Selective harvesting is acknowledged as a serious concern in efforts to conserve wild animal populations. In fisheries, most studies have focused on gradual and directional changes in the life-history traits of target species. While such changes represent the ultimate response of harvested animals, it is also well known that the life history of target species plastically alters with harvesting. However, research on the adaptive significance of these types of condition-dependent changes has been limited.
We explored the adaptive significance of annual changes in the age at sex-change of the protandrous (male-first) hermaphroditic shrimp and examined how selective harvesting affects life-history variation, by conducting field observations across 13 years and a controlled laboratory experiment. In addition, we considered whether plastic responses by the shrimp would be favourable, negligible or negative with respect to the conservation of fishery resources.
The age at sex-change and the population structure of the shrimp fluctuated between years during the study period. The results of the field observations and laboratory experiment both indicated that the shrimp could plastically change the timing of sex-change in accordance with the age structure of the population. These findings provide the first concrete evidence of adult sex ratio adjustment by pandalid shrimp, a group that has been treated as a model in the sex allocation theory.
The sex ratio adjustment by the shrimp did not always seem to be sufficient, however, as the supplement of females is restricted by their annual somatic growth rate. In addition, adjusted sex ratios are further skewed by the unintentional female-selectivity of fishing activity prior to the breeding season, indicating that the occurrence of males that have postponed sex-change causes sex ratio adjustment to become unfavourable.
We conclude that the plastic responses of harvested animals in selective fishing environments must be considered in efforts to conserve wild animal resources, because such responses can become maladaptive.
Mortality is a fundamental component that influences life-history trade-offs in nature and human harvesting such as hunting and fishing brings additional mortality to ecosystems. Harvesting is highly selective, however, and often causes unnatural change in the life-history traits of target species (e.g. Coltman et al. 2003; Edeline et al. 2007; Garel et al. 2007). Recent reviews caution that the effects of selective harvesting are relevant in efforts to sustain wild animal populations (Coltman 2008; Fenberg & Roy 2008; Allendorf & Hard 2009; Dunlop et al. 2009; Mysterud 2011). In fisheries, especially, it has been documented that size-selective harvesting causes phenotypic and genetic change in such traits as body size (Conover & Munch 2002; Carlson et al. 2007; Swain, Sinclair & Hanson 2007), reproductive scheduling (Olsen et al. 2004; Dieckmann & Heino 2007) and other parameters (Walsh et al. 2006; Biro & Post 2008). While such changes represent the ultimate response of harvested animals, it is also well known that the life history of target species plastically alters with harvesting (Beverton & Holt 1957; Rochet 1998). However, research on the adaptive significance of these types of condition-dependent changes has been limited. Although a plastic genotype may perhaps perform better than a less plastic one across varying conditions, evidence of adaptive phenotypic plasticity is still being explored, even in natural populations (Pigliucci 2005). Moreover, a model by Ernande, Dieckmann & Heino (2004) predicts that plasticity does not act as a buffer against the evolutionary pressure of the size-selective harvesting that might cause unfavourable genetic changes.
The selectivity of fishing activity affects not only body size but also other traits and ecological processes (Zhou et al. 2010). Adult sex ratio (the ratio of mature males to mature females) is often biased to one sex because of size-selective fishing (Rowe & Hutchings 2003), intentional sex-selective fishing (Abe 1992; Carver et al. 2005; Sato et al. 2007) and fishery stocking programmes (Kanaiwa & Harada 2008). As sex ratio directly affects mating behaviour and population dynamics (Emlen & Oring 1977; Shuster & Wade 2003), harvested animals would consequently suffer a burden in their need to respond to the artificially created skew in sex ratio. In particular, the sex ratio of sequential hermaphrodites (animals that change sex) is vulnerable as a result of size-selective fishing, because their sex-change is triggered by somatic growth and large individuals after changing sex are therefore targeted by the fishing (Sadovy 2001; Alonzo & Mangel 2004; Molloy et al. 2007; Sattar, Jorgensen & Fiksen 2008).
The pandalid shrimps (Decapoda, Pandalidae, Pandalus) are a group of protandrous (male-first) species that have been treated as a model to examine plastic life-history variations with respect to sex ratio fluctuation. By defining early or late sex-change as an alternative strategy of life history in a given population of Pandalus jordani Rathbun, Charnov, Gotshall & Robinson (1978) predicted that the optimal age (size) at sex-change is attained when the average fitness of males and females becomes equal. For example, when the proportion of females consisting of old (large) breeders is low among all breeders, earlier sex-change by young (small) breeders is facilitated as a conditional response, a so-called socially mediated sex-change (SSC), and sex ratio is adjusted as a result of this SSC process. Interestingly, the pandalid shrimp is the most exploited crustacean in cold waters (Gillett 2008), with most populations being primarily exposed to unintentional female-selective fishing due to the fact that females are larger than males (Bergström 2000). This implies that a sex ratio bias in selective fishing towards the first sex (i.e. male) may trigger SSC in populations of P. jordani. A similar phenomenon has also been observed in Pandalus borealis Krøyer, the most common commercial pandalid shrimp (Charnov 1981; Charnov & Anderson 1989).
Sex ratio adjustment by the pandalid shrimp has been successfully modelled as a part of the sex allocation theory that explains resource allocation to male and female functions (Charnov 1982). However, the model has not been tested in other animals, and SSC by pandalid shrimps still leaves some doubts due to lack of concrete empirical evidence. While SSC has been thoroughly studied in sequential hermaphrodites forming small mating groups comprising two to several individuals (Warner 1988; Munday, Buston & Warner 2006), it is not clear how flexible sex-change is in hermaphrodites such as the pandalid shrimp that mate at random within a population comprising over tens or hundreds of thousands of individuals. In addition, the SSC-like phenomenon in pandalid shrimps can be explained by accelerated sex-change resulting from growth enhancement after large individuals (i.e. old females) have been harvested (Hannah & Jones 1991; Koeller, Mohn & Etter 2000). In this explanation, sex-change is genetically regulated, with sex being determined by body size based on density-dependent processes, even though it appears to be adaptive sex ratio compensation. As a result, the adaptive significance of life-history variation in pandalid shrimps is still ambiguous, and hence, the evolutionary perspective has been ignored in the fishing resources management of pandalid shrimps.
While the effects of selective fishing on the mating systems of marine organisms have been cautioned in several landmark review articles (Rowe & Hutchings 2003; Hutchings & Rowe 2008; Allendorf & Hard 2009; Zhou et al. 2010), our knowledge of how target species plastically modify their life-history strategy to maximize fitness under unnatural mating conditions remains limited. Distinguishing between a density-dependent process and an adaptive process in plastic life-history changes would greatly improve our ability to conserve wild animal populations. For example, the former process could be mainly managed by quantitative biomass assessment, a standard approach in fisheries management, while the latter process would require additional management procedures that incorporate the mating system of target species.
In this study, we explored the adaptive significance of the annual changes in the age at sex-change of the pandalid shrimp Pandalus latirostris Rathbun (Fig. 1a) and examined how selective fishing activity affects life-history variation in this species, by conducting field observations across 13 years and a controlled laboratory experiment. Even though target species can respond plastically to environmental fluctuations, it may not always be adaptive under both natural and artificial environments, because the direction and intensity of natural and artificial selection processes are different. We therefore considered whether the plastic responses by this shrimp would be favourable, negligible or negative with respect to the conservation of fishery resources.
Materials and methods
As P. latirostris is a seagrass-dwelling species that does not have a planktonic larval period (Chiba, Goshima & Mizushima 2000), its populations are generally identified by the connectivity of seagrass areas. Analyses using microsatellite DNA makers clarified genetic differentiation between at least two populations of this species at the same latitude (Kawahara-Miki et al. 2011). Figure 1b provides a schematic of the life history of this shrimp species that inhabits Lake Notoro, a lagoon on the island of Hokkaido in northern Japan (Fig. 1c). Individuals of this species generally mature as males at age 1 and change sex to females at age 2. However, males that do not change sex at age 2, termed here as postponed-change males (PCM), are often observed in these populations. Moreover, males at age 0 and females at age 1, termed here as early-matured males (EMM) and early-changed females (ECF), respectively, also occur, although their proportions are relatively low (Chiba, Goshima & Mizushima 2000; Chiba & Goshima 2003). Such life-history variations are also common in other species of pandalid shrimps (Bergström 2000).
According to histological observations of P. latirostris (Onishi, Chiba & Goshima 2001), the sex of this shrimp is determined by the time the maturing season starts in early summer (late June) and does not change until the breeding season in autumn (from late August to late September) (Fig. 1b). During the breeding season, males of this shrimp patrol inside the seagrass areas and copulate with several unspecified females just after moulting (Chiba, Goshima & Shinomiya 2003). Females breed once a breeding season and carry eggs for about 8 months.
Pandalus latirostris is heavily exploited in Hokkaido, where landings of this shrimp have been recorded to gradually decline or become unstable in recent years (Hokkaido Research Organization 2011). In most fishing areas, P. latirostris is harvested after the completion of sex determination and before breeding (between early July and mid-August; Fig. 1b), with large individuals being selectively harvested by using cage traps. In the case of Lake Notoro (Fig. 1c), water salinity changed from brackish water to saltwater when bank revetments were built at the mouth of the lagoon in 1974, and extensive fishing of shrimp was initiated in 1978 (Fig. 1d). Figure 1d also shows the catch weight per cage trap, referred to as catch per unit of effort (CPUE), a statistic that is calculated from the catch weight, date, number of cages and number of fishing boats, which are statistics that have been recorded in shrimp fishing logbooks since 1983. Data on total annual catches and CPUE indicate that the abundance of shrimp increased after the change in water salinity, but numbers have been gradually declining over the past decade.
Shrimp were collected from Lake Notoro in late June or early July (early-maturing season) and mid-October (just after the breeding season) for 13 years from 1999 to 2011 (Fig. 1b). In both collection periods, five sampling stations were determined in seagrass areas with a bottom depth of about 2·0 m (Fig. 1c). After trawling a sled net (mouth opening 1·5 m × 0·5 m, length 1·8 m, mesh 3 mm) along the bottom for 100 m at a constant speed of about 17 m min−1 at each station, all the collected shrimp were divided into immature males, mature males and mature females based on the first and second pleopods (Fig. 1a) that have been previously confirmed to reflect gonad development (Onishi, Chiba & Goshima 2001), and they were then counted. The total number of collected individuals in each year was 272–5080 (2004, on average) when sampled during the maturing season and 1206–6712 (4029, on average) when sampled just after the breeding season.
Size-frequency distributions of the sexed shrimp based on 1-mm intervals in body length (BL, the length between the eye and the end of the telson; Fig. 1a) were constructed for each sampling season. When more than 100 shrimp of each sexual phase were collected at each sampling station, 100 individuals were randomly selected. Subsequently, the number of selected individuals in each size class was multiplied by the sampling rate that is the ratio of 100 to the total number of sampled individuals. A cohort separation procedure was used to estimate the age structure at each station, because crustaceans have no exact index of age. Here, we assumed that the observed size-frequency distribution consisted of mixed normal distributions, and we estimated the parameters of each normal distribution using a maximum-likelihood procedure based on Hasselblad (1966). We fitted models with different numbers of cohorts and then used the minimum Akaike's Information Criterion to determine the optimal model (Fig. 2). This model selection was performed using the solver routine of MS Excel 2003 for each model mixture. We then found that the smallest-BL cohorts in the maturing and breeding seasons were at age 1 and age 0, respectively, based on descriptions of their life histories (Chiba & Goshima 2003; Figs 1b and 2). It is possible that some cohorts consisting of age-3 groups in this procedure were not detected because of their small abundance (Chiba & Goshima 2003). Proving the identification of the age-3 group does not affect analyses of fluctuation in the relationship between age and sex, however, because all age-3 individuals become female in the maturing season (Chiba & Goshima 2003; Fig. 1b). Our previous laboratory observations showed that there was little difference in growth between immature and mature individuals at age 0 and between males or females at age 1 (Chiba, Goshima & Mizushima 2000; Chiba 2001). We therefore considered our cohort separation procedure to be reasonable for estimating the age groups of age 0, age 1 and age ≥ 2.
As these data collected at five sampling stations often deviated from normal distribution, the median was used to calculate annual proportions of the mature males (i.e. adult sex ratio), PCM, ECF, EMM and each age group in each collection period.
To examine the size and sex ratio of harvested shrimp, in the middle of the fishing season (late July or early August) from 1999 to 2011 except 2003 when fishing was banned, we randomly chose 100–197 harvested individuals collected by 10–15 cage traps and identified their sex after measuring their BL.
eld data analysis
The sex ratio of the shrimp population is determined by the abundance of females that bred as females in the previous season, virgin females that changed sex from male, and males that have newly matured or postponed sex-change. If the shrimp respond to the population structure in order to equalize the fitness of males and females (Charnov, Gotshall & Robinson 1978; Charnov & Hannah 2002), the occurrence of ECF and PCM can be expected to vary between years. For example, ECF will increase and PCM will decrease in a year that the proportion of age-1 group is high, because age-1 individuals are principal candidate males (Fig. 1b).
To examine the occurrence of ECF and PCM, the proportion of age-1 individuals and the sex ratio were calculated at the five sampling stations in each maturing season between 1999 and 2011. If the sex of the shrimp is determined genetically, the sex ratio in the maturing season well reflects the proportion of the age-1 group, because they are generally candidate males, while the others are candidate females (Fig. 1b). In other words, differences between the proportion of the age-1 group and the sex ratio denote the occurrence of ECF or PCM. To examine how the sex ratio differed from the proportions of the age-1 group, Spearman's rank correlation test was used to examine the relationship between the proportion of the age-1 group and the deviation of the sex ratio from the proportion of the age-1 group during the 13-year survey. Importantly, some individuals at age 0 show rapid growth and mature as males (i.e. EMM) by the breeding season (Fig. 1b). As the sampling gear does not collect individuals of age 0 during the maturing season (Chiba & Goshima 2003), by the breeding season the EMM might have caused a change in the sex ratio that was observed during the maturing season. The sex ratio was therefore corrected by adding the number of EMM collected during sampling in the breeding season for each year and then repeating the analysis of the relationship.
To detect factors affecting the occurrence of ECF and PCM using data sets collected during the maturing season (Fig. 1b), we established the generalized linear mixed model (GLMM) shown below.
where Pecfij and Ppcmij represent the proportion of ECF and PCM at station i in year j, respectively. αs and βs are model coefficients for the fixed effects. To confirm SSC by P. latirostris, we added Pf (the proportion of the age-1 group as first breeders) to the model, according to the model of Charnov, Gotshall & Robinson (1978) and Charnov & Hannah (2002). L1 and L2 are mean BLs in the age-1 group for Pecf and in the age-2 group for Ppcm, respectively. According to the size-advantage model (Warner 1975), body size strongly affects the timing of sex-change. Density (D) (the number of individuals per m2) and the interaction between L and D were also included in the model. Sampling year (Y) was included as a random effect in these models after confirming that there is no correlation of residuals between years, according to Manley, McDonald & Thomas (1993). Sampling station (St) is also treated as a random effect, because P. latirostris move between sampling stations. Binomial errors were considered in these models, because the response variables represented proportion data. In addition, we conducted the same analyses for the proportion of ECF and PCM using data collected during the breeding season (after the fishing season; Fig. 1b) to examine whether shrimp responded to mortality caused by the size- and sex-selective fishing. These analyses were performed using the package lme4 (Bates, Maechler & Bolker 2011) of freeware R 2.14.0 (R-Development-Core-Team 2011).
Although the estimation of age from cohort separation is a common method for crustaceans, it is difficult to completely remove errors in age estimates. We therefore conducted a controlled laboratory experiment to confirm the SSC of P. latirostris at the individual level. We hypothesized that SSC was facilitated by a relative size difference between breeders in a mating group, because age would not be a proximate cue for sex determination and because body size is a function of age.
On 19 October 2006, several hundred small (age 1) and large (age ≥ 2) individuals were collected from the five stations at Lake Notoro (Fig. 1c) and their sex was identified following the same procedure used in the field observations. To precisely classify the size of the collected shrimp, the length of the shrimp's carapace (from the base of the eye to the end of the carapace; Fig. 1a) was measured to within 0·01 mm using a digital calliper (CD-CX, Mitutoyo, Japan). The carapace length (CL) correlates closely with the BL (r2 = 0·99, n = 73). We chose 140 males (17·3–21·4 mm CL range) from the small size (age 1) group and 28 females (25·5–30·6 mm CL range) from the large size (age ≥ 2) group for the experiment. Here, small and large individuals denote small males and large females, respectively.
Two experimental mating groups were created to examine whether the size of small individuals at sex-change would vary with the mating group structure. We reared 10 small individuals in the presence of four large individuals as a mating group and 10 small individuals in the absence of large individuals as another group. This experiment was designed to compare the occurrence probability of PCM between the experimental mating groups. Although density (the total number of reared individuals in each tank) varied between the experimental treatments, our field observation demonstrated that density did not affect the occurrence of PCM (see 'Results').
This experimental rearing lasted for 8 months (from 30 October 2006 to 30 June 2007), covering the sex determination period of this species (Fig. 1b), to encourage as natural a response as possible. The shrimp of each mating group were reared in an opaque tank (130 L) during the experiment. Seven replications were set using different tanks for each group, and all tanks were independent of each other. All tanks contained refuges made with nylon nets and were supplied with air and ambient water at a constant rate (1 L min−1). The shrimp were exposed to a natural photoperiod and supplied with food (including mysids, small snails and commercial compounds) ad libitum each day. Large individuals were replaced only when rearing shrimp died. After the rearing period, the sex of the small individuals was examined both from the morphology of pleopods (Fig. 1a) and from the presence of eggs in their gonads.
Laboratory data analyses
As the occurrence of PCM was affected by both population structure and body size in the analysis of the field observation, we examined whether the same result could reappear in the laboratory experiment using the GLMM shown below.
where Skl is the sex of small individual k in mating group l of each experimental tank. γs are the model coefficients for fixed effects. G is the social structure of the mating group, that is, the experimental treatment where large individuals were present or absent, and L is the body size of the small individuals. We also added the effect of the tank (T) in the model as a random effect nested in the treatment effects. Binomial error was assumed in the model, because the sex of the reared individuals was binomial data. As the growth rate and mortality of reared individuals might have been affected by the density that differed between experimental treatments, the GLMM was also used to compare the body size and mortality of reared individuals between the experimental treatments. In these analyses, the body size of the small individuals or their mortality (i.e. dead or alive) at the end of the experiment represented a response variable. We used normal error and binomial error in the former and latter models, respectively. The GLMMs used in all laboratory data analyses were performed with the same package used in the analyses of eqns (1) and (2)
Population structure during the maturing season
Figure 3 shows parameters of the population structure of P. latirostris in the maturing season at Lake Notoro. In general, size-frequency distributions during this season were divided into two cohorts (Fig. 2a), and a third cohort was detected at two stations in 1999, 2010 and 2011 only, out of 65 stations in total across the 13-year survey. Although most cohorts estimated as age 1 were maturing males, some were maturing females, that is ECF (Fig. 1b). The mean proportion of ECF in the age-1 group for each year was 0·5%, with the proportion ranging from 0·0% to 11·0% (Fig. 3a). Most of the individuals belonging to the age- ≥ 2 group were maturing as females, but some of the group comprised males, that is PCM (Fig. 1b). The mean proportion of PCM in the age- ≥ 2 group for each year was 12·0%, with the proportion ranging from 0·0% to 32·6% (Fig. 3b). The cohorts estimated as age 3 were maturing females. Other parameters of the population, that is the proportion of age 1, body sizes at age 1 and age 2 and population density also fluctuated between years (Fig. 3c–f).
Age-0 groups generally consisted of immature males and were collected at the sampling in the breeding season (Fig. 2b). Although some of them became matured male, that is, early-maturing male (EMM; Fig. 1b), the mean proportion of EMM for each year was 0·3%, with the proportion ranging from 0·0% to 6·5%.
Adult sex ratio
The mean proportion of the age-1 group was 0·45 (45·1%), indicating that adult sex ratio (the proportion of mature males) had a female bias when it was assumed that all age-1 and all age- ≥ 2 individuals simply became males and females, respectively. However, the mean of the actual sex ratio was 0·53 (52·7%), indicating a male bias. The mean sex ratio corrected by the number of EMM remained 0·53 (53·0%). The mean sex ratio during the breeding season (after the fishing season) was 0·78 (78·0%), indicating that the sex ratio was further biased towards males after maturing.
The deviation of the sex ratio from the proportion of the age-1 group decreased as the proportion of the age-1 group increased (rs = −0·73, P = 0·01; Fig. 4), indicating a negative correlation between the proportion of the age-1 group and the supplement of PCM. This relationship did not change when we added the number of EMM (rs = −0·71, P = 0·01).
Factors governing sex determination
The proportion of ECF increased with an increase in the proportion of the age-1 group and body size at age 1 (Table 1). Moreover, significant interaction between size and population density was detected (Table 1). In contrast, the proportion of PCM decreased with an increase in the proportion of the age-1 group and body size at age ≥ 2 (Table 1). Population density did not affect the occurrence of PCM (Table 1). In the data set collected during the breeding season, however, neither the proportion of ECF nor that of PCM responded to the proportion of the age-1 group (Table 2).
Table 1. Results of the generalized linear mixed model (GLMM) analyses for the proportions of early-changed females (ECF) and postponed-change males (PCM) in the maturing season
Proportion of age 1
Body length at age 1 (L1)
L1 × D
Proportion of age 1
Body length at age 2 (L2)
L2 × D
Table 2. Results of the generalized linear mixed model (GLMM) analyses for the proportions of early-changed females (ECF) and postponed-change males (PCM) in the breeding season
Proportion of age 1
Body length at age 1 (L1)
L1 × D
Proportion of age 1
Body length at age 2 (L2)
L2 × D
In the laboratory experiment, although the probability of postponed sex-change decreased with body size increment in both treatments (body size, Z = −5·03, P = 0·00; Fig. 5), sex-change was constrained by the presence of large individuals (social structure of mating group, Z = 4·22, P = 0·00; Fig. 5). The mean body size (CL; Fig. 1a) of small individuals in the treatments with large individuals and without large individuals at the end of the experiment was 23·1 mm (±0·7 SD) and 23·4 mm (±0·5 SD), respectively. The mean survival rate of small individuals in the treatments with large individuals and without large individuals at the end of the rearing period was 84·3% (±0·3 SD) and 95·7% (±0·1 SD), respectively. There was no significant difference in body size (F1,123 = 2·3, P = 0·14) and survival rate (Z = 0·5, P = 0·63) between treatments. These results indicate that the sex determination of small individuals was affected by the social structure of the mating group.
Selectivity of shing
The mean BL of individuals harvested between 1999 and 2011 was 96·0 mm (±3·3 SD) of which most were relatively larger individuals in cohorts estimated to be of age ≥ 2 in each year. The mean sex ratio (the proportion of males) of harvested shrimp was 0·02 (2·2%), indicating that the sex of harvested shrimp was strongly biased towards females. The mean mortality of females between the maturing and breeding seasons during the observation period was 52·4%, although this was just 8·3% in 2003 when shrimp fishing was banned (Fig. 1d).
Sex ratio adjustment of Pandalus latirostris
Our analysis of the adult sex ratio indicates that individuals of P. latirostris can plastically change the timing of sex-change in accordance with the age structure by the time the maturing season starts. Differences in the sex ratio (0·53) and the proportion of the age-1 group (0·45) indicate that the sex ratio was not simply determined by age. In other words, the sex ratio was often attained based on the occurrence of early-change female (ECF) and PCM. In theory, the sex ratio of animals that exhibit sex-change is biased towards the first sex before changing sex (Charnov & Bull 1989), with the male-biased sex ratio (0·53) observed in our field data being similar to the typical sex ratio (0·57) in protandrous (male-first) sex-changing species (Allsop & West 2004). In addition, negative correlations between the proportion of the age-1 group and the supplement of PCM strongly suggest that the sex ratio of this shrimp population is determined by adjustment and is not a haphazard response.
Details concerning the process of sex ratio adjustment were demonstrated both in the field survey and in the laboratory experiment. ECF increased and PCM decreased as the proportion of the age-1 group increased in the field (Table 1), and this conditional response was closely confirmed in the laboratory, indicating that this shrimp conducts SSC by responding to the breeders age (size) structure. These findings provide the first concrete evidence of SSC in hermaphrodites that do not form any discrete mating groups, although SSC has been documented in species that form small mating groups, generally comprising two to several individuals (Warner 1988; Munday, Buston & Warner 2006). Our confirmation of SSC means that sex-change in this shrimp can be explained by adaptive sex ratio adjustment, which was first predicted by Charnov, Gotshall & Robinson (1978).
To fully understand the SSC of P. latirostris, however, it is important to keep in mind that body size is also related to the occurrences of ECF and PCM. ECF were relatively larger individuals in the age-1 group, whereas PCM were relatively smaller individuals in the age-2 group, suggesting that individuals of the age-1 group may not be able to grow to an adequate size as functional females, even though ECF are being adaptive each year. Yet, small individuals of the age-2 group are able to extend their male phase as PCM whenever this strategy is favourable, because they had already attained an adequate size as functional males in the previous breeding season. This would explain why the interaction between body size and density only affected the occurrence of ECF (Table 1). In fact, many empirical studies on pandalid shrimps mentioned that the annual growth rate varies the timing of sex-change (Hannah & Jones 1991; Koeller, Mohn & Etter 2000; Wieland 2004). Body size would be a strong restricting factor in sex-changing species, even though they can respond plastically to their environment (Baeza & Bauer 2004; Hamilton et al. 2007). We conclude that the sex ratio of a population of P. latirostris is adjusted by SSC, but it is not always perfect because becoming ECF can be restricted by annual somatic growth rates.
Effects of selective shing
The sex ratio was adjusted during the maturing season (0·53) and became further biased towards males during the breeding season (0·78), indicating that the mortality of females between the maturing and breeding seasons was much higher than that of males. In this instance, female-biased mortality is not natural but artificial. About 98% of the harvested shrimp were female in our observation, indicating that size-selective harvesting unintentionally causes a strongly male-biased sex ratio. In fact, there was a minimal decline in the proportion of females between the maturing and breeding seasons of 2003 when shrimp fishing was banned. Moreover, shrimp did not respond to the proportion of the age-1 group after the fishing season, indicating that their sex ratio adjustment by the time the maturing season starts is adversely affected by unnatural mortality, that is the unintentional female-selective fishing, just before breeding.
Plastic changes in life-history characteristics caused by fishing tend to be ignored by harvest fisheries, probably because it is generally believed that phenotypic plasticity is a more efficient response to a given environment (Pigliucci 2005). The response may not always be sufficient and/or adaptive, however, especially when the environment is unnaturally affected by human activities such as harvesting. In the case of P. latirostris, sex ratio adjustment reflects the expected adaptive response of existing theoretical models (Charnov, Gotshall & Robinson 1978; Charnov & Hannah 2002). However, the size restriction on becoming ECF causes insufficient sex ratio adjustment. Moreover, at the point of population growth, the occurrence of ECF does not completely compensate for the lack of large females due to female-selective fishing, because the fertility of small females would be lower than that of large females (Berkeley, Chapman & Sogard 2004; Palumbi 2004; Birkeland & Dayton 2005). In addition, the occurrence of PCM has more severe effects on population growth, because PCM become superfluous males as a result of the unpredictable mass mortality of females. The extension of male phases by SSC is not favourable even though this is an adaptive response under natural selection. A plastic response by harvested animals may therefore not be sufficient and/or adaptive in certain situations.
The effects of harvesting biased to one sex on population dynamics in aquatic animals, especially in invertebrates, receive less attention than they do in terrestrial animals such as mammals (Mysterud 2011). There is little doubt, however, that skewed sex ratios also affect population growth rates in aquatic ecosystems (Rowe & Hutchings 2003). In addition, we emphasize that the plastic responses of animals may become maladaptive in unnatural environments, an unfavourable situation that should be avoided through the application of appropriate resource management actions. In the case of P. latirostris, the skewed sex ratio could be eased by setting the fishing season prior to the period of sex determination. Moreover, flexible fishing management approaches based on the characteristics of populations would be needed in the future to achieve sustainability of fishing resources (Edwards & Plagányi 2011; Hamilton et al. 2011). For example, annual variation in shrimp age (size) structures and individual growth rates should be included in fishing plans for P. latirostris to mitigate the shortage of females. In conclusion, it is necessary to understand how harvesting affects the life history of target species and to estimate the impact of those activities on the conservation of wild animal populations.
Our manuscript has greatly benefitted from the comments and encouragement by E. L. Charnov and was further improved by the comments of an anonymous referee. We thank S. Sakazaki, K. Chida, T. Watanabe, K. Ishizaki, L. Aoki, Y. Shinomiya and S. Hashidzume for their cooperation with the field observations and facilities management. We also thank Y. Kogure and T. Yusa for their assistance with shrimp rearing. SC was financially supported by Kakenhi (21780184), ERTDF (E-1102), the Akiyama Life Science Foundation and the RIMI throughout the study period.