Issue 
Aquat. Living Resour.
Volume 35, 2022



Article Number  16  
Number of page(s)  17  
DOI  https://doi.org/10.1051/alr/2022016  
Published online  07 October 2022 
Research Article
Analyses of the composite yield per recruit model CYPR14 for inferring plausible fishing mortality targets of fish in the tropics
Fish and Wildlife Research Institute, Florida Fish and Wildlife Conservation Commission, 100 8th Avenue, SE, St. Petersburg, FL 33701–5095, USA
^{*} Corresponding author: Joseph.Munyandorero@MyFWC.com
Handling Editor: David Kaplan
Received:
1
December 2021
Accepted:
13
August 2022
Stocks' yield and size per recruit are widely used to provide fisheries management guidance. This study provides details for analyzing the composite (i.e. ageaggregated or stagestructured) yield per recruit (CYPR) model CYPR14, and proposes CYPR14 as a management tool for tropical fisheries. The fishing mortality rates maximizing CYPR (F_{CYPR}) and associated with the marginal increase in CYPR (F_{0.1}) and a target composite spawning potential ratio (CSPR; F_{35%CSPR} or F_{40%CSPR}) were suggested as candidate fishing mortality targets, provided assessments employ the delaydifferential model underlying CYPR14. Using Monte Carlo (MC) simulations relying on growth parameters and natural mortality of Lake Tanganyika's Lates stappersii and Lake Victoria's Lates niloticus, CYPR14 analyses involving maximum survivorship or declining survivorship were carried out to show how F_{CYPR}, F_{0.1}, F_{35%CSPR}, and F_{40%CSPR} could be generated, given an age of knifeedge recruitment (r). Baseline MC employed r = 1 year and yielded mean annual rates of F_{CYPR} = 0.52, F_{0.1} = 0.33, and F_{35%CSPR} = 0.51 for L. stappersii and F_{CYPR} = 0.23, F_{0.1} = 0.14, and F_{40%CSPR} = 0.16 for L. niloticus. CYPR14 with maximum survivorship produced CYPR isopleths such that the CYPR maximized at an infinite r and finite, higher F. For CYPR14 involving a declining survivorship, the CYPR declined with increased r and maximized with innermost closedloop contours at lower F and an optimal age. The CSPR isopleths from both types of CYPR14 analyses were first concave down, and the optimal age served as their inflection point. In terms of benchmarks based on the maximum sustainable yield and of proxies thereof, CYPR14 should be for its underlying delaydifferential model what the agestructured pool models are for agestructured assessment models.
Key words: Datapoor stocks / stock status / Lates stappersii / Lates niloticus / Monte Carlo simulations
© J. Munyandorero, Published by EDP Sciences 2022
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
1 Introduction
Since the 1980s, the status of tropical fisheries has often been assessed through length data analyses, essentially using the software called FiSAT II (Gayanilo et al., 2005). The status determination chiefly relied on two approaches. With the first approach, analysts sequentially (i) estimated growth parameters, (ii) calculated the rates of natural mortality (M) with Pauly's (1980) estimators, (iii) derived the rates of total mortality (Z) usually through lengthconverted catch curves (Pauly, 1983), (iv) subtracted natural mortality from total mortality to obtain the fishing mortality rates (F), and (v) derived the exploitation fractions as the ratios of fishing mortality to total mortality (F/Z). The results from (iv) and (v) were treated as indicators of, respectively, the fishing pressure and stock status for the time periods under review: analysts invariably concluded that the stocks were lightly (synonymized with weakly or under), optimally (synonymized with sustainably), or strongly (synonymized with over) exploited if the exploitation fractions were, respectively, less than 0.5, equal to 0.5, and greater than 0.5. In doing so, M was implicitly made equivalent to the fishing mortality producing the maximum sustainable yield (MSY), F_{MSY}: the stock status was evaluated by comparing F = Z − M with M. Thus, the stocks were judged to be lightly, optimally, or strongly exploited if F was less than M, equal to M, or greater than M.
The second approach simulated the relative yield per recruit (YPR) and the relative biomass per recruit (Pauly and Soriano, 1986). The biological reference points based on this approach include (i) the exploitation fraction maximizing the relative YPR, (ii) the exploitation fraction at which the marginal increase in the relative YPR is 10% of the marginal increase when the exploitation fraction is zero, and (iii) the exploitation fraction achieving 50% of the unfished relative biomass per recruit. To determine the stock status, scientists treated the exploitation fractions derived with the second method as targets (i.e. desirable levels) against which the exploitation fractions from the first approach were compared to determine whether the stocks were healthy (e.g. Bannerman and Cowx, 2002; Montcho et al., 2015).
Another type of analysis of potential management interest for tropical fisheries, and in general, for fisheries without age data irrespective of the fishery jurisdiction, may deal with composite (i.e. ageaggregated or stagestructed) yield per recruit (CYPR) models. CYPR models are derived from the governing equations of, and inherit assumptions associated with, delaydifference production models and delaydifferential production models under equilibrium (Munyandorero, 2012, 2015). Analyses employing CYPR models involve (i) an equation calculating the composite spawningstock per recruit (CSSR) and (ii) an equation calculating the CYPR as the product of the fishing mortality rate, average recruit during a stable fishing regime, and CSSR. Thus, unlike the age and lengthstructured YPR and spawningstock biomass per recruit, for which the selectivity and maturity schedules are often different, CYPR and CSSR are interrelated. The interrelation between CYPR and CSSR is due to the assumption that, consistent with the underlying delaydifference or delaydifferential production model, animals in the harvested population, including recruits, all are fully selected and equally reproductively mature, and the selectivity and maturity schedules are identical. CYPR models infer the reference points in terms of fishing mortality rates. But, like the age or lengthstructured spawningstock biomass per recruit, the absolute CSSRs are not typically used; the related fishing mortality rates must, if need be, be derived from the composite spawning potential ratios (CSPRs, i.e. the fished CSSRs divided by the unfished CSSRs).
This study is based on the CYPR model CYPR14. In fact, of the CYPR models identified by Munyandorero (2015), CYPR14 was recommended for use because (i) it integrates more growth parameters, (ii) it has a yieldlike curve (i.e. such a curve has a maximum whatever the age of recruitment lower than or equal to the maximum age in the population), (iii) the related maximizing fishing mortality corresponds to its counterpart derived from conventional surplusproduction models, and (iv) its maximizing fishing mortality was the most accurate estimator of F_{MSY} and proxy thereof as prescribed through singlespecies, nominally datarich stock assessments. CYPR14 is rooted in a continuous delaydifferential model (CDDM) that was developed by Walters (2011, 2020); see applications by Liao et al. (2016a and b) and Liao and Karim (2021). Like most CYPR models and consistent with the assumptions associated with the underlying delaydifference models and delaydifferential models (e.g. Quinn and Deriso, 1999), CYPR14 assumes (i) knifeedge selection or, equivalently, knifeedge recruitment to stable fishing regimes (because the ages of recruitment and selectivity are similar, these processes are henceforth used interchangeably), (ii) constant growth and natural mortality rates, and (iii) knifeedge maturity identical to the knifeedge selectivity. The latter assumption implies that in the CDDM and similar models, if spawning occurs early in this time step (e.g. this year) prior to the fishery and natural mortality, the total biomass of individuals at the start of this time step constitutes both the harvestable biomass and the spawning biomass, and comprises the biomass of this time step's recruits and the total biomass of adults that survive the previous time step: adults and recruits this time step all are fully selected and equally (100%) mature. Likewise, in CYPR14, the biomass of a recruit that enters a stable fishing regime represents the harvestable biomass and the spawning biomass.
The objective of this study was twofold. The first objective was to provide the CYPR14 details needed for analyses employing this model. The second objective was to propose a rationale of using CYPR14based fishing mortality targets (i.e. the fishing mortality rates that may potentially be desired by the fishery stakeholders for controlling perpetually the fishing effort in order to prevent overfishing and recruitment impairment). As illustration, stochastic analyses of CYPR14 were carried out and the plausible, candidate fishing mortality targets were generated for two fish species inhabiting each a different great lake in East Tropical Africa.
2 Methods
This section presents (i) the CYPR14 model (assumptions and deterministic equations, input requirements, and variations of the CYPR, CSSR, and CSPR with F) and (ii) CYPR14based fishing mortality targets (rationale for fisheries management and estimation), and (iii) stochastic illustrations of (i) and (ii) using lifehistory variables of two latid species of the African great lakes. Tables 1 and 2 provide, respectively, the notations and equations pertaining to CYPR14. The value of F maximizing the CYPR is denoted by F_{CYPR}.
Notations used in the CYPR14 analysis.
The CYPR14 model (equations without source are developed in this study).
2.1 The CYPR14 model
2.1.1 Assumptions and basic equations
For a singlespecies stock, consider a sequence of nominal F values (i.e. F values for fully selected fish) summarizing annual steps of piecewise stable fishing regimes: F (year^{−1}) = 0, …. By convention, each annual step begins at time 0, on 1 January, and ends at time 1, on 31 December. The age of knifeedge selection and the age of knifeedge maturity are equal for all stable fishing regimes, but each stable fishing regime is characterized by a unique nominal fishing mortality, assumed to be constant throughout the year. For each fishing regime, the previous selectivity and maturity patterns determine and constrain the CSSR, in that the CSSR calculation cannot integrate alternative, eventually different selectivity and maturity schedules.
Given the above considerations, CYPR14 is a CYPR model in which the CSSR is hereinafter represented by the function Ф of F, Ф(F) (Tab. 2, Eq. (1)) and the CYPR is described by the function γ of F, γ(F) (Tab. 2, Eq. (2a) or Eq. (2b)). In short, the CYPR14 model is a system of equations (1) and (2a) or (2b).
Following equation (2a), the CYPR is the CSSR multiplied by the number of fish caught, from one recruit (R = 1), between time 0 when that recruit entered a fishing regime and time 1 when that regime ended. Equation (2a) can also be represented by equation (2b). Equation (2b) says that the CYPR results from the application of the fishingregimespecific value of F to the product of the average recruit (Eq. (3)) and the CSSR; that product () quantifies the average, harvestable and spawning biomass produced by one recruit during a fishing regime. Note that is a function of F.
2.1.2 Input requirements
F is the only independent variable in the CYPR14 model. The CSSR, CYPR, average recruit, and their related quantities shown by equations (4a)–(4c) in Table 2 require natural mortality (M), asymptotic weight (W_{∞}), average weight of a recruit (w_{r}), and the rate of metabolic loss of body mass (µ) as defined by Walters (2011, 2020), but, whatever the fishing regime, these lifehistory variables are greater than zero, should be estimated externally, and are fixed.
Pauly's nonlinear empirical equation, without the variable temperature (Then et al., 2015), is herein used to estimate M: , where L_{∞} (cm) and K (year^{−1}) are the von Bertlanffy growth parameters; but note that M can be derived using any estimator that an analyst deems to be appropriate. The previous estimator of M is preferred over Pauly's (1980) estimators of M involving the variable temperature because recent studies found that the variable temperature was not statistically significant (e.g. Gislason et al., 2010; Then et al., 2015).
The asymptote weight is calculated as , where A is the scale and B is the exponent. Given the mean weight at age a (w_{a}) calculated as , where a_{0} is the age when the weight and length at age 0 are 0, the parameter µ is estimated using the nonlinear regression described by the equation derived by Huynh (2020): w_{a+1} = W_{∞} + (w_{a} − W_{∞}) e^{−μ}.
The average weight of a recruit (w_{r}) is w_{a} for an age a treated both as the age of knifeedge selectivity and as the age of knifeedge maturity.
2.1.3 Variations of the CSSR, CSPR, and CYPR with F
Munyandorero (2015) showed that the function Φ(F) for the CSSR varies similarly as the age and lengthstructured spawning stock biomass per recruit vary with F, and the composite spawning potential ratio (CSPR = the fished CSSR divided by the unfished CSSR) varies similarly as the static age and lengthstructured spawning potential ratios (SPR) vary with F. Regarding the function γ(F) for the CYPR, it exhibits a yieldlike curve that, whatever the value of r less than or equal to the maximum age in the population, approaches zero as F approaches infinity and always peaks at an F value (F_{CYPR}). These patterns contrast γ(F) with traditional age and lengthstructured per recruit models, for which the YPR has no peak if the age or length of recruitment exceeds the optimal age or the optimal length, i.e. the age or the length at which growth rate and biomass are maximum for an unfished cohort (Laurec and Le Guen, 1981; Froese et al., 2008). The previous patterns of CYPR and CSPR are illustrated below with the application species.
2.2 CYPR14based fishing mortality targets
2.2.1 Rationale for fisheries management
In fishery areas with datarich stocks and catch control rules (e.g. some fishery areas of the U.S.A), an F level associated with a target SPR (e.g. F at 40%SPR and denoted by F_{40%SPR}) is commonly treated as an F_{MSY} proxy and a fishing mortality target (i.e. F_{40%SPR} is believed not to jeopardize the spawning stock and to impair recruitment). By analogy to this management paradigm, an F level associated with a target percent CSPR (e.g. F at 40%CSPR or F_{40%CSPR}) can also be treated as a proxy of F_{MSY} that would be estimated once a CDDM is carried out, but the underlying SRR is untrustworthy, and the resulting MSYbased benchmarks are deemed unreliable for use. Such an F value can therefore be treated as a fishing mortality target to ensure healthy spawning stocks and avoid recruitment impairment in the tropics.
An F value maximizing the CYPR (F_{CYPR}) is also proposed as a candidate fishing mortality target for two reasons. First, such an F value theoretically corresponds to F_{MSY} as derived from conventional surplusproduction models (Munyandorero, 2015). Second, comparisons of F_{CYPR} to the accepted F_{MSY} or F_{MSY} proxies from nominally datarich stock assessments showed that F_{CYPR} has moderate to little bias (and, hence, moderate to high accuracy) visàvis most of those F_{MSY} or F_{MSY} proxies (Munyandorero' s (2015) Fig. 8a).
Another F level that can be set as a fishing mortality target is a value of F at which the marginal increase in CYPR is 10% of the marginal increase in CYPR when F = 0. Such an F value is denoted by F_{0.1} and, by definition, is such that . The application of F_{0.1} may ensure the optimum economic (and perhaps social) returns, minimal loss in maximum yield (Gulland and Boerema, 1973), and, compared to F_{CYPR}, improved recruitment.
Although F_{CYPR}, F_{0.1}, and an F value producing a target CSPR (F_{35%CSPR} or F_{40%CSPR} in this study) are suggested as alternative F_{MSY} proxies, their levels can be seen as indicative of different signals. Like the traditional SPR, the CSPR measures the proportion of the spawning population that remains after fishing and, as such, indicates the impact of fishing on a stock's reproductive capacity and so emphasizes protecting the spawning stock rather than maximizing yield (e.g. Gulf of Mexico SPR Management Strategy Committee, 1996; Munyandorero, 2015). Thus, like for a percent SPR (Gabriel and Mace, 1999), a percent CSPR lower than a target CSPR signals recruitmentoverfishing (i.e. a situation when the fishing mortality corresponding to that percent CSPR exceeds F at a target CSPR and reduced the biomass of mature individuals such that recruitment is impaired). There has also been a belief that F_{0.1} is a conservative reference point for defining recruitmentoverfishing (Mace and Sissenwine, 1993).
Nonetheless, F_{0.1} and F maximizing the traditional YPR have been alternatively treated as indicative of the fisheries productivity: they are fishing mortality rates beyond which the optimum growth is hindered, thereby depressing the potential yield and catch value (as a result of being based on animals that have not reached the optimal age or the optimal size) and, consequently, causing growthoverfishing (e.g. Rivard and Maguire, 1993; Gabriel and Mace, 1999; BenHasan et al., 2021). This paradigm is herein applied to F_{CYPR} and F_{0.1} from CYPR14. Note that both recruitmentoverfishing and growthoverfishing can happen simultaneously when the fishery is dealing with small and immature animals and the fishing mortality is high enough (BenHasan et al., 2021).
2.2.2 Estimation of F_{CYPR}, F_{0.1}, and F associated with a target CSPR
F_{CYPR}, F_{0.1}, and F associated with a target CSPR can be obtained through numerical search. F_{CYPR} can also be obtained either (i) through an optimization method of γ or (ii) by solving the zeroderivative of γ (γ′ = 0) for F; γ′ is given by equations (4a)–(4c).
3 Application of CYPR14 to two fish species of the African great lakes
3.1 Application species and input parameters
The equations found in Table 2 were applied to two latid species: (i) Lates stappersii exploited in Lake Tanganyika (3° 20'–8° 48' S, 29° 05'–31° 15' E (http://www.fao.org/fi/oldsite/ltr/gen.htm)) and (ii) Lates niloticus exploited in Lake Victoria (0° 20′ N–3° S, 31° 39′–4° 53′ E (Njiru et al., 2008)). These species were chosen solely because of the availability of lifehistory variables needed to address uncertainty in (hence to calculate plausible values of) various quantities, especially in the proposed fishing mortality targets, and because of their economic importance (Sarvala et al., 2006; Marshall and Mkumbo, 2011).
To analyze CYPR14 and infer plausible fishing mortality targets for Lake Tanganyika's L. stappersii and Lake Victoria's L. niloticus, the basic inputs were (i) the von Bertlanffy growth parameters L_{∞} and K (note: a_{0} = 0 because the growth parameters were derived using the electronic length–frequency analysis (ELEFAN) algorithm of the software FiSAT II, a method that precludes the estimation of a_{0}), (ii) the weight–length scales (A) and exponents (B), and (iii) the nominal fishing mortality rate F (F = 0, …, 2). L_{∞}, K, A, and B came from the literature and were treated as observed data (Supplementary Tabs. 1 and 2).
3.2 Characterizing uncertainty and stochasticity
Uncertainty was characterized under program R, version 4.1.0 (R Core Team, 2021), by performing random draws of L_{∞}, K, A, B, and M. Uncertainty in these parameters was propagated into other quantities (W_{∞}, w_{a}, w_{r}, µ, , CYPR, CSSR, CSPR, and fishing mortality targets) also using program R. As in Munyandorero (2018, 2020), it was considered (i) that the variability of the observed growth parameters reflected scientific uncertainty in their estimations and (ii) that no unique pairs (L_{∞}, K) and (A, B) were better than other such pairs for proper CYPR14 analyses. In other words, all compiled pairs (L_{∞}, K) and (A, B) were assumed to be equally reliable.
To incorporate growth parameters uncertainty, the values of L_{∞} and K on the one hand and of A and B on the other, were jointly resampled with replacement (5000 replicates) given each parameter's mean and without assuming a probability distribution error. The resampled parameter values were negatively related, strongly for B against A, and their boxplot distributions were approximately symmetrical (Supplementary Figs. 1a–d).
Uncertainty in M (Supplementary Figs. 2a, b) was introduced by sampling M values from a lognormal distribution error with a coefficient of variation (CV) of 0.5 and a mean, on log scale, obtained by inputting the random draws of L_{∞} and K in the equation . A CV of 0.5 was suggested by MacCall (2009) as a minimal default value for the variability of M on log scale.
Given the random realizations of L_{∞}, K, A, B, and M, 5000 Monte Carlo (MC) simulations were run employing various equations. For each MC sequence, a set of random values of L_{∞}, K, A, B, and M was used to derive : (i) asymptotic weight, (ii) mean weight at age, (iii) the parameter µ, (iv) the CSSR, CSPR, and F_{35%CSPR} for L. stappersii and F_{40%CSPR} for L. niloticus based on their presumed longevities (see below), (v) average recruitment, and (vi) CYPR, F_{CYPR}, and F_{0.1}. This way, the 5000 MC simulations generated different random values of W_{∞} (Supplementary Figs. 2c, d), w_{r} (Supplementary Figs. 2e, f), µ (Supplementary Figs. 2g, h), CYPR, F_{CYPR}, F_{0.1}, CSPR, F_{35%CSPR}, and F_{40%CSPR}. In the end, it was possible to describe uncertainty in various quantities.
3.2.1 Baseline analyses
Baseline MC runs employed the age of knifeedge recruitment and maturity r of 1 year for both species. Thus, baseline w_{r} was the mean weight at age 1 (Supplementary Figs. 2e, f). This assumption stemmed from the following facts. First, the L. stappersii fishery in Lake Tanganyika has operated with a variety of small meshsized and nonselective fishing gears and methods and has never been regulated in terms of size limits (Munyandorero, 2002, 2018). Second, even though the L. niloticus fishery in Lake Victoria has been regulated in terms of slot limits, it persistently operated with multiple, nonselective gear types and methods and compliance with slot limits may have been inefficient (Njiru et al., 2009; Munubi and Nyakibinda, 2020). Such fishery characteristics and management histories suggest that fishermen have always caught all available sizes and ages of Lake Tanganyika's L. stappersii and of Lake Victoria's L. niloticus. It should be noted that the age at 50% maturity for both species is attained in the first or second year; so, the assumption of knifeedge maturity at one year is slightly violated in baseline analyses.
3.2.2 Responses of CYPR and CSPR to various ages of knifeedge recruitment
In addition to age 1 used as the default age of knifeedge recruitment and maturity, sensitivity analyses involving different ages of recruitment (r = 0, …, 10 for L. stappersii, and r = 0, …, 16 for L. niloticus) were performed to evaluate the effects of r on CYPR14 results. The maximum lifespan of 10 years for L. stappersii came from Munyandorero (2002). The maximum lifespan of 16 years for L. niloticus was the mean of the ratios 2.996/K. The ratio 2.996/K as estimator of maximum lifespan is due to Taylor (1958).
Explicit in the evaluations of equation (1) and equations (2a) or (2b) by age r were (i) that w_{r} = w_{a} and (ii) that, for r > 1, the age of maturity became r. Implicit in those evaluations was that a recruit had a maximum survivorship to age r (i.e. the probability is 1 for that recruit to survive to age r in the absence of fishing). Because of this constraint of maximum survivorship to age r, technically due to defining the recruitment at age r, analyses and results were hereinafter referred to as CYPR14 analyses with maximum survivorship (CYPR14 m.s.), CYPR m.s., and CSPR m.s., etc.
In the end, sensitivity analysis results were displayed through isopleths of the summary statistics (mean, median, and the 2.5% and 97.5% percentiles, and hence, the 95% uncertainty envelop, UE) of CYPR and CSPR as functions of F and r, constructed with the R package akima (Akima and Gebhardt, 2021). The curves of fishing at candidate fishing mortality targets were superimposed on those isopleths to show how those mortality rates varied with r. In the CYPR isopleths thus constructed, the topographies to the right of the fishing curve at each fishing mortality target were treated as areas of growthoverfishing. For the CSPR isopleths, such topographies were treated as areas of recruitmentoverfishing.
3.2.3 Responses of CYPR and CSPR to declining survivorships to age r
A relatively meaningful sensitivity analysis of the effects of r on CYPR and CSPR would be (i) to consider that all members of a cohort graduate to the population at age r_{p} (the cohort size is herein 1 individual) but are not necessarily vulnerable and 100% reproductively mature and (ii) to calculate the CSSR. at successive values of r given that, from age r_{p} to age r − 1, the cohort endures natural mortality only and, at age r, also fishing mortality on top of natural mortality. Thus, there is a timelag between age r_{p} and age r (r − r_{p} years), during which the cohort is affected by natural survivorship, i.e. the probability that it will survive from r_{p} to r in the absence of fishing.
Denoting the survivorship at age r_{p} by l_{rp} = 1 (here, r_{p} = 1 year consistent with baseline analyses), the oneindividual cohort of age r_{p} is ultimately affected by its survivorship l_{r} to age r when it becomes fullyselected and 100% reproductively mature: l_{r} = exp[−(r − 1)M] (in fact, at age r_{p} = 1, l_{rp} is rewritten as l_{1} = exp(−0M) =1; if r = 2, l_{2} = exp(−0M) exp(−M) = exp(−M); if r = 3, l_{3} = exp(−M) exp(−M) = exp(−2M) and so on until when r = the maximum age in the population). Note that l_{r} can be obtained recursively using agespecific values of M: l_{r} = l_{r−1}exp(−M_{r–1}), but, for illustration purposes, M_{r–1} is set to M. In short, this approach initializes the recruitment of size 1, age r_{p}, and l_{rp =} 1 at the population level. When that recruit will enter a fishing regime at age r, it will, from age to age, have been cumulatively decremented by M such that the CSSR is represented by equation (5); because l_{r} decreases as r increases, it was hereinafter referred to as declining survivorship. The CYPR is calculated as usual using equations (2a) or (2b), but, in this case, employs equation (5) for the CSSR. Note (i) that if r = r_{p} (i.e. the youngest cohort in the population becomes 100% mature and fully selected), CYPR14 analyses reduce to baseline analyses, (ii) that the CSSR and CYPR involving the declining survivorship preclude use of ages r < r_{p}, because l_{r} cannot be greater than 1, and (iii) that l_{r} propagates additional uncertainty in the CSSR, CYPR, and CSPR.
4 Results
4.1 Baseline outcomes
In this section and Sections 4.2. and 4.3, the results relate to CYPR14 m.s. As expected, the curves of γ were domeshaped and varied asymmetrically with F (Figs. 1a, b). They peaked at mean F_{CYPR} = 0.52 year^{−1} (CV = 0.22) for Lake Tanganyika's L. stappersii and at mean F_{CYPR} = 0.23 year^{−1} (CV = 0.09) for Lake Victoria's L. niloticus (Tab. 3). The CSPR declined as F increased (Figs. 1c, d). Mean F_{35%CSPR} for L. stappersii was 0.51 year^{−1} (CV = 0.29) and mean F_{40%CSPR} for L. niloticus was 0.16 year^{−1} (CV = 0.10).
Figure 1 and Table 3 provide other statistics of F_{CYPR}, F_{0.1}, F_{35%CSPR}, and F_{40%CSPR}. Mean F_{CYPR} and mean F_{35%CSPR} for L. stappersii were close, and both estimates were substantially greater than mean F_{0.1} of 0.33 year^{−1} (CV = 0.23). For L. niloticus, mean F_{40%CSPR} < mean F_{CYPR} and was slightly greater than mean F_{0.1} = 0.14 year^{−1} (CV = 0.09). Note (i) that the CYPR and CSPR as well as F_{CYPR}, F_{0.1}, and F_{40%CSPR} for L. niloticus were more precise (narrower 95% uncertainty envelops, UEs, and interquartile ranges as well as lower CVs) than the CYPR, CSPR, F_{CYPR}, F_{0.1}, and F_{35%CSPR} for L. stappersii, (ii) that, for some reasons in L. stappersii, CYPR's 97.5% percentile peaked at a lower F than did CYPR's mean and lower percentiles, but various summary statistics peaked at almost the same F level for L. niloticus, (iii) that F_{CYPR} and F_{0.1} in L. stappersii were comparably precise but were more precise than F_{35%CSPR}, and (iv) that the precisions of L. niloticus' F_{CYPR}, F_{0.1}, and F_{40%CSPR} were comparable.
Fig. 1 Variations, against fishing mortality (F), of mean (black solid line), median (blue dashdotted line), and the 95% uncertainty envelops (gray area) of (a) and (b) the composite yield per recruit (CYPR) and (c) and (d) the composite spawning potential ratio (CSPR) based on baseline CYPR14 analyses (age of knifeedge recruitment = 1 year) for L. stappersii in Lake Tanganyika (left panels) and L. niloticus in Lake Victoria (right panels). The vertical, black dashed line, solid line, and dashed line represent, from left to right, the 25% percentile, mean, and 75% percentile of F_{CYPR} and F_{35%CSPR} (L. stappersii) or F_{40%CSPR} (L. niloticus); the vertical, green dashed line, solid line, and dashed line represent, from left to right, the 25% percentile, mean, and 75% percentile of F_{0.1}; the horizontal red dashed line, solid line, and dashed line on the CYPR plots represent, from bottom to top, the 25% percentile, mean, and 75% percentile of MSY per recruit; the horizontal red and solid line on the CSPR plots locate the levels of 35%CSPR and 40%CSPR. 
Summary statistics of candidate fishing mortality targets for L. stappersii in Lake Tanganyika and L. niloticus in Lake Victoria as obtained from base CYPR14 analyses (UI = uncertainty interval, CV = coefficient of variation).
4.2 Effects of r on the CYPR
For both application species, CYPR's summary statistics increased with F and mainly with r (Fig. 2). CYPR isopleths show (i) that CYPR medians and means were comparable, but CYPRs' 95% UEs were wide (e.g. at r = 10 years in L. stappersii, the 95% UE of CYPR ranged from >0.19 to >0.4 kg; at r = 16 years in L. niloticusi, the 95% UE of CYPR was about 28–60 kg) and (ii) that CYPRs' summary statistics were highest at the maximum age and were associated with some finite and relatively higher levels of F.
The variations of the candidate fishing mortality targets with r indicate that those mortality targets increased gradually from age 0 to age 4 in L. stappersii and to age 9 in L. niloticus, and then leveled off. For L. niloticus, the means, medians, and 2.5% and 97.5% percentiles of F_{40%CSPR} and F_{0.1} were marginally different, especially at lower values of r; all these summary statistics were lower than their F_{CYPR} counterparts and so were equally conservative (in terms of effort reduction), presenting less risk of growthoverfishing irrespective of r. For L. stappersii, (i) all summary statistics of F_{0.1} were less than the corresponding statistics of F_{CYPR} and F_{35%CSPR} and so were more conservative in terms of less risk to fishery yield, (ii) only the 2.5% percentile of F_{35%CSPR} was less than the 2.5% percentile of F_{CYPR} for all r values, (iii) the 97.5% percentile of F_{35%CSPR} was greater than the 97.5% percentile of F_{CYPR} whatever r, and (iv) the means and medians of F_{35%CSPR} and F_{CYPR} were similar from age 0 to age 4; for r > 4, mean and median F_{35%CSPR} exceeded mean and median F_{CYPR}. Thus, for r > 4 in L. stappersii, F_{CYPR} on average presented less risk of growthoverfishing than F_{35%CSPR}.
Fig. 2 Isopleths of, from top to bottom, the 2.5% percentile, median, mean, and 97.5% percentile of CYPR m.s. (kg) as a function of fishing mortality (F) and the age of knifeedge recruitment or of knifeedge selection for L. stappersii in Lake Tanganyika (left panels) and L. niloticus in Lake Victoria (right panels). The plots also show the eumetric fishing curves of the 2.5% percentile, median, mean, and 97.5% percentile of F_{CYPR} (blue solid line), F_{0.1} (orange solid line), as well as F_{35%CSPR} for L. stappersii and F_{40%CSPR} for L. niloticus (blue dashdotted line) against the age of knifeedge recruitment. Areas shaded by the red color, yellow color, and green color are, respectively, topographies of low CYPR, intermediate CYPR, and high CYPR. If F_{CYPR}, F_{0.1}, or F at a target CSPR is treated as the fishing mortality target, the region below and to the right of the corresponding lines represents growthoverfishing. 
4.3 Effects of r on the CSPR
At lower Fs, isopleths of the CSPR summary statistics were high and parallel to the yaxis, and, therefore, were insensitive to r (Fig. 3). With higher Fs, their isopleths were first concave down, increasing (up to age 4 in L. stappersii and age 9 in L. niloticus), then they suddenly became concave up, increasing. Furthermore, the CSPR response surfaces were precise, having similar means and medians; for L. stappersii, however, the lower and upper limits of mean CSPRs' 95% UEs were 5% lower and 10% greater, respectively, than mean CSPRs. No value of r in the F–r planes avoided CSPRs less than 35% for L. stappersii and less than 40% for L. niloticus.
The curves of the points (F_{CYPR}, r), (F_{0.1}, r), (F_{35%CSPR}, r), and (F_{40%CSPR}, r) were also superimposed on the CSPR isopleths (Fig. 3). They were described along with CYPR isopleths (Fig. 2). They noticeably vary similarly as the CSPR isopleths, suggesting that the CSSR may be their main determinant. The coordinates (F_{35%CSPR}, r) for L stappersii and (F_{40%CSPR}, r) for L niloticus lay perfectly on the 35% and 40% contours, except in the isopleth of mean CSPR of L. stappersii when r > 4 years, perhaps due to interpolation issues.
While the performance of F_{CYPR}, F_{0.1}, F_{35%CSPR}, and F_{40%CSPR} in CYPR isopleths was judged with respect to their risk to fishery yield, their performance in CSPR isopleths should deal with their risk to the spawning stock biomass. In this context, means and medians of F_{CYPR} and F_{35%CSPR} for L stappersii would equally avoid recruitmentoverfishing if r ≤ 4 years; for r > 4, F_{CYPR} would best protect the spawning biomass. Whatever the value of r of L. stappersii, mean and median F_{0.1} would outperform mean F_{CYPR} and mean F_{35%CSPR} in avoiding recruitmentoverfishing in that F_{0.1} summary statistics were associated with CSPR between 45% and 50%. For L niloticus, the spawning stock biomass would be at risk at mean or median F_{CYPR} whatever the age of knifeedge recruitment, but fishing at mean or median F_{40%CSPR} and, better, at mean and median F_{0.1}, would avoid recruitmentoverfishing.
Fig. 3 Isopleths of, from top to bottom, the 2.5% percentile, median, mean, and 97.5% percentile of CSPR m.s. (%) as a function of fishing mortality (F) and the age of knifeedge recruitment or the age of knifeedge selection for L. stappersii in Lake Tanganyika (left panels) and L. niloticus in Lake Victoria (right panels). The plots also show the variations of the 2.5% percentile, median, mean, and 97.5% percentile of F_{CYPR} (blue solid line), F_{0.1} (orange solid line), as well as F_{35%CSPR} for L. stappersii and F_{40%CSPR} for L. niloticus (blue dashdotted line) against the age of knifeedge recruitment. Areas shaded by the red color, yellow color, and green color are, respectively, topographies of low CSPR, intermediate CSPR, and high CSPR. If F_{CYPR}, F_{0.1}, or F at a target CSPR is treated as the fishing mortality target, the region below and to the right of the corresponding lines represents recruitmentoverfishing. 
4.4 Effects of declining survivorship to r on the CYPR
In this section and Section 4.5, the assumption of equal selectivity and maturity remains as in CYPR14 m.s., but one individual recruited to the population at age 1 and to fishing regimes at successive r values after being affected by the declining survivorship to age r (l_{r}).
For L. stappersii, the combination of l_{r} and r did not change the magnitude of F_{CYPR}, F_{0.1}, and F_{35%CSPR} and their variations with r. But that combination shifted the CYPR m.s. response surfaces from the topopenended orientations and increasing trends seen in Figure 2, to concavedown shapes at higher r values; those shapes became progressively vertical or oblique, with leftward orientations at lower r values (Fig. 4). The CYPR means and calculated percentiles increased toward innermost contours, whereby the topographies of higher CYPRs extended horizontally up to F = 2 year^{−1} and vertically up to r = 4 years. The highest CYPR was lower than the CYPR m.s. (Fig. 2), but contrary to the latter result, maximum CYPR occurred at low Fs and r = 1–4 years. For example, the greatest mean CYPR contour spanned r = 1 and r = 2 and was achieved by the lowest values of all candidate fishing mortality targets. In this context, it may be appropriate to apply F_{0.1} because it entails less fishing pressure while producing the highest yield.
The combination of l_{r} and r in L. niloticus also reshuffled the CYPR response surfaces and the patterns of F_{CYPR}, F_{0.1}, and F_{40%CSPR} versus r (Fig. 4). Use of l_{r} led to CYPR isopleths with leftward horizontal shapes and welldefined closedloop contours of maximum CYPR; the CYPR maximizing contours were centered at r = 5–9 years across narrow ranges of F. The highest CYPR was achieved by F_{CYPR} only, while the next highest CYPR was achieved by F_{40%CSPR} and F_{0.1}. The curves of the points (F_{CYPR}, r), (F_{0.1}, r), and (F_{40%CSPR}, r) peaked at r ≈ 7 years; F_{CYPR}, F_{0.1}, and F_{40%CSPR} were comparatively lower than their counterparts derived using CYPR14 m.s. Finally, the topographies of higher CYPRs did not stretch horizontally beyond F ≈ 1.2 year^{−1} but extended vertically up to r = 16 years.
Fig. 4 Isopleths of, from top to bottom, the 2.5% percentile, median, mean, and 97.5% percentile of CYPR involving a declining survivorship (kg) as a function of fishing mortality (F) and the age of knifeedge recruitment or of knifeedge selection for L. stappersii in Lake Tanganyika (left panels) and L. niloticus in Lake Victoria (right panels). The plots also show the eumetric fishing curves of the 2.5% percentile, median, mean, and 97.5% percentile of F_{CYPR} (blue solid line), F_{0.1} (orange solid line), as well as F_{35%CSPR} for L. stappersii and F_{40%CSPR} for L. niloticus (blue dashdotted line) against the age of knifeedge recruitment. Areas shaded by the red color, yellow color, and green color are, respectively, topographies of low CYPR, intermediate CYPR, and high CYPR. If F_{CYPR}, F_{0.1}, or F at a target CSPR is treated as the fishing mortality target, the region below and to the right of the corresponding lines represents growthoverfishing. 
4.5 Effects of declining survivorship to r on the CSPR
For L. stappersii, the CSPR isopleths and the variations of F_{CYPR}, F_{35%CSPR}, and F_{0.1} with r were insensitive to l_{r} (Fig. 5) in that they showed patterns and management performance comparable to those obtained with CSPR m.s. (Fig. 3). For L. niloticus, the CSPR isopleths were first concave down, then turned shortly upward (Fig. 5). Instead of showing vertical orientations seen with CSPR m.s isopleths. (Fig. 3), CSPR isopleths involving the declining survivorship were parallel with the curves of F_{CYPR}, F_{40%CSPR}, and F_{0.1} against r, and, like them, peaked between r = 5–9 years before trending leftward.
Fig. 5 Isopleths of, from top to bottom, the 2.5% percentile, median, mean, and 97.5% percentile of CSPR involving a declining survivorship (%) as a function of fishing mortality (F) and the age of knifeedge recruitment or the age of knifeedge selection for L. stappersii in Lake Tanganyika (left panels) and L. niloticus in Lake Victoria (right panels). The plots also show the variations of the 2.5% percentile, median, mean, and 97.5% percentile of F_{CYPR} (blue solid line), F_{0.1} (orange solid line), as well as F_{35%CSPR} for L. stappersii and F_{40%CSPR} for L. niloticus (blue dashdotted line) against the age of knifeedge recruitment. Areas shaded by the red color, yellow color, and green color are, respectively, topographies of low CSPR, intermediate CSPR, and high CSPR. If F_{CYPR}, F_{0.1}, or F at a target CSPR is treated as the fishing mortality target, the region below and to the right of the corresponding lines represents recruitmentoverfishing. 
4.6 On the tradeoffs between CYPR and CSPR
It may be useful to plot and read CYPR and CSPR isopleths together to determine management tradeoffs between CYPR and CSPR. Figure 6 presents mean CYPR and mean CSPR isopleths jointly, considering that the fisheries (hypothetically) operated at any candidate mean fishing mortality targets obtained from baseline analyses. Unlike in Schmalz et al. (2016), these fishing mortality targets were superimposed on isopleths because the application species lacked stock assessment results that may otherwise help judge the performance of CYPR14 results.
In the F–r dimensions of L. stappersii' s CYPR and CSPR, higher F for any value of r reduced CSPR irrespective of whether the calculations used the maximum or the declining survivorship (Fig. 6). For a given F > 0, the reduction of CSPR at F = 0 (i.e. 100%) was lesser as r was increased. For example, the CSPR associated with r = 2 years was reduced from 100% at F = 0 to 40–45% at F_{CYPR}, whereas at r = 6 years, the CSPR of 100% at F = 0 was reduced to 60–65% at F_{CYPR}. The same reduction pattern can be seen for L. niloticus' CSPR m.s. When, in L. niloticus, the calculations used the declining survivorship, the reduction of CSPR at F = 0 to a given F > 0 was highest for r associated with the inflection point of CSPR isopleths.
At any candidate fishing mortality target, higher r increased the CYPR m.s. and the CSPR m.s. for both application species (Fig. 6). But when the declining survivorship was accounted for in L. stappersii, the CYPR declined and the CSPR increased with higher r. Interestingly for this species, use of the declining survivorship suggested the possibility of achieving both maximum CYPR and CSPR > 40% at F_{CYPR} or F_{35%CSPR}. In L. niloticus, the CYPR involving the declining survivorship was maximum between r = 5–9 years, years below and above which the CYPR declined; the CSPR associated with F_{CYPR}, F_{0.1}, and F_{40%CSPR} varied little with r.
Fig. 6 Simultaneous presentation of tradeoffs between mean CYPR (red contours and figures) and mean CSPR (black contours and figures) as a function of fishing mortality (F) and the age of knifeedge recruitment or the age of knifeedge selection for L. stappersii in Lake Tanganyika (left panels) and L. niloticus in Lake Victoria (right panels), considering that the fisheries operate at mean F_{CYPR} (vertical, blue solid dash line), mean F_{0.1} (vertical, blue solid line), or mean F_{35%CSPR} for L. stappersii and mean F_{40%CSPR} for L. niloticus (vertical and light green dash line) as estimated from baseline analyses (Table 3). 
5 Discussion
The present study described details on the CYPR14 model and its analyses, with the goal to propose inclusion of CYPR14 in the tropical fisheries management toolkit. The populations of Lake Tanganyika's L. stappersii and Lake Victoria's L. niloticus were used to illustrate analyses that incorporated uncertainty in growth and natural mortality. CYPR14 may be suitable for datapoor stocks because it only requires growth parameters and natural mortality. CYPR14 can be the basis of various biological reference points (Munyandorero, 2015), but the focus herein was on potential fishing mortality targets in terms of CYPR14based F_{CYPR} and F_{0.1} and of an F value associated with a target CSPR.
Beside the simplest CYPR14 analyses, this study expanded on analyses of the stochastic response surfaces of CYPR and CSPR in F–r dimensions, whereby the calculated CSSR followed from CYPR14 m.s. or involved a declining survivorship. The latter analyses showed features that contrast CYPR and CSPR isopleths with isopleths of traditional length and agebased YPR and SPR or relative fecundity per recruit, RFR (i.e. isopleths of YPR and SPR or RFR as functions of F on the xaxis and of the length L_{c} or age t_{c} at first capture on the yaxis, assuming no fishinginduced mortality of animals below L_{c} or t_{c}). Likewise, the eumetric fishing curves associated with these two types of yield isopleths trended differently.
To grasp the differences between the YPR and CYPR and between the SPR or RFR and CSPR, first recall that the classical YPR isopleths tend to be horizontally Ushaped, opening at the right side; the YPR increases from the outer contours to the inner ones and maximizes at an infinite F level with the innermost, far right horizontal or oblique contour associated with the optimal age or the optimal length (e.g. Laurec and Le Guen, 1981; Annala and Breen, 1989; Goodyear, 1993; Froese et al., 2008; Schmalz et al., 2016; Mildenberger et al., 2017). Furthermore, the eumetric fishing curves (i) of the points (F_{max}, L_{c}) or (F_{max}, t_{c}) where F_{max} is the fishing mortality rate that maximizes the YPR for a given value of L_{c} or t_{c}, and (ii) of the points (F, L_{max}) or (F, t_{max}) where L_{max} or t_{max} is the length or the age that maximizes the YPR for a fixed value of F, have commonly been superimposed on the YPR isopleths (e.g. Laurec and Le Guen, 1981; Ault et al., 2019). Such curves increase and are concave down. On the other hand, the SPR or RFR isopleths increase with a downward concavity, trending asymptotically against F; the SPR or RFR is lowest at the inner and rightward contours, typically for small and young spawners, and highest otherwise, typically for large and old spawners (e.g. Laurec and Le Guen, 1981; Goodyear 1993; Froese et al., 2008; Schmalz et al., 2016; Munyandorero, 2018, Ault et al., 2019). While a contour of the target SPR or RFR can appear on the SPR or RFR isopleths, no literature describing such plots with eumetric fishing curves was found.
In contrast, CYPR m.s. isopleths are concave up with a bottomup increase of the CYPR that maximizes with an infinite r value at the innermost vertical contour associated with a finite F value. The reasons for the bottomup increase of the CYPR are that, following equations (1) and (2a) or (2b) and given a set of L_{∞}, K, A, and B, the values of W_{∞}, w_{a}, and µ do not change. As a result, higher r values drive up w_{r}, the CSSR, and CYPR so that, for F > 0, the CYPR is highest at the largest r value considered in the isopleth construction. Since, for a given value of F > 0, the age r maximizing CYPR (r_{max}) is unique, it may be superfluous to superimpose the eumetric fishing curves of the points (F, r_{max}) on CYPR isopleths. Regarding the curves of the points (F_{CYPR}, r), (F_{0.1}, r), (F_{35%CSPR}, r) for L. stappersii, and (F_{40%CSPR}, r) for L. niloticus, they increase and are concave up, varying similarly as the CSPR isopleths.
On the other hand, CYPR isopleths involving the declining survivorship, as well as the curves of the points (F_{CYPR}, r), (F_{0.1}, r), (F_{35%CSPR}, r), and (F_{40%CSPR}, r), show unique patterns that differ between the applications species, perhaps due to the differences in their longevity and other input parameters. Overall, the declining survivorship reduced the CYPR, such that, in L. stapperssii (a relatively shortlived and smallbodied species), higher r values were associated with the lowest CYPR, and the points (F_{CYPR}, r), (F_{0.1}, r), and (F_{35%CSPR}, r) varied similarly as in CYPR m.s. isopleths. Thus, unlike in CYPR m.s. isopleths, higher F_{CYPR}, F_{0.1}, and F_{35%CSPR} generated using the declining survivorship produced lower CYPR. In L. niloticus (a relatively longlived and largebodied species), the effects of the declining survivorship appear to be offset by the magnitude of w_{r} because higher CYPR can be seen at higher r values. For this species, not only F_{CYPR}, F_{0.1}, F_{40%CSPR} were low and spanned the topography of high CYPR across the r range, but also the curves of the points (F_{CYPR}, r), (F_{0.1}, r), and (F_{40%CSPR}, r) varied little.
Of the greater interest for CYPR isopleths involving the declining survivorship is the fact that maximum CYPR occurs at innermost contours across narrow ranges of low values of r and F. Because CYPR assumes, like the traditional YPR, that there is no fishinginduced mortality below r, and because the traditional YPR maximizes at the optimal age and the optimal length (Laurec and Le Guen, 1981; Froese et al., 2008), an analysis of biomass against age for an unfished cohort was carried out post hoc to check if the highest CYPR occurred at optimal ages. This analysis revealed that the optimal age associated with mean CYPR was 3 years for L. stappersii and 7 years for L. niloticus (Supplementary Fig. 3). Therefore, the declining survivorship caused CYPRs to maximize at r values in the vicinity of those optimal ages.
The patterns of CSPR m.s. isopleths and of CSPR and CYPR isopleths involving the declining survivorship were at glance puzzling. There was no a priori expected behavior of those CYPR isopleths. As for both types of CSPR isopleths, they were expected to resemble the traditional SPR or RFR isopleths, considering that the assumption of ages r of knifeedge selection implies that all fish below r and available in the fishing ground would not be vulnerable. Technically, this assumption is equivalent to a situation of a fishery in which various size or age limits have been imposed and fish below those limits, once caught, all have been released and have experienced zero fishinginduced mortality (100% postrelease survival). Yet the abovementioned CYPR and CSPR isopleths to some extent resemble the YPR and SPR isopleths found, for example, in Coggins et al. (2007), Goodyear (1993), and Munyandorero (2018) when species were subjected to catchandrelease and size limit regulations and to postrelease survival rates.
Regardless, a close examination of equations (1) and (5) along with their evaluation at each age r reveals that the rspecific CSSR is analogous to the agespecific biomass for an unfished cohort (Supplementary Fig. 3). In fact, in equations (1) and (5), the oneindividual cohort is affected by the natural survivorship to age r − 1 (l_{r−1}) only and, at age r, also by fishing and natural mortalities on top of l_{r}. It is worth recalling that, for equation (1), l_{r} = 1 to each age r; for equation (5), l_{r} decreases as r is higher (Tab. 2). In addition to l_{r}, another parameter varying with r is the mean weight w_{r}. It follows that, for each nominal F, the curve of the CSSR against r has an optimal age, acting as the inflection point for the curve in question (Supplementary Fig. 4 for the CSSR response surfaces). With the CSSR involving the declining survivorship, that inflection point is, irrespective of F, well defined at r = 2–3 years for L. stapersii' s mean CSSR and at r = 6 or 7 years for L. niloticus' s mean CSSR isopleths. The CSPR isopleths based on both types of survivorships and the CYPR isopleths involving the declining survivorship inherit the patterns of CSSR isopleths, including the inflection points at optimal ages, beyond and around which the CSPR and CYPR contours are reshuffled, taking various orientations and shapes apparently determined by the balance between l_{r} and w_{r}.
CYPR14 analyses involving the declining survivorship were primarily developed as sensitivity analyses. But they may be more realistic in evaluating the effects of r on CYPR and CSPR, for three reasons. First, the declining survivorship adjusts the CSSR m.s., through which it is counterintuitive that, in a multiage population, the yield from a cohort increases continually with r, owing to the implicit, strong assumption that, from age to age, the cohort experienced zero natural mortality since it recruited to the population. Second, the declining survivorship causes the CYPR to maximize at an optimal age, which is coherent with the traditional YPR analyses. Finally, for a fixed rate of F, higher r may be associated with lower CYPR and higher CSPR.
On a social note, and if the management goal is, for example, the maximization of yield, maximum CYPR based on the declining survivorship may be preferable over maximum CYPR m.s. In fact, stakeholders may be worried by the cost of high fishing pressure necessary for achieving the latter type of CYPR, whereas they would be relatively comfortable with the maximum CYPR calculated using the declining survivorship, because such an CYPR implies lesser fishing pressure.
A reviewer requested a sensitivity analysis on the impact of the assumption of equal (and differing) ages of maturity and selectivity on CSPR. The assumption of equal maturity and selectivity, which is typical of twostagestructured models of exploited population dynamics, is rigid; it strictly applies to a baseline analysis conditional on that assumption being justified. Exploring various ages r of recruitment/selectivity, which also are the ages of 100% maturity, implies a strong assumption that the schedules in question are shifted to and coincide at the explored values of r, because CYPR14 only considers the CSSR and the resulting CYPR and CSPR at an assumed age of knifeedge selectivity. In comparison to results employing the true age of maturity (typically in a baseline analysis), the previous paragraphs indicate (i) that greater r values in calculations with maximum survivorship inflate the CSSR and CYPR and (ii) that the combination of r and l_{r} brings about speciesspecific patterns of CSSR and CYPR. Regarding the effects of differing ages of maturity and selectivity on CSPR, they may be impossible to evaluate in the context of twostagestructured models.
The status of the stocks of Lake Tanganyika's L. stappersii and Lake Victoria's L. niloticus was not determined because this study was not accompanied with stock assessments aimed at such a task. The calculated values of F_{CYPR}, F_{0.1}, F_{35%CSPR} (for L. stappersii) and F_{40%CSPR} (for L. niloticus) were instead suggested as candidate fishing mortality targets, provided stock assessments or simulations (e.g. through management strategy evaluations) of Lake Tanganyika's L. stappersii and Lake Victoria's L. niloticus relied on CDDMs—this comment on the restrictive, joint use of CYPR14 and CDDM is fundamental in that it guards one against comparisons of CYPR14based fishing mortality targets to results from delaydifference/differential formulations other than the CDDM, because CYPR models and reference points therefrom or involving them (typically MSY benchmarks) must be consistent with the equations governing the parent stagestructured models (Munyandorero, 2015, Fig. 4 and Tab. 6). In that case, F_{CYPR} could serve in two ways. Like F_{0.1} or an F value associated with a target CSPR, F_{CYPR} would first be treated as an F_{MSY} proxy if the SRR and F_{MSY} resulting from a CDDMbased assessment are untrustworthy (note: the preference between F_{0.1}, F_{35%CSPR} or F_{40%CSPR}, and F_{CYPR} would depend on the management goal). In that case, for example, an estimate of F in the last year of the CDDM could be compared to F_{CYPR} to determine the stock status (provided this is the management objective). Second, since F_{CYPR} is a relatively accurate estimator of F_{MSY} (Munyandorero, 2015), it can be used to initialize the estimation of F_{MSY} in a CDDMbased assessment, consistent with the socalled managementoriented approach of exploited population dynamics (e.g. Martell et al., 2008).
6 Conclusions
Despite advances in developing CYPR models (Munyandorero, 2012, 2015), the present study may be the first one to provide detailed analyses specifically of CYPR14. The goal of the study was to propose CYR14 as a management tool for tropical fisheries, along with or alternatively to the commonly used lengthbased methods. More generally, it is useful to stress that, in terms of MSYbased benchmarks and proxies, CYPR14 is for CDDMs what the conventional agestructured pool models are for agestructured stock assessment models (e.g. Shepherd, 1982). Therefore, CYPR14, preferably involving the declining survivorship if the ages of recruitment to the population and to the fishery are different, would be fully useful if it were accompanied with CDDMbased stock assessments or simulations.
Although the previous goal and usefulness of CYPR14 were advocated for datapoor stocks in the tropics, they are valid for any datapoor stocks irrespective of the geographical area.
Conflict of interest
There are no known competing interests that could influence this study.
Data accessibility statement
The input parameters, tools (R scripts), and outputs that support this study are available at https://doi.org/10.17882/90241 (Munyandorero, 2022).
Acknowledgments
This work was prepared at the Florida Fish and It is dedicated to Carl Walters for his inspiring works Wildlife Research Institute. Reviewers are thanked for their comments. This research did not receive any grant from funding agencies in the public, commercial, or notforprofit sectors.
References
 Akima H, Gebhardt A. 2021. akima: Interpolation of Irregularly and Regularly Spaced Data. R package version 0.62.3. Available at https://CRAN.Rproject.org/package=akima. [Google Scholar]
 Annala JA, Breen PA. 1989. Yield and eggperrecruit analyses for the New Zealand rock lobster, Jasus edwardsii. New Zealand Journal of Marine and Freshwater Research 23: 93–105. [Google Scholar]
 Ault JS, Smith SG, Bohnsack JA, Luo J, Stevens MH, DiNardo GT, Johnson MW, Bryan DR. 2019. Lengthbased risk analysis for assessing sustainability of datalimited tropical reef fisheries. ICES J Mar Sci 76: 165–180. [CrossRef] [Google Scholar]
 Bannerman PO, Cowx IG. 2002. Stock assessment of the bigeye grunt (Brachydeuterus auritus, Val.) fishery in Ghanaian coastal waters. Fish Res 59: 197–207. [CrossRef] [Google Scholar]
 BenHasan A, Walters C, Hordyk A, Christensen V, AlHusaini M. 2021. Alleviating growth and recruitment overfishing through simple management changes: Insights from an overexploited longlived fish. Mar Coast Fish 13: 87–98. [Google Scholar]
 Coggins Jr. LG, Catalano MJ, Allen MS, Pine III WE, Walters CJ. 2007. Effects of cryptic mortality and the hidden costs of using length limits in fishery management. Fish Fish 8: 196–210. [CrossRef] [Google Scholar]
 Froese R, SternPirlot A, Winker H, Gascuel D. 2008. Size matters: how singlespecies management can contribute to ecosystembased fisheries management. Fish Res 92: 231–241. [CrossRef] [Google Scholar]
 Gabriel WL, Mace PM. 1999. A review of biological reference points in the context of the precautionary approach. Proceedings, 5th NMFS NSAW. NOAA Tech. Memo. NMFSF/SPO40. [Google Scholar]
 Gayanilo Jr. FC, Sparre P, Pauly D. 2005. FAOICLARM Stock Assessment Tools II (FiSAT II). Revised Version. User's Guide. FAO, Rome, 168 p. [Google Scholar]
 Gislason H, Daan N, Rice JC, Pope JG. 2010. Size, growth, temperature and the natural mortality of marine fish. Fish Fish 11: 149–158. [CrossRef] [Google Scholar]
 Goodyear CP. 1993. Spawning stock biomass per recruit in fisheries management: foundation and current use. Can Spec Pub Fish Aquat Sci 120: 67–81. [Google Scholar]
 Gulf of Mexico SPR Management Strategy Committee. 1996. An evaluation of the use of SPR levels as the basis for overfishing definitions in Gulf of Mexico finfish fishery management plans, Gulf of Mexico Fishery Management Council, Final Report. [Google Scholar]
 Gulland JA, Boerema LK. 1973. Scientific advice on catch levels. Fish Bull 71: 325–335. [Google Scholar]
 Huynh Q. 2020. Description of the delaydifference and delaydifferential models. Available at https://openmse.com/featuresassessmentmodels/1dd/. [Google Scholar]
 Laurec A, Le Guen JC. 1981. Dynamique des populations marines exploitées, Tome 1, Concepts et modèles. CNEXO/Centre Océanologique de Bretagne, Brest, France, 117 p. [Google Scholar]
 Liao B, Liu Q, Wang X, Baset A, Soomro SH, Memon AM, Memon KH, Kalhoro MA. 2016a. Application of a continuous time delaydifferential model for the population dynamics of winterspring cohort of neon flying squid (Ommastrephes bartramii, Lesueur 1821) in the northwest Pacific Ocean. Mar Biol Assoc 96: 1527–1534. [Google Scholar]
 Liao B, Liu Q, Zhang K, Baset A, Memon AM, Memon KH, Han Y. 2016b. A Continuous time delaydifferential type model (CTDDM) applied to stock assessment of the southern Atlantic albacore Thunnus alalonga. Chin J Oceanol Limnol 34: 977–984. [CrossRef] [Google Scholar]
 Liao B, Karim E. 2021. Evaluate the power of a modified continuous time DDM model, using BSPM and ASPM as benchmarks: a case study of a slowgrowing tuna species (Thunnus alalunga Bonnaterre, 1788). Iran J Fish Sci 20: 1740–1756. [Google Scholar]
 Mace PM, Sissenwine MP. 1993. How much spawning per recruit is enough? Can Spec Pub Fish Aquat Sci 120: 101–118. [Google Scholar]
 MacCall AD. 2009. Depletioncorrected average catch: a simple formula for estimating sustainable yields in datapoor situations. ICES J Mar Sci 66: 2267–2271. [CrossRef] [Google Scholar]
 Marshall BE, Mkumbo OC. 2011. The fisheries of Lake Victoria: Past present and future. Nature and Fauna 26: 8–13. [Google Scholar]
 Martell SJD, Walters CJ, Sumaila UR. 2008. Industryfunded fishing license reduction good for both profits and conservation. Fish Fish 9: 1–12. [CrossRef] [Google Scholar]
 Mildenberger TK, Taylor MH, Wolff M. 2017. TropFishR: An R package for fisheries analysis with lengthfrequency data. Meth Ecol Evol 8: 1520–1527. [Google Scholar]
 Munubi RN, Nyakibinda JN. 2020. Assessment of body size and catch per unit effort of Nile perch (Lates Niloticus) caught using different fishing gears at Magu district in Lake Victoria, Tanzania. Afr J Biol Sci 2: 73–83. [Google Scholar]
 Munyandorero J. 2002. The Lake Tanganyika clupeid and latid fishery system: indicators and problems inherent in assessments and management. Afr Stud Monogr 23: 117–145. [Google Scholar]
 Munyandorero J. 2012. A recruitmentmortality model in the precautionary management toolkit of African tropical inland, singlespecies fisheries. Fish Res 127128: 26–33. [CrossRef] [Google Scholar]
 Munyandorero J. 2015. Composite perrecruits: Alternative metrics for deriving biological reference points. Reg Stud Mar Sci 2: 35–55. [Google Scholar]
 Munyandorero J. 2018. Embracing uncertainty, continual spawning, estimation of the stockrecruit steepness, and sizelimit designs with lengthbased perrecruit analyses for African tropical fisheries. Fish Res 199: 137–157. [CrossRef] [Google Scholar]
 Munyandorero J. 2020. Inferring prior distributions of recruitment compensation metrics from lifehistory parameters and allometries. Can J Fish Aquat Sci 77: 295–313. [CrossRef] [Google Scholar]
 Munyandorero J. 2022. Analyses of the composite yield per recruit model CYPR14: Input parameters, R scripts, and outputs for the application species. SEANOE. https://doi.org/10.17882/90241. [Google Scholar]
 Njiru M, Kazungu J, Ngugi CC, Gichuki J, Muhoozi L. 2008. An overview of the current status of Lake Victoria fishery: Opportunities, challenges and management strategies. Lakes Reserv Res Manage 13: 1–12. [Google Scholar]
 Njiru M, Getabu AM, Taabu E, Mlaponi L, Muhoozi L, Mkumbo OC. 2009. Managing Nile perch using slot size: is it possible? Afr J Trop Hydrobiol Fish 12: 9–14. [Google Scholar]
 Pauly D. 1980. On the interrelationships between natural mortality, growth parameters, and mean environmental temperature in 175 fish stocks. J CIEM Mer 39: 175–192. [Google Scholar]
 Pauly D. 1983. Lengthconverted catch curves: a powerful tool for fisheries research in the tropics (part I). Fishbyte 1: 9–13. [Google Scholar]
 Pauly D, Soriano ML. 1986. Some practical extensions to the Beverton and Holt's relative yield per recruit model. In: Maclean JL, Dizon LB, Hosillo LV (Eds.), The first Asian fisheries forum, Asian Fisheries Society, Manila, Philippines pp. 491–496. [Google Scholar]
 Quinn II JT, Deriso RB. 1999. Quantitative Fish Dynamics. Oxford, Oxford University Press. [Google Scholar]
 R Core Team. 2021. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Available at https://www.Rproject.org/. [Google Scholar]
 Rivard D, Maguire JJ. 1993. Reference points for fisheries management: the eastern Canadian experience. Can Spec Pub Fish Aquat Sci 120: 31–57. [Google Scholar]
 Sarvala J, Langenberg VT, Salonen K, Chitamwebwa D, Coulter GW, Huttula T, Kanyaru R, Kotilainen P, Makasa L, Mulimbwa N, Mölsä H. 2006. Fish Catches from Lake Tanganyika mainly reflect changes in fishery practices, not climate. Verh int Ver Limnol 29: 1182–1188. [Google Scholar]
 Shepherd JG. 1982. A versatile new stockrecruitment relationship for fisheries, and the construction of sustainable yield curves. ICES J Mar Sci 40: 67–75. [CrossRef] [Google Scholar]
 Schmalz PJ, Luehring M, Rose JD, Hoenig JM, Treml MK. 2016. Visualizing tradeoffs between yield and spawners per recruit as an aid to decision making. N Am J Fish Manag 36: 1–10. [Google Scholar]
 Taylor CC. 1958. Cod growth and temperature. ICES J Mar Sci 23: 366–370. [CrossRef] [Google Scholar]
 Then AY, Hoenig JM, Hall NG, Hewitt DA. 2015. Evaluating the predictive performance of empirical estimators of natural mortality rate using information on over 200 fish species. ICES J Mar Sci 72: 82–92. [Google Scholar]
 Walters CJ. 2011. The continuous time SchnuteDeriso delaydifference model for agestructured population dynamics. Available at https://studylib.net/doc/7616513/thecontinuoustimeschnutederisodelay. [Google Scholar]
 Walters CJ. 2020. The continuous time SchnuteDeriso delaydifference model for agestructured population dynamics, with example application to the Peru anchoveta stock. the Institute for the Oceans and Fisheries, University of British Columbia. Working Paper number 202004. Available at https://fisheries.sites.olt.ubc.ca/files/2020/06/1ContinuoustimeSchnuteDerisomodelFinal.pdf. [Google Scholar]
Cite this article as: Munyandorero J. 2022. Analyses of the composite yield per recruit model CYPR14 for inferring plausible fishing mortality targets of fish in the tropics. Aquat. Living Resour. 35: 16
Supplementary Material
Table 1. Observed von Bertalanffy growth parameters used to analyze CYPR14.
Table 2. Observed weight–length scales (A) and exponents (B) used to analyze CYPR14 for L. stappersii in Lake Tanganyika (lengths were in mm and weight in g) and L. niloticus in Lake Victoria (lengths were in cm and weight in g).
Figure 1. Scatter plots and marginal boxplot distributions of the random draws of (a) and (b) the von Bertalanffy growth parameters and (c) and (d) weight–length scales and exponents for L. stappersii in Lake Tanganyika (left panels) and L. niloticus in Lake Victoria (right panels).
Figure 2. Frequency distributions of (a) and (b) natural mortality (M), (c) and (d) asymptotic weight (W_{8}), (e) and (f) mass of the average recruit (w_{r}), and (g) and (h) the rate of metabolic loss of body mass (μ) for L. stappersii in Lake Tanganyika (left panels) and L. niloticus in Lake Victoria (right panels). The dashed blue line represents the mean of the distribution.
Figure 3. Variations, against age, of mean (black solid line), median (blue dashdotted line), and the 95% uncertainty envelops (gray area) of biomass for an unfished cohort of L. stappersii in Lake Tanganyika and L. niloticus in Lake Victoria. The agespecific biomass Ba was given by B_{a} = l_{a}w_{a}, where w_{a} (kg) is mean weight at age a and la is the survivorship to age a: l_{a} = 1 if a = 0 year and l_{a} = l_{a}1exp(M) = exp(aM) if a > 0; M is natural mortality. Mean biomass, median biomass, the 95% lower biomass limit, and the 95% upper biomass limit of L. stappersii peaked at ages 3, 3, 2, and 4 years. Such summaries statistics for L. niloticus peaked at ages 7, 7, 6, and 7 years.
Figure 4. Isopleths of the mean composite spawningstock per recruit (CSSR) as a function of fishing mortality (F) and the age of knifeedge recruitment or the age of knifeedge selection r for L. stappersii in Lake Tanganyika (left panels) and L. niloticus in Lake Victoria (right panels). The CSSR of the top plots was calculated considering that the survivorship to each age r is 1 (Eq. (1), Tab. 2) and the CSSR of the bottom plots involved the declining survivorship to each age r (Eq. (5), Tab. 2).
(Access here)All Tables
Summary statistics of candidate fishing mortality targets for L. stappersii in Lake Tanganyika and L. niloticus in Lake Victoria as obtained from base CYPR14 analyses (UI = uncertainty interval, CV = coefficient of variation).
All Figures
Fig. 1 Variations, against fishing mortality (F), of mean (black solid line), median (blue dashdotted line), and the 95% uncertainty envelops (gray area) of (a) and (b) the composite yield per recruit (CYPR) and (c) and (d) the composite spawning potential ratio (CSPR) based on baseline CYPR14 analyses (age of knifeedge recruitment = 1 year) for L. stappersii in Lake Tanganyika (left panels) and L. niloticus in Lake Victoria (right panels). The vertical, black dashed line, solid line, and dashed line represent, from left to right, the 25% percentile, mean, and 75% percentile of F_{CYPR} and F_{35%CSPR} (L. stappersii) or F_{40%CSPR} (L. niloticus); the vertical, green dashed line, solid line, and dashed line represent, from left to right, the 25% percentile, mean, and 75% percentile of F_{0.1}; the horizontal red dashed line, solid line, and dashed line on the CYPR plots represent, from bottom to top, the 25% percentile, mean, and 75% percentile of MSY per recruit; the horizontal red and solid line on the CSPR plots locate the levels of 35%CSPR and 40%CSPR. 

In the text 
Fig. 2 Isopleths of, from top to bottom, the 2.5% percentile, median, mean, and 97.5% percentile of CYPR m.s. (kg) as a function of fishing mortality (F) and the age of knifeedge recruitment or of knifeedge selection for L. stappersii in Lake Tanganyika (left panels) and L. niloticus in Lake Victoria (right panels). The plots also show the eumetric fishing curves of the 2.5% percentile, median, mean, and 97.5% percentile of F_{CYPR} (blue solid line), F_{0.1} (orange solid line), as well as F_{35%CSPR} for L. stappersii and F_{40%CSPR} for L. niloticus (blue dashdotted line) against the age of knifeedge recruitment. Areas shaded by the red color, yellow color, and green color are, respectively, topographies of low CYPR, intermediate CYPR, and high CYPR. If F_{CYPR}, F_{0.1}, or F at a target CSPR is treated as the fishing mortality target, the region below and to the right of the corresponding lines represents growthoverfishing. 

In the text 
Fig. 3 Isopleths of, from top to bottom, the 2.5% percentile, median, mean, and 97.5% percentile of CSPR m.s. (%) as a function of fishing mortality (F) and the age of knifeedge recruitment or the age of knifeedge selection for L. stappersii in Lake Tanganyika (left panels) and L. niloticus in Lake Victoria (right panels). The plots also show the variations of the 2.5% percentile, median, mean, and 97.5% percentile of F_{CYPR} (blue solid line), F_{0.1} (orange solid line), as well as F_{35%CSPR} for L. stappersii and F_{40%CSPR} for L. niloticus (blue dashdotted line) against the age of knifeedge recruitment. Areas shaded by the red color, yellow color, and green color are, respectively, topographies of low CSPR, intermediate CSPR, and high CSPR. If F_{CYPR}, F_{0.1}, or F at a target CSPR is treated as the fishing mortality target, the region below and to the right of the corresponding lines represents recruitmentoverfishing. 

In the text 
Fig. 4 Isopleths of, from top to bottom, the 2.5% percentile, median, mean, and 97.5% percentile of CYPR involving a declining survivorship (kg) as a function of fishing mortality (F) and the age of knifeedge recruitment or of knifeedge selection for L. stappersii in Lake Tanganyika (left panels) and L. niloticus in Lake Victoria (right panels). The plots also show the eumetric fishing curves of the 2.5% percentile, median, mean, and 97.5% percentile of F_{CYPR} (blue solid line), F_{0.1} (orange solid line), as well as F_{35%CSPR} for L. stappersii and F_{40%CSPR} for L. niloticus (blue dashdotted line) against the age of knifeedge recruitment. Areas shaded by the red color, yellow color, and green color are, respectively, topographies of low CYPR, intermediate CYPR, and high CYPR. If F_{CYPR}, F_{0.1}, or F at a target CSPR is treated as the fishing mortality target, the region below and to the right of the corresponding lines represents growthoverfishing. 

In the text 
Fig. 5 Isopleths of, from top to bottom, the 2.5% percentile, median, mean, and 97.5% percentile of CSPR involving a declining survivorship (%) as a function of fishing mortality (F) and the age of knifeedge recruitment or the age of knifeedge selection for L. stappersii in Lake Tanganyika (left panels) and L. niloticus in Lake Victoria (right panels). The plots also show the variations of the 2.5% percentile, median, mean, and 97.5% percentile of F_{CYPR} (blue solid line), F_{0.1} (orange solid line), as well as F_{35%CSPR} for L. stappersii and F_{40%CSPR} for L. niloticus (blue dashdotted line) against the age of knifeedge recruitment. Areas shaded by the red color, yellow color, and green color are, respectively, topographies of low CSPR, intermediate CSPR, and high CSPR. If F_{CYPR}, F_{0.1}, or F at a target CSPR is treated as the fishing mortality target, the region below and to the right of the corresponding lines represents recruitmentoverfishing. 

In the text 
Fig. 6 Simultaneous presentation of tradeoffs between mean CYPR (red contours and figures) and mean CSPR (black contours and figures) as a function of fishing mortality (F) and the age of knifeedge recruitment or the age of knifeedge selection for L. stappersii in Lake Tanganyika (left panels) and L. niloticus in Lake Victoria (right panels), considering that the fisheries operate at mean F_{CYPR} (vertical, blue solid dash line), mean F_{0.1} (vertical, blue solid line), or mean F_{35%CSPR} for L. stappersii and mean F_{40%CSPR} for L. niloticus (vertical and light green dash line) as estimated from baseline analyses (Table 3). 

In the text 
Current usage metrics show cumulative count of Article Views (fulltext article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 4896 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.