Life histories

Specifying life histories (yes, invertebrates too!)

A life history describes the biological characteristics of a fish species, including growth, maturation, and rate of natural mortality. In the FishKit Data Repository, life histories are specified as a set of parameters (and associated equations) for a species. Members of any taxonomic group (e.g., finfish, invertebrates, elasmobranchs) can be specified as life histories if it is reasonable to describe their biological characteristics according to the required parameters (and associated equations). For example, length-at-age is mathematically modeled according to the von Bertalanffy growth curve in FishKit. The von Bertalanffy growth curve is a mathematical model that describes how an organism’s length increases over time. To utilize FishKit accurately, only species that are expected to grow according to the von Bertalanffy growth curve should be included in FishKit. While other growth relationships are well established in the scientific literature, they are not included in FishKit (Quinn and Deriso 1999).

Whether or not an invertebrate species can be included in FishKit is a typical question, and centers on the related question of whether its growth is reasonably represented using the von Bertalanffy growth curve. Users entering life histories must decide whether there is a reasonable species-specific expectation that the growth process can be approximated by the von Bertalanffy growth curve. For example, there is discussion in the literature on whether cephalopods grow according to a von Bertalanffy model; recent reviews recommend using the Schnute growth model (four-parameter model) to describe their growth, at least for squid (Arkhipkin and Roa-Ureta 2005). However, some octopus species have been modeled using the von Bertalanffy growth curve (Arreguín-Sánchez et al. 2000). Likewise, the process of growth through molting of many invertebrates species is frequently approximated by the von Bertalanffy model for the purposes of stock assessment, population dynamics modeling and other fisheries analyses.

Below is a summary of the parameters (and associated equations) used in specifying life histories in FishKit:

Length-weight conversion: length-weight conversion is used in calculating quantities related to biomass or weight in the catch.
Weight-at-age with parameters and is specified as an exponential function.
Von Bertalanffy growth: length-at-age is calculated according to the von Bertalanffy growth curve, where Linf is asymptotic length, K is the growth parameter, and t0 is the theoretical age at length zero.
Length at maturity: maturation follows a logistic function with parameters L50 and ΔL95, reflecting the length at which 50% of the population are mature and the length increment to which 95% of the population is mature, respectively. Thus, L95 = L50 and ΔL95.
Natural mortality: the natural mortality rate, M, should be specified as a constant instantaneous rate per year.
Protogynous hermaphroditic species: where applicable, transition from female to male is specified as a logistic function with parameters H50 and ΔH95, reflecting the length at which 50% of the population is male and the length increment to which 95% of the population is male, respectively. Specifying a species as a protogynous hermaphrodite requires input of both parameters, otherwise the species is assumed to be gonochoristic with a 50:50 sex ratio for all lengths and ages.

Relevant Modules:

Follow steps to create a new life history in the Data Repository

Confronting life history uncertainty

Utilize version control

FishKit is designed to accommodate uncertainty in life history parameters by allowing the user to enter multiple variants (versions) of a specified life history for a given record of species or stock. Uncertainty in life history parameters may arise for a given species due to differences in methodological approaches to parameter estimation, borrowing of parameters from other species, stocks or geographic locations, and other factors. Within the Data Repository, the Life History Repository utilizes version control functionality. Following the initial creation of a life history, editing can only take place through creation of additional versions. Versions are contained within a given life history, not as duplications as of a life history. Version control was implemented in FishKit to ensure that all analyses are reproducible, as any former or current version can be loaded by the Size Limit Builder or Stock Health Tracker. Users can take advantage of this workflow by using versions to represent multiple discrete descriptions of the life history, perhaps representing alternative scenarios, information sources, or points of view. Importantly, following the creation of multiple versions, the Size Limit Builder and Stock Health Tracker can be used to activate multiple versions and toggle between versions, creating opportunities to conduct rapid comparisons.

Relevant Modules:

Additional information sources

In addition to primary and secondary literature sources of life history information, FishKit provides direct access to parameter values and references contained in FishBase. When creating a new life history in the Data Repository, most steps involved in specifying life history parameters contain a tab called Supplementary Information, which assists the user in gathering life history information. For the life history characteristics of length-weight, growth, maturity, and natural mortality, an automated search of FishBase is conducted with available parameter estimates returned along with references and geographic information (Froese and Pauly 2011; Boettiger et al. 2012). For natural mortality, an additional helper algorithm is provided based on indirect correlative approaches for estimating natural mortality based on other life history parameters, including longevity-based (tmax) estimators.

Other useful online tools for life history information include:

Natural mortality tool: https://github.com/shcaba/Natural-Mortality-Tool
FishLife: https://github.com/James-Thorson-NOAA/FishLife

Relevant Modules:

Partial life histories

While filling all fields when creating a new life history is an ideal approach, this may not be possible in all instances. However, a variety of functions in the Size Limit Builder and Stock Health Tracker will remain available for partial life histories. Below, function availability is described relative to the availability of life history characteristics.

von Bertalanffy parameters: Both the Size Limit Builder and Stock Health Tracker will still offer full functionality when t0 is not specified. If left unspecified, FishKit analyses that utilize t0 will assume a value of zero. Specifying both Linf and K is required to use the Size Limit Builder. Specifying both Linf and K is required to use metrics in the Stock Health Tracker, with the exception of the LB-SPR assessment method. Specifying Linf is required for LB-SPR assessment, along with either K (and an independent estimate of M) or the ratio, M/K. As a good practice, if a reliable set of all three von Bertalanffy parameters are available, it is always useful to enter all these parameters.

Natural mortality: When available, preference should be given for specifying M and K separately to produce full functionality within the Size Limit Builder and Stock Health Tracker. When M and K are specified separately, these values are used to calculate the M/K ratio in the saved life history. However, if the M/K ratio is known but M is unknown, M/K can be specified instead of M, allowing some features in the Stock Health Tracker to remain accessible.

Maturity: Most functions in the Size Limit Builder and Stock Health Tracker require both parameters of the logistic maturity function. Sometimes, studies of fish maturation will report only L50, and some suggestions for addressing this complication are as follows. First, consider a setting ΔL95 equal to 1.0 cm to approximate immediate maturity at all lengths above L50. Second, consider a rule-of-thumb approach, such as setting ΔL95 equal to 15% of L50 to produce a maturity schedule that is consistent with a variety of fish families (Prince et al. 2015). Strictly speaking, where ΔL95 is not specified, the Size Limit Builder will not be available, but some functions in the Stock Health Tracker will remain active. To avoid this situation, consider alternative assumptions (versions) about ΔL95 and explore the effects of this uncertainty on the outputs of various analyses.

Relevant Modules:

Sexually dimorphic species

Sexual dimorphism in marine species is common among taxa, from invertebrates to marine mammals (Stamps 1993). It is commonly characterized by differences in behavior, morphological characteristics, and differences in life-history traits such as maturity schedules, growth patterns and rates, asymptotic size, natural mortality, and longevity between males and females (Parker 1992; Stamps 1993). These differences between sexes influence population structure and reproduction, and therefore, sexual dimorphism can make one sex more vulnerable to fishing, affecting the stability and sex proportion balance in the population (Hanson et al. 2008). For example, when one sex grows faster, it can be susceptible to the fishing gear for a longer proportion of its life span (i.e., spending more of its lifespan at sizes that are vulnerable to fishing) compared to the other sex, skewing sex ratios and potentially affecting reproductive success (Spirk 2012; Wszola et al. 2022). When sexes are incidentally exploited differently as a consequence of sexual dimorphism, management regulations should be based on the sex that is likely to be exposed to higher fishing mortality throughout its lifespan (Miethe and Dobby 2022).

Size Limit Builder

Where sexual dimorphism in growth is evident, size-selective fishing (e.g., size limits) may reduce disproportionally the number of fish of one sex (Tsai and Huang 2022). Additionally, sexual dimorphism in maturity and longevity, or both, may affect perceptions about the likely outcomes (or effectiveness in achieving management objectives) of management actions, particularly as it relates to protection of reproductive potential.

Alternative approaches for utilizing the Size Limit Builder for sexually dimorphic species are described below. For species that exhibit sex-specific life history parameters (e.g., growth or maturity, or both), some considerations are required when specifying life history parameters with intended use in the Size Limit Builder. Analyses carried out in the Size Limit Builder rely on a single-sex model of population dynamics (Harford 2024). This model means that a single set of life history parameters is assumed to be representative of both sexes; accordingly, the Data Repository accepts only a single set of life history parameters (not sex-specific parameters). Options for addressing sexual dimorphism are as follows. First, averaging parameters between sexes (or relying on a ‘combined fit’ of pooled data can be used to produce, for example, averaged growth or maturity parameters. This option should be used cautiously because misleading results for size could occur. In cases where the females of a species grow larger, they may also reach their maturity at proportionally larger size; therefore, an averaged growth curve may be misleading with respect to the true effects of size limits on reproductive potential. Second, using female life history parameters as inputs to size limit analysis may help to avoid overfishing females and ensure the population productivity and reproductive potential are protected (Kindsvater et al. 2017; Coscino et al. 2024). This option aligns with a conservative recommendation of using the life history data of the larger sex, typically females; where accordingly, females usually determine the reproductive potential of the population, and their larger sizes make them more vulnerable to fishing (Miethe and Dobby 2022). Further, the sustainability score (or spawning potential ratio) is a focal metric in the Size Limit Builder, in which case the female growth curve is likely to be a reasonable choice. Third, creating multiple representations of life history and evaluating size limits against each representation could produce transparency in the degree to which inputs (e.g., life history) influence outputs (e.g., performance of size limit options). The Data Repository is constructed with version control, which allows each life history record to contain multiple versions, each of which can be used in analysis. Version control can be utilized to produce multiple variants of a given life history, enabling exploration of the effects of various uncertainties on outcomes of analyses, like the sustainability score and catch score.

Relevant Modules:

Stock Health Tracker

Within the Stock Health Tracker, the assessment method known as Length-Based Spawning Potential Ratio (LB-SPR) assumes that both sexes share the same growth and maturity curve and that the sex ratio of the catch is balanced (Hordyk et al. 2015c). Consequently, resulting estimates of SPR produced by this assessment could be sensitive to alternative specifications of life-history parameters. LB-SPR should not be applied to species with sex-specific life history parameters without also having available sex-specific length frequency data. Relying solely on LB-SPR when parameters are sex-specific is likely to lead to erroneous management recommendations (Hordyk et al. 2015c; Coscino et al. 2024). Therefore, when there are sex-specific differences in life history parameters, this assessment method should only be applied only to female sub-set of length frequency data, using female life history parameters (Hordyk et al. 2015c; Coscino et al. 2024). In circumstances where sex-specific length frequency data are not available, running the assessment against alternative versions of a life history, and examining the level of agreement between results, could be worthwhile to capture a range of modeling results.

Relevant Modules:

Hermaphroditic species

Hermaphroditic species, where individuals change sex from female to male (protogynous hermaphrodites) or from male to female (protandrous hermaphrodites) as they grow, present some unique challenges for fisheries management. These challenges are evident under size-selective fishing practices (Easter and White 2016). In gonochoristic (separate-sex) species, size-selective practices are generally appropriate because both sexes tend to grow similarly (except in sexually dimorphic species), so size-selective fishing targets both sexes equally. For gonochoristic species, it is often assumed that females determine the reproductive potential of the population, as females produce eggs (considered the limiting gamete), and in many species, males can mate and spawn with multiple females (Heppell et al. 2006; Easter and White 2016). These assumptions need to be more carefully applied to species that change sex during their lifetime. Since size-selective fishing typically targets the largest individuals—disproportionately affecting males in protogynous hermaphrodites and females in protandrous hermaphrodites—intensively reducing one sex in the population and skewing sex ratios (e.g., leaving few males or females for reproduction). In extreme cases, this imbalance in sex ratios, such as when too many males are removed, may lead to sperm limitation as the limiting gamete, rather than egg production (Alonzo et al. 2008).

Size Limit Builder

Life histories for gonochoristic species or protogynous hermaphroditic species can be created using the Data Repository and later analyzed using the Size Limit Builder. Gonochoristic species are specified as having a 50:50 sex ratio for all lengths and ages. Protogynous hermaphroditism is specified using a logistic function for sex transition that defines the proportion of the population-at-length that is male. Because analyses carried out in the Size Limit Builder rely on a single-sex model of population dynamics (Harford 2024), separate male and female abundances are not available as model outputs. Instead, metrics of population reproductive output (including related quantities like spawning-biomass-per-recruit and spawning potential ratio) rely on mature female biomass. For gonochoristic species, this simply means that 50% of the total mature biomass in each age class is considered female. For protogynous hermaphroditic species, the proportion of total mature biomass in each age class that is considered female is derived from the logistic function for sex transition (converted to proportion at age).

If your species of interest is a protogynous hermaphrodite, use the Size Limit Builder with caution. Given the reliance on a logistic function for sex transition, the following caveats should be considered when the Size Limit Builder is applied to protogynous hermaphroditic species (e.g., wrasses or parrotfishes). First, the use of a logistic function for sex transition implies a fixed schedule of sex change at length or age. The Size Limit Builder does not account for potential environmentally- or behaviourally-induced flexibility in sex-change in any outputs of analysis. Second, all metrics related to reproductive output reflect female output. The Size Limit Builder cannot be used to address whether sperm limitation may produce risk of reproductive failure, and in fact does assume that female gametes are the limiting gamete of reproductive output.

For protogynous hermaphroditic species where the above stated caveats could concern the validity of the outputs of analysis, expert guidance and/or alternative analysis or decision-support should be sought. For example, a comprehensive approach that includes understanding sex change cues, defining appropriate size limits, and combining these size limits with other management strategies (e.g., spatial protection during reproduction) may offer more effective strategies for managing and conserving these sensitive and vulnerable populations or stocks, protecting reproductive females, mitigating sperm limitation risks, and helping to ensure sustainability (Armsworth 2001; Kindsvater et al. 2017). When information on the length at which the sex transition occurs is not available, the management of hermaphroditic species becomes more complex. In those situations, implementing conservative size limits and avoiding intense fishing of the largest individuals may be the most recommended approach to avoid altering sex ratios (Kindsvater et al. 2017).

Relevant Modules:

Highly migratory species

Can I use FishKit for my highly migratory species of interest? FishKit may be applied to a highly migratory species, but results will not be reliable unless (1) management actions like those examined Size Limited Builder are intended to be applied to the entire stock and (2) evaluations of stock health, as conducted using Stock Health Tracker, are conducted based on length sampling representative of the geographic extent of the stock.

Highly migratory species have complex life histories that take them through multiple jurisdictions, complicating their management (Lascelles et al. 2014). They travel across international boundaries and are targeted by multiple fleets from different countries, making it potentially challenging to coordinate management efforts (Carr 2018). These species could also be highly vulnerable to anthropogenic effects, as they depend on multiple habitats throughout their migratory journey (Shuter et al. 2011), making habitat loss an additional concern. When migration routes and timing are predictable, or when they form large aggregations (e.g., for reproduction), they are particularly susceptible to overexploitation, as they become easy targets for fleets (Shuter et al. 2011). For this reason, fishing mortality needs to be considered across multiple jurisdictions (Kadagi et al. 2022) since fishing mortality in one region can impact the entire stock (Kinney et al. 2020).

Highly migratory species, such as billfish, exhibit a range of life history traits, from fast growth and short lifespans (e.g., some marlins) to larger, long-lived species like swordfish (Fromentin and Fonteneau 2001; Pons et al. 2017). For example, some marlins (blue, white, and striped) are characterized by unique growth patterns and delayed maturity and are highly valued in recreational fisheries due to their large size (Pine Iii et al. 2008). These life history characteristics and the high market value of some species make these species more susceptible to overfishing (Pons et al. 2017). Additionally, some species are caught as bycatch from more valuable fisheries, creating further management challenges.

A common issue in highly migratory species fisheries is that data collected for these species varies in accuracy across regions (Dowling et al. 2019), as some areas may have more robust data collection programs than others. Due to these differences in data quality, life history parameters estimated for one region may vary significantly compared to another, potentially failing to reflect different life histories accurately. In some cases, these differences represent true spatial variability in life history (e.g., temperate and tropical tunas) (Fromentin and Fonteneau 2001), while in others, they may simply reflect differences in sampling designs or analytical methods (Kinney et al. 2020). Obtaining accurate life history information (e.g., growth and reproduction parameters) for species like tuna and billfish is challenging, and the inability to sample entire distribution ranges produces some “region-specific” parameters, which may complicate management, particularly when these life history traits do not align with the entire stock (i.e., the entire range is not represented) (Kinney et al. 2020).

To apply FishKit to highly migratory species, it is recommended to implement representative sampling across geographic extent of the stock of interest (Kinney et al. 2020). Additionally, combining multiple management measures may improve the sustainability of these species, such as implementing minimum size regulations and seasonal closures. Minimum size regulations, for example, can help protect juveniles and ensure that fish have the chance to spawn before being caught (Pons et al. 2017).

Short-lived species

Small pelagics, like sardines and anchovies, have short lifespans (less than 5 years), exhibit fast growth, and show large variability in recruitment. Their abundance may change rapidly over time due to boom-and-bust dynamics (Barange et al. 2009; De Moor et al. 2011). These boom-and-bust dynamics may be partly explained by variations in reproductive success (recruitment), which are sometimes triggered by oceanographic conditions (MacCall 2009; Essington et al. 2015; Jacobsen and Essington 2018). However, fishing has also played an important role in the boom-and-bust dynamics of these species (Beverton 1990; Essington et al. 2015; Jacobsen and Essington 2018). Like small pelagics, cephalopods are also characterized by short lifespans and rapid growth, but they are also semelparous species, meaning they reproduce once in their lifetime and then die (Arkhipkin et al. 2021). Many commercially exploited species’ life cycles range from 6 months to 2 years (Arkhipkin et al. 2015, 2021). Significant differences exist in growth and maturation rates between cohorts, and their growth shows high plasticity. Individuals from different cohorts can be of similar size, while individuals from the same cohort can exhibit a wide range of sizes (Arkhipkin et al. 2021).

Size Limit Builder

The Size Limit Builder may be applied to short-lived species if adjustments are made to the default display settings. The dials centrally located at the top of screen on “Step 2 Build size limits” provide equilibrium sustainability score and catch score values. These equilibrium per-recruit quantities are calculated on a monthly time step; thus, this analysis is suitable for short-lived species (e.g., those with lifespans of 5 years or less or natural mortality of 0.6 or greater). When using the Size Limit Builder, an activation is required that contains a life history. During activation, the size limit step size can be modified from its default of 1 cm and 1 inch and will affect the set of size limit options presented. Selection of the step size is necessary prior to activation, as per-recruit calculations are carried out based on this field. For most fishes, the default of 1 cm and 1 inch is a reasonable choice, but smaller intervals are available. Best results for short-lived species are likely produced when activation is specified using a size limit step of less than 1 cm and 1 inch.

The transitional dynamics analysis in FishKit should not be used for short-lived species. Unlike equilibrium per-recruit analysis, transitional dynamics are calculated on an annual time step, thus this analysis is not suitable for short-lived species (e.g., those with lifespans of 5 years or less or natural mortality of 0.6 or greater). Further, transitional dynamics occurs according to the assumption of constant recruitment. Therefore, the boom-and-bust dynamics of short-lived species are not effectively represented in transitional dynamics projections; thus, some caution should be taken in deciding whether these projections are a suitable basis for decision-support for short-lived species.

Relevant Modules:

Stock Health Tracker

The Stock Health Tracker is usually not appropriate for short-lived species, as management decisions typically do rely heavily on length-frequency distributions. Length composition data may not be the most appropriate type of information to monitor these populations, given their life history characteristics, and is seldom prioritized to monitor the health of coastal pelagic populations such as sardines and anchovies (or are there any examples around the world of this?). Furthermore, applying methods like LB-SPR can be challenging in small pelagics due to their rapid dynamics and high recruitment variability. Large and older fish are predominant in length composition data during prolonged poor recruitment. In these situations, LB-SPR may interpret the presence of large fish as an indicator of low fishing mortality, potentially underestimating fishing mortality and overestimating SPR (Garcia and Josse 1988; Coscino et al. 2024). The opposite effect may occur following strong recruitment, where a predominance of juvenile and immature fish in length composition data may create the false impression that the lack of mature fish is due to overfishing. In such situations, fishing mortality could be overestimated and SPR underestimated (Garcia and Josse 1988; Coscino et al. 2024).

For short-lived species, an additional caution is advised when relying solely on length composition data for monitoring population or stock condition, as length composition may not accurately reflect the length/age structure of the population (individuals of the same size may belong to different ages and cohorts) (Arkhipkin et al. 2021). In addition, many cephalopods, for example, exhibit protracted spawning periods (extended or often year-round spawning), with the simultaneous presence of cohorts or micro cohorts within the same area (Arkhipkin et al. 2021). This makes it challenging to interpret length-frequency data. Additionally, many species undertake extensive ontogenetic migrations, for example between spawning and feeding grounds, so changes in length composition may reflect migration patterns rather than changes to stock status.

Relevant Modules:

Species with low M/K & high L50/Linf

It is now commonplace to present and interpret the biological characteristics of a fish species in relation to the dimensionless life history ratios’ M/K and L50/Linf. For example, under equilibrium conditions (i.e., a balance between population growth and removals) and with certain assumptions (e.g., von Bertalanffy growth, constant M, asymptotic selectivity), the length composition is primarily determined by the M/K and fishing mortality-to-natural mortality (F/M) ratios (Hordyk et al. 2015b). Recognition of the importance of these dimensionless ratios in determining responses to fishing mortality began with work by Beverton and Holt in the 1950s and 1960s (Beverton and Holt 1957, 1959; Beverton 1963), and has continued with emphasis on taxonomic consistency and correlative relationships (Prince et al. 2015, 2023).

Prince et al. (2023) captures the breadth of variation in the ratios’ M/K and L50/Linf, as well as the correlation between these ratios. Generally, the extent of variation in these ratios across taxa can be specified as life histories in FishKit. As described by Prince et al. (2023), there is a negative correlation between M/K and L50/Linf, with a few fish families comprising low M/K and high L50/Linf: surgeonfishes (family: Acanthuridae), roughies (family: Trachichthyidae), Wreckfish (family: Polyprionidae), and sea chubs (families: Girellidae and Kyphosidae) (see Prince et al. (2023) for a complete description of relevant taxa).

Species of these families are expected to be dominated by larger, older individuals distributed around the asymptotic size, with relatively few smaller individuals. Juveniles typically approach Linf early (relative to life span), then persist at or near their asymptotic size for many reproductive cycles (Hordyk et al. 2015b, 2015a; Prince et al. 2023). Low M/K populations tend to have low turnover, meaning they replenish their populations more slowly than species with higher M/K ratios (Prince et al. 2023).

Size Limit Builder

When exploring size limit options for species with low M/K and high L50/Linf, it is worth noting some of the caveats that may emerge for low M/K species. These caveats include: (1) when a large portion of individuals persist at or near asymptotic length, the range of options for size limits may be narrow under the dual objectives of preventing capture of juveniles while maintaining a viable fishery of mostly asymptotic sized individuals; (2) be critical of rule-of-thumb approaches for size limits, like multiples of L50, in terms of lost fishing opportunities or high discard rates, or both; and (3) alternative management measures (e.g., seasonal closures, bag limits, etc.) should not be overlooked as a means to avoid overfishing while balancing stakeholder preferences, feasibility, fishery and market preferences, and potential for enforcement.

Relevant Modules:

Stock Health Tracker

The LB-SPR assessment method is sensitive to M/K values. Hordyk et al. (2015b) demonstrated, using simulations, that it worked well for species with M/K > 0.53 but produced more biased results as M/K values decreased. The LB-SPR method relies on detecting signals of fishing mortality from the right side of the length composition data (where larger individuals are represented). However, in low M/K species, the effect of fishing is less noticeable in the length composition until fishing mortality is very high, resulting in a very low SPR. This is particularly problematic when selectivity is dome-shaped (e.g., in gillnets), as the model may confuse the absence of larger individuals with high fishing mortality (Hordyk et al. 2015b). This misinterpretation results in an overestimation of F/M and subsequent underestimation of SPR. For this reason, methods like LB-SPR should be applied cautiously for species with low M/K ratios, as the effects of fishing may only become noticeable in length data when they are already excessively high.

Relevant Modules:

Quality scoring

In order to quantify the uncertainty in the life history parameters used in FishKit, a quality assessment framework designated “Quality scoring” has been developed. This innovative concept addresses the strengths and weaknesses in the knowledge base behind a parameter by critically reviewing the production process of the parameter and the scientific status and assumptions underpinning the number. Here, the term “quality” refers to the essential capturing of data quality by grading an information source against a pre-agreed set of criteria. This quality assessment framework can illuminate the transparency and scientific defensibility of life history parameters used in analyses that are central to FishKit.

The quality scoring is applied in the Data Repository to the length-weight relationship parameters, the von Bertalanffy growth function parameters, length at maturity parameters and natural mortality parameter. The framework is intended to support the evaluation of quality of pre-existing or original life history parameters that are input in the Data Repository. It is not intended to assess “synthetic” parameters that have been interpreted/modified/re-imagined from previous ones.

The quality scoring framework uses seven categories of information, i.e., spatial representativeness, temporal representativeness, taxonomic proxy, uncertainty, validation, specificity, and empirical basis, to diagnose and score the relative quality of parameters. A brief description of each category has been given in section “E. Quality scoring” in Steps for creating a life history. Further explanations on the rationale behind the scoring for each category are detailed below.

Spatial representativeness

Parameters calculated from data that originates from an area that constitutes a large proportion, i.e., greater than 75%, of the geographic range of the stock, are considered to be relatively high quality (score 4; Mazloumi & Nicol, 2023). Parameters like these are less influenced by local spatial variability induced by local variation in growth, predation and density; whereas parameters calculated from data that originated from outside of the geographic range of the stock in question (e.g., borrowed from a different stock), are considered of relatively low quality (score 0) because they are more likely to have been influenced by different ecological conditions (Cadrin et al., 2023). Parameters calculated using data from between 50 and 75%, between 10 and 49%, and less than 10% of the stock’s range, are considered intermediate to low in quality and receive scores of 3, 2 and 1, respectively.

Temporal representativeness

Parameters based on recently collected data are considered higher quality than older parameters due, in part, to the well-known tendency for populations to exhibit non-stationarity, particularly for data-limited approaches (Zhang et al., 2021; Berger, 2019; Johnston et al., 2015). A decadal step size was adopted for this category (Mazloumi & Nicol, 2023). Thus, parameters generated using data from the most recent decade are considered to have contemporary representativeness and have the highest temporal representativeness (score 4) while parameters generated using data from any longer than 40 years ago are considered weakest (score 0).

Taxonomic proxy

A parameter may need to be “borrowed” from some proxy stock or species. The quality score in this category refers to the taxonomic proximity of a substitute species to the actual species under consideration. The assumption here is that the highest quality parameters will be those that pertain to a species or subspecies that is taxonomically closer to that for which one seeks to score a life history. If a parameter, and presumably the data from which the parameter was estimated, pertains to a different stock of the same species, then the score would be 3. If the parameter is from the same genus or family as the species in question, then the scores would be 2 and 1, respectively. Lower taxonomic groupings result in scores of 0.

Uncertainty characterization

Uncertainty encompasses observation error (including measurement error and sampling error), demographic stochasticity (i.e., natural variability) and model misspecification (Mace & Sissenwine, 2002; Peterman, 2004). An informed distribution representing specified uncertainty in the parameter is scored a 4. If the uncertainty about the parameter being assessed is conveyed in the form of a uniform distribution, then it should be assigned a score of 3. If the parameter being assessed is a point estimate, meaning a parameter without a range of uncertainty specified, then it should receive the lowest available score of 2.

Validation

The parameter being scored should have been compared for consistency with other independent parameters, for example, those from other populations, different datasets and/or mathematical methods. If the parameter being assessed was demonstrated by those who produced it to be consistent with independent parameters from the same time period, and the same stock or population, then it will score high on this metric (score 4). Independent parameter(s) that were used to validate the parameter being assessed using older data from the same stock should be assigned a score of 3. If the independent parameter(s) used are from another stock, a lower score is expected because of the high potential for inter-stock variation in growth and ecology to bias the comparability of the parameters (score 2). If the parameter has been compared with independent parameter(s) and has been found to be inconsistent, then a lower score (score 1) should be given. If no comparisons were performed, then it should receive the lowest score of 0.

Specificity

Specificity refers to considerations given to whether the parameter being assessed varies during the life cycle of the species (variation by age, size or life stage). For example, specificity for natural mortality can be challenging in data-limited fisheries, so it is often assumed to be constant across age and size. Note that it is not possible to include parameter specification by life stage within the life history specification in FishKit, but the user should search for this evidence in the original source of information. When consideration has been given as to whether a parameter could vary with life stage and such detail is included in parameter specification (e.g. parameter specified by length or age, by sex, or juvenile vs adult), a score of 4 should be assigned. If a parameter is deemed invariant by life stage, a score of 4 should also be assigned. If a parameter likely varies by life stage and a coarse approximation was used (e.g. juvenile vs adult), then a score of 3 should be assigned. Constant parameters will result in a score of 2 or 1, depending if consideration was given on how likely a parameter varies by life stage. If there is no consideration given to parameter variation by life stage, it results in a score of 0.

Empirical basis

Empirical basis refers to both the degree to which direct observations, measurements, and statistics were used to estimate the parameter, and the degree to which the core assumptions about collection of data and computational methods were met. This category relies on subjective judgment of the user to score from very low confidence (score 0) to a maximum of very high confidence (score 4) on the methodology used to derive the parameter. It is recommended that the user reviews the guidance tables for scoring the empirical basis.
First, the user needs to identify and choose from the dropdown list which method has been used to derive the parameter. If the user finds a method that is not contained in the dropdown list, the user should choose “Other” and make a note of that in the text box. The length-weight relationship parameters are almost certain to have been produced using one method (linear regression on log-transformed variables). For the von Bertalanffy parameters, the methods can be length-frequency, length-age pairings, or mark-recapture. For the parameters related to length at maturity, methods for reliable identification of mature and immature fish are microscopic identification, macroscopic identification and gonadosomatic index (GSI). For producing estimates of the parameters related to length at maturity, the following methods are most commonly used: logistic model; length-weight relationship; and life history ratios. For natural mortality (referred from here onward as M), the guidance is structured according to direct methods versus indirect methods. Direct methods refer to using direct observations, measurements and statistics. Indirect methods rely on comparative life history studies to borrow strength from many species (Hamel & Cope, 2022; Pauly, 1980). Indirect methods can use Tmax (maximum age); K and/or Linf; and GSI. Direct methods can measure Z (total mortality) in an unexploited stock using catch curve analysis; relate Z from catch curve analysis to actual fishing effort and extrapolate to zero fishing effort; measure both Z and exploitation rates and solve for M and F (fishing mortality); mark-recapture and telemetry studies; and estimate M internally in an integrated stock assessment model.
Second, the user should review the section in the guidance tables corresponding to the method identified, reflect if the assumptions are met and based on the user’s expert beliefs about the degree of confidence in the data and methods to produce the estimate, select the most appropriate score. The user should write a brief justification for the score in the corresponding text box (i.e., which assumptions are not met and any concerns about the method used).

Overall quality score

The maximum quality score that can be produced for a given life history parameter is 28, i.e., seven information categories times maximum score of four per category. The total quality score is shown as a percentage. Thus, the relative quality of any parameter can be systematically compared to the quality of an alternate parameter, with the highest scoring assumed to be of relatively better quality. An overall life history quality score is generated by averaging the parameter-specific scores that were recast as percentages. This calculation yields the overall life history quality score for the included set of parameters. In this way, it will be possible to compare the quality of an entire set of life history parameters across species or life history versions.

Guidance tables for Quality scoring system

Table 1. Guidance on empirical basis scoring for length-weight parameter estimates (a and b). Start with a score of 4. If assumption(s) are not met, consider downgrading the score to maximum cumulative downgrade of four points (resulting in empirical basis score of zero for that estimate).

Method	Guiding questions and related assumptions. Maximum empirical basis score is 4
Length-weight data
Data collection and quality	*What is the degree of confidence in the quality of length-weight data?* Data collection: Gear used for collecting specimens do not introduce bias with respect to length or weight, meaning that samples are representative of the population (Froese, 2006). Sampling accounts for patterns in population segregation that may lead to biased length-weight data; segregation patterns can include aggregation of individuals according to sex, habitat/location. Weight of a fish at a given length varies along the year, so analysis has taken this into account (Kimmerer et al., 2005). Equal numbers of randomly selected small, medium-size and large specimens have been included (Froese, 2006). Data quality: Number of specimens measured, range and type of length measurements and units of measurement have been noted (Froese, 2006). Plots of outliers/dispersion of data and plots of raw and log-transformed data have been checked to identify spurious data (Gerritsen & McGrath, 2007).
Estimation of parameters
Linear regression on log transformed variables	*What is the degree of confidence that assumptions in the use of linear regression were met?* Assumption of independent observations; if observations are not independent, this has been accounted for in the analysis (e.g. cluster sampling means that samples within each cluster are correlated; estimate length-weight relationship for each cluster if sample size allows; (Gerritsen & McGrath, 2007). Log transformation shifts the basis of the regression line from the mean to the geometric mean (i.e. median), so a correction factor needs to be applied if geometric mean is used (Gerritsen & McGrath, 2007) or parameters need to be back-transformed to natural scale. Assumption of linear relationship between log-transformed variables. Diagnostics to check assumptions: Normally distributed residuals; Homogeneity of variances; Observed vs fitted values.

Table 2. Guidance on empirical basis scoring for von Bertalanffy parameter estimates. If assumptions not met, consider score downgrade by one point, to maximum cumulative downgrade of four points (resulting in empirical basis score of zero).

Method	Guiding questions and related assumptions to be met
Length-frequency	What is the degree of confidence in length-frequency approach to produce robust growth parameter estimates?
	Biological characteristics are suitable for application of this methodology (Pauly, 1987). Requires distinct spawning season to follow progression of length-classes through time. Typically, better suited for shorter-lived species and/or those that cannot be aged, also where overlap in length-at-age is minimal compared to long-lived species. The von Bertalanffy function is suited to describing the average growth trajectory
	Sampling design produces adequate observations for growth estimation Failure to conduct sampling that accounts for patterns in population segregation within may lead to biased growth estimates. Segregation patterns can include aggregation of individuals according to age, sex, habitat (ontogenetic shifts or size-driven segregation), size-driven schooling behavior (Miranda & Colvin, 2017). Failure to collect length measurements over regular growth intervals can lead to biased parameter estimates (Laslett et al., 2004). Biased parameter estimates can occur when fishing gear is size selective and/or where fishery-dependent sampling is not representative of the biological population (Laslett et al., 2004; Pauly, 1987).
	Problems related to model fitting are not evident Bin size applied to length-frequency data can influence accuracy of growth parameter estimation and should be adequately considered in analysis (Taylor & Mildenberger, 2017; Wang et al., 2020). Several sources of variation can produce challenges in model fitting, although see alternative modeling approaches to address this complication (Batts et al., 2019; Laslett et al., 2004; Zhou et al., 2022).
Length-age pairings	What is the degree of confidence in mark-recapture approaches to produce robust growth parameter estimates?
	Sampling design produces adequate observations for growth estimation (Miranda & Colvin, 2017) Failure to conduct sampling that accounts for patterns in population segregation within may lead to biased growth estimates. Segregation patterns can include aggregation of individuals according to age, sex, habitat (ontogenetic shifts or size-driven segregation), size-driven schooling behavior. Sampling bias can occur when fishing gear is size selective and/or where fishery-dependent sampling is not representative of the biological population, resulting in exclusion of size classes and inevitably also age classes. Gear selectivity issues can be avoided by sampling with a variety of gears or gear configurations. Bias can occur when only a segment of the fish population is sampled. Sampling bias can occur when designs are not standardized; that is, where observers are inconsistent, resulting in incompatible and potentially non-repeatability of data. Sampling bias can occur when an insufficient number of samples is collected. While life history will affect sample size requirements, Kritzer et al. (2001) suggests 7 to 10 length-age pairs per age class, noting that improvement of precision of von Bertalanffy parameter estimates tends to be negligible beyond 300 samples.
	Problems related to model fitting are not evident (Ogle et al., 2017). Length-age pairing may exhibit multiplicative error structure (variance in size increases or decreases with age). Failure to address this issue can lead to problems in model fitting. Patterns in residual plots suggest modeling assumptions have not been met, including structural forms of growth equations (Pardo et al., 2013; Schnute, 1981). Fitting algorithm may fail to find global minimum, as evidenced by converge failure, or multiple starting values producing different parameter estimates. Parameter estimates for most growth models will be correlated; however, absolute correlations > 0.9 suggest ambiguity in parameter estimates (Motulsky & Christopoulos, 2003). Unreasonable parameter estimates are likely to be obtained without sampling the full range of lengths and ages. Missing smaller individuals tends to poorly estimate intercept and curvature (e.g., von Bertalanffy t0 and K), while missing larger individuals tends to poorly estimate curvature and asymptotic length (e.g., von Bertalanffy L∞) Fitting to back-calculated lengths-at-age should involve hierarchical modeling (i.e., mixed effects modeling) with sufficient consideration of the mathematical relationship between otolith width and body length, including individual differences in length – radius relationships (Ogle et al., 2017; Vigliola & Meekan, 2009).
	Problems related to age validation are not evident Validation refers to accuracy and verification refers to a process that ensures age estimates achieve acceptable levels of accuracy and precision. An absence of reasonable expectation of accuracy and of a quality control standards for ageing of fishes can lead to biased estimates (as opposed to unavoidable random ageing error) (Buckmeier et al., 2017; Campana, 2001; Choat et al., 2009).
Mark-recapture	What is the degree of confidence in mark-recapture approaches to produce robust growth parameter estimates?
Mark-recapture	Problems related to model fitting are not evident Failure to account for individual variation in growth can result in biased parameter estimates. Alternatives to Fabens (1965) original formulation take individual growth variability into account (Dureuil et al., 2022; Eveson et al., 2007; Zhang et al., 2007).

Table 3. Guidance on empirical basis scoring for maturity parameter estimates (L50 and L95). Start with a score of 4. If assumption(s) are not met, consider downgrading the score to maximum cumulative downgrade of four points (resulting in empirical basis score of zero for that estimate).

Method	Guiding questions and related assumptions Maximum empirical basis score is 4
Maturity data
Microscopic identification	What is the degree of confidence in microscopic identification for reliable identification of mature/immature fish? Mature/immature fish are correctly distinguished from identification of maturity stages (particularly important for hermaphroditic species; (DeMartini & Howard, 2016). Histological maturity coincides with functional maturity (size species begins shifting from juvenile to adult habitat; (J. Prince et al., 2022). Additionally, see estimation of parameters (below)
Macroscopic identification	What is the degree of confidence in macroscopic identification for reliable identification of mature/immature fish? Needs validation by more than 1 trained technician for assessing maturity stages to minimize bias from individual perception Sampling needs to be conducted at a time when there is less fish in resting stage (after spawning), because there is a high degree of misclassification between resting/inactive and immature fish (Ferreri et al., 2009; Klibansky & Scharf, 2015; Min et al., 2022). Macroscopic maturity coincides with functional maturity (Min et al., 2022) Difficult to apply for species with skipped spawning (Min et al., 2022), or species with fractional spawning (multiple spawning events over the year (Ferreri et al., 2009). Additionally, see estimation of parameters (below).
GSI (Gonadosomatic Index) (Flores et al., 2019)	What is the degree of confidence in using GSI for reliable identification of mature/immature fish? Cut-off value for distinguishing mature and immature fish needs to be carefully defined (Soares et al., 2020). Sampling needs to be conducted at a time when there is less fish in resting stage (after spawning), because there is a high degree of misclassification between resting/inactive and immature fish (Flores et al., 2019; Soares et al., 2020). Difficult to apply for species with skipped spawning (Min et al., 2022), or species with fractional spawning (multiple spawning events over the year; (Ferreri et al., 2009))
Estimation of parameters
Logistic model	*What is the degree of confidence in the logistic model for producing robust estimates of L50/L95?* Linear relationship between logarithmic transformed variables if using a GLM with a logit link function (ICES, 2008). Individual observations of maturity need to be independent Diagnostics conducted to verify above assumptions (e.g. residual analysis, observed vs fitted values).
L-W relationship (Fontoura et al., 2010)	What is the degree of confidence in the LWR method for producing robust estimates of L50/L95? Change points (inflection points) in allometric growth are evident, allowing to distinguish growth stages and inferring L50 (Soares et al., 2020). Changes in growth patterns are a reflection of reaching sexual maturity and not another life trait (Fontoura et al., 2010). Only applicable for species with polyphasic growth (Hashiguti et al., 2019)
Life history ratios (L50 = 0.66 Linf; (Jensen, 1997))	What is the degree of confidence in using life history ratios for producing robust estimates of L50/L95? Needs estimation of Linf (see Table S2 guidance) Ratios likely vary by taxonomic hierarchy (e.g., family, genus, species) (J. D. Prince et al., 2023)

Table 4. Guidance on empirical basis scoring for indirect and direct methods for estimating M. Start with a score of 4. If assumption(s) are not met, consider downgraded the score to maximum cumulative downgrade of four points (resulting in empirical basis score of zero for that estimate).

Method	Guiding questions and related assumptions to be met
Indirect methods
Tmax Maximum empirical basis score is 4	What is the degree of confidence that Tmax is representative of maximum age? Very old fish exist in the stock or existed during sampling of fish ages. Age truncation, resulting in loss of very old fish, is more likely where a stock is exploited (Barnett et al., 2017; Stewart, 2011).
	Sampling of age structure of catches or population is unbiased against capture of very old fish. Fishing gear or fishery-independent survey gear may fail to select for fish of the maximum age (Hamley & Regier, 1973). Survey design, including spatial and temporal extent, may affect encounters of fish of maximum age (Cailliet et al., 2001; Reis & Pawson, 1992).
	The Tmax age estimation method was validated and verified. Validation refers to accuracy and verification refers to a process that ensures age estimates achieve acceptable levels of accuracy and precision. An absence of reasonable expectation of accuracy and of a quality control standards for ageing of fishes can lead to biased estimates (as opposed to unavoidable random ageing error) (Buckmeier et al., 2017; Campana, 2001; Choat et al., 2009). Unvalidated ageing of older slow growth individuals can underestimate true age (Cailliet et al., 2001; Nesslage et al., 2022). Validation methods can include 1) tag–recapture, often combined with oxytetracycline injection and analysis in growth-zones of bone upon recapture; (2) analysis of growth-zones over time; and (3) radiometric approaches utilizing a known radioactive decay series as an independent chronometer in otoliths from bony fishes (Cailliet et al., 2001); (4) identification of strong year-classes, (5) length frequency analysis, and (6) examination of the edge of a particular structure throughout the year to show that only one annulus is formed (Chilton & Beamish, 1982).
	Sample size is sufficient to likely detect fish of maximum age. Small sample size, even from an unfished stock, may fail to detect fish of the maximum age (Nesslage et al., 2022).
K and/or L∞	If K and/or L∞ are used in M estimation, obtain scoring from Table 4 von Bertalanfy Growth function parameters
GSI (e.g., M = 0.03 + 1.68WGSI)	What is the degree of confidence in the observed GSI?
	Gonad weight observations are carefully timed. Failing to align sampling of somatic and reproductive tissues with reproductive cycle can bias GSI estimates (Gunderson & Dygert, 1988).
	Dry weights rather than weight weights are used in GSI calculation Wet weights can cause inaccuracies (Gunderson & Dygert, 1988).
	Sampling approach minimized the effects of interannual variability in reproductive effort. GSI can be biased due to failure to account for temporal variability in growth and reproductive output (e.g., gonad and oocyte size) (Gunderson & Dygert, 1988).
	The histological stage of the oocytes is consistent among observed gonadal weights Degree of oocyte maturation can affect gonadal weight observation, and thus, bias GSI estimates (Gunderson & Dygert, 1988).

Direct Methods
Measure Z in an unexploited stock using catch curve analysis	What is the degree of confidence in the catch curve analysis?
	The stock has been unexploited for multiple generations. Only for unexploited stocks does Z equal M (Maunder et al., 2023). Recovery times must be considered if stock within an MPA is to be considered “unexploited” (Barceló et al., 2021).
	The age structure of the stock is stable. Recruitment variability can bias estimates of Z and therefore M (Smith et al., 2012). Multiple years of catch-at-age data are likely to produce more robust estimates of Z than a single sampling event (Smith et al., 2012).
	Population dynamics are stationary. Where multiple years, in particular multidecadal time-series, of data are used in analysis, non-stationarity in survival can produce biased and/or imprecise estimates of Z (Chapman & Robson, 1960).
	All individuals in the stock have an equal probability of capture and sampling is representative of the entire stock. Failure to constrain cohort analysis to only fully recruited age-classes can produce erroneous estimate of Z (Smith et al., 2012).
Relate Z from catch curve analysis to actual fishing effort and extrapolate to zero fishing effort (Pauly, 1984)	What is the degree of confidence in the catch curve analysis and measurement of fishing effort? For degree of confidence in the catch curve analysis refer to above assumptions for Z in an unexploited stock using catch curve analysis?
	Fishing effort is targeting the stock in question. Effort that is not targeting the stock in question is not representative.
	Fishing effort is accurately measured and recorded The measure of effort needs to be representative of that deployed in the fishery (e.g., trip vs. hours fished). Inaccurate recording of effort will produce biased estimates of unexploited Z (McCluskey & Lewison, 2008).
	There is interannual contrast in fishing effort. Lack of interannual contrast in fishing effort can produce biased estimates of unexploited Z (Windsland, 2015).
Measure both Z and exploitation rates and solve for M and F (e.g., Hewitt et al., 2007)	What is the degree of confidence in the catch curve analysis and measurement of exploitation rate? For degree of confidence in the catch curve analysis refer to above assumptions for “Z in an unexploited stock using catch curve analysis”.
	The commercial catch is accurately measured and recorded. Inaccurately recorded catch can lead to erroneous exploitation rates and biased estimate of M (Hewitt et al., 2007).
	The abundance estimates are accurate. Inaccurately abundance estimates can lead to erroneous exploitation rates and biased estimate of M (Hewitt et al., 2007).
	The assumptions of the Baranov’s catch equation, i.e., the population is in a steady state over time, and the instantaneous rates of fishing and natural mortalities of fish are constant over time and age, are valid (Liu & Heino, 2014).
Mark-recapture and telemetry studies (e.g., Hoenig et al., 1998)	What is the degree of confidence in the mark-recapture or telemetry study design and analysis?
	Tag loss rates are known and reasonable. Tag loss can be conflated with natural mortality, thus failure to measure tag loss can lead to erroneous estimates of M (Chapman, 1961).
	Tagging mortality is known and is acceptably low. Tagging mortality can be conflated with natural mortality, thus failure to measure tag loss can lead to erroneous estimates of M (Arnason & Mills, 1981; Hoenig et al., 1998).
	Migration is accounted for in the estimation model. Failure to correctly apply closed vs. open population models can lead to biased estimates of M (Ricker, 1975).
	Sufficient receivers are deployed in minimize possibility of incomplete detection Incomplete detection can lead to biased estimates of M (Peterson et al., 2021).
Estimate M internally in an integrated stock assessment model (Aldrin et al., 2021; Jiao et al., 2012; Johnston et al., 2015; Maunder & Punt, 2013).	What is the degree of confidence in the M estimate from the stock assessment model?
	Substantial historical catch-at-age data are available. Integrated stock assessment models rely on historical catch-at-age data (Fournier & Archibald, 1982).
	Model misspecification is assessed. Failure to account for parameter and/or model misspecification and uncertainty can lead to biased or imprecise estimates of M (Gudmundsson & Gunnlaugsson, 2012).
	Informative or non-informative priors on M are vetted and defensible. Failure to examine underlying assumptions of prior on M can lead to biased or imprecise posteriors.