Selection equilibrium for hatchery and wild spawning fitness in integrated breeding programs

Daniel Goodman
Can. J. Fish. Aquat. Sci./J. Can. Sci. Halieut. Aquat. 62(2): 374-389 (2005)

Abstract: Some salmon hatchery programs intentionally integrate the wild and hatchery population by taking naturally spawned fish as some fraction of the broodstock and allowing hatchery progeny to constitute some fraction of the adults spawning in the wild. This circumvents some ecological concerns about the effects of hatchery fish on the "wild" population while still reaping some of the benefits of increased potential for harvest, but it increases some genetic concerns. Here, we model phenotypic evolution in the integrated population to investigate the effects on natural spawning fitness at the joint selection and demographic equilibrium. We find a potential, but not a certainty, depending on quantitative aspects of the management interacting with biological characteristics of the stock, for substantial erosion of natural spawning fitness, compared with the original wild population, including the possibility of runaway selection driving natural spawning fitness effectively to zero. The vulnerability to such evolutionary deterioration increases with the magnitude of the contribution of hatchery breeding to the total production and increases with harvest. The response of the selection equilibrium to increasing contribution of hatchery progeny to the broodstock can exhibit a catastrophic discontinuity.

389

Introduction

Hatchery husbandry practices have advanced to the point where hatchery propagation of salmon can be an effective means to produce smolts for release to the wild, where they will rear, grow, and mature and return for harvest. While the smolt to adult return rates of hatchery salmon are often lower than those of the corresponding wild fish, the egg to smolt survival rates are generally enough higher in the hatchery, compared with the wild, to more than compensate. This results in a net numerical life cycle advantage for the hatchery stock in hatchery propagation, allowing a higher harvest rate. A salmon hatchery program that is successful from the standpoint of production for harvest may, nevertheless, be judged a liability because of harmful effects on wild

salmon sharing the same natural habitats.

The negative ecological effects of hatchery production on a wild stock with which it interacts can include competition and predation (Waples 1999; Levin and Williams 2000; Levin et al. 2001). An unselective harvest managed for the higher productivity of the hatchery stock will overharvest the wild stock (Hilborn 1992).

Straying of hatchery adults into the wild spawning grounds raises the potential for harmful genetic effects as well. Hatchery propagation of a hatchery stock gives rise to domestication

selection, which results in accumulation of adaptations to hatchery conditions, with attendant deterioration of performance under natural conditions during the corresponding portion

of the life cycle. Further, the postrelease life history that evolves to mesh with the hatchery rearing may not prove adaptive when fish from the hatchery stock spawn in the wild.

There is also a potential for inbreeding depression and loss of genetic diversity if the effective population size of an isolated hatchery stock is too small. Introgression between a domesti-

cated or inbred hatchery stock and a wild stock will depress the fitness of the wild stock.

Concerns about the harmful effects of hatchery salmon stocks on their wild counterparts have prompted consideration of integrated production programs, where there is no importation of hatchery broodstock from other basins, where adults of local natural spawning origin are included among the hatchery broodstock in each generation, and a substantial number of adults of hatchery spawning origin are allowed to migrate to the natural spawning grounds. If the hatcheryorigin

fish spawn successfully with the natural-origin fish in the natural spawning grounds, the hatchery- and naturally spawning populations will converge genetically. In the end, there will be just one integrated population, with individuals that are derived from both hatchery and natural spawning in the previous generation and where fish of hatchery or natural spawning origin may cross over to the respective alternative spawning environment in any generation.

The hope, in advocating integrated programs, is that the merging of the hatchery and wild stocks will reduce the degree of domestication and inbreeding, which is a reasonable expectation, since a totally distinct hatchery stock is unlikely to diverge under these conditions and the effective population size is that of the integrated aggregate. It is still meaningful, however, to compare the properties of the population before and after such a program is initiated. In particular, we might inquire into changes in natural spawning fitness by comparing the natural spawning fitness in the absence of a hatchery influence with the natural spawning fitness that obtains at the selection equilibrium when a defined integrated production program is in effect. The present analysis will

address that question by means of a model of demographic dynamics and phenotypic selection in a managed integrated production program.

Several versions of integrated salmon hatchery programs have been proposed, and several are underway (e.g., Carmichael and Messmer 1995; Dauble and Watson 1997; Fast and Craig 1997). Variations have to do with whether the intervention is intended to be temporary or indefinite, with

the intended management of harvest, and with intended quantitative limits on the fraction of the naturally spawned run that may be taken as hatchery broodstock, the fraction of the broodstock that may be progeny of hatchery spawning, and the fraction of the adults on the natural spawning ground that may be progeny of hatchery spawning. The programs are variously called supplementation, augmentation, or supportive breeding, but the definitions are not consistent, and

these terms are sometimes applied also to more conventional hatchery production programs that are still thought to serve some conservation role.

The question of the effects of integrated production on natural spawning fitness has been articulated previously (Campton 1995; Currens and Busack 1995; reviewed in Bilby et al. 2003). The theoretical possibility of the phenomenon has been treated explicitly in a number of modeling

analyses. These models include one-locus two-allele classical genetic models of essentially antagonistic selection in the respective hatchery and natural environments (Reisenbichler

1984; Byrne et al. 1992; Harada 1992), single-trait quantitative genetics models specifically for a trait that can be conditionally deleterious in the respective hatchery and natural environments (Adkison 1995; Hard 1995; Ford 2002), models that also consider mutation and drift (Lynch and O’Hely 2001), and a more complicated quantitative model assuming some degree of coadaptation of traits (Emlen 1991) in the respective environments. With the exception of the analysis by Lynch and O’Hely (2001), the more comprehensive of these analyses were carried out by simulation or numerical solution for selected parameter values. These models show that evolution towards lower natural spawning fitness is indeed a theoretical possibility in integrated production programs.

In the present analysis, we focus exclusively on the effects of selection in the integrated population, ignoring the possible drift and inbreeding effects attributable to small population

size. Further, we will assume that the set of geographic locations where broodstock is collected and where hatcheryphase products are released are chosen so that the hatchery phase does not merge multiple locally adapted stocks or inadvertently take broodstock selectively. Finally, we will not consider populations so near extinction that an immediate genetic or demographic rescue is needed.

The analysis will proceed in two main sections. First will be analysis of a model just of the demographic dynamics of a harvested integrated hatchery production system, operating

under defined management protocols, primarily to obtain solutions for demographic equilibrium. The model will then be extended to consider the evolutionary equilibrium in such a system when the population encompasses a range of possible genotypes with different fitnesses in both the hatchery and natural spawning phases (Goodman 2004). The goal is to develop a comprehensive analysis that considers fitness as expressed in population dynamic terms in the context of a

model that includes the effects of harvest and the controls on the breeding and release program.

In the interests of generality, the representation of particular genetic mechanisms will be bypassed in favor of a fitness set analysis that simply considers the spectrum of possible genetically determined phenotypes (Levins 1962; Schluter and Nychka 1994) and solves for the selection equilibrium as an evolutionarily stable state (Maynard Smith and Price 1973). This approach has the advantage of requiring fewer assumptions and fewer parameters. Basically, the evolutionary

model is phenotypic, not genotypic, although it assumes heritability of phenotype. The assumptions of the model are unexceptional: (i) it postulates that evolution will maximize

the overall effective fitness in the system, (ii) it diagnoses, via the demographic model, the respective contributions of hatchery spawning phase and natural spawning phase fitness

to the overall fitness, and (iii) it assumes that there are inevitable trade-offs between achieved hatchery spawning fitness and natural spawning fitness. The analysis will dissect the

relationship between the management controls of an integrated breeding protocol and the resulting relative weights that hatchery spawning fitness and natural spawning fitness will have in their respective contributions to overall fitness.

These weights then determine the selection forces that will emphasize one fitness component at the expense of the other. The generality of the approach does exact some cost in

the range of properties about which the model can make predictions. Because the model makes no assumptions about genetic mechanism, the results only can characterize the selection equilibrium. The analysis is silent about questions of the speed of the selection response, or the speed of the possible reverse evolution when supplementation is terminated. Further, to promote generality, and hopefully to crystallize an efficient representation of the factors that control the fitness

effects of selection in an integrated production system, the emphasis will be on analytical solutions, where possible, and where analytical solutions are not available, the results of systematic numerical explorations will be displayed to identify critical regions of the parameter space and to confirm robustness of the results.

Discussion

Integrated hatchery breeding programs have been proposed as a possible solution to the problems posed by the negative effects of hatchery stocks on the wild stocks with which they interact. Once the integrated breeding program has been in place, the wild stock will no longer be distinct,

so there is no point to asking whether the hatchery phase of the integrated population has a negative effect on the fitness of the natural spawning phase. This does not, however, preclude

comparison of the characteristics of the original wild stock, before it came under hatchery influence, with the characteristics of the population that results from the integrated hatchery production. The present modelling analysis conducts such a comparison, where the characteristic of interest is the intrinsic replacement rate realized from natural spawning.

The adverse selection phenomena are the result of selection for traits that are adaptive in the hatchery-reared life cycle but are not optimal for a natural spawning life cycle. A

significant component of the selection during the hatcheryreared life cycle may occur as life history selection after release, acting on life history consequences of the individual’s state at release. Thus, strong selection, different from the selection on natural spawning fish, can occur during the course of a hatchery-reared generation, even if mortality within the hatchery is slight. Assuming that the wild population’s original life history is at an adaptive peak, other selection on life history traits will move the phenotype off that peak, diminishing performance under conditions experienced by the wild population.

By representing the available phenotypes as a fitness set characterized by biologically possible combinations of hatchery replacement rate and natural spawning replacement rate, and combining this with the demographic model of the integrated production system, the analysis here has obtained a closed-form solution for the selection equilibrium that results. We find that integrated breeding has the potential for large depression of natural spawning fitness, potential for runaway selection driving the natural spawning fitness to zero, and potential for a discontinuous knife-edge response to gradual increases in the contribution of hatchery-spawned fish to the hatchery production of the integrated population.

The previous literature on evolutionary models of integrated production shows a potential for selection to reduce natural spawning fitness (Adkison 1995; Lynch and O’Hely 2001; Ford 2002). The present analysis generalizes and extends those results, showing a spectrum of selection responses, including runaway selection and a knife-edge response to increasing supplementation, and revealing the parameter ranges for which they occur. All of the models are in agreement that a reduction in natural spawning fitness is possible and that the reduction could be large, which lends credence to the concerns about large-scale implementation of integrated production programs and raises grave questions about the suitability of supplementation as a conservation

measure for weak stocks that are still self-sustaining (Goodman 2004).

The capacity of the model developed here to solve explicitly for the selection equilibrium as a function of the management controls motivates us to inquire whether we can, with present knowledge, reasonably predict, quantitatively, the expected amount of depression of natural spawning fitness in a defined integrated breeding program.

The model of demographic equilibrium used here is sufficiently specified by six straightforward quantities: adult to adult replacement rates, harvest rates, and broodstock removal rates for both the hatchery spawned and naturally spawned fish. These are readily measurable, and while they

have not all been measured systematically in well-controlled actual implementations of integrated production, we have enough experience with such measurements to know the range of values to expect for the initial condition (before selection has proceeded to a substantial extent).

The model for the selection equilibrium requires one more term, α, describing the negative correlation, at a selection limit, between natural spawning and hatchery spawning replacement

rates. This term reflects an aspect of the set of biological possibilities for the stock. We have found that the results for any particular scenario examined with the model are strongly affected by this term. In contrast with the demographic quantities in the model, this trait correlation has

never been measured for salmon in an integrated breeding program, and in fact, except for theoretical assurance of a negative correlation, we do not have a basis for expecting values in one range of magnitudes or another across the spectrum of values that give rise to quite different quantitative results for the selection equilibrium.

There are a modest number of stocks for which natural spawning and hatchery spawning replacement rates have both been measured. The collection of such empirical (Ra,Rw) pairs would indicate what the range of possibilities may be for a range of stocks over a range of environments, but the slope of the border of this collection of points is not the slope of the trade-off within a given stock for a given environment, and it is the latter that we need to characterize α

quantitatively.

The local magnitude of the trait correlation term could in principle be back-calculated from measurements of the demographic quantities at the selection equilibrium, if we had them. But deliberate integrated production is a relatively new idea, and systematic empirical measurements have never been undertaken to monitor the evolution of natural spawning replacement rates, over many generations, in a program following a defined protocol. Practically, such measurements

would also need to be made over enough brood years to average out the effects of environmental variation.

Nevertheless, if supplementation is going to be deployed on an “experimental” basis, these are the quantities that need to be measured for the experiment to be informative.

The modelling results from this paper show that this experiment has the potential for a significant downside. It would be prudent to proceed cautiously in implementing integrated production with careful experimental design and systematic monitoring to determine how frequent and severe

the depression of natural spawning fitness really is. In particular, since the modelling shows that the depression of natural spawning fitness will increase with the magnitude of the hatchery contribution to total production, it would be good to determine empirically whether specific policy caps on the amount of hatchery contribution can limit the fitness erosion to a tolerable level.