaries from largely

continuous processes, it is inevitable that there will be situations in which the population status

of a group of fish cannot neatly be assigned. There will often be quasi-reproductively isolated

groups of fish within a population, referred to as subpopulations (discussed in “Spatial

Structure,” p. 18). Deciding whether a group of fish is a marginally independent population or a

significantly distinct subpopulation within a larger population will not always be straightforward.

Extinction risk models can be utilized that explicitly allow for any level of reproductive

isolation, so from a modeling perspective (assuming the degree of reproductive isolation is truly

known), the distinction between a population and a subpopulation is reduced to one of semantics.

However, it is possible that the management implications of how population substructure is

defined could be much greater, depending on how the VSP concept is applied in policy.

Another scenario in which a group of fish will not fit neatly into our definition of a VSP

is when the group is demographically independent of other groups, and its “natural” probability

of extinction within 100 years is more than “negligible.” Some independent populations may not

be viable, even under pristine conditions. It is important to recognize that naturally non-viable

independent populations are possible. The implications of these types of populations for ESU

viability are discussed in “Populations not meeting VSP guidelines” (p. 27).

Relationship of the Population Definition to the ESU Definition

An ESU is defined by two criteria: 1) it must be substantially reproductively isolated

from other conspecific units, and 2) it must represent an important component of the

evolutionary legacy of the species (Waples 1991c). Our population definition is based on a

single criterion: it must be sufficiently reproductively isolated from other conspecific units so

that its population dynamics or risk of extinction are substantially independent of other units over

a time frame of at least 100 years (“Definitions,” p. 2). Thus, the two definitions share a

common requirement for substantial reproductive isolation; but an ESU must also represent an

important component of the species' evolutionary legacy. Consequently, ESUs are generally

more reproductively isolated over a longer period of time than are the populations within them.

Subpop.

Pop.

ESU

Subpop.

Pop. Pop.

ESU

No biologically based groupings.

Distance

Biologically based groupings.

Figure 1. This figure illustrates why subpopulations, populations, and ESUs are likely to have a biological

basis. Each vertical line represents a panmictic (completely interbreeding) group of fish. If the

probability of mating between two individuals is simply a function of distance and the fish are

arranged as in “A,” there will be some biological basis for grouping fish into subpopulations,

populations, and ESUs. If the fish are arranged as “B” depicts, the probability of mating may

still decline with distance, but there are no biologically obvious groupings. The homing

tendencies of Pacific salmon—combined with spatial structure of freshwater spawning

habitat—suggest that most salmon species will resemble the scenario depicted in “A” rather

than that in “B.” The distance measure in this figure may represent simple Euclidean distance

or a more complex measure, such as a metric involving migration barrier permeability.

No population, as it is defined here, would ever be a member of more than one ESU, but a single

ESU may contain multiple populations.

Population Definition and Artificial Propagation

The stated purposes of the ESA are to provide a means whereby the ecosystems upon

which endangered and threatened species depend may be conserved, to provide a program for

conserving such species, and to take the steps needed to achieve these purposes (ESA sec. 2[b]).

The ESA's focus is on natural populations and the ecosystems upon which they depend.

Artificial propagation of a listed salmonid species is not a substitute for eliminating the factors

causing or contributing to a species' decline (NMFS 1993).

There are hundreds of artificial propagation programs for salmonids in Washington,

Oregon, Idaho and California. Collectively, they released several hundred million juvenile fish

in the late 1990s (Beamish et al. 1997). Whether by design, as in a supplementation program, or

through unintentional straying, hatchery fish often spawn with natural fish in the wild.

In cases

where hatchery fish interbreed with natural fish on spawning grounds and a substantial number

of the spawners are fish of hatchery origin, the naturally spawning component cannot be

considered demographically independent of the hatchery component. In such cases, hatchery

and wild spawning fish are part of the same population. A population that depends upon

naturally spawning hatchery fish for its survival is not viable by our definition (see discussion in

the Appendix sections “Population size,” p. 53 and “Population growth rate and related

parameters,” p. 64). In contrast, it is possible for hatchery-origin and naturally-produced adults

to spawn in the same stream but not be demographically linked to one another. In such cases, the

natural- and hatchery-origin groups of fish constitute separate populations. The natural fish

could be considered a viable population if they meet the VSP criteria.

For the purposes of this document, hatchery fish are defined as fish whose parents were spawned in a hatchery,

regardless of parental lineage, and natural fish are defined as fish whose parents spawned in the wild, regardless of

parental lineage. These are definitions for clarity only, and imply nothing about the risks or benefits of hatchery

programs.

PARAMETERS FOR EVALUATING POPULATIONS

Introduction to Parameters

Population Parameters

Four parameters form the key to evaluating population status. They are: abundance,

population growth rate, population spatial structure, and diversity. NMFS focuses on these

parameters for several reasons. First, they are reasonable predictors of extinction risk (viability).

Second, they reflect general processes that are important to all populations of all species. For

example, many factors influence abundance, (e.g., habitat quality, interactions with other species,

harvest programs, etc.). Many of these factors are species- or ESU-specific. By focusing on

abundance, we can seek general conclusions about an ESU's extinction risk even in the absence

of detailed, species-specific information on all of the factors that influence abundance. Third,

the parameters are measurable. The Appendix discusses specific methods of estimating

population status in the context of each parameter.

Several potential parameters, notably habitat characteristics and ecological interactions,

are not components NMFS uses to define population status, even though they are unquestionably

important to salmonid population viability. The reason these attributes (and others) are not part

of the viability criteria is that their effects are ultimately reflected in the four primary parameters

we do examine. Whenever possible, we discuss how these factors influence a specific

parameter. For example, a population’s spatial structure is to a large degree dictated by habitat

structure, and the spatial structure guidelines reflect this fact. Habitat characteristics and

ecological interactions both tend to be very species-specific, thus, it is well beyond the scope of

this present document to provide guidelines for these factors for every species and life-history

type. However, during the recovery planning process, it will be necessary to explore the explicit

relationships between habitat characteristics, ecological interactions and population parameters

within each ESU.

Guidelines for Each Population Parameter

In order to use the previously mentioned four population parameters to make viability

assessments, NMFS has developed a series of guidelines for each parameter. The guidelines are

drawn from a survey of the conservation biology and salmonid literature. These guidelines are

crude in the sense that they do not take into account the specifics of any particular species or

population. However they are also practical because in many situations, population-specific data

is not available, or a decision about pending action needs to be made before a detailed analysis

can be completed. In these situations, using guidelines may be the best way to evaluate a

population's status. To present the guidelines as concisely as possible, this section includes only

a brief overview of the rationale behind each of them. The bulk of the data, reasoning, and

examples used to create the guidelines are contained in the Appendix. It should be emphasized

that these guidelines are based on our interpretation of currently available data and literature. As

with all scientific endeavors, these guidelines can be modified as new data, more rigorous

analysis and clearer interpretations are generated.

Population Size

Small populations face a host of risks intrinsic to their low abundance; conversely, large

populations exhibit a greater degree of resilience. A large part of the science of conservation

biology involves understanding and predicting the effects of population size. All else being

equal, small populations are at greater risk of extinction than large populations primarily because

several processes that affect population dynamics operate differently in small populations than

they do in large populations. These processes are deterministic density effects, environmental

variation, genetic processes, demographic stochasticity, ecological feedback and catastrophes

(Appendix section “Population size,” p. 53). Deterministic effects of population density fall into

two opposing processes: compensation (an increase in productivity with decreasing density) and

depensation (a decrease in productivity with decreasing density). Compensation occurs because

there is an increasing need to compete for limited resources as a population expands to fill (or

exceed) available habitat. The negative relationship between productivity and abundance

observed under compensation can give a population substantial resilience. This resilience occurs

because any decline in abundance is offset by an increase in productivity, which tends to restore

a population to some equilibrium level.

A diverse suite of processes can cause depensatory density effects at small population

sizes. These include the inability of potential mates to find one another and increased predation

rates when predators are unsatiated (see Appendix section “Population growth rate and related

parameters,” p. 64). Depensatory processes at low population abundance (also termed “Allee”

effects) result in high extinction risks for very small populations because any decline in

abundance further reduces the population's average productivity, resulting in a steep slide toward

extinction. Environmental variation can cause small populations to go extinct when chance

events reduce survival or fecundity to low levels for an extended time. The genetic processes

that may negatively affect small populations include diversity loss, inbreeding depression and the

accumulation of deleterious mutations. Demographic stochasticity refers to random events

associated with mate choice, fecundity, fertility, and sex ratios that can create higher extinction

risks in small populations relative to large populations. Ecological feedback is similar to

density-dependent processes, but it emphasizes the role salmon play in modifying their physical

and biological environment and it usually operates at time lags absent from density-dependent

processes. Examples include the contribution of salmon carcasses to riparian zone nutrient

cycles, and the effect of spawning salmon on spawning gravel quality. Both of these processes

can contribute to the success of future salmon generations, but they are only significant at

relatively high population densities. Catastrophes are environmental events that severely reduce

a population size in a relatively short period of time. Because catastrophic events often affect

more than one population and the extinction risks associated with catastrophic failure can be

relatively independent of population size, the effects of catastrophes are considered in the section

on ESU-level viability.

We developed the following guidelines in order to assess population viability in light of

the abundance parameter. Note that the ESA=s primary focus is on natural populations in their

native ecosystems, so when we evaluate abundance to help determine VSP status, it is essential

to focus on naturally produced fish (i.e., the progeny of naturally spawning parents). Because

extinction risk depends largely on specific life-history strategies and the local environment,

setting fish abundance criteria will require application of species or population specific

information. For this reason, the following guidelines prescribe factors that need to be

considered but do not provide specific numerical criteria.

Two sets of described guidelines that follow are: Viable Size Guidelines and Critical Size

Guidelines. (Note that these levels are not equivalent to the ESA concepts of Asurvival@ and

Arecovery”; see Appendix section “Applying VSP in the regulatory arena,” p. 33, for more

discussion). A population must meet all of the viable population guidelines to be considered

viable with respect to this parameter. If a population meets even one critical guideline, it would

be considered to be at a critically low level. Also, note that different guidelines are likely to

dominate decisions for different populations. For example, environmental variation (Viable

Guideline 1) will often dictate a larger minimum population size than would genetic concerns

(Viable Guideline 3), but for some populations genetic concerns may predominate.

Population Growth Rate and Related Parameters

Population growth rate (productivity

) and factors that affect population growth rate

provide information on how well a population is “performing” in the habitats it occupies during

the life cycle. These parameters, and related trends in abundance, reflect conditions that drive a

population’s dynamics and thus determine its abundance. Changes in environmental conditions,

including ecological interactions, can influence a population's intrinsic productivity or the

environment's capacity to support a population, or both. Such changes may result from random

environmental variation over a wide range of temporal scales (environmental stochasticity). In

this section, however, we are most concerned with measures of population growth and related

parameters that reflect systematic changes in a population's dynamics.

We focus on population growth rate and related parameters as integrated indicators of a

population’s performance in response to its environment. Specific characteristics of a

population’s environment that affect its dynamics, while likely to be similar across populations,

are necessarily deferred to individual case studies. In most cases we are concerned with

estimating a mean parameter that describes some aspect of population dynamics (such as long-

term population growth rate) and with estimating the variance of this parameter. Depending on

the question or parameter of interest, estimates of variance may contribute to descriptions of

uncertainty in parameter estimates, and consequences of decisions based on such estimates may

play an integral role in evaluating the viability of a population. While it is intuitively sensible to

We use the terms “population growth rate” and “productivity” interchangeably when referring to production over

the entire life cycle. We also refer to “trend in abundance” which is simply the manifestation of long-term

population growth rate.

Viable Population Size Guidelines

1. A population should be large enough to have a high probability of surviving

environmental variation of the patterns and magnitudes observed in the past and

expected in the future. Sources of such variation include fluctuations in ocean conditions

and local disturbances such as contaminant spills or landslides. Environmental variation and

catastrophes are the primary risks for larger populations with positive long-term average

growth rates.

2. A population should have sufficient abundance for compensatory processes to provide

resilience to environmental and anthropogenic perturbation. In effect, this means that

abundance is substantially above levels where depensatory processes are likely to be

important (see Critical Guideline 1 as follows) and in the realm where compensation is

substantially reducing productivity. This level is difficult to determine with any precision

without high quality long-term data on population abundance and productivity, but can be

approximated by a variety of methods.

3. A population should be sufficiently large to maintain its genetic diversity over the long

term. Small populations are subject to various genetic problems, including loss of genetic

variation, inbreeding depression, and deleterious mutation accumulation, that are influenced

more by effective population size than by absolute abundance.

4. A population should be sufficiently abundant to provide important ecological functions

throughout its life-cycle. Salmonids modify both their physical and biological

environments in various ways throughout their life cycle. These modifications can benefit

salmonid production and improve habitat conditions for other organisms as well. The

abundance levels required for these effects depend largely on the local habitat structure and

particular species’ biology.

5. Population status evaluations should take uncertainty regarding abundance into

account. Fish abundance estimates always contain observational error, and therefore

population targets may need to be much larger than the desired population size in order to be

confident that the guideline is actually met. In addition, salmon are short-lived species with

wide year-to-year abundance variations that contribute to uncertainty about average

abundance and trends. For these reasons, it would not be prudent to base abundance criteria

on a single high or low observation. To be considered a VSP, a population should exceed

these criteria on average over a period of time.

Critical Population Size Guidelines

1. A population would be critically low if depensatory processes are likely to reduce it

below replacement. The specific population levels where these processes become important

are difficult to determine, although there is theory on mate choice, sex-ratios, and other

population processes that may be helpful in placing a lower bound on safe population levels.

In general, however, small-population depensatory effects depend largely on density rather

than absolute abundance. A species= life-history and habitat structure play large roles in

determining the levels at which depensation becomes important.

2. A population would be critically low if it is at risk from inbreeding depression or

fixation of deleterious mutations. The most important genetic risks for very small

populations are inbreeding depression and fixation of deleterious mutations; these effects are

influenced more by the effective breeding population size than by absolute numbers of

individuals.

3. A population would be critically low in abundance when productivity variation due to

demographic stochasticity becomes a substantial source of risk. Demographic

stochasticity refers to the seemingly random effects of variation in individual survival or

fecundity that are most easily observed in small populations. As populations decline, the

relative influences of environmental variation and demographic stochasticity changes—with

the latter coming to dominate in very small populations.

4. Population status evaluations should take uncertainty regarding abundance into

account. Fish abundance estimates always contain observational error and therefore

population targets may need to be much larger than the desired population size in order to be

confident that the guideline is actually met. In addition, salmon are short-lived species with

wide year-to-year abundance variations that contribute to uncertainty about average

abundance and trends. For these reasons, it would not be prudent to base abundance criteria

on a single high or low observation. To be considered critically low, a population would fall

below these criteria on average over a short period of time.

use population growth rate as an indicator of risk and viability, the issue of how to do so in a

quantitative way is still an area of active research.

Estimates of population growth rate (i.e., productivity over the entire life cycle) that

indicate a population is consistently failing to replace itself, are an indicator of increased

extinction risk, no matter what the cause. Some evidence suggests that the major extinction risk

for Pacific salmonids does not arise from stochastic processes but rather from processes (such as

habitat degradation or overharvest) that exert a sustained detrimental effect on a population and

result in a chronically low population growth rate and a negative trend in abundance (Emlen

1995, Ratner et al. 1997). Under this scenario, small population size is a transient stage toward

deterministic extinction. While stochastic processes certainly affect the time to extinction, they

do not affect the likelihood of the outcome.

Although our overall focus is on population growth rate over the entire life cycle, estimates of

stage-specific productivity (particularly productivity during freshwater life-history stages) are

also important for comprehensive evaluation of population viability. Although declines in stage-

specific productivity may not immediately manifest in reduced abundance if offset during other

portions of the life cycle, they may indicate reduced resilience to variation in productivity

elsewhere in the life cycle. As examples, estimates of smolt production provide a measure of

both a population’s potential to increase in abundance (should the recent poor ocean conditions

abate) and a population’s ability to weather future periods of poor ocean conditions. Along

similar lines, changes or shifts in traits that are clearly related to productivity (such as size-at-

return of spawners) may contribute to evaluations of population viability. Such ancillary data

may provide an important source of information that supports more rapid detection of changes in

conditions affecting population growth rate in salmonid populations.

Other measures of population productivity, such as intrinsic productivity and the intensity

of density-dependence, may provide important information for assessing a population’s viability.

For example, a population's intrinsic productivity (defined as its maximum growth rate when free

of density-dependent limitation) partially determines the abundance at which demographic

stochasticity begins to play an important role in determining the fate of the population. Intrinsic

productivity is also a measure of a population's ability to rebound from short-term environmental

or anthropogenic perturbations (resilience). Analyses of extinction models have highlighted the

importance of elucidating the existence and intensity of compensatory density-dependence for

estimating a population's extinction risk (Ginzberg et al. 1990). However, obtaining unbiased

estimates of a population’s intrinsic productivity and the intensity and functional form of

density-dependence affecting a population can be difficult due to autocorrelation and observation

error in the data. Such difficulties hinder the use of these parameters as independent measures of

a population’s viability.

Not surprisingly, the guidelines for population growth rate and related parameters are

closely linked with those for abundance; indeed, the following guidelines are often conditioned

on a population's status in terms of abundance.

Population Growth Rate and Related Parameters Guidelines

1. A population’s natural productivity should be sufficient to maintain its abundance above the

viable level. A population meeting or exceeding abundance criteria for viability should, on average,

be able to replace itself. That is, spawner: spawner ratios or cohort-replacement ratios should

fluctuate around 1.0 or above. Natural productivity is typically measured as the ratio of naturally-

produced spawners born in one broodyear to the number of fish spawning in the natural habitat during

that broodyear; population abundance estimates at other life-history stages may also be used,

provided such estimates span the entire life cycle (e.g., smolt to smolt estimates).

2. A viable salmonid population that includes naturally spawning hatchery fish should exhibit

sufficient productivity from naturally-produced spawners to maintain population abundance at

or above viability thresholds in the absence of hatchery subsidy. In a strict sense, this guideline

suggests that the mean Natural Return Ratio (NRR) for a viable population should fluctuate around

1.0, indicating negligible hatchery influence on the population. In a practical sense, the requirement

that a viable population be demographically independent of a hatchery population suggests that a

viable population’s mean NRR not be less than approximately 0.9, but this estimate neglects other

issues related to the influence of hatchery fish on natural production. A viable population should not

exhibit a trend of proportionally increasing contributions from naturally spawning hatchery fish.

3. A viable salmonid population should exhibit sufficient productivity during freshwater life-

history stages to maintain its abundance at or above viable thresholds—even during poor ocean

conditions. A population’s productivity should allow it both to exploit available habitat and exhibit a

compensatory response at low population sizes. When spawner abundance is below the long-term

mean, there should be a corresponding increase in per capita smolt production, even though such an

increase may not suffice to offset declines in marine survival.

4. A viable salmonid population should not exhibit sustained declines in abundance that span

multiple generations and affect multiple broodyear-cycles. “Sustained” declines are those that

continue longer than the typical lag in response associated with a population’s generation time. Thus,

sustained declines differ from rapid transitions between one stable level and another (e.g., changes in

abundance related to large-scale, low frequency environmental forcing such as those related to

oceanic regime shifts). They also differ from short-term, severe perturbations in abundance, such as

those related to strong El Niño events that are followed by relatively rapid recovery.

5. A viable salmonid population should not exhibit trends or shifts in traits that portend declines

in population growth rate. Changes in such traits, such as size and age of spawners, that affect

population growth rate are often more easily and precisely quantified than are changes in abundance

and thus, may provide earlier indication of declining population growth rate. For example, reduced

size of mature individuals in a population may indicate reduced fecundity, lessened ability to reach

spawning grounds, a decreased capacity for constructing redds that are deep enough to resist bed

scour, or other factors that contribute to reduced production of offspring. Likewise, increasing age-at-

return may reduce a population’s intrinsic productivity by exposing adults to greater pre-reproductive

spawning risk.

6. Population status evaluations should take into account uncertainty in estimates of population

growth rate and productivity-related parameters. To estimate long-term trends and spawner-

recruit ratios, it is important to have an adequate time series of abundance. Unfortunately, such time

series, when they exist at all, are often short, contain large observational errors, or both. These

constraints may greatly limit the power of statistical analyses to detect ecologically significant trends

before substantial changes in abundance have occurred.

Spatial Structure

When evaluating population viability, it is important to take within-population spatial

structure needs into account for two main reasons: 1) because there is a time lag between

changes in spatial structure and species-level effects, overall extinction risk at the 100-year time

scale may be affected in ways not readily apparent from short-term observations of abundance

and productivity, and 2) population structure affects evolutionary processes and may therefore

alter a population’s ability to respond to environmental change (Appendix section “Spatial

Structure,” p. 90).

A population’s spatial structure is made up of both the geographic

distribution of individuals in the population and the processes that generate that distribution.

Spatially structured populations are often generically referred to as “metapopulations,” though

the term metapopulation has taken on a number of different meanings. A population’s spatial

structure depends fundamentally on habitat quality, spatial configuration, and dynamics as well

as the dispersal characteristics of individuals in the population.

Because many of the processes that affect small population extinction risk (e.g., genetic

risks and demographic stochasticity) depend primarily on breeding structure, we will focus on

spawning group distribution and connectivity. Restricting the discussion to spawning groups is

practical for the purposes of evaluating viability, but spatial structure exists during all life-history

stages. Although we focus on breeding structure, it should be noted that with little or no

modification, many of our guidelines could apply to non-breeding spatial structure.

Because freshwater habitat is often quite heterogeneous, salmonids may experience

spawning habitat as discrete “patches” whose quality can vary from highly productive to

unusable. If straying among patches is limited, the population may be divided into

subpopulations. (“Straying” occurs when an individual spawns in a different patch from the one

in which it hatched.

) Straying rates form a continuum and, by definition, levels of straying

among subpopulations are higher than those found between populations. The spatial scales that

define habitat patches and subpopulation boundaries are not strictly defined because such

determinations are likely to be species- and population-specific.

As one example of how a degraded spatial structure can threaten the viability of a population, consider a

population divided into subpopulations. A population with a high subpopulation extinction rate can persist only if

new subpopulations are founded at a rate equal to the rate at which subpopulations naturally go extinct. If human

activity interferes with the formation of new subpopulations by restricting straying patterns or destroying habitat

patches suitable for colonization, the population will ultimately go extinct as subpopulations blink out one by one.

However, there will be a time lag between the disruption of spatial processes and reductions in the abundance or

productivity of the population because abundance will not necessarily decline until subpopulations start going

extinct.

The term “straying” is commonly used in the salmonid literature as we have defined it here. Despite the negative

connotations of the word “straying,” there is no reason to assume that straying is evolutionarily maladaptive. For

non-salmonids, the terms “dispersal” or “migration” are generally used to describe movement away from natal

habitats, resulting in breeding in a non-natal location. However, in the salmonid literature dispersal and migration

often refer to any movement among habitats and straying is the only term restricted to breeding structure.

A number of different population structures are possible—depending on habitat patch

physical stability, subpopulation demographic stability, and the levels of straying among patches

(Figure 2). Although there are few data on salmonid breeding spatial structure, what data are

available suggest that the habitat patch size and quality vary greatly within a population. This

indicates that source-sink or island-mainland dynamics might be common. In contrast, there are

few data to support a panmictic or classical metapopulation view of salmonid population spatial

structure (see Appendix section “Spatial Structure,” p. 90 for definitions of these terms). It

should be emphasized, however, that salmonid spatial structure is not well understood, and there

is currently no scientific consensus on what a “typical” spatial structure is. The following

guidelines focus on key processes that are likely to be important in maintaining a viable spatial

structure, regardless of population type.

Diversity

Several salmonid traits exhibit considerable diversity within and among populations.

This variation has important effects on population viability (Appendix section “Diversity,” p.

101). Some of these varying traits are anadromy, morphology, fecundity, run timing, spawn

timing, juvenile behavior, age at smolting, age at maturity, egg size, developmental rate, ocean

distribution patterns, male and female spawning behavior, physiology and molecular genetic

characteristics. Of these traits, some (such as DNA or protein sequence variation) are completely

genetically based, whereas others (such as nearly all morphological, behavioral, and life-history

traits) usually vary as a result of a combination of genetic and environmental factors.

In a spatially and temporally varying environment, there are three general reasons why

diversity is important for species and population viability. First, diversity allows a species to use

a wider array of environments than they could without it. For example, varying adult run and

spawn timing allows several salmonid species to use a greater variety of spawning habitats than

would be possible without this diversity. Second, diversity protects a species against short-term

spatial and temporal changes in the environment. Fish with different characteristics have

different likelihoods of persisting—depending on local environmental conditions. Therefore, the

more diverse a population is, the more likely it is that some individuals would survive and

reproduce in the face of environmental variation. Third, genetic diversity provides the raw

material for surviving long-term environmental changes. Salmonids regularly face cyclic or

directional changes in their freshwater, estuarine, and ocean environments due to natural and

human causes, and genetic diversity allows them to adapt to these changes.

Any actions that affect basic demographic and evolutionary processes (e.g., patterns of

mutation, selection, drift, recombination, migration, and population turnover) have the potential

to alter a species’ diversity. For example, straying and gene flow among populations strongly

influence diversity within and among populations. Gene flow refers to the movement of genes

from one population to another and results from strays that successfully reproduce. There are a

number of ways in which human actions could substantially alter patterns of straying and,

therefore, potentially alter patterns of diversity and adaptation among salmonid populations. For

example, blocking migration corridors with dams and dewatering rivers can prevent salmonids

from homing and thus increase the rate of straying into other populations.

homogeneous panmictic population

patchy panmictic population

“classical”

metapopulation

Low survival years -

single panmictic

population

Moderate survival year -

source sink population

High survival years -

structured population

Panel B: Temporal changes in

population structure

Panel A: “Traditional”

classification of population

structure

Panel C: Static vs. dynamic habitat

Static habitat

Dynamic habitat

Time 1 Time 3Time 2

structured populaton

with no extinctions

source-sink population

island-mainland population

Figure 2. Theoretical types of spatially structured populations. Panel A shows a “traditional” type

classification scheme that does not consider correlated environmental effects that impact all

subpopulations nor does it explicitly consider the physical dynamics of the habitat patches

themselves. The circles indicate habitat patches, with the size of the circle indicating the size or

capacity of the patch, and the degree of shading indicating the density of the subpopulation—

white indicating an empty patch and black indicating a high density patch. The arrows indicate

levels of migration, with thick arrows indicating high migration; thin arrows moderate migration,

and dashed arrows indicating intermittent migration. Panel B shows how spatial structure may

oscillate over time as a result of correlated environmental changes in survival or productivity

among subpopulations. Correlated environmental changes might result, for example, from annual

variation in ocean survival that affects all subpopulations. Panel C shows two potential habitat

patterns. In a static habitat, the location of suitable patches remains constant over time, though

patches may or may not always be occupied. In a dynamic habitat, the location of suitable habitat

continually changes, and so the location of subpopulations must also change.

Spatial Structure Guidelines

1. Habitat patches should not be destroyed faster than they are naturally created. Salmonid

habitat is dynamic, with suitable habitat being continually created and destroyed by natural processes.

Human activities should not decrease either the total area of habitat OR the number of habitat patches.

This guideline is similar to the population growth rate criterion—i.e., a negative trend has

deterministically negative affects on viability—though the relationship between decreasing number of

patches and extinction risk is not necessarily linear.

2. Natural rates of straying among subpopulations should not be substantially increased or

decreased by human actions. This guideline means that habitat patches should be close enough

together to allow appropriate exchange of spawners and the expansion of the population into under-

used patches, during times when salmon are abundant (see Guideline 3). Also, stray rates should not

be much greater than pristine levels, because increases in stray rates may negatively affect a

population’s viability if fish wander into unsuitable habitat or interbreed with genetically unrelated

fish.

3. Some habitat patches should be maintained that appear to be suitable or marginally suitable,

but currently contain no fish. In the dynamics of natural populations, there may be time lags

between the appearance of empty but suitable habitat (by whatever process) and the colonization of

that habitat. If human activity is allowed to render habitat unsuitable when no fish are present, the

population as a whole may not be sustainable over the long term.

4. Source subpopulations should be maintained. Some habitat patches are naturally more productive

than others. In fact, a few patches may operate as highly productive source subpopulations that

support several sink subpopulations that are not self-sustaining. Protecting these source patches

should obviously be of the highest priority. However, it should be recognized that spatial processes

are dynamic and sources and sinks may exchange roles over time

5. Analyses of population spatial processes should take uncertainty into account. In general, there

is less information available on how spatial processes relate to salmonid viability than there is for the

other VSP parameters. As a default, historic spatial processes should be preserved because we

assume that the historical population structure was sustainable but we do not know whether a novel

spatial structure will be.

Diversity Guidelines

1. Human-caused factors such as habitat changes, harvest pressures, artificial propagation, and

exotic species introduction should not substantially alter variation in traits such as run timing,

age structure, size, fecundity, morphology, behavior, and molecular genetic characteristics.

Many of these traits may be adaptations to local conditions, or they may help protect a population

against environmental variation. A mixture of genetic and environmental factors usually causes

phenotypic diversity, and this diversity should be maintained even if it cannot be shown to have a

genetic basis.

2. Natural processes of dispersal should be maintained. Human-cased factors should not

substantially alter the rate of gene flow among populations. Human caused inter-ESU stray rates

that are expected to produce (inferred) sustained gene flow rates greater than 1% (into a population)

should be cause for concern. Human caused intra-ESU stray rates that are expected to produce

substantial changes in patterns of gene flow should be avoided.

3. Natural processes that cause ecological variation should be maintained. Phenotypic diversity can

be maintained by spatial and temporal variation in habitat characteristics. This guideline involves

maintaining processes that promote ecological diversity, including natural habitat disturbance regimes

and factors that maintain habitat patches of sufficient quality for successful colonization.

4. Population status evaluations should take uncertainty about requisite levels of diversity into

account. Our understanding of the role diversity plays in Pacific salmonid viability is limited.

Historically, salmonid populations were generally self-sustaining, and the historical representation of

phenotypic diversity serves as a useful “default” goal in maintaining viable populations.

Integrating the Parameters and Determining Population Status

The key parameters used to determine whether a population is likely to enjoy long-term

viability are abundance, population growth rate, population spatial structure, and diversity. As a

rule, these four population attributes cannot be viewed entirely independently. For example, the

value ranges for population growth rate considered necessary for a viable population clearly

depend on the population’s abundance (Appendix section “Population growth rate and related

parameters,” p. 64). Making an overall determination of population status will usually require

some method of integrating the parameters.

We believe the guidelines for each of the four parameters (p. 14, 15, 17, and 21) describe

the actions needed to maintain salmonid population viability over a 100-year period.

Nevertheless, because the guidelines we provide are qualitative, modeling tools (such as

population viability analyses) may prove useful in exploring tradeoffs between parameters such

as population size and productivity. For example, a model could be constructed that examines

the extinction dynamics of a salmonid population under different population size scenarios and

spawner-to-recruit ratios in order to combine a population’s characteristics and estimate their

effects on viability. Similarly, a structured modeling approach could explore the effects on

population viability that different subpopulation numbers, abundance, and distribution would

have.

ESU VIABILITY

Introduction

To help understand basic salmonid biology and formulate priorities for salmon

management, it is useful to explore the extinction risks that individual population’s experience.

However, in many cases it is also important to consider how these individual population risks

relate to sustainability of larger conservation units such as Genetic Diversity Management Units

(GDMU, defined by WDFW), Major Ancestral Lineages (defined by WDFW), or Gene

Conservation Groups (defined by ODFW). The ESA is ultimately concerned with the extinction

of any entity that qualifies as a “species” under the ESA, which for vertebrate animals includes

species, subspecies, or “distinct population segments.” According to NMFS policy, groups of

salmon populations that represent Evolutionarily Significant Units (ESUs) of the species as a

whole are considered distinct population segments and hence, “species” under the ESA. The

goal of this section of the document is to discuss factors that should be evaluated when

determining the numbers and distribution of VSPs needed to sustain larger conservation units

such as ESUs. Like the section on population viability, this section describes only qualitative

guidelines for determining ESU viability. The assumption is that in order to determine the

necessary population numbers and distribution in a given ESU, case-specific information will be

required. Appendix section “ESU Viability Guidelines” (p. 127) provides a more detailed

rationale for the guidelines.

Number and Distribution of Populations in a Recovered ESU

By definition, a VSP has a negligible risk (over a time scale of 100 years) of going

extinct as a result of genetic change, demographic stochasticity, or normal levels of

environmental variability. If these were the only factors influencing viability, a single VSP

would be enough to ensure the survival of an entire ESU. However, three additional factors need

to be considered in relating VSPs to viable ESUs: 1) catastrophic events, 2) long-term

demographic processes, and 3) long-term evolutionary potential. Catastrophic events are sudden,

rare occurrences that severely reduce or eliminate an entire population. These events fall outside

the scope of the “normal” temporal and spatial scales of environmental variation considered

when evaluating VSPs. Long-term demographic processes involve extinction and recolonization

of entire populations at time scales greater than 100 years. The concern about long-term

evolutionary potential centers on the role diversity plays in ESU viability over time scales that

are generally greater than 100 years.

In addition to biological considerations, the ESA defines an endangered species as a

species that is “…in danger of extinction in all or a significant portion of its range.” The ESA

does not define “significant portion of the range” and there is no indication that the phrase is

meant to be (or can be) defined entirely in scientific terms. As noted in the appendix section

“Applying VSP in the regulatory arena” (p. 33), it will probably be necessary to define “a

significant portion of the range” in both scientific and policy terms. Scientifically, a significant

ESU Viability Guidelines

1. ESUs should contain multiple populations. If an ESU is made up of multiple populations, it is less

likely that a single catastrophic event will cause it to become extinct. Also, ESUs may function as

“metapopulations” over the long term and the existence of multiple populations would be necessary

for the operation of sustainable population-level extinction/recolonization processes. In addition,

multiple populations within an ESU increase the likelihood that a diversity of phenotypic and

genotypic characteristics will be maintained, thus allowing natural evolutionary processes to operate

and increasing the ESU’s viability in the long term. Obviously, this guideline does not apply to ESUs

that appear to contain a single population (e.g., Lake Ozette sockeye). In ESUs containing a single

population Guideline 6 becomes increasingly important.

2. Some populations in an ESU should be geographically widespread. Spatially correlated

environmental catastrophes are less likely to drive a widespread ESU to extinction. This guideline

also directly relates to the ESA mandate of protecting a species in a “significant portion of (its)

range.”

3. Some populations should be geographically close to each other. On long temporal scales, ESUs

may function as “metapopulations” and having populations geographically close to one another

facilitates connectivity among existing populations. Thus, a viable ESU requires both widespread

(Guideline 2) AND spatially close populations.

4. Populations should not all share common catastrophic risks. An ESU containing populations that

do not share common catastrophic risks is less likely to be driven to extinction by correlated

environmental catastrophes. Maintaining geographically widespread populations is one way to

reduce risk associated with correlated catastrophes (Guideline 2), but spatial proximity is not the only

reason why two populations could experience a correlated catastrophic risk.

5. Populations that display diverse life-histories and phenotypes should be maintained. When an

ESU’s populations have a fair degree of life-history diversity (or other phenotypic diversity), the ESU

is less likely to go extinct as a result of correlated environmental catastrophes or changes in

environmental conditions that occur too rapidly for an evolutionary response. In addition, assuming

phenotypic diversity is caused at least in part by genetic diversity, maintaining diversity allows

natural evolutionary processes to operate within an ESU.

6. Some populations should exceed VSP guidelines. Larger and more productive (“resilient”)

populations may be able to recover from a catastrophic event that would cause the extinction of a

smaller population. An ESU that contains some populations in excess of VSP threshold criteria for

abundance and population growth rate is less likely to go extinct in response to a single catastrophic

event that affects all populations. It is important to note that the abundance guidelines do not take

catastrophes into account. This guideline is particularly relevant if an ESU consists of a single

population.

7. Evaluations of ESU status should take into account uncertainty about ESU-level processes. Our

understanding of ESU-level spatial and temporal process is very limited. ESUs are believed to have

been historically self-sustaining and the historical number and distribution of populations serves as a

useful “default” goal in maintaining viable ESUs.

portion of the range will be determined by evaluating the risks to ESU persistence at a number of

time scales, including those longer than 100 years.

Populations Not Meeting VSP Guidelines

After conducting a viability analysis, a group of fish considered to be an independent

population under the VSP definition might not appear to meet the VSP guidelines. There are a

number of reasons why a population may appear non-viable. Some of these reasons involve a

correct assessment of population status, while others involve some error in assessment.

Understanding why a population is not considered viable, and the potential mistakes in reaching

that conclusion is important in assessing ESU viability.

A group of fish could be mislabeled as not viable if it has been misidentified as an

independent population and should actually be considered to be a subpopulation of a larger

group. Population and subpopulation designations fall along a continuum of reproductive

isolation, and in borderline cases it is difficult to determine population structure for a group of

fish (see “Identifying populations,” p. 51). Ambiguity about population structure may often be a

challenging reality in evaluating ESU-level status.

Even if population boundaries are appropriately identified, a population may still be

inaccurately characterized as inviable. Evaluating population status with limited data is a

difficult task and errors are inevitable. This is particularly true if a population is in some way

unusual and comparing it to other salmonid populations would be misleading.

Despite challenges assessing population boundaries and population viability, the

conclusions that a population is not viable may be correct. In many areas, human activities have

so ffected individual survival and fecundity that it is difficult or impossible for an independent,

viable population to become established.

An interesting situation exists when a population is independent, but would not be

considered viable (i.e., it would not have a negligible risk of extinction in 100 years), even under

pristine conditions. There are a number of reasons why independent populations might not be

viable, even under pristine conditions. For example, a population may be particularly prone to

catastrophic events or large environmental fluctuations, as would be expected on the margins of

a species’ range. Alternatively, a population may be so small naturally that it is unlikely to

persist for 100 years. Truly independent but inviable populations may periodically be generated

by strays from viable populations, and these relatively ephemeral populations would be part of

ESU-level metapopulation processes. These types of populations will have to be taken into

account on a case-by-case basis when an ESU’s status is being evaluated. They may be quite

important to an ESU’s viability if it functions as a large metapopulation with population turnover

occurring at relatively short time scales. In an ESU consisting primarily of populations that

would be viable under pristine conditions, the potentially viable populations would likely be

given highest conservation priority.

IMPLEMENTING THE VSP GUIDELINES

Introduction

ESA recovery plans are required to provide objective, measurable criteria for determining

when delisting is warranted (ESA Sec. 4[f]). Delisting criteria are also useful in performing

other ESA related activities, such as consultations, permits, habitat conservation plans, and

regulations (see “Applying VSP in the regulatory arena,” p. 33). The VSP guidelines presented

in this document are intended to guide the development of specific delisting/recovery criteria for

Pacific salmon ESUs.

The guidelines presented here are intentionally general so they can be applied equally

across the wide spectrum of life-history diversity, habitat conditions, and meta-population

structures represented by Pacific salmon. It would be irresponsible to propose specific numerical

criteria to be applied uniformly across all Pacific salmon ESUs; specific critieria should take into

account species-specific life-history traits and local habitat structure. For example, population

structure and abundance criteria for sockeye salmon spawning in a north Washington coastal

lake would be expected to differ substantially from those for steelhead spawning in intermittent

streams on the Southern California coast. In addition, the type and amount of information

available varies considerably by species and region, which will influence the emphasis placed on

the various guidelines in specific applications.

Practical Application

Given the generality of these guidelines, it is important to consider how they will be

applied in practice to develop biological delisting criteria. NMFS envisions this to be a three-

step process:

1) Define populations within each ESU.

2) Identify VSP criteria for each population.

3) Identify ESU-wide delisting criteria based on the VSP population criteria.

These steps are discussed in a draft guidance document for technical recovery teams (NMFS

2000a).

Uncertainty, Precaution, and Adaptive Management

A main concern in translating the guidelines into specific criteria will be the degree of

uncertainty in much of the relevant information. There are two widely recognized principles for

approaching resource conservation under uncertainty that should be applied in application of

VSP: the precautionary approach and adaptive management.

The precautionary approach requires managers to implement conservation measures even

in the absence of scientific certainty regarding risks. This approach is widely recommended in

harvest management. For example, NMFS' National Standard Guidelines for the Magnuson-

Stevens Fishery Conservation and Management Act (50 CFR Part 600, FR 63:24212-24237)

specifiy use of a precautionary approach both for the specification of optimum yield (OY,

National Standard 1) and in managing bycatch (National Standard 9). The precautionary

approach is also clearly defined in the FAO Code of Conduct for Responsible Fisheries (see

discussion in Ch. 5 of Committee on Ecosystem Management for Sustainable Marine Fisheries

1999). We believe this approach to be equally appropriate in other management areas. An

application to habitat is discussed briefly by the Committee on Protection and Management of

Pacific Northwest Anadromous Salmonids (1996). In practical terms, a precautionary approach

results in shifting the burden of proof, putting that burden in favor of resource conservation

rather than direct economic benefit. For quantitative criteria, this can be accomplished through

careful consideration of appropriate levels of statistical confidence and power in comparing

estimated population parameters to criteria.

Adaptive management encourages the design of management policies designed to

improve the resource knowledge base through active or passive experimentation (Holling 1978,

Walters 1986). General principles for designing adaptive policies are described by Walters

(1986). The Committee on Protection and Management of Pacific Northwest Anadromous

Salmonids (1996) discussed specific applications to salmon management under the ESA.

Interim Application

We have emphasized the role of technical recovery teams in establishing specific criteria

for listed ESUs. However, management actions still need to be taken while recovery teams are

being formed and recovery goals are being developed. It may be useful for agencies involved in

such actions to establish interim criteria that will be used until recovery plans are adopted. Such

interim criteria should be based on a careful consideration of the precautionary principle and

adaptive management methods. Often, a full review of population structure will not be possible

within management time frames. In such situations, agencies might adopt population definitions

already available in state wild fish conservation reports or similar stock-based management

plans, provided these definitions are reasonably consistent with the VSP population definition.

Interim population definitions and criteria should be based on precautionary application of the

VSP criteria. Strong precaution at this stage will help ensure a rapid start along the road to

recovery and that no significant parts of the ESU are lost before the full recovery plan is

implemented. At the ESU level, we suggest that during this interim period, actions should be

taken such that all populations with the ESU retain the potential to achieve viable status. This

would ensure that all parts of the system are maintained until a final plan establishes ESU-level

criteria. Adaptive management should be emphasized to provide improved information and to

allow for changing goals and management strategies as recovery plans are developed.

Examples

The VSP guidelines are currently being applied in formal ESA recovery planning by the

Puget Sound and Willamette/Lower Columbia technical recovery teams, but products from these

groups are not yet available. (The only example of applying these guidelines is the work of a

multi-agency work group preparing a “Quantitative Analytical Report” for upper Columbia

River steelhead and spring chinook salmon [Ford et al. 1999a].)

APPENDIX

Applying VSP in the Regulatory Arena

This document focuses on scientific/biological aspects of populations and more complex

conservation units, but it is also intended to be useful for a wide variety of applications in the

conservation and management of salmonids—at local, state, national, and international levels.

Typically these applications require relating biological principles to regulatory language

involving terms such as “survival,” “recovery,” and “extinction.” For example, implementation

of the federal ESA requires determining when listed species are at risk of extinction or

endangerment, when they have reached a level at which they can be delisted, and when they are

likely to be jeopardized by a proposed action. This document defines another term, population

viability, as a negligible probability of extinction over a 100-year time frame. Although 100

years is a commonly-used time horizon for evaluating extinction risk (or persistence probability)

in the field of conservation biology, it does not relate directly to any particular regulatory

language described previously. It is beyond the scope of this document to make this formal

linkage. Nevertheless, it is useful to briefly review the key regulatory language for federal ESA

implementation and to comment in general on how such linkages might be made.

Listing Criteria

The ESA (“Number and distribution of populations in a recovered ESU,” p. 25) defines

an “endangered” species as “any species which is in danger of extinction throughout all or a

significant portion of its range.” A “threatened” species is “any species which is likely to

become an endangered species within the foreseeable future throughout all or a significant

portion of its range.” From the perspective of applying VSP to entire ESUs, the relevant terms

are “likely to become,” “foreseeable future,” and “significant portion of its range.” Neither

NMFS nor the U.S. Fish and Wildlife Service has defined any of these terms in a policy context.

Thompson (1991) reviewed existing extinction models and rules of thumb for evaluating

population status and noted that although selecting values for extinction probability (p) and time

horizon (t) is only partially a biological exercise (and therefore there is no intrinsically “right”

answer), conservation biologists commonly choose the values p = 0.5 and t = 100 years to

represent an endangerment threshold. Thompson also suggested that a reasonable interpretation

of a “likely” event is one that has at least a 50% chance of occurring. Defining “foreseeable

future” is less intuitive. Thompson suggested that something on the order of 10 years might be

appropriate, but a number of other interpretations are possible.

Because ESUs are considered “species” under the ESA, and a species can be listed if it is

threatened or endangered in “all or a significant portion of its range”, it is essential to take the

meaning of this phrase into consideration when applying VSP guidelines to entire ESUs. The

common scientific usage of “statistical significance” does not appear to be pertinent here; rather,

the relevant meaning of “significant” must be more along the lines of “important; of

consequence” (Random House Dictionary, 2nd Edition). “Range” has an obvious geographic

interpretation, and the sections of this document that discuss population abundance and

geographic distribution are relevant in this context. In addition, we believe it is important to

consider other aspects of the “range” concept when evaluating ESU viability; these might

include ecological diversity, life-history diversity, and genetic diversity. The “Diversity” section

(p. 19) discusses long-term ecological and evolutionary processes and thus is directly relevant to

this concept of “range.” In summary, because the process of determining what constitutes a

“significant portion” of a species’ range is only partially based on biological considerations, the

technical definitions of these terms are of limited use.

Recovery

NMFS and the U.S. Fish and Wildlife Service define recovery under the Act to be

“improvement in the status of a listed species to the point at which listing is no longer

appropriate” under the ESA (50 CFR S402.02). This indicates that there is a strong connection

between listing and delisting criteria. Thus, the biological criteria used to make listing

determinations should also be used to evaluate recovery. However, there are several reasons

why listing and delisting criteria values should not be identical. First, if simple threshold values

were used for setting criteria, a species that fluctuates around the critical value might require

repeated listing and delisting actions even though its status had changed relatively little. Second,

listing and delisting criteria require that population trends (and other factors) are considered in

addition to abundance, and trends are expected to differ between declining and recovering

populations. Finally, delisting should occur only in conjunction with an approved,

comprehensive recovery plan that lays out conservation measures that address the factors that led

to the initial decline and those that impede recovery. The preceding discussion applies to

recovery as the ESA defines it. In addition, the NMFS, states, tribes, and many other

stakeholders have an interest in recovering salmon populations to the point at which they can

support sustainable harvest or other “broad sense” recovery goals (e.g., to produce fully

functional ecosystems). Recognizing this, NOAA has made the following commitment (Garcia

1998):

It is our policy that the recovery of salmonid populations must achieve two goals: 1) Restore

salmonid populations to the point where they no longer require the protection of the ESA, and

2) restore salmonid populations to a level that allows meaningful exercise of tribal fishing rights.

We see no conflict between the statutory goals of the ESA and the federal trust responsibilities to

Indian tribes. Rather, the two federal responsibilities complement one another.

Furthermore, NOAA has an obligation under the Sustainable Fisheries Act to restore

depleted populations to optimal levels of abundance and productivity. The VSP concept can be

used to inform management decisions in this context. If, for example, it were a management

objective to ensure a population's sustainability while providing for a specified level of harvest,

VSP guidelines could be used to help determine the population abundance, productivity,

diversity, and structure that would be required to achieve this objective.

Jeopardy

Federal agencies cannot undertake or authorize an action that is “likely to jeopardize the

continued existence” of a species listed under the ESA (ESA Section 7). Joint NMFS-USFWS

regulations define “jeopardize the continued existence of” to mean “to engage in an action that

reasonably would be expected, directly or indirectly, to reduce appreciably the likelihood of both

the survival and recovery of a listed species in the wild by reducing the reproduction, numbers or

distribution of that species” (50 CFR S402.02). In the context of jeopardy, “survival” is “the

condition in which a species continues to exist into the future while retaining the potential for

recovery” (USFWS and NMFS 1998). The precise meaning of “into the future” is not defined,

but it clearly represents something short of the time frames associated with ESA recovery. In

contrast, the concept of “long-term survival,”, as considered in long-term conservation plans,

does not differ substantially from the concept of ESA recovery because both require a high

probability that the species will persist for substantial periods of time (Waples 1997). Thus

“jeopardy” currently lacks a precise biological definition; instead, it is defined in a legal context

that introduces several secondary terms that do not have precise biological definitions.

Therefore, it is not surprising that it has been difficult to develop and apply jeopardy standards.

We will need clearer definitions or interpretations of these secondary terms before we can forge

a formal link between VSP and jeopardy determinations. Nevertheless, it is useful to consider

the risks that affect listed species’ survival because that is often the major factor in jeopardy

determinations.

For purposes of this discussion, we assume that risks to a species may constitute jeopardy

if the risks pose threats to short- or long-term species survival. Some jeopardy evaluations have

made use of “critical” thresholds that trigger strong management actions if exceeded. In most

cases, a “critical” status means that a population has a non-negligible probability of going extinct

over a relatively short time period (e.g., 10 years). A problem with implementing jeopardy

standards has been the difficulty in assessing the cumulative effects of a number of actions that

have impacts which are necessarily evaluated at small spatial and temporal scales. Using the

VSP concept as a framework should help determine the net effects “jeopardy” actions have on

population or ESU viability.

Relationship of VSP to Other Concepts

Relationship to Minimum Viable Population Concepts

The VSP concept is closely related to the concept of a minimum viable population

(MVP) (Shaffer 1981). Soulé (1987) defines an MVP as a population that is sufficiently

abundant and well adapted to its environment that it will persist in the long term without

significant artificial demographic or genetic manipulations. Meffe and Carroll (1994) define an

MVP as “the smallest isolated population size that has a specified percent chance of remaining

extant for a specified period of time in the face of foreseeable demographic, genetic, and

environmental stochasticities, plus natural catastrophes.” The MVP concept has been used in a

number of conservation applications, from reserve design to extinction risk analyses (see reviews

in Soulé 1987).

Though the VSP concept shares many features with the MVP concept, the two differ in

several important ways. First, abundance has historically been the primary factor in defining an

MVP (Soulé 1987). A VSP, on the other hand, is defined by a variety of parameters, including

population growth rate, population spatial structure, and diversity. Second, although we suggest

minimal thresholds for a VSP (see appendix sections “Population growth rate and related

parameters” through “Viable ESUs,” p. 64-127), any population that meets or exceeds these

thresholds is considered viable. In other words, VSPs include not just minimally viable

populations, but more robust populations as well. Finally, the VSP concept is specifically

tailored for use with Pacific salmonids and thus emphasizes parameters and criteria that are

particularly relevant to this group of species.

Relationship to Quantitative Population Viability Analysis

Population viability analysis (PVA) is a widely-used tool for estimating extinction

probabilities (Soulé 1987, Caughley and Gunn 1996) and it is being increasingly applied to

salmon populations (Spencer 1999). However, there are presently no models that completely

represent the various risks facing salmonid populations. The VSP concept is intended to provide

useful benchmarks for evaluating actions, such as harvest or artificial propagation, that directly

affect natural populations and for which incremental increases in extinction risk may be difficult

or impossible to accurately quantify. The VSP concept is not intended to replace quantitative

risk models in situations where these models can be appropriately used, and the concept could be

used in conjunction with quantitative models in some cases. For example, the effects of harvest

on abundance (one parameter of a VSP) may be relatively easy to quantify and model, but the

simple life-cycle models usually used to evaluate and set harvest levels, often are inadequate for

accurately estimating extinction risk. The VSP concept, by determining the level of abundance

(among other parameters) necessary for long-term survival, could be used in conjunction with a

quantitative life-cycle harvest model to determine if a specific harvest action is likely to cause a

population to fall below VSP parameters. Until extinction risk can be accurately estimated under

a variety of scenarios for Pacific salmonid populations and ESUs, the VSP concept will be of

great help in assessing actions that directly or indirectly affect population viability.

Relationship to Properly Functioning Conditions

Although viable salmonid populations clearly require high quality freshwater habitat, this

document focuses on population processes and does not attempt to establish the relationship

between particular habitat attributes and population viability. This is appropriate given the

purpose of this document, which is to provide a foundation for setting recovery goals of listed

ESUs. These goals will be based on current and projected status of the fish populations, not on

the presence or absence of particular habitat attributes. Of course, assessments of the current and

future status of the population will take habitat attributes into consideration, but the fish

themselves are the ultimate indicator of whether or not the population continues to be in danger

of extinction.

Understanding the relationship between freshwater habitat quantity and quality and

population viability is critical in developing recovery plans and in determining the impact of

proposed land use activities on fish survival. To assess the effects of actions that may adversely

modify a species' habitat, NMFS uses the concept of “Properly Functioning Condition” (PFC,

NMFS 1999). PFC defines, based on currently available knowledge, the freshwater spawning

and rearing habitat conditions necessary for the long-term survival of Pacific salmon

populations. To evaluate the effects of specific habitat actions, NMFS uses analytical tools (e.g.,

the Matrix of Pathways and Indicators, NMFS 1996) to determine whether an action will

maintain, restore, or degrade the values of the parameters that describe properly functioning

conditions. The incremental increase in extinction risk from a habitat action is not estimated per

se. Instead, PFC describes the freshwater habitat conditions needed for long-term species

survival, and subsequent actions are evaluated based on how they affect the habitat conditions.

Defining PFC is an ongoing process that will continue to undergo revision, as more scientific

data become available. A particular challenge in developing PFC guidelines is relating habitat

actions at a variety of spatial scales to population-level responses.

Relationship to the Sustainable Fisheries Act (SFA) and Maximum

Sustainable Yield (MSY)

In addition to ESA responsibilities, NMFS has responsibility for administering marine

fisheries under the Magnuson-Stevens Fishery Conservation and Management Act, as amended

by the Sustainable Fisheries Act of 1996 (SFA). Important concepts in the SFA that relate to

population viability include optimum yield (OY), overfishing, and essential fish habitat (EFH).

The Pacific Fishery Management Council recently introduced Draft Amendment 14 to the

Pacific Coast Salmon Plan (PFMC 1999), which incorporates the SFA provisions into Pacific

salmon ocean fisheries management. In developing the VSP concept, we have made no

assumptions regarding the allowable level of harvest for listed salmon populations. In this

regard, harvest is treated similarly to other management factors that influence salmon

populations (i.e., hatcheries, habitat, and hydropower). However, we believe the VSP concept is

consistent with the intent of the SFA provisions because populations achieving viable status will

likely provide greater ocean and freshwater harvest opportunities.

The VSP guidelines have no explicit relationship to maximum sustainable yield (MSY)

analyses for a number of reasons. First, MSY generally deals only with population numbers and

VSP takes into account other parameters that affect population viability such as spatial structure

and diversity. Also, VSP is concerned with extinction risk, not with setting harvest levels. In

addition, VSP analyses focus on estimating extinction risk and examining stochastic processes at

small population sizes, issues that are generally not covered in MSY estimates. Assuming MSY

is actually being achieved, a wild population harvested at MSY is, by definition, sustainable

(VSP)—provided that the time horizon of MSY is the same as VSP and the MSY estimate takes

into account all the factors affecting viability, such as genetic diversity and spatial structure.

Relationship to Other Conservation Assessment Approaches

A number of other approaches are used to assess risks at the species and population

levels. One of the most widely used is that of the International Union for the Conservation of

Nature (IUCN) for its Red Book (IUCN Species Survival Commission 1994). The IUCN rates

species on the basis of five general criteria: Population reduction, limited extent of occurrence

combined with decline or fluctuation, low population abundance with continuing decline,

extremely low abundance, and quantitative viability analysis. These criteria differ from ours in

three respects: 1) IUCN criteria are applied to a species as a whole, not to individual populations,

2) some of the IUCN criteria link abundance or geographic range with decline (analogous to our

“population growth rate”) while our criteria treat abundance and productivity separately, and

3) we explicitly treat population structure and diversity as separate parameters. Allendorf et al.

(1997) proposed criteria for Pacific salmon that were similar to the IUCN criteria, but adapted

them specifically to the purpose of ranking threats to individual populations rather than

classifying them as individual risk categories (see also Currens et al. 1998, Wainwright and

Waples 1998). The specific criteria values proposed here differ somewhat from both the IUCN’s

and those developed by Allendorf et al. because they serve a different purpose: we define criteria

that are appropriate for classifying populations as viable, sub-viable, or critical within the context

of broader ESUs and in concert with the requirements of the ESA. Our criteria differ further

from the IUCN’s because we restrict them to a single genus, rather than developing them for all

organisms.

Identifying Populations

Introduction to Identifying Populations

Conceiving of and defining a population is relatively straightforward in a theoretical

sense. It is another matter to identify populations in nature. In practice, information is limited

about the distribution, local abundance, and migratory patterns of a species during their life

cycle. For Pacific salmonids, whose life history occurs in such diverse habitats as freshwater

tributary streams and ocean environments, the challenges associated with delimiting population

boundaries are many. Nevertheless, a number of different types of information can be used to

indicate the geographic or temporal boundaries of a salmon breeding population. This section

has two main parts. First we briefly outline the kinds of information that can be used to help

identify salmon populations. Second, we review approaches to identifying salmonid populations

used in management at state, tribal, and federal levels.

Types of Information Used in Identifying Populations

The different types of population boundary indicators discussed in this section apply to

all levels of structure, from substructure within populations to the structure of populations within

ESUs. Most of the methods described in this section can be used to generate a nested hierarchy

of spatial scales over which some level of indicator subdivision can occur. There is spatial

pattern at many scales in the biological and geographical data relevant to salmon, the challenge

to scientists managing salmon under the Endangered Species Act is to identify the appropriate

groupings of salmon that are most useful for predicting the long-term persistence of populations

and ESUs. Making the leap from identifying distinct groups of fish based on similarity of

characteristics to identifying distinct populations depends critically on how independent the

groups of fish are. As we have defined populations in this document, a group of fish is

considered an independent population if migrants from other groups do not appreciably affect the

population dynamics or extinction probability of the focal group (see discussion in “Population

Concepts,” p. 4).

Evidence for independent populations

The best evidence for identifying populations comes from information on a group’s

extinction probability and the degree to which its population dynamics are independent of those

of any other group. Such evidence could come from direct observations of trends in abundance

or productivity from groups of fish with known inter-group stray rates. If the dynamics or

extinction probabilities found in one group of fish is not appreciably affected by strays from any

other group, the focal group can be considered an independent population. In rare cases, quasi-

natural “experiments” may be available to test the effects inter-group straying has on population

dynamics and extinction. Such an experiment may occur where a naturally spawning run is

locally extirpated or where a hatchery-derived population is taken out of production. In such

cases, the effect on population dynamics of removing migrants from a system can be observed.

Because such experiments require good, long-term abundance or productivity information before

and after removing a population’s neighboring groups, opportunities to test population

independence in this manner are rare. In lieu of empirical information, modeling efforts can be

used to test the probability that groups of fish connected by a particular stray rate have

independent population or extinction dynamics. Such modeling exercises have been used to

explore metapopulation persistence for a number of species (Hanski and Gilpin 1997). In

addition, Hastings (1993) showed analytically that under certain conditions, population dynamics

remained independent as long as the inter-population migration rate was less than 10% (see also

Kendall and Fox 1998).

It is critical to understand the distribution of stray rates between spawning groups

in order to model the effects of reproductive isolation on salmonid population trends.

Demographic estimates of dispersal include radio-tagging studies, data from coded-wire tag

retrievals, and studies of stray rates from wild and hatchery fish (Giger 1972, Lister et al. 1981,

Quinn et al. 1991, Labelle 1992, Quinn 1993, Vanderhaegen and Doty 1995). There are few

estimates of stray rates for Pacific salmonids (Table A1). Those that do exist indicate that there

is high variability within and among species in terms of the percentage of fish that return to

streams

Table A1. Estimates of straying (the percentage of marked fish returning to a location other than that in

which it was marked) for Pacific salmonids. Straying estimates for some species have been made

for fish initially marked in their natural habitats (natural) and for fish reared in hatchery facilities

(hatchery).

Species % Straying

Geographic scale of

straying

Natural (N) or

Hatchery (H)

origin

Geographic region Reference

Sockeye

salmon

0.6 - 1.5 N/A N

Cultus Lake (Fraser R.), British

Columbia

Chum salmon 2.2 - 10 350-2000 km N Hokkaido, Japan 2

Chum salmon 5.2 - 5.4 ≈10 km H British Columbia 3

Pink salmon 0.1 - 62 ≈ 400-800 km H Soviet Union 4

Coho salmon

0.5 – 67

0 – 3.9

1 – 65

0 - 67

California: 10 km;

B C: 9 – 159km;

Puget Sound: <150km;

WA coast: <150 km

Scott and Waddell Creeks in

California; British Columbia,

Puget Sound, Washington coast

5, 6, 12

Coho salmon

0.0 - 27.7

1 –7

<0.5 – 4

0 – 12.4

0 – 100*

0 – 99*

B.C.: 7-58 km

Puget Sound: <150km;

WA coast: <60 km;

Columbia: <150 km

Hood Canal: <50 km

Grays Harbor: <25 km

British Columbia, Puget Sound,

Washington coast, Columbia R.

basin, Hood Canal

6, 12, 13

Steelhead 2 - 3 10 km N

Scott and Waddell creeks in

California

Chinook

salmon (fall)

3.2

> 60 km but w/in

Columbia basin

N Columbia River (Lewis R.) 7

Chinook

salmon (fall)

89.7

4.6 – 5.7

7 - 86

< 3

2 – 25

5 - 95

Lewis: < 30 km,

Lewis: w/in Columbia

basin;

Sacramento: 48-336 k

Puget Sound: <150 km

WA coast: <150km

Columbia R: <400 km

Lewis R., Columbia R. basin,

Sacramento R. basin, Puget

Sound, Washington coast

7, 9, 11,

Chinook

salmon (spring)

0.3 - 3.6

98.3% w/in 50 km,

1.7% out of Columbia

basin

H Columbia River (Cowlitz R.) 8

Coastal

cutthroat trout

0 - 30 70-150 km H Oregon coastal rivers 10

*: straying estimate is expressed as the percentage of adult fish in a stream that originated from another location

References:

1. Foerster 1968 (in Quinn 1993)

2. Sakano 1960

3. MacQuarrie and Bailey 1980 (in Quinn 1993)

4. Glubokovsky and Zhibbotovski 1989 (in Quinn 1993)

5. Shapovalov and Taft 1954

6. Labelle 1992

7. McIsaac 1990 (in Quinn 1993)

8. Quinn and Fresh 1984

9. Quinn et al. 1991

10. Giger 1972

11. Cramer 1989

12. Vanderhaegen and Doty 1995.

13. Ruggerone 1997

other than those in which they were born. The percentage of fish straying from streams in which

they are marked can be as high as 95%, but the higher estimates are typically based on smaller

sample sizes, so their associated confidence levels are lower (Table A1). It is difficult to make

generalizations about the magnitude of stray rate variation among and within species because

there are so few estimates and the geographic scales over which they have been made vary

greatly. In addition, because of logistical considerations, many of the stray rate estimates are

based on movements of marked hatchery fish, and the relative propensities of hatchery and wild

fish to stray is not well understood (Quinn 1993).

Genetic estimates of salmonid straying suggest that the gene flow rate is less than the

straying rate (see “Diversity,” p. 19). Only strays that successfully spawn and produce viable

offspring contribute to gene flow. In addition, both ecological and genetic methods have been

used to more directly estimate the distribution of dispersal distances between parent and

offspring. The dispersal curve generated is the frequency distribution of offspring (i.e.,

spawners) as a function of the distance from where they were produced (i.e., where the spawner

hatched). These distributions define the area within which mating is expected to occur, or the

area encompassing a population. One such method is to estimate neighborhood size, which is a

function of the variance in parent-offspring dispersal distributions (Wright 1946, Crawford

1984). Using empirical estimates of dispersal distances between parents and offspring,

neighborhood sizes have been estimated for a few bird, plant, and insect species (Barrowclough

1980, Crawford 1984, Levin 1988, Ruckelshaus 1996, Fig. A1a). There are no such estimates

for salmonids. Another method for estimating parent-offspring dispersal distributions is to use

genetic markers to track individuals of known parentage. This method involves genotyping and

spatially mapping parents and their offspring. The dispersal distributions can then be generated

by quantifying the distances over which offspring dispersed from their parents (Meagher 1986,

Meagher and Thompson 1987, Grosberg 1991, Fig. A1b). There are no such estimates of parent-

offspring dispersal distributions for salmonids. Dispersal distributions could be estimated with

tagging studies—such as those using Coded Wire Tags (CWTs)—in which juveniles are marked

and then monitored to see where they return to spawn (e.g., Quinn and Fresh 1984, Quinn et al.

1991, Labelle 1992). In contrast, by using molecular markers, parents and their offspring can be

observed at the exact location of hatching (CWTs can only be employed after a fish may already

have migrated some distance from its natal location.) As might be expected, the sample size

requirements of such studies will be very large. However, the approach using molecular markers

to track individuals of known parentage may be feasible in some salmonid systems.

Indicators of population structure

Empirical information on salmonid stray rates coupled with long-term population

abundance data is rarely available. As proxy evidence for identifying a population, spawning

groups can be clustered based on similarity of a number of characteristics, and then the degree of

independence of the clusters can be estimated using additional information such as the likely

stray rates among clusters. Formal clustering algorithms based on distance measures can be used

to combine spawning groups (Hartigan 1975). Using any one of these indicators to interpret the

degree of reproductive isolation among groups requires an assumption that “all else is equal.”

For example, if two spawning groups exhibit a similar characteristic—such as an abundance

Figure A1a. Distributions of dispersal distances of eelgrass (Zostera marina) based on pollen (left graph) and

seed (right graph) dispersal. By combining these distributions, the neighborhood area was estimated

from Na = 4π(σ

/2 + σ

), where σ

and σ

are the variance in pollen- and seed-dispersal

distributions, respectively. Results from this population indicated that random mating among

individuals occurred on average within a circle of area 524 m

(from Ruckelshaus 1996).

Figure A1b. Frequency distributions of pollen, seed and combined pollen and seed dispersal estimated by

identifying and mapping seedlings and parents. Seedlings were assigned to parents based on

genealogy reconstruction using 11 polymorphic enzyme marker loci (from Meagher and Thompson

1987).

trend—they would be grouped together in a cluster analysis. Assuming that all else is equal

between the groups would lead to the conclusion that those spawning groups are similar in

abundance over time because they are demographically linked. However, all else is not equal.

For instance, the fish could occur in correlated environmental conditions at any stage of their life

history, a situation that would lead to correlated trends in abundance even in the absence of

demographic connections (Box A1, Hanski and Gilpin 1997, Kendall et al. 2000). Because it is

very difficult to validate such simplifying assumptions, it is better to use multiple, independent

indicators to explore similarities among spawning groups. This method offers more confidence

that groupings are biologically reasonable and that they actually reflect the degree of

reproductive isolation.

Geographic and habitat indicators: A salmon population can be identified, in part, by the

likely spatial distribution of its spawning habitat. Physical features such as a river basin’s

topographical and hydrological characteristics dictate to a large degree where and when salmon

can spawn and delimit the spatial area over which a single group of fish can be expected to

interact. For example, a group of fish returning to spawn in the upper tributaries of a large river

basin such as the Umpqua River on the Oregon coast are not expected to be part of a population

that includes another Oregon coastal river drainage, such as the Alsea River. The long distances

the fish need to travel to spawning habitat in a large river basin combined with generally accurate

homing tendencies make it very unlikely that a single population would encompass multiple

basins or large tributaries.

The geographic characteristics of river mouths and estuaries through which salmon

migrate can also be instrumental in indicating groups of salmon that are likely to be in the same

population. Timing and direction of water flow and spatial distributions of feeding, rearing, and

refuge habitats affect salmonid migratory patterns. For example, the freshwater plume produced

by the Columbia River is a prominent hydrographic feature along the western coasts of Oregon

and Washington. The conditions of the plume vary seasonally; therefore, the timing of the

juvenile salmonid downstream migration affects the salinity, temperature, nutrients, and

sediment load conditions the fish experience in the plume. These conditions, in turn, affect the

likely direction of juvenile fish migrations, their food sources, and their energetic balance.

Geographical and hydrological differences among river drainages imply that ecological

characteristics also differ. For example, some river systems consist of many spring-fed streams,

and others can be made up of streams whose origins are mostly from glacier- or snowmelt-run

off. The productivity, flora, and fauna of glacier- and spring-fed streams are likely to be very

different. Information on freshwater and marine habitat characteristics such as temperature, flow

regimes, prey, pathogens, and predator species can help define the boundaries of salmon

populations. For example, the frequency and types of parasites found in mixed-stock ocean

fisheries have been used to identify salmon populations by determining where the parasite

species occur in freshwater spawning and rearing habitats (Groot and Margolis 1991). In

addition, a population may be identified by examining data on patterns of use of freshwater and

marine feeding, rearing, and spawning habitats; these data are derived from stream surveys,

angler catch, and ocean sampling.

Box A1. Assigning “resident” and anadromous salmonids to populations.

One example of the importance of understanding the genetic basis for phenotypic variation when

defining population boundaries is the difficult issue of deciding whether individuals who spend

all of their life history in fresh water (often termed “resident” fish) should be included in the same

population as anadromous fish. This challenge arises in populations of O. mykiss (steelhead and

rainbow trout), O. clarki clarki (coastal cutthroat trout), and O. nerka (sockeye and kokanee) in

many river basins throughout their geographic range (Shapovalov and Taft 1954, Foerster 1968,

Barnhart 1986, Trotter 1989). These distinct life-history forms would be considered to be in the

same population under the following scenarios. First, it could be that the spawning groups of the

two forms are sympatric in time and space—thus allowing interbreeding between the life-history

types. If, in this case, the progeny segregate into alternative life-history forms, it would indicate

that the phenotypic variation occurs in a single population. Assigning fish to a population is

relatively clear in this instance. Alternatively, there could be relatively little inter-breeding

between resident and anadromous salmonids within a stream, but there could be enough lability

in life-history expression that resident parents give rise to a fraction of anadromous offspring each

generation. Those anadromous offspring of resident parents could then interbreed with other

anadromous fish. In this case, whether an individual fish remains in fresh water throughout its

life cycle or migrates to the ocean would be determined largely by environmental conditions.

Under this second scenario, the key information needed to determine whether the different life-

history forms are part of a single population is the degree to which genes are exchanged between

alternative life-history types. Understanding the genetic (as opposed to environmental)

determinants of life-history variation would help in assigning fish with distinct life-history types

to appropriate populations. If the life-history forms are primarily genetically determined and the

forms do not interbreed, fish with distinct life-history types would not be part of the same

population.

There is little empirical evidence supporting either of the above scenarios for any species and, as

expected, the degree to which distinct life-history forms appear to interbreed varies

geographically. It is clear that for all three species there have been periods when the two life-

history forms would have been considered part of the same population; this is because the

“resident” form appears to have arisen multiple times evolutionarily (rev. in Busby et al. 1996,

Gustafson et al. 1997, Johnson et al. 1999). What this means in practice is that it is not possible

to generalize about the extent to which life-history forms within a species are reproductively

isolated. In some cases, selective divergence of the two forms may be maintaining them as

separate populations (e.g., sockeye and kokanee in British Columbia—Wood and Foote 1996). In

other cases, periodic interbreeding during years of high relative abundance of one life-history

form may suggest that they are part of a single interbreeding group (e.g., steelhead and rainbow

trout in the Deschutes River, Oregon—ODFW 1998). The lack of generality is seen in genetic

data for steelhead and “resident” O. mykiss suggesting that in some streams, the two life-history

forms are not distinguishable, but in others, they are genetically distinct (Wilson et al. 1985,

Currens et al. 1987, Leider et al. 1995). Similarly, genetic data for sockeye and kokanee have

shown some groups within a lake system to be genetically similar, and others to be distinct (Foote

et al. 1989, Craig 1995, Taylor et al. 1996).

In general, the collective biotic and abiotic characteristics of occupied salmonid habitat

help define a population because we expect that those ecological characteristics constitute the

selective environment in which the salmon exist. If different groups of salmon experience

different selective environments and there is very little migration between those environments,

we expect those groups’ phenotypic characteristics to diverge.

Demographic indicators: To identify a salmonid population, it is necessary to understand the

extent to which it is reproductively isolated from other groups of fish. Demographic

characteristics such as birth and death rates and fecundity determine the population dynamics of

a group of fish. The level of inter-group dispersal influences the degree to which demographic

trajectories are correlated among individual groups, depending on the degree to which the

environmental conditions are correlated. Estimates of adult dispersal in both freshwater and

marine habitats are of critical importance in defining a population unit (see “Evidence for

independent populations,” p. 39).

Long-term abundance data are another type of demographic information that can help

identify a salmon population. Because one of the criteria for defining a population is that it be

largely demographically independent from other groups, abundance data may help show whether

the demographic trajectories of two proposed populations are largely independent of one another.

In cases where the dynamics of two groups of salmon are not correlated with one another, it

might be reasonable to conclude that the two groups are not part of the same population—

provided environmental variation across the range of the groups can be ruled out as a cause of

the uncorrelated dynamics. If the demographics of two groups of salmon do have correlated

dynamics, they might still not be part of the same population. In this latter case, it is possible

that two groups of organisms occur in environments with correlated conditions—resulting in

correlated dynamics between two groups that are not demographically linked (e.g., Grenfell et al.

1998, Kendall and Fox 1998; Fig. A2).

By exploring patterns of abundance among salmonid species within the same geographic

area, it is possible to gather additional information on the degree to which correlated

environmental conditions drive abundance correlations between spawning groups. Assuming

that different species of salmon exhibit different stray rates between spawning sites, any

similarities in population dynamics would most likely be due to similarities in common

environmental conditions the species are experiencing. Therefore, if more than one species

exhibit similar trends in abundance among spawning sites, it would indicate that correlated

environmental conditions are the most likely explanation for abundance correlations among

populations within a species. Teasing apart the effects of correlated environmental conditions

and dispersal to determine the primary cause of population dynamics is a complicated but

important exercise (Kendall and Fox 1998, Kendall et al. 2000).

There are few studies that explore the extent of spatial correlation in salmonid population

dynamics. Rieman and McIntyre (1996) found weak, but statistically significant positive

correlations in bull trout redd counts between streams that were close to one another. The

positive correlations in redd counts were stronger between closer streams, just as one would

expect if the correlated redd counts were due to correlated environmental conditions or dispersal

b b

Figure A2. The figure shows why demographic independence is theoretically a useful factor in

designating populations. The circles represent functionally panmictic groups of fish, such as

might be found at a single river reach. (In a panmictic group, all mature individuals are equally

likely to mate with one another.) The arrows represent migration among the breeding groups—

heavy arrows represent high migration rates and the dashed arrows represent low migration rates.

High migration causes demographic coupling among the groups; low migration causes them to

become demographically uncoupled. Breeding groups labeled with the same lower case letter are

considered part of the same population. The different patterned areas represent regions with

different environmental dynamics. As the text illustrates, the simple scenarios depicted in this

figure have important implications for 1) our definition of a population, and 2) how we estimate

the parameters for assessing population status.

Scenario A represents the easiest case to interpret, because the demographic and environmental

processes have the same boundaries. The two groups labeled “a” are part of the same population

because the high rate of migration between them means that any extinction of a single group

would only be temporary because the remaining group could recolonize the other—thus rescuing

it from extinction. The “a” groups and “b” groups are not considered to form a single population

because there are not enough migrants from “b” to prevent the “a” population from going extinct,

and vice versa. Because population and environmental boundaries coincide in this scenario,

demographic processes and environmental forcing will work in concert to create similar

abundance fluctuations in all the groups within a population and independent abundance

fluctuations among different populations.

In contrast, both populations in scenario B share the same environmental dynamic. If

environmental forcing were strong (i.e., abundance is strongly influenced by varying

environmental conditions), then fish abundance would vary in a similar fashion in each

population. However, “a” and “b” would still be considered independent populations because the

risk of extinction due to demographic processes in “a” is independent of the risk in “b.”

In scenario C, all four groups experience different environmental dynamics, and if environmental

forcing were strong, the abundance variation would be independent for all four groups. However,

“a” is still a legitimate population because of the high inter-group migration rate. Thus for the

“a” groups, individual reaches within a single population may temporarily go “extinct,” but when

the population is considered as a whole, as long as a single reach remains extant, the population

will survive. These types of substructured populations are discussed in detail in “Spatial

Structure,” p. 18.

In scenario D, we show groups of fish in a hypothetically constant environment. In a constant

environment, abundance changes would be driven completely by demographic and genetic

processes. Even if they were to start with the same population size, the abundance patterns for

population “a” would deviate from the abundance pattern of population “b.” Our definition

focuses on the extinction risk that independent populations face. As a consequence, separate

populations could abstractly be described as being sufficiently isolated, but in a constant

environment they would have largely uncorrelated demographic trajectories.

among nearby spawning areas. Milner et al. (1993) explored correlations in salmonid densities

between sites within and between tributaries in a river in North Wales. Atlantic salmon and

brown trout fry and parr densities were compared at several sites within each of three tributaries

to the River Conwy. For both trout and salmon, weakly significant positive density correlations

were found both within and between the tributaries, though the variance synchrony was stronger

within tributaries than it was at larger spatial scales (Milner et al. 1993). McKinnell (1999)

reported on synchronized freshwater catch and recruitment dynamics for Atlantic salmon in

Swedish rivers. And, like the other studies, he found a weak relationship between the survival of

salmon from different hatcheries and the distance between them. Fish from hatcheries that were

closer together were slightly more correlated in terms of marine survival than were fish from

hatcheries that were separated by greater distances (McKinnell 1999). Bradford (1999) found

positive covariation in coho smolt abundance for streams in Western North America that were

separated by distances less than 20-30 km. In summary, the evidence for spatial correlation in

salmonid population numbers suggests that abundance variations among fish in streams that are

closer to one another tend to be significantly positively correlated, but the correlations are often

not very strong.

Genetic indicators: Genetic characteristics are useful in identifying salmon populations because

they indicate the extent of reproductive isolation among groups. Molecular genetic markers such

as allozymes and nuclear or mitochondrial DNA markers can be used to statistically describe an

organism’s population structure based on their genetic similarities. In addition, studies of

quantitative trait variation (e.g., life-history characteristics) can often help to distinguish

salmonid groups. Salmon populations, stocks, and ESUs have been identified with the help of

genetic markers and life-history traits (WDF et al. 1993, ODFW 1995, Weitkamp et al. 1995,

Busby et al. 1996, Myers et al. 1998, Johnson et al. 1999). The spatial distribution of particular

alleles and their frequencies can also be used to define the spatial scale(s) over which groups of

salmon are structured. The nature of genetic data allows for a hierarchical exploration of the

geographic structuring of salmon populations. Nevertheless, just as it is difficult to use

geographic and abundance indicators to identify populations, the extent to which genetic markers

can be used to identify independent populations still needs to be determined. Translating the

genetic structure of a salmonid species into a set of independent populations is not a simple task,

because of the assumptions about population structure that must be made when estimating allele

frequencies. As with other indicators of spatial structure, it is necessary to estimate the

significance of the genetic structure in terms of population and extinction dynamics in order to

identify populations.

For the most part, genetic markers are straightforward indicators of genetic variation

among groups. However, there are some instances where genetic data can be uninformative or

misleading (Whitlock and McCauley 1999). For example, if there is insufficient variability at

marker loci or only a small number of marker loci are used, genetic markers may indicate little or

no population structure when it is in fact present. In addition, using molecular genetic markers to

estimate degree of reproductive isolation may require certain assumptions (e.g., that the groups

of organisms sampled are in drift-migration equilibrium). As discussed elsewhere (see

“Diversity,” p. 19), salmon populations are dynamic in space and time, and in many cases their

evolutionary dynamics are not likely to be at equilibrium. For example, a subpopulation founded

by a few individuals during a colonization event is likely to diverge genetically from the

founders’ original subpopulation because of the genetic drift that occurs during the sampling

from the parent subpopulation. This would be true even if, over the course of many generations,

the two subpopulations were connected by high enough levels of migration to be considered part

of the same population. Because extinction and recolonization processes are thought to occur

relatively frequently in some salmon species in some locations (Mangel and Tier 1994, Dunham

and Rieman 1999), it is not a simple matter to attribute causes (e.g., low levels of gene flow vs.

recent founding event) to observed population structure. This may also present a problem when

attempting to identify populations that have experienced chronic declines in abundance, if, for

example, a previously contiguous population has become fragmented into what now appear to be

several isolated populations. Rigorous modeling or empirical comparisons to patterns of genetic

variation in healthier systems may help overcome some of these potential problems (Currens et

al. 1996). Finally, although salmon populations are likely to be commonly perturbed from

equilibrium conditions, theoretical work has shown that under some circumstances estimates of

inter-population gene flow obtained from population structure statistics approach their

equilibrium values fairly rapidly (Slatkin and Barton 1989, Slatkin 1993, Crow and Aoki 1984).

Recently developed likelihood methods for estimating gene flow and divergence time may

provide more accurate estimates of gene flow than F

based estimates, especially in non-

equilibrium situations (e.g., Nielsen and Slatkin 2000).

It is appealing to use molecular genetic markers to survey and identify salmon

populations because of the large sample sizes that can be obtained with relatively little effort.

Variation in genetically-based life-history traits can also be used to help determine population

boundaries. However, gaining an understanding of the distribution of genetically-based life

history and other quantitative traits is much more labor-intensive and, as a result, this

information is usually lacking for most natural populations. In practice, phenotypic variation in

traits such as run timing, size at migration, and ocean and estuarine migratory patterns has been

used to help delineate salmonid groups. This approach is probably most useful at relatively large

geographic scales—such as ESUs (e.g., Busby et al. 1996, Myers et al. 1998)—because the

potentially confounding effects of environmentally-induced phenotypes are less likely to occur

over large geographic regions. Technically, only those phenotypic traits based on underlying

genetic variation (rather than environmentally-induced variation) can help define a population

because populations are defined on the basis of reproductive isolation and, ultimately,

demographic independence (see Figure A2). However, in combination with other population

indicators, spatially-structured phenotypic variation within an ESU could be used as a proxy for

genetically-based variation.

Rigorous, direct methods for estimating the genetic components of quantitative trait

variation involve controlled breeding designs and rearing offspring to life stages that express

quantitative traits of interest (Falconer and Mackay 1996). Because of the relatively long

generation times for Pacific salmonids, there are many challenges associated with rearing

offspring from controlled matings. Indirect evidence for the extent of genetic (vs.

environmental) sources of variation in phenotypic traits can be obtained from studies of traits

within maternal or paternal families, an approach that is somewhat easier to use for salmon,

especially in hatchery populations (Silverstein and Hershberger 1992, Beacham and Evelyn

1992, Heath et al. 1994, Herbert et al. 1998, Mousseau et al. 1998, Hard et al. in press). A few

studies of quantitative trait variation in salmon have been conducted and they have found that a

significant portion of the observed phenotypic variation in some life-history and meristic traits is

genetically based (see “Diversity,” p. 19). Such results suggest that describing the spatial

distribution of life-history variation (and other quantitative traits) for salmon may be a good

indicator of population structure, even if the actual genetic basis for trait variation is not known

in most cases.

In a comparative review, Healy and Prince (1995) showed that in general, phenotypic

variation among salmon populations is much greater than allozyme frequency variation.

Molecular markers are ultimately a better tool for grouping fish than phenotypic variation for at

least two reasons. First, selection is expected to be greater on quantitative traits than on

molecular loci (which are presumed to be neutral). Estimating reproductive isolation from

divergence of selected characters requires making a number of restrictive assumptions that are

not needed when using neutral markers (Hartl and Clark 1989). Second, some portion of

phenotypic variance is due to environmental variation and is therefore of no use in estimating

reproductive isolation among groups (Falconer 1981). Nevertheless, where it is difficult to

obtain useful data, a useful proxy (e.g., phenotypic variation) is better than no information at all.

The paucity of information on quantitative trait variation among salmon populations has

unfortunate consequences in terms of our ability to evaluate population and ESU viability. This

is partly because such variation is a good indicator of the extent and nature of a population’s

local adaptations. The degree to which a population exhibits local adaptation is an important

indicator of its response to environmental changes and, ultimately, the probability of its

persistence. More generally, it is useful to know the distribution of locally adapted traits among

groups of salmon in order to describe diversity patterns at a number of spatial scales. As

“Diversity” (p. 19) discusses, genetic diversity is a critical factor in long-term persistence at the

population, ESU, and species levels. It also is an important descriptor of the biological

components of the evolutionary lineages that are the focal points for conservation.

Identifying populations—combining the evidence

Without evidence of independent population dynamics or extinction probabilities,

identifying populations involves combining stray rate estimates with any information on

indicators of population structure discussed previously. Ford et al. (2000) summarizes methods

for identifying populations using available data for spring-run chinook salmon and steelhead in

the upper Columbia River, Washington, region.

Existing Approaches to Identifying Salmon Populations or Groups

States and other groups that manage salmon have, for management purposes, employed a

number of different definitions for particular groups of fish and many of those definitions are

similar to the population definition we propose in this document. The Washington Department

of Fish and Wildlife (WDFW 1998) defines a stock using Ricker’s (1972) definition; although

they do acknowledge that smaller production units may be important and that considerable

uncertainty arises when this definition is applied to groups of salmon (WDFW 1998). In

practice, salmon stock boundaries have been identified using information on such aspects as run

timing, size and age at migration, ocean migration patterns, spawning locations, and genetic

similarities. For the purposes of reporting the status of coastal cutthroat trout, the WDFW

introduced the concept of a stock complex, or a “group of stocks typically located within a single

watershed or other relatively limited geographic area” (WDFW 1998). The WDFW found that

identifying individual stocks of coastal cutthroat trout was more challenging than identifying

salmon and steelhead stocks. This was due, in part, to the wide range of migratory behaviors that

cutthroat trout exhibit and because there is a great deal of uncertainty about the genetic

relationships among different life-history forms (WDFW 1998). The WDFW also uses various

genetic, geographic, and life-history data to delineate Genetic Diversity Units (GDUs), and

Major Ancestral Lineages (MALs), both of which attempt to define conservation or management

units at taxonomic levels higher than populations but lower than species. For example, the

WDFW defines an “Upper Columbia Summer + Fall, Snake Fall and Mid- and Lower-Columbia

Chinook” Major Ancestral Lineage within which there are 7 Gene Diversity Units (Marshall et

al. 1995). Three of these GDUs are contained within the Lower Columbia River chinook salmon

ESU (Myers et al. 1998). In other cases, such as Puget Sound chinook salmon, the MAL and the

ESU boundaries are equivalent (Marshall et al. 1995, Myers et al. 1998).

Along similar lines, the Oregon Department of Fish and Wildlife (ODFW) also uses a

definition of “population” similar to Ricker's (1972) definition. They identify Gene

Conservation Groups (GCGs), which are defined as genetically distinct clusters of one or more

populations within a taxonomic species that arise when gene flow between the cluster and other

populations of the same species is zero (or very low) over a sufficiently long period of time

Kostow et al. 1995, OAR 635-07-501). The GCGs are identified using molecular genetic and

geographic information to define groups of spawning populations that have similar genetic and

life-history traits (Kostow et al. 1995). The ODFW often uses genetic data to delineate GCG

boundaries, but if such data are not available, geographic proximity is used as a guide in

identifying groups (Kostow et al. 1995).

Although the WDFW/tribal and the ODFW’s population definitions are based on the

same Ricker stock definition, in practice, the agencies tend to reach different conclusions as to

what constitutes a “population.” The WDFW and tribal (SASSI) stocks (WDF et al. 1993)

generally encompass larger geographic areas than what the ODFW would consider a population

(Kostow et al. 1995). Part of this difference may result from geographic differences between

two states. Nevertheless, a large part of the difference is likely due to different interpretations of

the “substantial degree do not interbreed” portion of Ricker’s definition. By adding independent

population dynamics and extinction risk to Ricker’s definition, we are attempting to arrive at the

clearest possible interpretation of the word “substantial.”

The Pacific Fishery Management Council's Pacific Coast Salmon Fishery Management

Plan recognizes numerous “stocks” of chinook and coho salmon for purposes of ocean harvest

management. The plan has no specific criteria for designating stocks, but in general they cover

larger areas than do the populations defined here. The proposed draft Amendment 14 to the plan

restructures the management stocks slightly to provide more comprehensive coverage of the

ESUs defined in NMFS' ESA listings, but retains the larger geographic scale of the present plan

(PFMC 1999). We expect that each PFMC stock would be made up of one or more populations

as they are defined here.

The state of California delineates separate runs of salmonids based on genetic distinctness

and life-history differences such as spatial and temporal variation in run timing and juvenile

outmigration (CDFG 1998). They manage these distinct runs (e.g., Central Valley spring vs. fall

chinook salmon) separately from one another in setting harvest levels and escapement goals.

Their “runs” would probably be composed of several populations as we have defined them in

this document. In addition, Barnhart (1994) has used geographic distribution, life history,

straying, and genetic data to identify populations and meta-populations of several salmonid

species in the Klamath-Trinity Basin of California.

Population Size

Introduction

Small populations face a host of risks intrinsic to their low abundance; conversely, large

populations exhibit a greater degree of resilience. A large part of the conservation science

literature is directed toward understanding and predicting the effects of population size (Soulé

1986, 1987, Meffe and Carroll 1994, Caughley and Gunn 1996). Population abundance is an

important determinant of risk, both by itself and in relationship to other factors. There are a

variety of risks associated with the demographics of small populations, including both directional

and random effects. Directional effects include two forms of population density-dependence:

compensation and depensation. Random effects on population dynamics include demographic

stochasticity, environmental stochasticity, and catastrophes (Gilpin 1987, Lande 1993). (Note

that only demographic and local environmental stochasticity are considered at the population

level here; large-scale environmental effects and catastrophes are considered at the ESU level,

see “Viable ESUs,” p. 125.) Lande (1993) used a simple population dynamics model to examine

the relative risks associated with these three effects as a function of carrying capacity and long-

run population growth rate. He concluded that, for populations with a positive long-run average

growth rate, risks from demographic stochasticity are substantial only at very low population

sizes, while environmental stochasticity and catastrophes are the dominant risks at higher

population sizes. For populations with negative average growth rate, risks from all three types of

variation change nearly exponentially with capacity, and thus all may be important.

Small populations also lose adaptive variation and gain maladaptive variation at higher

rates than large populations (Mills and Smouse 1994, Lande 1995, Schultz and Lynch 1997),

both phenomena generate lower average fitness and a higher probability of extinction. Finally, if

populations are declining due to deterministic effects (e.g., habitat degradation or overharvest),

population abundance combined with measures of trends and productivity can be an indicator of

how long it will take the population to reach a critically low abundance level (Caughley 1994).

Given these considerations, it is clear that risk tends to vary inversely with abundance if

other factors are held constant. This means we can, in theory, define simple numeric thresholds

to delineate different risk categories ranging from healthy down to extinct. In the following

discussion, we focus on two such thresholds: 1 ) a Aviability@ threshold above which populations

have negligible risk of extinction due to local factors, and 2) a Acritical@ threshold below which

populations are at relatively high risk of extinction in the near future. However, putting numbers

on such thresholds is a difficult task for three reasons. First, other factors are not constant and

often correlate with abundance levels; this means a pure abundance threshold must be

conditional on particular conditions, especially those relating to population trend and variability.

Second, the theory relating abundance to extinction risk is based on relatively simple models of

population dynamics, and it is not clear how to relate this theory to the complexities of real

populations. Third, our ability to accurately measure population abundance and density is quite

limited, so there is a great deal of estimation error in any analysis of abundance or risk. For all

these reasons, there is a great deal of uncertainty associated with any thresholds that may be

established.

All else being equal, small populations are at greater risk of extinction than large

populations primarily because several processes that affect population dynamics operate

differently in small populations than in large populations. These are: deterministic density

effects, environmental variation, genetic processes, demographic stochasticity, ecological

feedback, and catastrophes. The first five of these processes are discussed in more detail.

Catastrophes (environmental events that severely reduce population abundance in a relatively

short amount of time) affect more than one population over a large geographic area, so the

effects of catastrophes are considered in the section on ESU-level viability (“Viable ESUs,” p.

125) rather than here.

Density effects

Population ecologists typically divide processes that affect birth and/or death rates into

two types: density-dependent effects, and density-independent effects (for example, Nicholson,

1933, Wilson and Bossert 1971, Ricklefs 1973, Hilborn and Walters 1992). Because of the

structure of the differential-equation models that describe these different effects, density-

independent effects are often termed “linear” effects, while density-dependent effects are termed

Anon-linear@. Density-dependent effects on population productivity arise from two competing

processes: compensation (a decrease in productivity with increasing density) and depensation (an

increase in productivity with increasing density). Compensation is a stabilizing influence. With

it, an increase in population is offset by a subsequent decline in productivity; it is most important

at higher abundance levels. Depensation is a destabilizing influence and is important mainly at

very low abundance levels.

Figure A3 illustrates the typical shape of curves relating parental density to offspring

density for deterministic populations with positive intrinsic growth rate exhibiting no density-

dependence, compensation only, and compensation plus depensation. In this illustration, it is

clear that the population with linear dynamics will increase in each generation regardless of

abundance and that the population with compensation only will reach a stable equilibrium at a

positive abundance where the curve crosses the replacement line. The population with both

compensation and depensation has three equilibrium points (points where it crosses the

replacement line): stable equilibria at zero (extinction), high abundance, and an unstable

equilibrium at intermediate abundance. If population abundance is above the unstable

100

200

300

400

500

600

700

800

900

1000

0 200 400 600 800 1000

Parental Abundance

Offspring Abundance

Linear

Compensation

Depensation

Replacement

Figure A3. Typical shape of parent-offspring (or stock-recruit) curves for populations with no density-

dependence (“Linear”), compensation only, and compensation plus depensation. The

replacement line (where offspring equal parents) is included for reference.

equilibrium, the population will more than replace itself and move toward the upper stable point;

if it is below the unstable point, it will fail to replace itself and move toward zero.

Compensation at high abundance occurs because there is increasing competition for limited

resources as a population expands to fill (or exceed) available habitat. Compensation=s negative

relationship between productivity and abundance can give a population substantial resiliency

because any decline in abundance is offset by an increase in productivity that will tend to restore

the population to some equilibrium level. The role of compensation is widely recognized in

fishery management through the use of compensatory production models, usually following the

form of Ricker or Beverton-Holt recruitment models (Ricker 1975, Hilborn and Walters 1992).

For salmon, there is extensive literature discussing models and evidence for compensatory

processes in salmon (for example, Ricker 1954, Skud 1973, Solomon 1985, Peterman 1987).

In contrast, when abundance is very low, depensation will accelerate declines and slow

recovery. Depensatory processes at low population abundance (also termed “allele effects” or

“inverse density dependence”) will result in high extinction risks for very small populations

because any decline in abundance further reduces a population=s average productivity, resulting

in a spiraling slide toward extinction. A variety of processes can lead to depensatory population

behavior, and such effects are documented for a wide variety of organisms (reviewed by

Courchamp et al. 1999). The most commonly cited of these are the uncertainty that mates will

be able to find one another in sparse populations (Dennis 1989), randomly skewed sex ratios at

low abundance (Gabriel and Bürger 1992), non-linear response of predators to shifting prey

abundance (Peterman 1987), and scaling effects of random variation among individuals (Lande

1998). Synergistic interactions among these effects may also increase risks at low abundance

(Stephens and Sutherland 1999). Fundamental population theory clearly requires that

depensation become significant at some low abundance level in all populations, but the exact

level where this occurs, and whether depensation will overcome compensation, depends on the

behavior and ecology of individual populations. For example, population density must be high

enough to provide sufficient temporal and spatial overlap for males and females to find each

other to spawn, but the level of population abundance necessary for this to occur depends on the

life-history characteristics of the population in question as well as aspects of the population=s

habitat. In consequence, populations that return to spawn over long periods of time or inhabit

large geographic areas may need to be more abundant to be considered viable than those with

more limited spatial or temporal spawning distributions.

In contrast to compensation, fish population models rarely include depensation, although

that situation is changing (Ludwig 1998). The direct evidence for depensation in salmonid

populations is mixed. Peterman (1987) summarized evidence from a few studies indicating

depensatory predation mortality among juvenile salmon, and used a simple population model to

show that such mortality could explain observed population dynamics. In contrast, Myers et al.

(1995) found significant evidence of depensation in population data for only 2 of 11 populations

of salmon examined. Liermann and Hilborn (1997) found only one salmon population out of 27

that clearly exhibited depensation; it was uncertain whether depensation was evident in the

remaining 26. Because of the great uncertainty surrounding depensation parameter distributions

in their study and the consequences of ignoring depensation if it really is significant, Liermann

and Hilborn recommended that fishery analyses employ models that incorporate depensation.

Translating these effects into specific criteria for “viable” or “critical” abundance levels

is no simple task. The specific form and function of both compensatory and depensatory

processes depend on species- and population-specific life-history characteristics that are difficult

to measure accurately. Both types of effect may vary either abruptly or smoothly with changing

abundance (for example, Getz and Haight 1989, Dennis 1989), and either pattern could cause

abrupt thresholds in population dynamics. When these deterministic effects are combined with

stochastic effects, the situation becomes even more intractable. However, some general

guidance can be noted. First, it is obvious that a population that is unable to replace itself is in

trouble, so we can say that a population is in critical condition when it is so low that depensatory

processes are expected to reduce average production below replacement. Similarly, we can say

that a population that is near its compensatory stable equilibrium abundance is likely to be safe

from deterministic extinction risks; however, such a level may be well above the actual lower

threshold of viability as we have defined that term.

Precisely estimating critical population levels based on these principles is impossible

without extensive observations of population dynamics across a wide range of abundances;

however, a number of approaches can be used to provide reasonable estimates of such levels.

Estimates of habitat capacity can be used to define a likely range for compensatory processes.

Similarly, theory can be used to establish upper and lower limits for depensatory processes based

on mate-finding, predator-prey interactions, sex-ratio stochasticity, etc.

Environmental variation

Environmental variation can cause small populations to go extinct when chance events

frequently reduce survival or fecundity to low levels over an extended time. There is an

extensive amount of literature on the relationships among extinction risk, persistence time,

population abundance and level of variation in demographic parameters (for example,

MacArthur and Wilson 1967, Tuljapurkar and Orzack 1980, Goodman 1987, Belovsky 1987,

Lande and Orzack 1988, Thomas 1990, Dennis et al. 1991, Lande 1993).

One obvious conclusion from this literature is that there are no simple, generic abundance

levels that can be identified as viable; the form of the population model used (e.g., linear or non-

linear, simple or age-structured), average parameter values, and degree of environmental

variation all combine with abundance to determine extinction risk. In general, model results

suggest that numbers greater than 1,000 to 10,000 are often needed to protect a population from

moderate-to-high levels of environmental variation (Belovsky 1987, Goodman 1987, Thomas

1990). Thomas (1990) examined both empirical evidence and model results to determine “safe”

(i.e., low extinction risk) abundance levels. Thomas concluded that Asafe@ levels vary depending

on the degree of natural variation in population abundance. However, Thomas did provide

guidance regarding appropriate population abundance (geometric mean abundance per

generation) for general classes of organisms, suggesting that 1,000 is adequate for species of low

variability, and 10,000 is usually adequate for the most variable bird and mammal species. For

extremely variable species, such as many insects and the most variable of vertebrates, Thomas

recommended increasing these values by an order of magnitude. Salmon have relatively high

fecundity (thousands of eggs per female), and therefore have the potential for rapid changes in

population abundance. However, unlike insects, salmonids exhibit age-structure and overlapping

generations that may reduce the variability of adult populations. Thus, salmonids are probably at

the high end of the variability spectrum for typical vertebrate populations. Thomas= guidance

might therefore translate to a recommended Asafe@ population abundance of somewhere between

1,000 and 10,000 spawners per generation, depending on the degree of variability in abundance

of a given population. Thomas= approach is difficult to apply in practice, because it assumes:

1) a complete census of all individuals in a generation, 2) a population that is stable about some

mean value (i.e., no upward or downward trend), and 3) that there are no density-dependent

effects. In 1995, NMFS considered using this method to set recovery abundance goals for Snake

River salmon populations (NMFS 1995b), but found there was not enough data to categorize

populations into Thomas= variability categories except for a number of spring/summer chinook

salmon stocks; these stocks exhibited relatively low variability during a thirteen-year period.

From this, and considering that the period for which there were data was one of relatively stable

environmental conditions, NMFS concluded that variability for these populations was probably

low to moderate and thus that safe abundance levels would fall between 1,000 and 5,500

spawners per generation. Considering that spring-summer chinook exhibit a 4- or 5-year mean

generation time, these 1,000 to 5,500 spawners per generation would translate to 200-250 to

1,100-1,375 spawners per year.

Genetic processes

There is a great deal of theoretical work relating population size to genetic diversity and

the degree of inbreeding within populations (e.g., Wright 1938, Crow and Kimura 1970). There

is also a body of theoretical work on the relationship between population size and the

accumulation of deleterious alleles and how this affects population viability (Lynch 1990, Lande

1994, 1995, Schultz and Lynch 1997), and on the relationship between the amount of genetic

diversity in a population and the population's fitness (e.g., Burger and Lynch 1995). Many of the

classical population genetic models that have been used to explore the relationships between

population size and level of inbreeding or genetic drift do not accurately correspond to typical

salmon life-histories (i.e., semelparous overlapping generations). Waples and Teel (1990) and

Waples (1990a, 1990b) have updated some of these models to take into account typical salmon

life-history patterns. One aspect of salmon biology that has not received a great deal of attention

in population viability models is the fact that salmonids are partially tetraploid (reviewed by

Allendorf and Thorgaard 1984). Partial tetraploidy may reduce the severity of inbreeding

depression in comparison to the amount seen in diploid organisms because, all else being equal,

tetraploids are less homozygous than diploids. However, inbreeding depression has been

observed in Pacific salmonids, so the benefits of tetraploidy might not be substantial.

One important concept from population genetic theory is that of effective population size

, Wright 1931, Crow and Kimura 1970). The effective size of a population is defined as the

size of an idealized population that would produce the same level of inbreeding or neutral

variance in allele frequency (genetic drift) seen in an observed population in which one is

interested (see Hartl and Clark 1989, Caballero 1994 for reviews). Attributes of such an

idealized population typically include discrete generations, equal sex ratios, binomial variance of

reproductive success, random mating, constant breeding, population size, and non-selective

gamete-to-adult mortality. Violating any of these attributes usually increases the rate of

inbreeding or drift in comparison to the idealized case, and therefore reduces effective size of the

population of interest. Effective size can be defined with respect to either inbreeding or genetic

drift, as well as other population attributes (reviewed by Crow and Kimura 1970). For example,

the inbreeding effective size refers to the size of an ideal population that would produce the same

rate of inbreeding as an observed population. The variance effective size refers to the size of an

ideal population that produces the same rate of allele frequency change due to genetic drift as an

observed population. The inbreeding and variance effective sizes are usually nearly identical

when populations are of constant size, but they generally differ when populations are growing or

shrinking (e.g., Crow and Kimura 1970, Ryman et al. 1995). Because almost no natural

populations are “ideal,” the effective size of a population is almost always smaller than the

observed number of breeding individuals (reviewed by Frankham 1995). Effective size is a

useful concept because it is a parameter that can be estimated from real populations, and

therefore, acts as a bridge between the complexities of real populations and the necessarily

simplified assumptions made in population genetic models. The effective size of a population

generally refers to the population size of an entire generation, whereas the effective number of

breeders (N

) refers to the effective number of individuals in a single year (Waples and Teel

1990, Waples 1990a). For species or populations with multiple overlapping age classes (i.e., all

the Pacific salmon species expect for pink salmon and perhaps some coho salmon populations),

the effective population size per generation is approximately equal to the geometric mean of the

effective number of breeders per year multiplied by the number of years in a generation (itself

approximately equal to the mean age at reproduction). For example, a population with a four-

year generation time and a geometric mean N

of 100/year would have a N

of ~400/generation

(Waples 1990b).

Several examples from fruit flies (reviewed by Wright 1977) and Tribolium (Rich et al.

1979) in laboratory settings give empirical evidence of the relationship between population size

and the rate of loss of genetic diversity. In natural populations, there are several examples of low

levels of genetic diversity associated with known severe bottlenecks; these include elephant seals

(Lehman et al. 1993), prairie chickens (Bouzat et al. 1998a, b), and koalas (Houlden et al. 1996).

Quinn et al. (1996) found lower levels of protein and DNA heterozygosity in New Zealand

chinook salmon than in the Californian founder population, a result consistent with a known

founder effect. Waples and Teel (1990) discuss year-to-year changes in protein allele

frequencies in Oregon hatcheries that they attributed to genetic drift due to small broodstock

sizes. In a literature survey of genetic variation in many species, Nei and Graur (1984) found a

small but statistically significant correlation between the level of protein heterozygosity and

species-wide abundance. Latter et al. (1995) measured the rate of inbreeding at small effective

populations sizes in Drosophila and found that inbreeding as measured by homozygosity at

specific marker loci increased somewhat more slowly than predicted by theory, although

substantial inbreeding depression was still observed.

The empirical evidence to date points to a complicated relationship between levels of

genetic diversity (as measured by protein or DNA heterozygosity) and population fitness or

viability. O'Brien et al. (1983, 1985) suggested that very low levels of heterozygosity in the

South African cheetah were correlated with poor fitness (low sperm counts, susceptibility to

disease), but other authors have disputed these conclusions (e.g., Pimm 1991). In the case of the

prairie chickens cited, a population that experienced a severe reduction in abundance was less fit

than two other populations that maintained large population sizes (Bouzat et al. 1998a, b). Leary

et al. (1983, 1985a, b) found a correlation between bilateral asymmetry and level of protein

heterozygosity in rainbow trout and several studies have found correlations between protein

heterozygosity and fitness in marine invertebrates (e.g., Koehn et al. 1988). Saccheri et al.

(1998) found a significant correlation between extinction and level of protein heterozygosity

(itself correlated with inbreeding level in this study) in natural populations of butterflies in

Finland. Quattro and Vrijenhoek (1989) found a significant correlation between protein

heterozygosity and fitness in a Sonoran topminnow. Frankham (1995) summarized studies of

genetic variation in a large number of endangered species and found lower than average levels of

genetic variation in most of them. The predominance of sexual species and the relative rarity of

old asexual lineages also argue for the general importance of genetic variation to animal

populations (reviewed by Judson and Normark 1996). Although these and other studies certainly

suggest that there is relationship between genetic diversity and population fitness across a wide

variety of organisms, there are notable examples where a lack of genetic diversity has not

stopped a population from persisting or growing, at least in the short term. In the case of the

elephant seal cited, for example, the species has recovered from about 100 individuals in the late

1800's to 125,000 in 1989 (discussed in Caughley and Gunn 1996) despite its near lack of

measurable genetic diversity. E.H. Bryant and co-authors have published a series of papers on

the quantitative genetic effects of repeated bottlenecks in populations of houseflies and found

that additive genetic variation in the morphological traits they measured generally was not

reduced by even very severe bottlenecks (e.g., Bryant and Meffert 1990). Finally, in many cases

it is difficult to untangle the effects of genetic diversity on population fitness from the effects of

inbreeding on population fitness, because genetic diversity and inbreeding are often correlated.

A reduction in fitness associated with inbreeding (inbreeding depression) has been

demonstrated for a wide variety of organisms in captivity (reviewed by Wright 1977, Thornhill

1993, Falconer and Mackay 1996, Lynch and Walsh 1998); these include rainbow trout and

Atlantic salmon (reviewed by Tave 1993). Further examples of an association between

inbreeding and a reduction in fitness or extinction in natural habitats include the evening

primrose Clarkia pulchella (Newman and Pilson 1997), white-footed mice (Jimenez et al. 1994),

and Finnish butterflies (Saccheri et al. 1998) and Sonoran topminnows (Quattro and Vrijenhoek

1989). Frankham (1998) estimated inbreeding coefficients in 210 populations found on islands

and concluded that inbreeding depression was sufficiently high to be a contributing factor in the

higher rates of extinction that island populations experience in comparison to mainland

populations. Several studies have shown that the rate of inbreeding (as opposed to the final level

achieved) has an effect on inbreeding depression. For example, Latter et al. (1995) and Latter

(1998) found that when Drosophila were inbred at a slow rate (~1%/generation), the resulting

inbreeding depression was less than the level of depression observed in other studies where the

rate of inbreeding was higher. About 25% of the populations that inbred at this slow rate

nevertheless went extinct due to genetic causes within a period of 200 generations.

The conservation literature contains a number of recommendations on guidelines for the

minimum population sizes needed to avoid deleterious genetic effects. All of these

recommendations are based on effective number of breeders per generation and for a number of

reasons must only be considered as rough guidelines. In order to avoid inbreeding depression,

Franklin (1980) suggested, based on empirical data from animal breeding studies, that an

effective size of 50 was reasonable. Franklin (1980) and Soulé (1980) suggested that an

effective population size of 500 is necessary to retain sufficient genetic variation for long-term

population persistence. This value is based on estimates of the rates at which variation in an

isolated population is added by mutations, and removed by drift and selection. Recently Lande

(1995) has pointed out that because ~90% of new mutations are unconditionally deleterious (and

therefore do not contribute to future evolution), a more realistic long-term minimum viable

effective population size may be closer to 5,000 per generation. Using similar arguments, Lynch

(1996) has suggested that an effective size of ~1,000 is usually large enough for a population to

maintain genetic variation. Based on the probability of losing rare alleles, Waples (1990a) has

suggested that 100 effective breeders/year is necessary to maintain genetic variation in salmon

populations in the short-term. These general recommendations have some limitations that must

be understood before they are used to help determine levels of abundance necessary for viable

salmonid populations. First, they are based on models of single, reproductively isolated

populations. As the term is used in this document, a population is substantially reproductively

isolated, but may receive low levels of gene flow from other populations. Gene flow, like

mutation, is a source of genetic variation, so it is likely that populations connected by gene flow

will have somewhat smaller minimum genetically viable population sizes than completely

isolated populations. Second, the genetic parameters that form the basis for the Franklin (1980)

and Lande (1995) recommendations were estimated from data obtained from only one species

(Drosophila melanogaster), and must therefore be regarded as preliminary. Nonetheless, these

recommendations may be reasonable as starting points for determining the minimum abundance

necessary for long-term genetic viability, especially in the absence of additional information.

In order to convert these recommendations of effective population size per generation to

salmon spawning abundance per year, it is necessary to know the ratio of the effective number of

breeders to the observed number of breeders (N

/N ratio) and the generation time for the

population in question. Several studies suggest that a N

/N ratio of 0.3 is approximately correct

for salmon and steelhead in general (see following AAssessment Methods@). With this ratio, the

recommended minimum long-term genetically viable population sizes discussed above range

from 1,670/generation (Franklin 1980 and Soulé 1980) to 16,700/generation (Lande 1995). The

minimum spawning size recommended by WDFW (1997) falls in this range (3,000/generation).

For populations that spawn at multiple age classes, the spawners/generation value must be

divided by the generation length (median age of reproduction) to obtain the corresponding

numbers of spawners per year. For example, many chinook salmon populations have about a

four-year generation time (reviewed by Healey 1991). A range of ~417 to ~4,170 breeders per

year, therefore, may be reasonable minimum values for maintaining sufficient genetic diversity

to ensure long-term persistence of chinook salmon populations. Based on genetic evidence,

Allendorf et al. (1997) concluded that salmon populations with N

below 500 (or N below 2,500)

per generation would be at high risk and populations with N

below 50 (or N below 250) per

generation would be at very high risk. Wainwright and Waples (1998) noted that if demographic

factors were included, thresholds for these categories would be higher, but they did not suggest

specific values.

Demographic stochasticity

Demographic stochasticity refers to the effects of apparently random events associated

with mate choice, fecundity, fertility, and sex ratio and tend to cause higher extinction risks in

small populations than in large ones. Mathematical theory for extinction due to random variation

in birth and death rates among individuals in a population was first developed in the 1930s (see

Goodman 1987, Gabriel and Bürger 1992), but was first widely applied in developing island

biogeography theory (MacArthur and Wilson 1967, Richter-Dyn and Goel 1972). The

importance of sex-ratio stochasticity has only recently been recognized and assessed (Gabriel

and Bürger 1992). Most theory addressing demographic stochasticity has made the simplifying

assumptions of density-independence and discrete generations with no age structure (e.g.,

Goodman 1987, Belovsky 1987, Lande 1993), although some has included density-dependence

(e.g., Tier and Hanson 1981, Gabriel and Bürger 1992). In isolation, both birth/death rate

variation and sex-ratio variation are important factors only at very low abundance (less than

about 30 individuals), but in combination they become important at somewhat higher abundance

(Gabriel and Bürger 1992), and even more importantly, when combined with environmental

stochasticity (Tier and Hanson 1981, Goodman 1987). Lande (1998) demonstrated that

demographic stochasticity reduces the mean rate of population increase (a form of depensation,

as discussed previously), and this effect becomes stronger as population size declines.

Thus, it is clear that these effects are predominant at very low population abundance, so

they are probably more important in establishing critical abundance levels than viable levels

(which are more influenced by environmental stochasticity). Effects of demographic

stochasticity increase roughly exponentially as population size declines (Goodman 1987, Lande

1993, 1998), but there is no clear critical point at which one can separate high-risk from

moderate risk for a particular species or population. However, demographic stochasticity should

be considered a risk factor for any population with abundance below a few hundred individuals.

Ecological feedback

Salmon are also important components of their ecosystems, and salmon population size is

expected to have an effect on ecosystem function, which, in turn reflexively affects salmon

abundance. In a sense, ecological feedback is similar to density-dependent processes, but it

emphasizes the role of salmon in modifying their physical and biological environment. In

contrast to classical density-dependence, ecological feedback typically involves time lags on the

order of a generation or longer. Examples of such feedback include the role of salmon carcasses

in riparian zone nutrient cycles, and the effect of spawning salmon on the quality of spawning

gravels. Both these processes contribute to the success of future salmon generations, but are

only significant at relatively high population densities.

Consider first the role of salmon in riparian zone nutrient dynamics. Salmon spawning

areas are often nutrient poor, and salmon carcasses are an important pathway by which marine

nutrients (important for both juvenile salmon and other species) enter both freshwater and

terrestrial ecosystems (Kline et al. 1990, 1993, 1997, Bilby et al. 1996, 1998, Wipfli et al. 1998).

There may be a critical average abundance level below which salmon production is nutrient

limited, but this level is likely to vary according to local habitat conditions, and is clearly

difficult to quantify. Similarly, redd-construction activities affect the size-distribution and

stability of spawning-area sediments (Montgomery et al. 1996, Committee on Protection and

Management of Pacific Northwest Anadromous Salmonids 1996). Through this mechanism,

abundance in one generation can affect the survival of future generations.

These processes are only two among many interactions between salmonids and their

environment. Populations with abundance levels at which these interactions become effective in

maintaining natural ecosystem processes have been termed Aecologically functional populations@

(Connor 1988). Maintaining enough salmon to provide this level of function in freshwater,

estuarine, and marine environments will promote ecosystems with a full suite of population

regulatory mechanisms that may be absent when populations are artificially depressed.

Assessment Methods

Salmon population abundance can be estimated in a variety of ways. Adult dam and weir

counts, redd counts, spawner and carcass surveys, harvest estimates, and juvenile counts have all

been used to estimate population abundance (see “Diversity,” p. 19, for relevant references). The

quantity and quality of data vary considerably across species and geographical regions (e.g.,

Weitkamp et al. 1995, Busby et al. 1996, Myers et al. 1998) and accurate abundance estimates

are only available for a small fraction of the salmon populations inhabiting the Pacific Northwest

and California (e.g., WDF et al. 1993, Kostow et al. 1995). In some cases, estimates may be

available for all of the populations in a river basin as a group, but not for each population

individually (e.g., adult counts of steelhead at Columbia River dams). In other cases, accurate

abundance data may be only available for a small number of index streams or stream reaches,

and these estimates are used as proxies for other, presumably similar, populations or are

expanded to produce basin-wide estimates. Although breeding population size estimates are

usually most relevant for assessing status, juvenile abundance estimates may be the only

information available for some populations. While juvenile data can be used to assess status,

translating that data to adult abundance is not straightforward and requires substantial knowledge

of local conditions—including typical survival rates and limiting factors. It is also important to

recognize that the ESA and NMFS policy both focus on natural population abundance and the

presence of naturally spawning hatchery fish can greatly complicate the interpretation of

abundance.

Directly assessing a population's genetically effective size is usually more complicated

than assessing abundance because effective size is based not only on the number of spawning

fish but also on their variance in reproductive success (see Geiger et al. 1997). There are

typically two ways in which a population's current effective population size can be directly

measured. The first, reviewed by Caballero (1994), involves estimating effective population size

from demographic parameters such as abundance, sex ratios, mating systems, and the fecundity

distribution among individuals. The second general method involves estimating effective

population size from patterns of genetic variation (reviewed by Waples 1991b). Neither method

is routinely easy to apply, and it is likely that for practical purposes a population=s effective size

will often be inferred from the number of its spawners. In order to do this, it is necessary to have

a general idea of the relationship between the effective number of breeders (N

) and the actual

number of spawners in a population. N

/N ratios estimated from six populations of Snake River

spring chinook and one population of Sacramento River winter chinook range from a low of

0.013 (Bartley et al. 1992) to a high of 0.7 (Waples et al. 1993, R. Waples

) and average

approximately ~0.2 to ~0.4. The large range is most likely due both to large sampling errors in

estimating N

and real biological differences among populations.

Robin Waples, NWFSC, 2725 Montlake Blvd. E, Seattle, WA 98112. Pers. comm.

Guidelines

In light of the preceding discussion, we have developed a set of guidelines (following) for

assessing adequacy of a population=s abundance. Note that the ESA=s primary focus is on natural

populations in their native ecosystems (e.g., Waples 1991c, Hard et al. 1992). When evaluating

abundance for the purposes of determining VSP status it is essential to focus on naturally

produced fish (i.e., the progeny of naturally spawning parents). Because risk to populations

depends largely on specific life-history strategies and the local environment, it is not possible to

use simple population theory to rigorously determine adequate abundance levels. For this

reason, the following guidelines prescribe factors that need to be considered, but do not provide

specific numerical criteria.

The following guidelines are separated into two general categories: 1) those defining a

viable population, and 2) those identifying critically low abundance. A population would be

considered to have viable abundance only if all of the viability guidelines are met, i.e., the

guideline requiring the highest abundance in any given situation will effectively determine

whether that population is considered viable. A population would be considered to be at

critically low abundance if any of the critical guidelines were met. Also, note that different

elements are likely to dominate decisions for different populations—for example, viability

Guidelines 2 and 4 (Box A2) would be most restrictive for populations with extensive habitat

available, while 1 and 3 would be most restrictive where habitat capacity is limited. In addition,

environmental variation (Guideline 1) will often dictate a minimum population size that is larger

than genetic concerns would (Guideline 3).

Population Growth Rate and Related Parameters

Introduction

In this Appendix we discuss population growth rate (productivity

) and factors that affect

population growth rate as parameters for assessing population viability. We also review how

estimates of population growth rate and related parameters are considered in the context of

processes that occur at larger spatial and temporal scales, and in relation to other parameters

discussed in this document. Our purpose here is not to explore the myriad mechanisms which

influence population growth rate and drive changes in population size, but rather to emphasize

the importance of detecting such changes and how estimates of population growth rate and

We use the terms “population growth rate” and “productivity” interchangeably when referring to production over

the entire life cycle. We also refer to “trend in abundance” which is the manifestation of long-term population

growth rate.

Box A2. Viable Population Size Guidelines.

1. A population should be large enough to have a high probability of surviving

environmental variation of the patterns and magnitudes observed in the past and

expected in the future. Sources of such variation include fluctuations in ocean conditions

and local disturbances such as contaminant spills or landslides. Environmental variation and

catastrophes are the primary risks for larger populations with positive long-term average

growth rates.

2. A population should have sufficient abundance for compensatory processes to provide

resilience to environmental and anthropogenic perturbation. In effect, this means that

abundance is substantially above levels where depensatory processes are likely to be

important (see following Critical Guideline 1, Box A3) and in the realm where compensation

is substantially reducing productivity. This level is difficult to determine with any precision

without high quality long-term data on population abundance and productivity, but can be

approximated by a variety of methods.

3. A population should be sufficiently large to maintain its genetic diversity over the long

term. Small populations are subject to various genetic problems (including loss of genetic

variation, inbreeding depression, and deleterious mutation accumulation) that are influenced

more by effective population size than by absolute abundance.

4. A population should be sufficiently abundant to provide important ecological functions

throughout its life-cycle. Salmonids modify both their physical and biological

environments in various ways throughout their life cycle. These modifications can benefit

salmonid production and improve habitat conditions for other organisms as well. The

abundance levels required for these effects depend largely on the local habitat structure and

particular species biology.

5. Population status evaluations should take uncertainty about abundance into account.

Fish abundance estimates always contain observational error, and therefore population

targets may need to be much larger than the desired population size in order to be confident

that the guideline is actually met. In addition, salmon are short-lived species with wide year-

to-year abundance variations that contribute to uncertainty about average abundance and

trends. For these reasons, it would not be prudent to base abundance criteria on a single high

or low observation. To be considered a VSP, a population should exceed these criteria on

average over a period of time.

Box A3. Critical Population Size Guidelines.