Aim: Patterns of population renewal in marine fishes are often irregular and lead to volatile fluctuations in abundance that challenge management and conservation efforts. Here, we examine the relationship between life history strategies and recruitment variability in exploited marine fish species using a macroecological approach. Location: Global ocean.

Time period: 1950-2018.

Major taxa studied: Bony and cartilaginous fish.

Methods: Based on trait data for 244 marine fish species, we objectively extend the established Equilibrium-Periodic-Opportunistic (E-P-O) life history classification scheme to include two additional emergent life history strategies: “Bet-hedgers” (B) and Salmonic (S) strategists. B strategists include Rockfishes and other species inhabiting patchy benthic habitats with life histories that blend characteristics of E and P species; they combine very long lifespans with elevated investments in both parental care and fecundity. S strategists are comprised of mostly salmonids that share life history characteristics with E and O species: elevated investments in parental care reminiscent of E strategists, but with reduced fecundity and short lifespans characteristic of O species. We analyzed how the E-B-P-O-S life history classification mapped onto patterns of recruitment variability observed in population time series data (n = 156 species).

Results: Generalized linear models suggest that life history strategy explains a modest, yet significant amount of recruitment variability across species. Greater predictive power arose after controlling for increased recruitment variance associated with variable fishing pressure, with O strategists showing the strongest sensitivity. B strategists were similarly susceptible to exploitation as P stocks, but their longer times to maturity make them particularly vulnerable to overfishing. Main conclusions: A broader recognition of the distinct ecology of Salmonic and Bet-hedger groups is important when studying life history strategies in marine fish. More generally, our results stress the importance of considering life history strategies for understanding patterns of recruitment variability across fish stocks.

Life history trait data were collected from FishBase (Froese & Pauly 2000; accessed Sep 24 2019) on a stock-specific basis for 244 exploited marine fish species. FishBase was accessed using the package rfishbase (version 3.0.4; Boettiger et al. 2015) for the software R (version 4.0.3; R Development Core Team, 2016). All life history traits were the mean (if numeric) or most frequent (if categorical) value reported for each stock of each species. "tmax" is the maximum time to maturity in years. "Fecundity" is the logarithmic mean of minimum and maximum fecundity. "PCI" is the Parental Care Index that weighted quantitative and categorical data on the mode of fertilization, Balon’s (1990) reproductive guilds, the presence/absence of any kind of parental care, and the duration of the gestation period. Missing values of maturity and fecundity for a small fraction of stocks (6% and 19% respectively) were imputed using closely related traits belonging to that stock and traits from other stocks using additive regression and bootstrapping techniques implemented with the aregImpute function in the Hmisc package (Harrell et al., 2017) of the software R. Separate imputations were performed on the taxonomic groups: Elasmobranchii, Scorpaeniformes, and non-Scorpaeniform teleosts.

Time series of recruitment (R), spawning stock biomass (SSB), fishing rate (F), and exploitation rate (ER) were retrieved from the RAM Legacy Stock Assessment Data (RAM SAD) version 4.44 assessment data only (http://ramlegacy.org). Data gathered prior to 1950, stocks with less than 10 years of data, and stocks with an unknown assessment method were excluded. Exploitation rate was used to estimate the fishing rate for stocks where it was missing using the Baranov equation. Variance in the recruitment time series was calculated as normalized recruitment deviations from expectations based on (1) a simple density dependent survival model (R/SSB), (2) a normal compensation (Beverton-Holt) model, and (3) an over compensation (Ricker) model.

RAM SAD stocks were matched to FishBase stocks using information on regional location from both databases. If a RAM SAD stock was not represented by one FishBase stock, then the mean of one or more FishBase stocks were used. These means are indicated with fractional FishBaseStockCodes.