Expected rate of fisheries-induced evolution is slow

Abstract

Commercial fisheries exert high mortalities on the stocks they exploit, and the consequent selection pressure leads to fisheries-induced evolution of growth rate, age and size at maturation, and reproductive output. Productivity and yields may decline as a result, but little is known about the rate at which such changes are likely to occur. Fisheries-induced evolution of exploited populations has recently become a subject of concern for policy makers, fisheries managers, and the general public, with prominent calls for mitigating management action. We make a general evolutionary impact assessment of fisheries by calculating the expected rate of fisheries-induced evolution and the consequent changes in yield. Rates of evolution are expected to be ≈0.1–0.6% per year, and the consequent reductions in fisheries yield are <0.7% per year. These rates are at least a factor of 5 lower than published values based on experiments and analyses of population time series, and we explain why the published rates may be overestimates. Dealing with evolutionary effects of fishing is less urgent than reducing the direct detrimental effects of overfishing on exploited stocks and on their marine ecosystems.

The intensification of trawl fisheries after the Second World War resulted in an increase in fishing mortality (1), and selection pressure therefore also increased significantly for many commercially exploited fish stocks. Strong selection may lead to an evolutionary response that reduces productivity and is detrimental for the yield of the fishery (2). An explicit theoretical calculation demonstrating that fisheries-induced evolution can lower the age of maturation did not appear until recently (3). The focus since that work appeared has been on statistical data analysis of time series of size or age at maturation and growth rate, which show rapid changes in some stocks. These changes have been attributed to genetic change and therefore interpreted as a signature of fisheries-induced evolution (4, 5). Experimental evidence showing a strong and rapid response to fisheries-like selection pressure lends credibility to this interpretation (6). That selection leads to evolution is not in doubt, however, the relevant question in relation to management is how detrimental and rapid fisheries-induced evolution is compared with the direct detrimental effects of overfishing (7). The answer is a critical element of any formal evolutionary impact assessment of fishing activity (4, 8). Theoretical studies of fisheries-induced evolution have focused on concepts and on finding the optimal value of a trait under selection (3, 9–11). We are aware of only 3 studies that have attempted theoretical quantifications of the rate of evolution (the selection response) (11–13). All were targeted toward specific cod stocks and found selection responses on the size at maturation smaller than −0.3% per year.

Here, we generalize a simple quantitative genetics calculation of the selection response to fishing for a given stock (11) to work for any stock, by formulating the selection response as a function of life-history parameters, most notably size at maturation and the ratio between adult mortality and von Bertalanffy growth rate M/K. This allows us to calculate the expected selection response and the expected impact on fisheries yield for fish in general. The basis of the method is a calculation of the lifetime expected reproductive output R₀ as a function of a trait φ (see Materials and Methods). From this the rate of evolution of the trait (selection response) R can be approximated as (11):

where h² is the heritability of the trait, and σ is the spread of the distribution of the trait. R₀′ is the derivative of R₀ with respect to the trait (the selection gradient), and both terms are evaluated at the current value of the trait φ. The traits being considered are size at maturation, consumption (leading to changes in growth rate), and investment in gonads. We calculate the differential selection response, which is the difference between the selection responses with and without fishing mortality. The differential selection response is normalized by the value of the trait and the generation time: (R(F) − R(0))/(φT), where T is the generation time. This relative differential selection response is a measure of the rate of evolution of a trait and will henceforth be referred to as the selection response for brevity. A description of the detailed calculation of R₀ as a function of life history parameters is given in Materials and Methods.

The selection response of size at maturation is negative, meaning that fishing is expected to reduce size at maturation (Fig. 1A). The selection response is strongest for large fishes, but is not expected to exceed −0.1% per year for any fish stock. For small fishes (size at maturation <1 kg) the selection response of the growth rate is negative, but for large fishes, it is positive. This positive response of growth rate to fishing is contrary to general expectation and is due to fish being caught before maturity. In this case, a higher growth rate increases the probability that an individual will reach maturation before being caught in the fishery. Were fishing to target larger individuals, the selection response on growth rate would be negative [supporting information (SI) Fig. S1], but the rate would remain on the order of 0.1% per year. Investment in gonads shows the strongest response, with expected increases for small fishes of up to 0.6% per year. An increase in investment in reproduction leads to lower somatic growth, lower asymptotic (maximum) size and a lower yield to the fishery. Fishing on fish larger than the size at maturation increases the selection response of growth and investment in gonads (Fig. S1). However, fisheries targeting mature fish predominantly take place during easily accessible spawning aggregations. The resultant selection pressure leads to a response for maturation and growth that is opposite to that due to a pure size-based fishery (Fig. 2 and Fig. S2). Thus, a combination of a spawner fishery targeting only mature individuals and a fishery outside the spawning season targeting both juvenile and mature individuals may lead to mutually canceling selection responses for maturation and growth. The expected loss in yield is dominated by the strong selection response of investment in reproduction and is largest for small fishes (Fig. 1B). For large fishes (size at maturation >1 kg), the change in yield is small, and for the largest fishes, it is positive, due to the expected increase in growth rate.

Fig. 1.

Expected selection response and yield changes due to fishing. (A) Expected relative differential selection response to fishing for changes in size at maturation (blue), growth rate (green), and investment in gonads (red) as functions of size at maturation. (B) Expected relative change in yield given changes in growth and size at maturation for changes in maturation, growth, and investment in gonads seen in isolation (blue, green, and red) and all combined (black). The dashed lines represents ±1 standard deviation calculated from Monte Carlo simulations with random parameter values drawn from distributions specified in Table S1.

Fig. 2.

Expected selection response as a function of fishing pattern. At the left edge the fishery is purely size based, and at the right edge, it is only targeting mature individuals. Size at maturation is 2.5 kg (thick lines) and 250 g (thin lines). Line colors are as defined in Fig. 1.

Can the results of the calculation be trusted? Insofar as we trust basic quantitative genetics, the methods used are fairly standard. The calculation relies on the basic “breeders equation,” which is not accurate for populations with overlapping generations, like most fish species with a size at maturation >10 g. This is countered by using the time to reach maturity to normalize the selection response rather than the average generation time, leading to a small overestimate of the selection response. Traits may be correlated, which will affect the calculated rates either negatively or positively. Because we have no knowledge of the correlation of growth-related traits, it was not possible to take this effect into account. The key aspect of the calculation is the identification of the relevant tradeoffs. For size at maturation and investment in reproduction, these tradeoffs are well established and follow directly from the bioenergetic budget (14). The tradeoffs connected with somatic growth are less well established (15). In the current calculation, it has been assumed that increased growth rates trade off neutrally with higher natural mortality, i.e., that growth divided by mortality is constant (16, 17). Because growth is probably not neutral, this assumption leads to an overestimate of the selection pressure from fishing. The same applies to investment in reproductive effort, where it is assumed that there are no extra costs either in terms of growth (apart from what follows directly from bioenergetic considerations and from increased natural mortality due to slower somatic growth) or mortality associated with changing the reproductive effort. This is probably not the case and, again, leads to an overestimate of fisheries-induced evolution. The most reliable estimate is therefore the selection response on size at maturation; rates of changes in growth and investment in reproduction are likely to be overestimated.

Selection responses have been estimated for some wild-fish populations yielding rates of evolutionary change in the range of 0.5.32 kDarwins (4), with a median on the order of 10 kDarwins, corresponding to a selection response of ≈1% per year (see SI Text for a definition of Darwins). These rates are an order of magnitude larger than our theoretically expected selection responses for size at maturation and growth rate. The question is then why the theoretically expected rates are apparently much smaller than those observed in some collapsing fish stocks (18–21) and also much smaller than those from experimental studies (6). Reconciling such different estimates from independent sources (theoretical, field observations, experiments) is a powerful way to improve our understanding of both the science and its implications for management. In our view, some of the assumptions that underlie the field and experimental estimates need to be reevaluated because they may lead to overestimation of rates of evolution.

The NW Atlantic stocks of American plaice (Hippoglossoides platessoides) and cod (Gadus morhua) show the highest rates of evolutionary change (7–32 kDarwins with a median of ≈11) compared with the other stocks (0.5–21 kDarwins with median ≈6) (4). The estimated rates for the NW Atlantic may be unreliable and overestimated because they do not take proper account of (i) geographic variability, (ii) environmental variability, and (iii) timing of changes in fishing mortality. A prolonged period of cooling from 1980 to 1995 had a great impact on the marine ecosystems from northern Labrador to Georges Bank, which are affected by the Labrador Current. It was a major factor in the decline of cod and other species in these areas, because it affected their productivity (22, 23). It also caused substantial southward shifts in distribution of cod (24). Olsen et al. (19) show that geographic differences in the midpoint of the probabilistic maturation reaction norm (PMRN) between stocks are as big as the difference observed over time during the period of stock collapse. For example “the reaction norm midpoints of 6-year-old females from southern Newfoundland were 70 cm as compared with only 50 cm for females from southern Labrador.” Without entering the debate on whether a southward shift in cod distribution represent migration across stock boundaries or not (25), it is evident that geographic effects may have contributed to the observed decline in PMRN midpoint in the NW Atlantic. The breakdown of spatial structure that accompanied the collapse of the cod stocks has also been investigated (26) and raises other doubts about population trends estimated from the trawl surveys. The need to eliminate any possible phenotypic effects becomes critical during such periods of environmental change and the time course of changes in fishing mortality needs to be matched to apparent evolutionary responses.

The experiments by Conover and Munch (6) provide a strong demonstration of the potential effect of fisheries-induced evolution on the growth rate of fish. They performed selection experiments on Atlantic silverside (Menidia menidia), which is a small (W = 10 g) semelparous species. The selection was such that 90% of the 90% fastest growing individuals were removed in one go before spawning. The selection pressure applied in the experiments was presumably intended to resemble a fishery, but is in fact quite different (13). In the experiment the selection is “knife-edge,” acting directly on the trait (here, growth rate), whereas in an actual fishery, the selection is on size and is unlikely to be knife-edge. Even for a short-lived species like the silverside, fishing takes place throughout a season, and slow growers will therefore also be affected, although by a smaller amount than the fast growers. Additionally the size-selectivity of trawls is not sharp, and a range of large and small individuals will be caught. For larger species, the fishery will target several generations, which, again, means that slow growers will also be targeted, albeit to a lesser extent. These experiments demonstrate that rapid selection is possible but greatly overestimate the magnitude of the selection response induced by actual fisheries.

Our theoretical calculation is an attempt at a general evolutionary impact assessment of fishing (4, 8). It demonstrates that strong selection pressure by commercial fishing is expected to induce modest rates of changes in growth-related traits. The discrepancy between the predictions of life-history theory and empirical measurements points to inadequacies in either life-history theory or in the indirect statistical inferences of genetic change or in the underlying data and the assumptions concerning the data. We have pointed to the latter explanation for 1 prominent example and to uncertainties in the tradeoffs used in the life-history calculations. From a fisheries-management perspective, the expected evolutionary changes are slow, compared with the direct effects of exploitation. Our conclusion is therefore that the priority for management should be to control the direct effects of fishing to maximize yields and also protect marine ecosystems. Neutralizing or reversing the effects of fisheries-induced evolution are a worthy longer-term consideration because, from a conservationist perspective, no anthropogenically induced evolution may be considered slow.

Materials and Methods

The calculations rest on the assumption that the population is at ecological equilibrium. For the calculations of the selection response of size at maturation, it is assumed that density dependence acts before maturity, which is a commonly accepted assumption for marine fish (27). For the calculations of selection response on growth rate and yield, the assumption is that density dependence is either multiplicative or acts at the larval stage (28). The calculation of the selection response in Eq. 1 requires a model for growth and reproduction and a specification of mortality. Growth is modeled according to the principles of what has become the standard growth model for life-history analysis of fish (14, 29). Mortality is split into contributions from predation (natural mortality) and fishery.

Juvenile growth rate is hwⁿ − δw, where w is individual weight, hwⁿ is the assimilated consumption, and δw is the amount of energy used for respiration. At the size of maturation m, the individual furthermore invests in reproduction kw. The model leads to a theoretical asymptotic (maximum) size W = ((δ + k)/h)^1/(n−1) = m/η, where η is known to be constant across species (17, 30). The total mass specific energy available for respiration and reproduction can therefore be written as δ + k = h(m/η)ⁿ⁻¹. The division of energy between δ and k is assumed to be such that a fraction 1 − ε of available energy h(m/η)ⁿ⁻¹w is used for respiration and the rest for reproduction. The evolved traits are described with factors varied around a basic value of 1: size at maturation φ_m, growth φ_h, and investment in gonads φ_k. The growth rate can then be written as g(w) = φ_hhwⁿ − [ε + φ_k(1 − ε)H(w − φ_mm)]h(m/η)¹⁻ⁿw, where H is a Heaviside function switching from zero to 1 at the size of maturation φ_mm (Fig. S3A).

Mortality μ depends on individual size and has 2 components: predation mortality, which decreases with size μ_p = α_pwⁿ⁻¹ (31), and fishing mortality starting at size η_FW: μ_F = FH(w − η_FW) (Fig. S3B). A spawner fishery is simulated by an additional component to the fishing mortality: F_sH(w − m). The constant α_p in the natural mortality can be related to the commonly used natural mortality at maturation M = α_pmⁿ⁻¹ (32). The relation between M and growth is customarily expressed as the M/K ratio (17, 30), where K = φ_hhWⁿ⁻¹/3 is the von Bertalanffy growth parameter. This is used to express the mortality constant through the life-history parameter M/K as α_p = (M/K)φ_hhη¹⁻ⁿ/3 (32).

The calculation of expected lifetime reproductive outcome (measured in weight per time) follows the usual procedure (16), with the only change that integration is performed over individual weight and not age (11): R₀ = P_w0→m∫_m^W̃P_m→wφ_kkw/g(w)dw, where the probability of surviving from size a to size b is P_a→b = exp[−∫_a^bμ(w)/g(w)dw], the realized maximum size is W̃ = ((δ + φ_kk)/(φ_hh))^1/(n−1), and k = εh(m/η)ⁿ⁻¹. The generation time used for normalization of the selection response is the expected time to reach maturation, which can be found by integration of g(w) as T = (m/η)¹⁻ⁿln(1 − η¹⁻ⁿ)/(h(n − 1)).

Yield from the fishery is calculated by assuming that recruitment is constant as Y = ∫_ηF^W̃WP_w0→wwμ_F(w)/g(w)dw. The change in yield is found by using the chain rule as: ∂Y/∂t = ∂Y/∂φ∂φ/∂t, with derivatives of yield being evaluated numerically as finite differences. The parameters and their values have been estimated from established cross-species life-history analyses of fish (Table S1). To account for deviations from the “average” life history, Monte Carlo calculations of the selection responses and change in yield have been performed with random values of the life-history parameters sampled from the distributions specified in Table S1.