Cryptic temporal changes in stock composition explain the decline of a flounder (Platichthys spp.) assemblage

Abstract Unobserved diversity, such as undetected genetic structure or the presence of cryptic species, is of concern for the conservation and management of global biodiversity in the face of threatening anthropogenic processes. For instance, unobserved diversity can lead to overestimation of maximum sustainable yields and therefore to overharvesting of the more vulnerable stock components within unrecognized mixed‐stock fisheries. We used DNA from archival (otolith) samples to reconstruct the temporal (1976–2011) genetic makeup of two mixed‐stock flounder fisheries in the Åland Sea (AS) and the Gulf of Finland (GoF). Both fisheries have hitherto been managed as a single stock of European flounders (Platichthys flesus), but were recently revealed to target two closely related species: the pelagic‐spawning P. flesus and the newly described, demersal‐spawning P. solemdali. While the AS and GoF fisheries were assumed to consist exclusively of P. solemdali, P. flesus dominated the GoF flounder assemblage (87% of total) in 1983, had disappeared (0%) by 1993, and remained in low proportions (10%–11%) thereafter. In the AS, P. solemdali dominated throughout the sampling period (>70%), and P. flesus remained in very low proportions after 1983. The disappearance of P. flesus from the GoF coincides in time with a dramatic (~60%) decline in commercial landings and worsening environmental conditions in P. flesus’ northernmost spawning ground, the Eastern Gotland Basin, in the preceding 4–6 years. These results are compatible with the hypothesis that P. flesus in the GoF is a sink population relying on larval subsidies from southern spawning grounds and the cause of their disappearance is a cessation of larval supply. Our results highlight the importance of uncovering unobserved genetic diversity and studying spatiotemporal changes in the relative contribution of different stock components, as well as the underlying environmental causes, to manage marine resources in the age of rapid anthropogenic change.


| INTRODUC TI ON
Unobserved genetic diversity, whether in the form of within-species genetic structuring or of the presence of cryptic species (morphologically indistinguishable biological groups incapable of interbreeding), is of major importance for the conservation and management of global biodiversity in the age of rapid environmental changes (Bickford et al., 2007;Dirzo et al., 2014). Biodiversity assessment at the level of morphospecies (species that can be distinguished by quantifiable morphological characteristics) is likely to greatly underestimate species diversity across taxonomic groups and biogeographic regions (Pfenninger & Schwenk, 2007), and thus hamper the estimate of biodiversity loss due to environmental changes (Bálint et al., 2011). Unveiling hidden genetic structure resulting from local adaptation (Benestan et al., 2015;Bradbury et al., 2013;Lamichhaney et al., 2012;Momigliano et al., 2017a;Prunier, Laroche, & Bousquet, 2011) and cryptic speciation (Bradbury et al., 2014(Bradbury et al., , 2010Fennessy et al., 2016;Herbert et al., 2004;Momigliano, et al., 2017b) is important in the delineation of conservation and management units for exploited or threatened species (Funk et al., 2012), as well as for the identification of suitable protected areas that maximize biodiversity representativeness and complementarity (Cook, Page, & Hughes, 2008). Cryptic species complexes that consist of rare or threatened taxa are especially vulnerable, as each taxon can be even rarer or more threatened than the one nominal unit being considered and may respond differently to environmental change (Schönrogge et al., 2002). Unveiling cryptic species and populations also enables more accurate estimates of taxon-specific biological traits (e.g., Feckler et al., 2014) that determine resilience to environmental change and exploitation (Hilborn, Quinn, Schindler, & Rogers, 2003). The presence of cryptic taxa poses serious challenges to understanding the causes and effects of population change, since it can confound our understanding of past and present demographic changes, source-sink dynamics, population connectivity, and trophic interactions.
Unobserved genetic diversity has especially important implications for the management of exploited marine resources. Fishery management has been based on the assessment of stocks, usually defined as genetically homogeneous and demographically independent populations which were often unrealistically assumed to have clear and temporally stable geographic boundaries (Begg, Friedland, & Pearce, 1999). The past three decades of fishery research demonstrate that this is seldom the case: single-species fisheries may exploit multiple, demographically independent but morphologically indistinguishable stocks that at times co-occur in the same geographic location (Bonanomi et al., 2015;Campana, Chouinard, Hanson, & Fréchet, 1999;Jónsdóttir, Marteinsdottir, & Campana, 2007;Lindegren, Waldo, Nilsson, Svedäng, & Persson, 2013;Schuchert, Arkhipkin, & Koenig, 2010). A failure to recognize this unobserved diversity can lead to overestimated maximum sustainable yields, and therefore to overharvest of the more vulnerable stock components (Hutchinson, 2008;Sterner, 2007).
Examples of mixed-stock fisheries include the cod fisheries in the Gulf of St. Lawrence (Ruzzante, Taggart, Lang, & Cook, 2000) and the Kattegat (Lindegren et al., 2013), which were traditionally considered as single-stock fisheries but turned out to be mixtures of stocks that aggregate during feeding but segregate during spawning.
Differences between stock components may be behavioral (e.g., different spawning ecotypes) and/or genetic (locally adapted populations) instead of clearly morphological and are therefore likely to be overlooked by fishery managers and fishermen (Reiss, Hoarau, Dickey-Collas, & Wolff, 2009). Furthermore, cryptic species or unrecognized populations of the same species within a mixed-stock fishery may respond differently to fishing pressure and to natural and anthropogenic environmental change, leading to undetected spatiotemporal shifts in their relative contribution to local stocks (Bonanomi et al., 2015). Such processes may lead to dramatic fishery collapse, as demonstrated in a recent study on the West Greenland cod fishery (Bonanomi et al., 2015). Using diagnostic SNPs genotyped from archived samples, Bonanomi et al. (2015) reconstructed the spatiotemporal relative contribution of two biologically distinct (but morphologically indistinguishable) major stock components: the local and cold-adapted West Greenland offshore cod population and the Iceland offshore cod population, the contribution of the latter being correlated with higher sea surface temperature (SST). The authors revealed that the collapse of the West Greenland cod fishery in the 1970s was the result of the gradual disappearance of the local West Greenland offshore population (due to overfishing), and the subsequent decline of Iceland offshore cod due to a period of unfavorable (colder) SST.
The fact that separate stock components may respond very differently to environmental change and fishing pressure is of concern in the age of rapid climate change (Hoegh-Guldberg & Bruno, 2010), coastal eutrophication (Smith, Tilman, & Nekola, 1999), and overfishing (Pauly, Christensen, Dalsgaard, Froese, & Torres, 1998).
Spatiotemporal tracking of the contribution of different genetic populations to mixed-stock fisheries (Bonanomi et al., 2015;Dahle, Johansen, Westgaard, Aglen, & Glover, 2018;Ruzzante et al., 2000) may therefore play a pivotal role in future adaptive management aimed at avoiding further fishery collapse that may be only partially dependent on local fishing pressure and may be affected by environmental changes occurring in spawning areas hundreds of kilometers away from the fishing grounds.
Flounders (Platichthys spp.) in the Baltic Sea show two distinct reproductive strategies: offshore spawning of pelagic eggs and coastal spawning of demersal eggs (Nissling, Westin, & Hjerne, 2002;Solemdal, 1967). Pelagic spawning occurs exclusively in deep offshore basins of the southern and central Baltic Sea (viz. the Arkona Basin, the Bornholm Basin, and the Eastern Gotland Basin) where salinity is sufficiently high (>11 psu, see Figure 1a) for eggs to achieve neutral buoyancy (Nissling et al., 2002). Flounders with demersal eggs can spawn successfully in salinities as low as 6 psu (Nissling et al., 2002), in conditions routinely encountered in coastal waters of the Northern Baltic Proper (NBP) and the Gulf of Finland (GoF).
In a recent study, Momigliano et al. (2017b) demonstrated that Baltic Sea flounders with pelagic and demersal eggs are a pair of closely related species arising from two independent colonization events of the Baltic Sea from the same ancestral population following the end of the last glaciation. The Baltic Sea flounders with demersal eggs have since been officially described as a new species: the Baltic flounder Platichthys solemdali (Momigliano, Denys, Jokinen, & Merilä, 2018). Based on studies from the past decade, European (P. flesus) and Baltic (P. solemdali) flounders are considered parapatric, with both species co-occurring in some areas of the central Baltic Sea. For example, the species meet around Gotland, which is assumed to be the northern limit of P. flesus' distribution (Florin & Höglund, 2008;Hinrichsen et al., 2017;Nissling et al., 2002), and at the southwestern entrance of the Gulf of Riga, where P. flesus and P. solemdali co-occur in similar proportions (Momigliano et al., 2017b).  Bendtsen, Söderkvist, Dahl, Hansen, and Reker (2007). Areas colored in red, yellow, and green are theoretically suitable spawning habitats for P. flesus. The three major spawning grounds of P. flesus in the Baltic Sea are indicated by black arrows. Red filled squares represent sampling areas in the AS (SD 29) and the GoF (SD 32). The large red rectangle identifies the main study area, including the EGB, the AS, and the GoF. available (1980)(1981)(1982)(1983)(1984), catch estimates were above 1,500 t/y and started to decline in the second half of the 1980s. The most dramatic decline in landings took place in the GoF: In the early 1980s, more than 1,000 t/y were landed, whereas in the last ten years, landings averaged below 100 t/y (ICES, 2017). While there are some doubts on how reliable catch estimates in the 1980s were, particularly from the USSR, catch per unit effort (CPUE) data from fish surveys on the Finnish coast confirm a decrease of over 60% from 197560% from to 199560% from , and over 90% decrease from 197560% from to 201260% from (Jokinen et al., 2015, suggesting that the temporal decline in landings was not entirely a reflection of inaccurate reporting or lower fishing effort. Momigliano et al. (2018Momigliano et al. ( , 2017b) discovered that P. flesus can, in fact, be found at low density throughout the NBP (including the AS) and the GoF, and that they do not interbreed with P. solemdali (both studies found no evidence of hybridization in >300 individuals sampled throughout the Baltic Sea). The co-occurrence of both species has been likely overlooked because P. flesus and P. solemdali can only be distinguished based on genetic data, gamete morphology, and physiological differences that cannot be routinely assessed by fishers and managers (Momigliano et al., 2018). One explanation for the occurrence of P. flesus in the AS and GoF, where there are no suitable spawning grounds (Figure 1a), is that these locations are "sinks" (Pulliam, 1988;Underwood & Fairweather, 1989), sourced through either larval subsidies or adult spillover from spawning grounds in the central Baltic Sea-as is the case for cods in the Gulfs of Finland and Riga (Aro, 1989;Casini et al., 2012). Here, we used genetic data from archival (otolith) samples to reconstruct the temporal genetic makeup of a mixed flounder stock in the Åland Sea (AS) and the Gulf of Finland (GoF), which has hitherto been managed as a single stock of the European flounders (P. flesus) (ICES, 2017), but was recently revealed to consist of two closely related species (Momigliano et al., 2018(Momigliano et al., , 2017b. Our main aim was to test for possible temporal changes in stock composition in coastal areas of the Northern Baltic Sea. In particular, we explored the hypothesis that P. flesus from the EGB are a source population seeding the GoF, and that past fluctuations in stock size were influenced by environmental changes in geographically distant source populations. Specifically, we wanted to know how the relative proportions of P. flesus and P. solemdali have changed over time in the study region, and how these changes relate to RV in the EGB. To this end, we reconstructed the relative proportion of P. flesus and P. solemdali over the past four decades in two locations, one likely to receive larval subsidies from the EGB (GoF), and another that is less likely to receive larval subsidies (AS) given the prevailing water current patterns in the Baltic (Maslowski & Walczowski, 2002). We hypothesized that if larval subsidies of P. flesus from the central Baltic are an important factor in the demographics of flounders in the GoF: (a) the proportion of P. flesus in the GoF, but not in the AS, should be correlated

| The otolith collection
To determine the relative contribution of P. solemdali and P. flesus to local populations, we used selected samples from a collection of more In 1999-2011, the sampling scheme was changed by an internationally coordinated sampling program in order to standardize sampling between countries and areas. As before, the whole daily flounder catch/fisherman was recorded, and details of gillnets and their properties were recorded. Sample size varied between 10 and 50 kg per fisherman. A normal length-based stratified sampling strategy was used for otoliths. The whole catch was weighted and specimens measured according to 1 cm classes. For age determination, the aim was to collect 10 otoliths/each cm length class per year quarter (quarters 3 and 4).
Whole sagittal otoliths were used for age determination. All

| Sampling areas
Two areas, representing the Åland Sea (the western Åland Archipelago; Finnish National Square 49) and the Gulf of Finland (the Helsinki area; Finnish National Squares 54/53), were selected for the study (henceforth referred to as "AS" and "GoF," respectively;

| Time points and cohorts
For both areas, four time points were sampled (Table 1)

| Assignment tests
We employed a Bayesian approach where the probability of a fish to be assigned to one species is calculated as a function of its genotype and prior belief that the sample belongs to a given species; this probability is then updated as more genetic tests are performed in succession. The very same Bayesian approach has been used by Toli, Calboli, Shikano, and Merilä (2016) for sex identification after multiple consecutive genetic tests, and the full conceptual framework and the mathematical formulation of this approach are detailed in that paper. Because every sample can be drawn from two possible species, calculating the probability that a sample has been drawn from one immediately provides also the probability that the sample has been drawn from the other, because p (P. solemdali) = 1 -p (P. flesus).
Briefly, we determined the allelic and genotypic frequencies at five loci for P. flesus and P. solemdali based on 235 samples (128 P. solemdali and 107 P. flesus) whose species identities had been previously determined with certainty (see Momigliano et al., 2017b; data available from Momigliano et al., 2017c); this information provided us with the specificity and sensitivity of each genotype for attributing each sample to one (or the other) species using the approach described by Toli et al. (2016). We chose a conservative (uninformative) prior of p = 0.5, and we used the genotype of the sample at a first locus and its relative frequency in the two species to calculate a posterior probability that the sample was drawn from P. solemdali.
This posterior probability was then used as the prior for the next test, based on the sample's genotype at the next marker and its relative frequency in the two species. The result of this calculation was then an updated posterior probability that the sample belongs to P. solemdali. This iterative process was carried for all subsequent genetic tests, allowing us to update and refine the posterior probability. Samples whose final posterior probability of being P. solemdali was 1 (or close to one) can be considered as the demersal-spawning species with a high degree of confidence, and samples whose final posterior probability of being P. solemdali was 0 (or close to 0) could be confidently identified as P. flesus.

| Environmental drivers
To test whether environmental conditions in the EGB could ex- All statistical analyses were performed using the R computing environment (R Core Team, 2017).

| RE SULTS
A total of 444 individuals were successfully genotyped, out of which 433 were identified to species level with more than 99% probability and 439 with more than 95% probability ( F I G U R E 2 (a) Bayesian assignment test based on five outlier SNPs. Each bar represents an individual. Y-axis represents the assignment probability to the demersal-spawning P. solemdali (red) and pelagic-spawning P. flesus (blue). On the x-axes are given the sampling location, sampling year, and the birth year (sampling year minus fish age estimated from otolith). The map shows sampling locations (AS and GoF) and the potential "source" population (EGB), arrows show dominant current patters (Maslowski & Walczowski, 2002). (b-e) Relationship between reproductive volume in the EGB (Ustups et al., 2013) and proportion of P. flesus. There is no linear (b) or logistic (c) relationship in the AS, while in the GoF (d), reproductive volume explained 57% of the variation in the proportion of P. flesus. (e) the same data analyzed using a logistic regression to determine whether there is a "threshold" in reproductive volume after which most of the individuals are likely to be the pelagic-spawning species  1970 1971 1977 1 978 1986 1987 1997 1998 1998 1973 1975 19771978 1979 1986 1987 1997 1998 2006 2007 GoF A previous study demonstrated that larval abundance of flounders in the EGB is correlated with both spawning stock biomass (SSB) and RV (Ustups et al., 2013). The authors found no correlation between larval supply and recruitment in the EGB, suggesting that recruitment was regulated at the post-settlement stage (Ustups et al., 2013). However, only a small proportion of the larvae released in the EGB would eventually be transported to the GoF, and hence, it would not be surprising if recruitment of P. flesus in the GoF was supply-limited. Recruitment in flatfish is usually a reflection of density-independent factors affecting egg and larvae at the local scale (Leggett & Frank, 1997), and shows higher variation close to the species' distribution margin. Density-dependent processes in the phase immediately following settlement, such as competition for space and resources in overcrowded nurseries, may also play an important role in regulating recruitment and dampening inter-annual variability (Beverton, 1995). However, when larval supply is low, density-dependent mortality within nurseries becomes less important and recruitment is mostly a result of density-independent mortality during the ics (Caley et al., 1996;Doherty, 1982;Underwood & Fairweather, 1989). Changes in environmental conditions in spawning grounds for P. flesus could affect larval supply to the GoF, but would not affect local populations of P. solemdali, and therefore provide a likely explanation for the patterns observed in this study. However, environmental conditions in the EGB are only one of the factors that are likely to shape recruitment of P. flesus to the GoF. Fluctuations in SSB in the EGB and changes in deep-water currents at the time of spawning are also expected to affect, respectively, the number of larvae produced and the proportion of larvae that would reach the GoF. Unfortunately, SSB estimates for the EGB are only available from some of the years from which we have sampled cohorts (Hinrichsen et al., 2017;Orio et al., 2017;Ustups et al., 2013).
Therefore, we were unable to explore the possible relationship between SSB and the decline of the proportion of P. flesus in the GoF. Similarly, it is possible that deep-water currents have changed through time, affecting larval transport to the GoF, but we do not have access to data on water velocity preceding 1993. Hence, the readers should be aware of the fact that these potentially important factors may also have changed over time and be correlated with the RV in the EGB. Therefore, while the hypothesis put forward in this study is plausible and supported by the available data, further work will be needed to reach a definitive conclusion on the causes of the cessation of P. flesus larval subsidies to the GoF.
The timing of the shift from P. flesus-dominated to P. solemdalidominated assemblages in the GoF closely mirrors the rise and fall in cod biomass in the Gulf of Riga (GoR) (Casini et al., 2012). As the GoF for the P. flesus, the GoR is a "true sink" habitat for cod as there are no suitable spawning grounds. Cod biomass in the GoR increased in the decade 1977-1987 as a result of a combination of larval supply from the Baltic proper as well as by active migration of juveniles and adults, but as soon as environmental conditions in the Baltic Proper became unfavorable, cod almost entirely disappeared from the GoR (Casini et al., 2012). A similar pattern was also seen in the GoF cod (e.g., Aro, 1989). Cod eggs require higher salinity than P. flesus' eggs (>13 vs. This scenario bears resemblance to cases of fishery collapse caused by the successive disappearance of previously unobserved, demographically independent stock components, such as the collapse of the West Greenland cod fishery (Bonanomi et al., 2015). It is not entirely clear whether the low proportion of P. flesus and the low commercial landings and CPUE recorded in the past three decades in the GoF are symptoms of a transient phase of unfavorable environmental conditions or are entirely a result of anthropogenic habitat degradation. Nevertheless, these results raise concern over the future of P. flesus and P. solemdali in the GoF.

| CON CLUS ION
What had been previously considered as a single-stock fishery is in reality composed of two distinct species of flounders, each of which can dominate local assemblages at different points in time. The use of historical samples allowed us to discover a hidden turnover of flounder assemblages and to demonstrate this turnover coincided temporarily both with the decline in flounder stock and with the decline of environmental conditions in the northernmost spawning grounds of P. flesus. In the GoF, P. flesus has almost completely disappeared before scientists and managers even noticed their presence. The decrease in flounder abundance in the GoF does not reflect the gradual decline of a single population, but rather the disappearance of P. flesus and the successive decline of P. solemdali, which in turn are determined by the degradation of environmental conditions at local and regional scales.
Our results highlight the importance of studying spatiotemporal changes in the relative contribution of different stocks to mixed-stock fisheries, and the underlying environmental causes, to manage marine resources in the age of rapid anthropogenic change. Of particular concern is the fact that the decline of P. flesus in the northern Baltic Sea could have led to stronger fishing pressure on P. solemdali, which itself is demonstrated to have suffered from anthropogenic environmental changes (Jokinen et al., 2015(Jokinen et al., , 2016. Furthermore, as the conditions for pelagic reproduction will likely continue to deteriorate as a result of ongoing environmental changes (Vuorinen et al., 2015), there are also concerns over possible local extinctions of P. flesus within the Baltic Sea (Momigliano, et al., 2017b). Altogether, our results call for an immediate re-assessment of the conservation status of the two flounder species in the Baltic Sea by the International Union for the Conservation of Nature. The genetic test developed by Momigliano et al. (2018) and further refined in this study provides the means for monitoring both flounder species independently where they co-occur, and could be employed for real-time mixed-stock analyses of the catch (Dahle et al., 2018). This will enable to estimate demographic changes, resilience to climate change and exploitation, and responses to management for each species separately, creating the bases for the effective adaptive management of P. flesus and P. solemdali in the Baltic Sea.

ACK N OWLED G EM ENTS
We thank Jacqueline DeFaveri, for discussions and useful comments on earlier versions of this manuscript.

CO N FLI C T O F I NTE R E S T
None declared.