Age is not just a number – senescence affects how fish populations 2 respond to different fishing regimes

21 The presence of senescence in natural populations remains an unsolved problem in 22 biology. Described as an age-dependent increase in natural mortality (known as 23 actuarial senescence) and an age-dependent decrease in fecundity (known as 24 reproductive senescence), the role of senescence in nature is still poorly understood. 25 Based on empirical estimates of reproductive and actuarial senescence, we explored 26 how senescence affects the population dynamics of Coregonus albula , a small, 27 schooling salmonid fish. Using an empirically-based eco-evolutionary model, we 28 investigated how the presence or absence of senescence affects how the fish 29 population responds to pristine, intensive harvest, and recovery phases. Our results 30 showed that at an individual level, the presence of senescence was accompanied by 31 life-history trade-offs, i.e. lower asymptotic length and smaller size and younger age at 32 maturity, both in the presence and absence of fishing. At the population level, the 33 response to different fisheries selection patterns depended on the presence or absence 34 of senescence. Importantly, the results indicate that through the lifehistory trade-offs 35 between early reproduction and late life survival, the young and small individuals can 36 have an important role in population recovery, especially when senescence is present. 37 Since most life-history and fisheries models ignore senescence, they may be over- 38 estimating reproductive capacity and under-estimating natural mortality. Our results 39 highlight the need for increasing biological realism in these models to ensure the 40 successful management of our natural resources. 41


44
Senescence is considered a fundamentally fitness decreasing trait, and its presence 45 and role in natural populations remains an unsolved problem in biology (Monaghan et  kill them before the consequences of aging would commence (Medawar 1952). Today, 52 it is well known that this is not the case and evidence for senescence across taxa is  Life-history trade-offs may take place between functions as well as within the same 62 function. For instance, increased growth rate and increased reproductive effort early in 63 life and higher natural mortality rate later in life are known to be linked (Kirkwood and 64 Rose 1991; Lester et al. 2004). Similarly, investment in future reproductive effort is 65 thought to be of lesser value in terms of fitness benefits than current reproductive effort, 66 4 mainly due to the uncertainty of future reproduction (Zhang and Hood 2016). Following 67 the close link between life history characters and senescence, it is therefore likely, as 68 hypothesised by Benoît et al. (2018), that assuming that increased allocation to 69 reproduction in early life leads to an increased rate of ageing later, fishing-induced 70 changes in maturation age and senescence could actually be linked. Indeed, the natural 71 mortality rate of many fish populations is thought to have increased in the recent 72 decades (Gislason et al. 2010). 73 While most fishes express indeterminate growth and high longevity (Carey and Judge,74 2000), it has been suggested that fish experience delayed senescence relative to birds 75 and mammals, facilitated in part by the capacity for increasing fecundity with age 76 (Reznick et al. 2002). Indeed, female body size and reproductive output in fish are 77 known to be positively correlated, indicating that the older and larger the fish, the higher 78 its reproductive output. Most fisheries and fish population models (Beverton-Holt, 1957; and life-history models, the natural mortality of fish is often assumed to be independent 86 of the age or size of the fish (Gislason et al. 2010). Models with increasing fecundity 87 with age and size, and age-and size-independent natural mortality essentially describe 88 fish as having no reproductive or actuarial senescence at all. While rarely included in fisheries and life-history models, senescence in fish was first 90 documented over 60 years ago (Gerking 1957;Comfort 1960Comfort , 1963Woodhead and 91 Ellett 1966, 1967, 1969a   The eco-evolutionary model includes five main components (Fig 1a -e). These are four 148 dependent sets of variables: growth, fecundity, survival, population demographics, and 149 an independent variable: senescence.

150
Life history traits such as size and age at maturity are thought to be controlled by many 151 loci (Roff 2002). In fishes, the correlation of size at maturity and ∞ is a well-known life-152 history invariant (Charnov 1993). Thus, in the growth component (Fig. 1a), we utilised 153 empirical length-at-age data back-calculated from fish scales to model the ∞ . The ∞ 154 was set to be an evolving trait so that the genotype coding ∞ of each individual was 155 described by 10 diploid loci with two alleles in each. The alleles were inherited in the 156 classic Mendelian way, so that each offspring received one randomly drawn allele from the mother, and one from the father. Each allele was coded as 0 or 1 and the sum of 158 alleles across the ten loci was coupled with a normally distributed random number 159 (mean zero) to describe phenotypic variability, and then the sum was linearly translated more on k see below "Model parametrisation"). We used the empirical data to determine 164 that the maturation size threshold was at 67% of their ∞ (mean size at 2 years of age) 165 and no earlier than on their second autumn, which is in line with literature (Jensen 166 1998; Karjalainen et al. 2016). This way, we ensured that the fish in the model will 167 mature once they reach 67% of their ∞ , but never before they reach their second 168 autumn. Thus, fish younger than two years old, or two-year-olds smaller than 67% of 169 their ∞ could not yet reproduce.

170
The fecundity component (Fig. 1b) is based on a length-weight relationship, which was 171 specifically calculated for C. albula from Lake Puulavesi. Using this length-weight   The population component (Fig. 1d) describes density dependency so that at 75% of 182 the population carrying capacity, the individual growth is reduced to 50% of that 183 predicted by the individual's vB growth curve (its L ∞ and k parameters). Additionally, egg 184 production was set to be density dependent so that the closer the population was to its 185 carrying capacity, the fewer eggs were produced.   The empirically collected weight data (N = 27) and the growth trajectories calculated 224 above were used to calculate the length-weight relationship = × (Ricker 1975).  An increase in natural mortality following sexual maturity and reproduction is an reproduction was estimated to be the increase in mortality rate from age group 1 to age 285 group 2 as per Marjomäki (2005), and this was applied once in every scenario, whether 286 actuarial or reproductive senescence was modelled or not. its peak). For simplicity hereafter, when we discuss trawling, a logistic selection curve is 308 assumed, and when we discuss gillnetting, a dome-shaped selection curve is assumed.

309
Regardless of the fishing method, the fishing mortality (F) of the fully selected size class 310 was set to 0.7, which is considered a realistic level of magnitude for intensively fished 311 populations (Viljanen 1986). The fishing mortality in terms of biomass was kept identical consistently differed from those that did not include actuarial senescence. These 322 differences were seen before, during and after fishing in all parameters investigated.

323
Given that actuarial senescence appeared to be the major cause of the differences (Fig.   324 2), likely due to the relatively low reduction in reproductive output with age (Fig. S1, S2), 325 we focus most of the present work on two instead of four scenarios: a scenario with 326 both reproductive and actuarial senescence and a scenario with no senescence.

327
Additionally, reproductive and actuarial senescence are known to be linked (Kirkwood 328 and Shanley 2010), so exploring either both types of senescence together or none at all 329 is biologically more relevant than separating the senescence types.

330
Asymptotic length 331 The populations with senescence had a consistently lower ∞ than those with no 332 senescence. For both the senescent and non-senescent scenarios, fishing caused a 333 decline in ∞ (Fig. 3a, b), and the decline caused by trawling was larger than the decline 334 caused by gillnetting, regardless of the presence of senescence. However, the type of 335 fishing played a role in the relative change within a scenario. The senescent scenario had 336 a smaller decline in ∞ than the non-senescent scenario when trawled (Fig. 4a). The 337 opposite occurred when gillnetting was applied: the senescent scenario had a larger drop 338 in ∞ as a result of dome-shaped fishing compared to the non-senescent scenario (Fig.   339 4b). When fishing was ceased after 100 years, ∞ started to increase slowly in all 340 scenarios, but in none of the scenarios did the ∞ recover back to the level prior to fishing.

341
Associated changes in the vB growth parameter k, and average size and age at 342 maturation are shown in supplementary material (Fig S3 a, b, c, d, respectively).

343
Biomass 344 In the absence of fishing, whether pristine or recovery phase, the scenario with 345 senescence produced a lower biomass than the scenario without senescence (Fig. 3c,   346 d). When fishing pressure was applied the biomass of both populations declined.

347
Trawling (Fig. 3c) caused a larger drop than gillnetting (Fig. 3d). However, when 348 trawled, the population with senescence maintained a higher biomass than the one 349 without senescence (Fig. 3c). In the gillnetting scenario, the senescent population had a 350 slightly lower biomass during fishing compared to the non-senescent population (Fig.   351 3d).

352
Regardless of the type of fishing, the relative drop in biomass for populations with 353 senescence was smaller than for those with no senescence (Fig. 4 c, d). When trawling 354 was applied, the level of biomass stayed relatively constant for both senescent and non-355 senescent scenarios (Fig. 4c). However, gillnetting caused a sharp decline in biomass 356 then a sharp increase and then a slow, continuous decline for both scenarios (Fig. 4d).

17
The decline did not level off at any point during hundred years of fishing. When fishing 358 was ceased, all scenarios experienced a rapid increase in biomass with an initial peak 359 (that exceeded the level prior to fishing in gillnetting scenario), and then a sharp drop. The number of individuals (N) was consistently higher for populations with senescence, 366 than for those without, regardless of fishing type or the presence or absence of fishing 367 (Fig. 3e, f). The start of trawling caused a rapid initial decline in the number of fish, but 368 as trawling continued, the N increased in both senescent and non-senescent scenarios, 369 however it never reached the pre-fishing level (Fig. 3e). This was different from 370 gillnetting, which caused a steady increase in the N during fishing, above the pre-fishing 371 levels (Fig. 3f). As fishing was ceased, the populations that were trawled experienced a 372 rapid initial increase in N, and then a declining trend. When gillnetting was ceased, it 373 caused a slow decline in the number of fish. No scenario reached the pre-fishing level in 374 two hundred years of recovery.

375
In both trawling and gillnetting scenarios, the relative change in the number of fish was 376 smaller for the senescent scenario, compared to the non-senescent scenario (Fig. 4e,   377 f). However, as the fishing continued, the difference between the senescent and non- 3a) than did gillnetting (Fig. 3b) for both senescent and non-senescent populations.

423
However, in the trawling scenario the asymptotic length of the senescent population 424 declined proportionately less than that of the non-senescent population (Fig. 4a). The 425 20 opposite occurred in the gillnetting scenario (Fig. 4b), where the asymptotic length of  In absolute terms, the senescent population maintained a higher biomass during 444 trawling than the non-senescent population (Fig. 3c). Our simulation allowed for control 445 over the fishing mortality, and the catch in terms of biomass was set identical for the 446 senescent and non-senescent populations. Before the fishing started, the population 447 with senescence had a lower biomass than that of the non-senescent population. Since 448 21 the absolute biomass of the catch is the same in both populations, this means that the 449 proportional catch from the senescent population (with initially lower biomass) is higher 450 than the catch from the non-senescent population (which had a higher biomass initially).

451
Therefore, the lower asymptotic length of the senescent population (Fig. 3a)  and the non-senescent population to gillnetting in terms of biomass did not differ as 472 much as they did to trawling. However, the drop in biomass from pristine phase to 473 fishing phase was relatively smaller for the senescent population (Fig. 4d), despite the 474 larger relative drop in asymptotic length (Fig. 4b). Like the situation in trawling scenario, 475 the enhanced reproduction of younger and smaller fish is likely to drive the relatively 476 higher biomass of the senescent population.

477
The present study showed a decline in the age and size at maturity for senescent . Therefore, for species that undergo 535 senescence, estimates of fecundity that ignore senescence may prove to be incorrect.

536
As discussed above, the smaller and younger fish may have an important role in 537 population recovery.

538
Similarly, due to lack of age-specific natural mortality data, typical fisheries models 539 assume a constant rate of natural mortality regardless of the age and size of the fish, or