Vol. 535: 129–144, 2015 MARINE ECOLOGY PROGRESS SERIES Published September 15 doi: 10.3354/meps11423 Mar Ecol Prog Ser

OPENPEN ACCESSCCESS

Population ecology of atlantica (, Siphonophora) in the Western English Channel

Michael Blackett1,3,*, Cathy H. Lucas1, Rachel A. Harmer2, Priscilla Licandro3

1Ocean and Earth Science, National Oceanography Centre, University of Southampton Waterfront Campus, European Way, Southampton, SO14 3ZH, UK 2Plymouth Marine Laboratory, Prospect Place, The Hoe, Plymouth, PL1 3DH, UK 3Sir Alister Hardy Foundation for Ocean Science, The Laboratory, Citadel Hill, Plymouth, PL1 2PB, UK

ABSTRACT: Recent observations suggest that the siphonophore is expanding its geographical distribution. The mechanisms behind this expansion remain unclear due to our limited knowledge of the ’ ecology. We modelled the functional relationship between the 2 main life-cycle stages of M. atlantica over a 5 yr period (2009−2013) in the Western English Channel. Our aims were to determine the key features of the species’ population dynamics and the influence of local environmental conditions on its population development. Our results high- lighted a strong coupling between the timing of specific environmental conditions and the devel- opment of the M. atlantica population, thereby explaining interannual differences in the phenol- ogy of its blooms. Population development commenced with the initiation of eudoxid production by the overwintering polygastric stages. This reproductive event was linked to the onset of a spring temperature threshold, suggesting a critical basal limit of 10°C for eudoxid production. Interannual variability in the timing of this threshold modulated the degree of mismatch between the developing M. atlantica population and the availability of prey. Unusually cold con- ditions in the spring of 2010 and 2013 limited the capacity for M. atlantica to initiate eudoxid pro- duction leading to poor trophic phasing and the production of single autumn cohorts. In contrast, warmer conditions during spring 2009, 2011, and 2012 facilitated earlier population development, optimal trophic phasing and the production of both summer and autumn cohorts. These findings represent an important addition to our understanding of the ecology of M. atlantica in the North- east Atlantic.

KEY WORDS: Muggiaea atlantica · Jellyfish · Population dynamics · Population ecology · Phenology · Phenological shift · Trophic mismatch

INTRODUCTION opportunistic life history traits (high reproductive potential and high feeding rates) that enable rapid Cnidarian jellyfish (medusae and siphonophores) population increase in response to favourable envi- are a ubiquitous component of the pelagic ecosys- ronmental conditions (Lucas & Dawson 2014). The tem. These play a key ecological role as seasonal abundance of these species fluctuates dra- predators and competitors that can modulate the matically, with dense accumulation separated by structure and dynamics of marine communities (Mills periods of absence or rarity (Lucas & Dawson 2014). 1995, Pitt et al. 2009). Most cnidarian jellyfish have Consequently, their structuring effect on the pelagic

© The authors 2015. Open Access under Creative Commons by *Corresponding author: [email protected] Attribution Licence. Use, distribution and reproduction are un - restricted. Authors and original publication must be credited. Publisher: Inter-Research · www.int-res.com 130 Mar Ecol Prog Ser 535: 129–144, 2015

community is complex and dynamic (e.g. Matsakis & 2012, Batistić et al. 2013), probably in re sponse to Conover 1991). Knowledge of the mechanisms that hydroclimatic changes. Large blooms of M. atlantica regulate gelatinous predator populations is funda- have caused dramatic ecological (Greve 1994, Kršinić mental to our understanding of the functioning of & Njire 2001) and economic (Fosså et al. 2003, Baxter marine ecosystems (Condon et al. 2013). et al. 2011) ramifications, particularly in novel habi- The population dynamics of cnidarian jellyfish are tats. Knowledge of the population dynamics and the result of life cycle and life history adjustments environmental requirements of this species is needed (sensu Boero et al. 2008), modulated by myriad envi- to fully appreciate its role in coastal marine ecosys- ronmental factors. Most cnidarian jellyfish have com- tems. However, our understanding is hampered by a plex, metagenic life cycles alternating between sex- paucity of data on its different life stages. ual and asexual generations. Meroplanktonic species The present study is based on observations of the typically employ life cycle adjustments (qualitative polygastric and eudoxid life stages of M. atlantica changes at the individual level) and may persist and the congeneric Muggiaea kochi in the Western locally as sexual pelagic or asexual benthic life English Channel between 2009 and 2013. We inves- stages at different times (Boero et al. 2008). In con- tigated seasonal changes in the abundance of these trast, holoplanktonic species tend to adopt life history key life-cycle stages to better understand the dynam- adjustments (quantitative changes at the population ics that characterise their population development. level) and may undergo alternating periods of high Identifying the main biological and physical factors and low absolute abundance (Boero et al. 2008). that influenced population development allowed us Improving our knowledge of the inter-relationships to explore environmental cues that influence their between different cnidarian jellyfish life-cycle stages phenology in the Western English Channel. and the influence of environmental factors would enable a better understanding of their population dynamics (Lucas 2001, Lucas & Dawson 2014). MATERIALS AND METHODS Calycophoran siphonophores are holoplanktonic colonial cnidarian jellyfish (Mackie et al. 1987). Sampling station These typically small and active species are ambush predators that release a network of tentacles to Data analysed in the present study were collected entrap prey (Mackie & Boag 1963). The as a part of ongoing zooplankton research conducted life cycle of the Calycophora is characterised by an by the Western Channel Observatory (www.western alternation between an asexual polygastric stage and channelobservatory.org.uk). Data were collected a sexual eudoxid stage. Polygastric colonies asexu- weekly at a coastal station, L4 (50° 15’ N, 4° 13’ W) ally produce a chain of cormidia (each cormidium located 7.5 nautical miles (~13.9 km) southwest of comprising single nutritive, reproductive and buoy- Plymouth, UK, in proximity to the 50-m isobath. Stn ant zooids) along a central stem. As the chain grows L4 is characterised by high seasonal variability. successive cormidia are released as autonomous Intense primary productivity occurs during the sexual eudoxid colonies. The eudoxid stage buds spring and autumn (Widdicombe et al. 2010), with successive reproductive gonophores that, via exter- the summer period characterised by stratified and nal fertilisation, recruit new polygastric colonies nutrient-depleted waters (Smyth et al. 2010). Pat- through 2 short-lived larval stages, the planula and terns of water circulation are typically characterised calyconula (Carré & Carré 1991). Simultaneous asex- by an eastward coastal trajectory, delivering waters ual and sexual reproduction, high fecundity, rapid of a more northerly or southerly origin, dependent growth rates and short generation times enable the upon the prevailing wind patterns (Pingree & Grif- rapid development of large multivoltine blooms fiths 1980). Strong tidal influence (Pingree 1980) and (Mackie et al. 1987). periodic fluvial input from the Tamar estuary (Smyth In temperate biomes of the 3 great oceans the caly- et al. 2010) are also dominant hydrographic features. cophoran siphonophore Muggiaea atlantica repre- sents a major component of neritic gelatinous zoo- plankton (Mapstone 2009, 2014). Recent observa- Physical data tions suggest that M. atlantica has expanded its geo- graphical distribution, establishing new populations A SeaBird SBE 19plus Profiler was used to obtain in the Western English Channel (Blackett et al. 2014) depth profiles of a range of physical and chemical and the Mediterranean and Adriatic (Licandro et al. environmental parameters (temperature, salinity, Blackett et al.: Population ecology of Muggiaea atlantica 131

water column stratification and dissolved oxygen) Numerical analyses (see Smyth et al. 2010 for further methodological details). Temperature (°C) and salinity data were Data preparation integrated be tween the surface and 50 m depth.

Water column stratification was assessed in terms Tostabilisevariance,biologicaldatawerelog10(x+1) of mixed-layer depth (MLD), estimated as the transformed. For both physical and biological vari- depth at which water density (kg m−3) exceeded a ables, randomly spaced missing data (~12% of the near-surface (10 m) reference value by a set thresh- total 260 weekly observations) were interpolated old (0.003 kg m−3) (de Boyer Montégut et al. 2004). using the eigenvector filtering with missing data Dissolved oxygen data were not included in the method (Ibañez & Conversi 2002). All analyses were analysis because they were highly correlated with programmed using MATLAB (R2014b 8.4.0.150421). temperature (r = −0.86).

Eudoxid stage identity Muggiaea and other zooplankton data Since the eudoxid stage is produced directly (asex- Zooplankton samples were collected in duplicate ually) by the polygastric stage (Carré & Carré 1991), by vertical hauls from the sea floor to surface using a and eudoxid production rates are linearly related to WP2 net (mesh size = 200 µm, mouth area = 0.25 m−2) the size of the polygastric colony (Purcell 1982), a (UNESCO 1968). To reduce inter-sample variability, tight coupling of their abundance could be ex pec ted. duplicate samples were averaged and data repre- Indeed, previous observations have shown that the sented as the number of individuals or colonies per sexual and asexual stages of both M. atlantica and M. cubic metre (see Eloire et al. [2010] for a detailed kochi typically covary in phase (Dowidar 1992). Con- description of the sampling methodology). sequently, the relative contribution of M. atlantica Siphonophore colonies are fragile and are rarely and M. kochi to the total abundance of Muggiaea collected intact by nets. However, the number of spp. eudoxid stages was estimated using simple lin- Mug giaea atlantica and Muggiaea kochi necto- ear regression. Muggiaea spp. eudoxid stage abun- phores provided a direct estimate of their polygastric dance (response variable) was regressed on M. abundance as species of the genus Muggiaea de - atlantica and then M. kochi polygastric abundance velop only a single nectophore (Mackie et al. 1987). (explanatory variables) to quantify the strength of Enumeration of the Muggiaea spp. eudoxid stage each species association with the total abundance of was based upon the number of detached bracts and Muggiaea spp. eudoxid stages. The regression mo - intact colonies. The eudoxid stages of the genus del assumptions of linearity, homogeneity of vari- Muggiaea are morphologically indistinguishable ance, normality and independence of residuals were (Kirkpatrick & Pugh 1984) and were represented by carefully monitored. Standard Newey-West hetero- their total abundance. scedasticity and autocorrelation consistent covari- Specific zooplankton taxa were selected as poten- ance estimators (HAC) were used to provide robust tial sources of prey using size-range (100−1000 µm) estimates of regression parameter variance and facil- (Purcell 1981) and prey-suitability (Purcell 1982, itate sound statistical inference (Andrews & Mona- Mapstone 2009) criteria (Table 1). han 1992).

Table 1. Percentage relative contribution of different taxa to the total potential prey availability for M. atlantica at the sampling location. Taxa were selected following size range and prey suitability criteria derived from the literature (Purcell 1981, 1982, Mapstone 2009)

Season Prey taxa Calanoida Calanoida Cyclopoida Harpacta- Poecilo- Copepoda Bivalvia Cirripedia (CVI) (CI−CV) coida stomatoida (napulii)

Winter 22.0 17.1 13.4 0.6 17.4 1.7 0.9 2.6 Spring 20.1 20.0 10.3 0.8 2.8 0.7 0.1 16.1 Summer 22.4 24.5 9.8 0.5 2.8 3.1 0.6 0.7 Autumn 25.9 20.4 3.8 1.6 14.9 1.3 1.2 0.5 132 Mar Ecol Prog Ser 535: 129–144, 2015

Seasonal variability for autocorrelation following the modified Chelton method proposed by Pyper & Peterman (1998). All The dominant modes of temporal variability were data were standardised at zero mean and unit devia- extracted for the biological and physical variables tion (z-scored) for correlation analyses. using the eigenvector filtering (EVF) method (Cole- brook 1978, Ibañez & Etienne 1992). The EVF me- thod involves a principal component analysis per- Critical environmental ranges formed on an autocovariance matrix composed of the original series lagged progressively; the number of For environmental factors that were significantly lags (8−11 wk) was defined as the time lag at which correlated with the PC2 of M. atlantica and residuals the autocorrelation function of the original series first of the expected abundance of eudoxids, we identi- passed zero (Ibañez & Etienne 1992). The EVF tech- fied critical environmental ranges using quotient plot nique decomposes a time series into successive sig- analysis (van der Lingen et al. 2001). Quotient plot nals of decreasing variance. A cumulative variance analysis quantifies the association of species abun- threshold of 80% was used to determine the number dance with categories of environmental parameters, of principal components (PC) retained (Ibañez & providing an estimate of positive and negative asso- Conversi 2002). The main periodicities that charac- ciations. terised the PCs were then estimated using the auto- Quotient values (Qi) were computed as: correlation function (Legendre & Legendre 2012). M i Qi = (1) Ei

Population dynamics where Mi is the percentage of the total abundance of M. atlantica in environmental category i and Ei is the The functional relationship between the abun- percentage frequency of occurrence of environmen- dance of eudoxid and polygastric stages was investi- tal category i. Qi = 1 indicate a random association of gated using simple linear regression. Following the species abundance and related environmental cate- procedure described for eudoxid stage identity, re - gories, whereas Qi > 1 indicate a positive association gression analyses were computed between the PCs and Qi < 1 indicate a negative association. The statis- of the polygastric (explanatory variable) and eudoxid tical significance of associations was evaluated using (response variable) life-cycle stages and the studen- a randomisation test with 10 000 permutations to tised residuals were retained for analyses. The resid- compute CI with α = 0.05. For this analysis the nega- uals represent deviations from the typical linear rela- tive and positive residuals were analysed separately. tionship between the abundance of eudoxid and This procedure involved first inverting the negative polygastric stages. Positive residuals indicate that in- residuals, then re-scaling both the positive and ne - creased abundance of the eudoxid stage was ob- gative residual between 0 and 1. This procedure served relative to that which would be expected from enabled the association of both positive and negative the linear eudoxid−polygastric relationship, while residuals with environmental categories to be asses- negative residuals indicate the inverse. We consid- sed. The intervals of environmental categories were ered residuals greater than ±1 SD as significant devi- chosen by a compromise between the resolution of ations from the expected abundance of the eudoxid the profiles and the number of data points within stage. each category. To ensure representation of data within each category, copepod abundance scores in excess of the 95th percentile were omitted from this Correlations with environmental variables stage of the analysis (< 7% of the 260 observations).

Hypothetical links between variability of M. atlantica and physical environmental factors (as the RESULTS PCs) were evaluated using Pearson product moment correlation analyses. Links with biological variables Eudoxid stage identity (as the PCs) were explored using partial correlation analyses, allowing the effects of controlling physical Inspection of the raw data showed that the abun- variables to be removed. For all correlation analyses, dance of the polygastric stage of M. atlantica and the effective degrees of freedom (dfa) were corrected the eudoxid stage of Muggiaea spp. fluctuated in Blackett et al.: Population ecology of Muggiaea atlantica 133

2009 2010 2011 2012 2013

3.0 A B C D E +1]) x [ 10 2.0 (log –3 1.0

0.0 Colonies m 0 25 50 0 25 50 0 25 50 0 25 50 0 25 50 Week

Fig. 1. Temporal fluctuation in the abundances of Muggiaea atlantica polygastric stage (solid black line), Muggiaea kochi polygastric stage (dash-dotted grey line), and Muggiaea spp. eudoxid stage (solid grey line) at Stn L4 in the Western English Channel between 2009 and 2013. Data are weekly abundance values that have been smoothed with a 3-week moving average to improve legibility

phase (Fig. 1). Polygastric M. kochi were less abun- Seasonal variability dant and did not vary in the same way (Fig. 1, Table 2). Linear regression analysis confirmed a Eigenvector filtering (EVF) decomposed the poly- significant relationship between the raw abundance gastric and eudoxid stage M. atlantica time series into (decimal logarithm) of the polygastric stage of M. 2 principal components: PC1 and PC2 (Fig. 2). PC1 atlantica and the Muggiaea spp. eudoxid stage (β = accounted for 58% and 62% of the respective total 1.15, r2 = 0.80, F = 989, df = 258, p < 0.001). The polygastric and eudoxid stage variability, while PC2 number of M. kochi polygastric colonies was not explained a further 25% and 24%. These components associated with eudoxid stage variability (β = 0.27, represented harmonics of the seasonal cycle, with r2 = 0.003, F = 1.4, df = 258, p > 0.05) nor the resid- perio dicities of 54 wk (PC1) (Fig. 3A) and 22 wk (PC2) uals from the M. atlantica-eudoxid model (β = 0.33, (Fig. 3B). PC1 represented broad-scale changes in r2 = 0.03, F = 10.4, df = 258, p > 0.05). These results the absolute magnitude and distribution of sea sonal provided a clear indication that the eudoxid popu- abundance (Fig. 2A−E), while PC2 represented the lation was composed primarily of M. atlantica. underlying fine-scale dynamics that drove this vari- Whilst the possibility of a small contribution from ability (Fig. 2F−J). M. kochi cannot be completely eliminated, it was Inspection of the broad-scale seasonal variability considered negligible. Total eudoxid abundance (PC1) revealed that in 2010 and 2013 polygastric and was considered to represent M. atlantica eudoxid eudoxid stage M. atlantica exhibited restricted peri- abundance and M. kochi was not included in fur- ods of abundance that occurred relatively late in the ther analyses. year, during the autumn weeks (Fig. 2B,E). This pat- tern was in contrast to the other 3 years analysed (2009, 2011 and 2012), −3 Table 2. Mean ± SD and 95th percentile abundance (colonies m ) of Mug- during which periods of high abun- spp. at Stn L4 in the Western English Channel between 2009 and 2013 giaea dance occurred earlier (summer) and were more extensive. Compared to 2009 2010 2011 2012 2013 the other years analysed, the magni- Muggiaea atlantica tude of peak abundance of the Mean 39 ± 69 25 ± 60 8 ± 12 14 ± 20 32 ± 76 eudoxid, and particularly the poly- 95th percentile 229 123 35 69 222 gastric stage, was comparatively low Muggiaea kochi in 2011 and 2012. The winter minima Mean 3 ± 16 1 ± 5 1 ± 3 0 ± 0 0 ± 0 separating these 2 years was also less 95th percentile 0 6 7 0 0 extreme (Fig. 2C,D). Muggiaea spp. eudoxid Fluctuation at the fine-scale (PC2) Mean 136 ± 212 128 ± 305 55 ± 66 66 ± 74 165 ± 344 95th percentile 505 567 186 241 1145 revealed the underlying dynamics behind patterns of seasonal variabil- 134 Mar Ecol Prog Ser 535: 129–144, 2015

2009 2010 2011 2012 2013 3.0 A B C D E 2

1

+1]) 2.0 x [ 10 0 1.0 residual 1 (log

–1 PC1 PC

0.0 –2 0 25 50 0 25 50 0 25 50 0 25 50 0 25 50

2.0 F G H I J 2

+1]) 1.5 1 x [ 10 1.0 0 residual 2 (log

0.5 –1 PC2 PC

0.0 –2

0 25 50 0 25 50 0 25 50 0 25 50 0 25 50 Week Fig. 2. (A−E) Broad temporal scale (PC1), and (F−J) fine temporal scale (PC2) population dynamics of Muggiaea atlantica be- tween 2009−2013, showing polygastric stage (black lines) and eudoxid stage (grey lines) abundances. Orange and blue shaded areas represent, respectively, positive and negative residuals of the expected eudoxid stage abundance

relatively high. Very low amplitude peaks of abun- 1.0 1.0 A B dance in the winter were also a common feature, 0.5 0.5 although this peak was less apparent in 2012.

0.0 0.0

–0.5 –0.5 Population dynamics Autocorrelation –1.0 –1.0 Linear regression analyses revealed strong rela- 0 15 30 45 60 0 15 30 45 60 tionships between the PC1 of polygastric and eudoxid Lag (weeks) stage abundance (β = 1.23, r2 = 0.92, F = 2783, df = Fig. 3. Periodicity of (A) broad-scale and (B) fine-scale fluc- 258, p < 0.001). The residuals derived from this re- tuation in the abundance of Muggiaea atlantica polygastric gression model oscillated with a periodicity of 44 wk (black lines), and eudoxid (grey lines) stages derived from (Fig. 4A) and represented broad-scale changes in the the autocorrelation function absolute magnitude of eudoxid stage abundance rel- ative to that of the polygastric stage (Fig. 2A−E). In ity (Fig. 2F−J). During the years 2010 and 2013, 2011 and 2012, PC1 residuals were consistently posi- which were characterised by restricted periods of tive, indicating that these years were characterised seasonal M. atlantica abundance, single high-ampli- by high relative abundance of eudoxid stage com- tude peaks of abundance were observed in Septem- pared to the other years studied (Fig. 2C,D). ber (weeks 34 and 38, respectively). This pattern was The residuals derived from the linear regression of in contrast to the other 3 years analysed. In 2009 and PC2 eudoxid and polygastric stage abundance (β = 2012, 2 discrete low-amplitude peaks of abundance 1.19, r 2 = 0.87, F = 1127, df = 258, p = < 0.001) fluctu- were evident, the first occurring in week 27 (June) ated regularly (Fig. 2F−J) with a periodicity of 26 wk and the second during weeks 42−44 (October). In (Fig. 4B). The PC2 residuals represented deviations 2011, a low-amplitude peak of abundance occurred from the expected abundance of the eudoxid stage in week 27 (June), after which abundance remained associated with cycles of production and decline. Blackett et al.: Population ecology of Muggiaea atlantica 135

1.0 1.0 trast, the PC1 abundance of the eudoxid stage de- B A creased considerably during the 2 coldest winters 0.5 0.5 (2010 and 2013). The PC2 residuals were also positively 0.0 0.0 correlated with temperature (Table 3), confirming the importance of temperature to the phenology of M. at- –0.5 –0.5 lantica. Depth-integrated salinity also emerged as a

Autocorrelation potentially important variable, being correlated posi- –1.0 –1.0 tively with PC1 of both eudoxid abundance and the 0 15 30 45 60 0 15 30 45 60 residuals (Fig. 6A−E, Table 3). The analysis showed Lag (weeks) that certain prey items were potentially important co- Fig. 4. Periodicity of (A) broad-scale and (B) fine-scale resid- variates after the effects of temperature were con- uals of the expected abundance of the eudoxid stage of Mug- trolled for using partial correlation analysis (Table 4). giaea atlantica, computed using the autocorrelation function. Positive correlations were found between the PC1 The residual eudoxid abundance is the studentised residuals and PC2 residuals and the abundance of calanoid derived from the linear regression of the eudoxid and poly- gastric stages and indicates deviations from the expected (particularly copepodid stages CI−CV) and cyclopoid abundance of eudoxids based on their typical relationship (Stage CVI) (Fig. 6F−J, Table 4). The abun- dance of polygastric and eudoxid stages was positively correlated with the abundance of bivalve larvae at the High positive residuals were typically observed in fine-scale (PC2) (Fig. 5F−J, Table 4). the initial phases of population increase (Fig. 2F−J). These spikes of higher than expected eudoxid abun- dance suggested periods of accelerated eudoxid pro- Critical environmental ranges duction. However, in 2010 the period of population increase was characterised by negative residuals, Quotient plot analyses revealed critical thermal suggesting that during this year the rate of eudoxid ranges associated with the development and pheno - production was comparatively low. Otherwise, nega- logy of the M. atlantica population (as PC2). The ana - tive residuals were typically observed following lysis did not reveal any associations with specific peaks of abundance (Fig. 2F−J), which indicated that conditions of salinity or water column stratification. the abundance of the eudoxid stage typically de - For both the polygastric and eudoxid stages, high clined more rapidly than the polygastric. However, in PC2 abundance was significantly associated with 2010 the period of population decline was charac- temperatures above 12.9°C, while a negative associ- terised by positive residuals (Fig. 2G), suggesting ation was found with temperatures below 12.8°C (p < that during this period the eudoxid stage was more 0.001) (Fig. 7A,B). High positive PC2 residuals, indi- persistent than in the other years. cating accelerated eudoxid production, were signifi- cantly associated initially with a spring increase in

Correlations with environmental Table 3. Pearson product moment correlations for Muggiaea atlantica poly - variables gastric, eudoxid, and residuals of eudoxid abundance as a function of physical environmental variables at broad (PC1) and fine (PC2) temporal scales. T: tem- The results of correlation analysis perature; MLD: mixed layer depth. dfa: degrees of freedom after correction for autocorrelation. Bold values indicate statistical significance (p < 0.05). ns: not between the principal components of significant environmental factors and M. atlanti ca variability are presented in Tables 3 & Polygastric Eudoxid Residual 4. Both the PC1 and PC2 of M. atlan - r p dfa r p dfa r p dfa tica polygastric and eudoxid stage abundance were positively correlated PC1 0.87 0.01 7 0.88 0.01 7 0.19 ns 19 with depth-integrated temperature T Salinity 0.12 ns 53 0.31 0.03 51 0.67 0.01 13 (Fig. 5A−E, Table 3). The less extreme MLD −0.35 ns 11 −0.28 ns 11 0.21 ns 17 minima observed in the PC1 signals of temperature and M. atlantica abun- PC2 dance in 2011 and 2012 indicated T 0.57 0.00 23 0.68 0.00 20 0.41 0.02 32 Salinity 0.15 ns 27 0.21 ns 24 0.19 ns 39 increased survivorship during the MLD 0.01 ns 258 0.00 ns 258 0.00 ns 258 warmest winter of the 5 years. In con- 136 Mar Ecol Prog Ser 535: 129–144, 2015

Table 4. Partial correlation analyses of Muggiaea atlantica polygastric, eudoxid, temperature to a critical range of and residual eudoxid abundance as a function of biological environmental variables 10.1−11.0°C (p < 0.001) (Fig. 7C). at broad (PC1) and fine (PC2) temporal scales. Partial correlation analysis allowed Subsequent periods of accelerated the effect of temperature to be controlled for. dfa: degrees of freedom after correc- tion for autocorrelation. Bold values indicate statistical significance (p < 0.05). ns: eudoxid production in autumn not significant were significantly associated with temperatures of 13.8−15.6°C (p < Polygastric Eudoxid Residual 0.001) (Fig. 7C). Periods of negative r p dfa r p dfa r p dfa PC2 residuals, indicating acceler- PC1 ated eudoxid decline, were signifi- Copepoda cantly associated with a winter Calanoida (CVI) 0.09 ns 33 0.17 ns 34 0.17 ns 17 decreaseintemperaturetoacritical Calanoida (CI−CV) 0.16 ns 22 0.20 ns 22 0.27 0.03 17 Cyclopoida 0.22 ns 26 0.32 ns 27 0.41 0.03 28 thermal range of 8.3−9.5°C (p = Poecilostomatoida −0.05 ns 10 0.02 ns 10 0.26 ns 17 0.001) (Fig. 7C). Harpacticoida −0.28 ns 13 −0.26 ns 14 0.03 ns 23 Nauplii 0.35 ns 13 0.32 ns 13 −0.03 ns 32 The analysis also showed a sig- Bivalvia 0.27 ns 11 0.28 ns 11 0.07 ns 26 nificant association be tween the Cirripedia −0.17 ns 17 −0.17 ns 17 −0.04 ns 41 re siduals of eu doxid abundance PC2 (as PC2) and specific ranges of Copepoda Calanoida (CVI) 0.12 ns 21 0.30 ns 23 0.44 0.05 21 abundance of copepods and cope - Calanoida (CI−CV) 0.12 ns 22 0.27 ns 24 0.51 0.02 20 podids (Calanoida and Cyclo poida) Cyclopoida 0.22 ns 36 0.24 ns 41 0.25 0.10 46 (Fig. 8). Positive residuals (eudoxid Poecilostomatoida 0.00 ns 23 0.04 ns 22 0.11 ns 23 Harpacticoida −0.11 ns 50 −0.01 ns 58 0.12 ns 35 production) were significantly as- Nauplii 0.13 ns 122 0.12 ns 136 0.02 ns 113 sociated with the maximum (as the Bivalvia 0.36 0.02 42 0.29 0.05 47 −0.13 ns 39 95th percentile) abundance of cope - Cirripedia −0.01 ns 63 0.02 ns 77 0.10 ns 49 pods (4392−4768 ind. m−3), while

2009 2010 2011 2012 2013 2.5 A B C D E

1.2 +1]) x [ 10 0.0

–1.2 PC1 (logPC1

–2.5 0 25 50 0 25 50 0 25 50 0 25 50 0 25 50 3.0 F G H IJ

1.5 +1]) x [ 10 0.0

PC2 (log –1.5

–3.0 0 25 50 0 25 50 0 25 50 0 25 50 0 25 50 Week Fig. 5. Synchronous fluctuation of the abundance of Muggiaea atlantica polygastric (black lines) and eudoxid (grey lines) stages and environmental variables (red lines) at different temporal scales. (A−E) Broad-scale variability (PC1) of polygastric and eudoxid abundance with depth-integrated temperature (r = 0.87, p = 0.01, and r = 0.88, p = 0.01, respectively). (F−J) fine- scale fluctuation (PC2) of polygastric and eudoxid abundance with the PC2 abundance of bivalve larvae (r = 0.38, p = 0.01, and r = 0.31, p = 0.05, respectively) Blackett et al.: Population ecology of Muggiaea atlantica 137

2009 2010 2011 2012 2013 2.5 A B C D E

1.2

0.0

PC1 anomalyPC1 –1.2

–2.5 0 25 50 0 25 50 0 25 50 0 25 50 0 25 50 3.0 F G H IJ

1.5

0.0

PC2 anomaly –1.5

–3.0 0 25 50 0 25 50 0 25 50 0 25 50 0 25 50 Week Fig. 6. Synchronous fluctuation of Muggiaea atlantica residual eudoxid abundance (black lines) and environmental variables (red lines) at different temporal scales. (A−E) broad-scale variability (PC1) of residual eudoxid abundance with depth- integrated salinity (r = 0.67, p = 0.01). (F−J) fine-scale fluctuation (PC2) of residual eudoxid abundance with the PC2 abun- dance of calanoid copepodids (r = 0.51, p = 0.02)

1.4 1.4 1.4 1.4 A B A B 1.2 1.2 1.2 1.2

1.0 1.0 1.0 1.0

0.8 0.8 0.8 0.8

0.6 0.6 quotient Abundance 810121416 810121416 0.6 0.6 0.6 1.4 2.2 2.8 3.6 4.4 0.4 0.8 1.2 1.6 2.0 2.4 2.8 1.4 1.4 0.05 C D Prey availability (103 ind. m3) 1.2 1.2

Abundance quotient Abundance Fig. 8. Critical ranges of copepod and copepodid (Calanoida 1.0 1.0 and Cyclopoida) prey availability associated with the residual Muggiaea atlantica eudoxid abundance identified by quo- 0.8 0.8 tient plot analysis. (A) Positive residuals, indicating eudoxid production, and (B) negative residuals, indicating eudoxid de- cline. Bars represent abundance quotients within each cate- 0.6 0.6 810121416 810121416 gory of temperature. Solid line denotes quotient value 1 (in - Temperature (°C) dicating random selection), quotients > 1 indicate positive association, and quotients < 1 indicate negative association. Fig. 7. Thermal preferences of Muggiaea atlantica identified Dotted lines denote 95% CI by quotient plot analysis. (A) Polygastric stage abundance, (B) eudoxid stage abundance, (C) positive residuals, indica- ting accelerated eudoxid production, and (D) negative resid- negative residuals (eudoxid decline) were signifi- uals (inverted), indicating accelerated eudoxid decline. Bars cantly associated with the minimum abundance of represent abundance quotients within each category of tem- copepods (53−449 ind. m−3). perature. Solid line denotes quotient value = 1 (indicating Our analysis identified a significant influence of random selection), quotients > 1 indicate positive associa- tion, and quotients < 1 indicate negative association. Dotted temperature and the abundance of copepod prey lines denote 95% CI (particularly calanoid cope podid stages CI−CV) on 138 Mar Ecol Prog Ser 535: 129–144, 2015

2009 3.0 the residuals of M. atlantica eudoxid abundance. 16 A >eudoxid abundance Interannual differences in the seasonal dynamics of 14

PC2 anomaly of copepodid prey was at the seasonal minimum

Temperature (°C)8 –3.0 when the critical thermal threshold was reached in late spring; under these conditions accelerated eudoxid 2012 decline was observed, with accelerated eudoxid pro- 16 3.0 D duction occurring subsequently, concurrent with in - 14 creased copepodid prey abundance.

12 0.0 10 DISCUSSION 8 –3.0 Here we modelled the functional relationship be - 2013 tween the 2 main life-cycle stages of Muggiaea at - 3.0 lan tica, revealing key features of its population 16 E dynamics. Our results highlighted a tight coupling 14 be tween the timing of specific environmental condi- tions and the development of the popu- 12 0.0 M. atlantica lation, thereby explaining interannual differences in 10 the phenology of its blooms in the Western English 8 Channel. –3.0 Both M. atlantica and the congeneric Muggiaea 0 10 20 30 40 50 kochi occur in the Western English Channel (Black- Week ett et al. 2014, present study). However, our results only demonstrated a link between the abundance of Fig. 9. Representation of the population dynamics of M. at- lantica in response to interannual variability in temperature the polygastric stage of M. atlantica and the uniden- (solid black line) and food availability (abundance of calanoid tified Mug giaea spp. eudoxid stage. The congeners copepodid prey [as the PC2 anomaly]; dash-dotted grey line) M. atlan ti ca and M. kochi, have slightly different in the Western English Channel between 2009 and 2013. Ver- thermal tolerances and are considered cool-temper- tical red lines denote the timing of the onset of the critical ate and warm-temperate analogues, respectively thermal threshold for eudoxid production. Shaded areas rep- resent positive (orange) and negative (blue) residuals of the (Alvariño 1971). Blackett et al. (2014) have shown expected abundance of the eudoxid stage that M. atlantica is resident in the Western English Blackett et al.: Population ecology of Muggiaea atlantica 139

Channel, whereas M. kochi is a transient non-resi- with the occurrence of a specific temperature range dent, probably restricted by low winter temperatures (10.0−11.1°C). Laboratory experiments on the con- in the Channel. These factors provide strong support generic M. kochi in the Mediterranean have shown for our decision to identify the eudoxids recorded in that polygastric colonies exist in a suspended repro- the present study as M. atlantica. ductive state at typical winter temperature, but suc- We extracted the dominant modes of temporal vari- cessfully reproduce at characteristic spring tempera- ability that characterised the M. atlantica population. ture (Carré & Carré 1991). Since polygastric colonies This procedure allowed us to focus our analysis on typically die after asexual reproduction, Carré & seasonal changes in abundance (PC1), and also, the Carré (1991) proposed that reproductive dormancy underlying fine-scale dynamics (PC2) that drove this facilitates overwintering of the polygastric popula- variability. Our results revealed 2 distinct phenologi- tion. Our results support this hypothesis and suggest cal patterns: (1) early spring population development a critical basal limit of 10°C for asexual reproductive leading to an extended period of abundance with activity of M. atlantica in the Western English Chan- population maxima in both summer and autumn, and nel. This thermal limit is in agreement with the (2) late spring population development resulting in a results of Blackett et al. (2014) who showed that local restricted period of abundance with a population development of the M. atlan ti ca polygastric popula- maximum in autumn. Phenological variability in caly - tion at an open-shelf station in the Western English cophoran populations is common at coastal locations Channel only occurred when the sea surface temper- in temperate regions (Mackie et al. 1987 and refer- ature was above 9°C. ences therein). In some areas this variability appears In addition to the thermal requirements, siphono - to be related to localised changes in seasonal hydro- phore reproduction is also dependent on food avail- graphic conditions (Mackie et al. 1987), the ef fects of ability (Purcell 1982). Our analysis identified a posi- which are often difficult to quantify. Indeed, many tive link between deviations from the expected cnidarian jellyfish populations seemingly appear and abundance of eudoxids and the abundance of cala - disappear irregularly (Graham et al. 2001, Boero et al. noid and cyclopoid copepods. Field studies have 2008). Whilst a multitude of en viron mental factors shown that copepods and their developmental stages undoubtedly contribute, at different levels, to the dy- represent the dominant dietary component of M. namic of a population (e.g. Hutchinson 1957), our atlantica (Purcell 1981, 1982). Therefore, the avail- analysis revealed regularity in the phenological re- ability of copepod prey concurrent with the onset of sponse of the M. atlantica population to localised suitable temperature conditions seems to be an im - variability in temperature and food availability. portant factor influencing the initiation of asexual Knowledge of the population dynamics of calyco- reproductive activity in spring. Indeed, high rates of phoran siphonophores is sparse. Cyclical changes in eudoxid production were associated with the maxi- the relative abundance of the polygastric and eu - mum abundance of copepod prey, while peak eu- doxid stage have been used as an indicator of the life doxid decline was associated with the minimum span of generations (Moore 1949, 1952) and as sig- abundance, echoing the experimental results of Pur- nals of reproductive cycles (Patriti 1964, Mills 1982). cell (1982). However, copepod availability rapidly In the present study we used the residuals from lin- declined following the inception of population devel- ear regression analysis to detect when the eu doxid opment, which suggests that other prey taxa must be stage was more or less abundant than ex pected. Con- important during subsequent population growth. We sidering the fine-scale (PC2) population fluctuation, identified a positive relationship between the popu- periods of higher than expected eudoxid abundance lation density of M. atlantica and the abundance of were recorded at the onset of M. atlantica population bivalve larvae. Bivalve larvae represent a common growth in spring. Large polygastric colonies typically prey resource for many other small cnidarian jellyfish persist through the winter (Hosia & Båmstedt 2008) (Hansson et al. 2005), and Batistić et al. (2013) re- and eudoxid production rates are positively linked ported a similar link with M. atlantica in the Adriatic. to the size of the polygastric colony (Purcell 1982). The spring pulse of eudoxid production provides These factors suggest that periods of higher than the kernel for population development. Sexual repro- expected eudoxid abundance recorded in spring rep- duction by the eudoxid stage supplies new recruits to resent the initiation of eudoxid production by the the polygastric population that in turn increases the overwintering polygastric population. source of eudoxids. During the years 2009, 2011, and The spring pulse of M. atlantica eudoxid produc- 2012, the critical thermal limit for eudoxid production tion recorded in the present study was associated (10°C) was reached in early spring, coincident with 140 Mar Ecol Prog Ser 535: 129–144, 2015

the primary maxima of copepod abundance. These tion was coincident with the autumn copepod popu- conditions provided high food availability leading up lation maxima, suggesting that relaxation of food lim- to the onset of eudoxid production by the overwinter- itation could have been the trigger for secondary pop- ing polygastric stage, likely increasing their repro- ulation development. Subsequent population growth ductive capacity (Purcell 1982). Then as temperature was again correlated with the abundance of bivalve and the abundance of bivalve larvae continued to larvae. The similarity between population develop- increase, the high feeding rates and reproductive ment in spring and autumn lends weight to our capacity of M. atlantica (Purcell 1982, Carré & Carré hypothesis that M. atlantica population development 1991) allowed rapid population growth that culmi- is in phase with regular cycles of prey and tempera- nated in summer population maxima. ture conditions. Field studies have indicated an upper thermal limit The United Kingdom experienced exceptionally of 24°C for the survival of M. atlantica (Marques et al. cold conditions during the winter of 2009−2010 and 2008, Batistić et al. 2013). Therefore high tem per a - the spring of 2013 (www.metoffice.gov.uk/climate/ ture was unlikely to be a limiting factor in the present uk/ summaries). These cold conditions extended study as the maximum average temperature in the throughout central and northern Europe (Cattiaux et Western English Channel is ~16.5 °C (Smyth et al. al. 2010, Andrews 2013) with effects on populations 2010). Decline of the M. atlantica population follow- of various aquatic (Edwards et al. 2013) and terres- ing summer maxima was instead correlated with trial (Glądalski et al. 2014) animals. In the Western the re duced availability of larval bivalve prey, sug - English Channel the critical thermal limit for eudoxid gesting that food limitation was the trigger for popu- production (10°C) oc cur red 4−7 wk later than during lation decline. Food limitation is considered a major the other years. These cold conditions caused a delay cause of population decline for many species of to the initiation of M. atlantica population develop- cnidarian jellyfish (Pitt et al. 2014 and references ment, which had the effect of disrupting the phasing therein). Indeed, the effect of food limitation on the of eudoxid production with food availability. This abundance of M. atlantica eudoxids was pro nounced, trophic mismatch resulted in the development of with the period following the summer maxima being single autumn population maxima, rather than the characterised by lower than expected abundance. summer and autumn peaks recorded during the This decline is probably explained by the combined warmer years. effect of the cessation of eudoxid production in re- During 2010 the late onset of M. atlantica popula- sponse to food limitation (Purcell 1982) and the rapid tion development was characterised by a lower rate mortality of existing eudoxids that usually have a of eudoxid production than was identified during the shorter life span than the polygastric stage (Carré & other years analysed. During this year the copepod Carré 1991). A similar mechanism was responsible population displayed an atypical phenology, charac- for the decline of the M. atlantica population in win- terised by a summer, rather than spring, primary ter, although the decline was compounded by low win- maximum. This pattern resulted in an extended ter temperatures. Temperatures <9.5°C represented period of low copepod prey availability during the a limit be low which the abundance of eudoxids rap- winter and spring. Food limitation and low tempera- idly de clined, potentially signalling the temperature tures during this period may have reduced survivor- at which the polygastric stage enters reproductive ship and limited the growth potential of the overwin- dormancy. tering polygastric stages. The reduced eudoxid The significance of food availability is further production at the onset of the critical thermal limit demonstrated by the initiation of secondary popula- likely reflects the impact of these negative factors on tion growth during the years characterised by early the reproductive capacity of the overwintering poly- spring development (2009, 2011, and 2012). Second- gastric population. ary population growth (in autumn) developed in the The disruption of phasing between food availabil- same way as documented in spring, starting with a ity and eudoxid production was more severe in 2013. pulse of higher than expected eudoxid abundance. During this year the critical thermal limit for eudoxid This sequence probably represents the initiation of production occurred after the primary (spring) cope- eudoxid production by polygastric colonies remain- pod population maxima; instead coinciding with ing from the summer cohort. Given the thermal toler- their mid-season minimum. Under these conditions, ances of M. atlantica previously documented, a tem- despite the onset of suitable temperature, eudoxid perature cue for this secondary population develop- production was restricted by insufficient food avail- ment seems unlikely. The pulse of eudoxid produc- ability. As a result, population development was de - Blackett et al.: Population ecology of Muggiaea atlantica 141

layed and only commenced once food availability nel represents the most northerly recorded habitat increased with the autumn copepod peak. The sub- supporting a self-sustaining population of M. atlantica sequent population maximum occurred in late in the Northeast Atlantic (Blackett et al. 2014 and ref- autumn, 5 weeks later than the time of the 2010 pop- erences therein). At this latitude, M. atlantica is likely ulation maximum. at the extreme of its thermal tolerance limits. How- Dramatic seasonal variation of upper-ocean envi- ever, the thermal requirements of cnidarian jellyfish ronmental conditions is characteristic of temperate are usually location-specific and different populations regions (Longhurst 1998). Zooplankton must be able can adapt to different temperature regimes (Lucas & to adapt to variability in the timing and extent of Dawson 2014). For instance, the M. atlantica popula- these changes to optimise the phasing of their life tion on the coast of the cycles with favourable environmental conditions successfully reproduces at temperatures of 8−10°C (Mackas et al. 2012). The degree of mismatch be - (Purcell 1982). tween the environmental conditions and the species’ The original match-mismatch hypothesis (Cushing environmental requirements influences fitness (Cush- 1990) states that a trophic mismatch between predator ing 1990, Durant et al. 2007), with potentially pro- and prey leads to poor growth, survival and subse- found effects on the size and phenology of zooplank- quent recruitment. In our study a trophic mismatch ton populations (Ji et al. 2010, Mackas et al. 2012). with the copepod prey population in the Western Our results highlight the fundamental importance of English Channel resulted in dramatic phenological temperature in shaping cnidarian jellyfish popula- shifts of M. atlantica; however the effect on population tions. Changes to the seasonal temperature cycle in size was less extreme. Cold spring conditions re- the Western English Channel produced dramatic stricted M. atlantica population development, leading shifts in the phenology of the M. atlantica population to trophic mismatch and the development of single by modulating the degree of trophic mismatch expe- autumn population maxima with low mean an nual rienced during the initial phases of population devel- abundance. However, the amplitude of these single opment. Temperature variability has been shown to autumn maxima was higher than peaks during those modify the timing of important developmental events years where both summer and autumn maxima were of numerous other species of zooplankton (Beau- recorded. This disparity highlights the resilience and grand et al. 2002, Edwards & Richardson 2004). With adaptability of M. atlantica, and cnidarian jellyfish in direct and indirect effects on virtually all aspects of general. Their opportunistic life history traits and marine ecosystems (e.g. biological, physical and chem- physiology enable high feeding rates, rapid growth ical components), temperature consistently emerges rates, high fecundity and short generation times (Pur- as the most common pheno logical correlate (Mackas cell et al. 2007, Acuña et al. 2011). These traits afford et al. 2012). cnidarian jellyfish populations great flexibility, en- Marine organisms with long life spans, such as fish, abling rapid conversion of available food energy into often employ mechanisms that lead to fixed seasonal large population biomass. The negative impact of a timing for reproduction (Cushing 1990). This strategy trophic mismatch can be quickly overcome in re- ensures that, over the long term, reproduction will be sponse to subsequent favourable conditions. in phase with favourable conditions at least during Despite optimal phasing between the onset of eu- some years, despite interannual variability in the tim- doxid production and copepod prey availability, dur- ing of environmental conditions (Cushing 1990, Du- ing 2011 and 2012 the M. atlantica population den- rant et al. 2007). Because most zooplankton have short sity was lower than during the other years analysed. (<1 yr) life spans, a fixed seasonal timing window in- This low abundance was particularly pronounced in creases the risk that in the event of a ‘mismatch’, the the polygastric stage, indicating that the high resid- population will not survive to reproduce in a subse- ual abundance of the eudoxid stage was probably a quent year (Mackas et al. 2012). Therefore, zooplank- consequence of reduced polygastric stage production ton species tend to reproduce episodically, in response during these 2 years. This result was not linked to a to environmental cues (Mackas et al. 2012). Muggiaea reduction in copepod or bivalve larvae prey avail- sp. typically have a life cycle in the order of weeks to ability (data not shown). However, the years 2011 months (Carré & Carré 1991). Although we have im- and 2012 were characterised by warmer and more plied that M. atlantica initiates reproduction in re- saline conditions than the other years analysed. sponse to a fixed thermal threshold (10°C), we do not Although many Calycophora thrive under warm and consider this temperature as a thermal cue, but rather high salinity conditions (e.g. Buecher 1999), M. at lan - a physiological limitation. The Western English Chan- tica displays euryhaline characteristics and is known 142 Mar Ecol Prog Ser 535: 129–144, 2015

to tolerate salinities outside the extremes recorded in Acknowledgements. This work was conducted as part of a the present study (Blackett et al. 2014 and references PhD Project by M.B. and co-funded by the Natural Environ- ment Research Council and the Sir Alister Hardy Founda- therein). As changes in the salinity conditions could tion for Ocean Science. We thank Dr. Aino Hosia and 3 reflect modification to water circulation patterns in anonymous reviewers for their valuable comments and the Western English Channel (Pingree 1980), it is suggestions on the manuscript. possible that advection and migration of M. atlantica could explain the low densities recorded in 2011 and LITERATURE CITED 2012. The seasonal dynamic of the M. atlantica popula- Acuña JL, López-Urrutia Á, Colin S (2011) Faking giants: tion revealed a recurrent low amplitude peak of the evolution of high prey clearance rates in jellyfishes. abundance in winter (weeks 5−10). It is conceivable Science 333:1627−1629 Alvariño Á (1971) Siphonophores of the Pacific with a that this peak could reflect local reproduction during review of the world distribution. Bull Scripps Inst the relatively warm winter conditions of 2013. How- Oceanogr Univ Calif 16: 14−32 ever, recurrent reproductive activity in winter would Andrews DWK, Monahan JC (1992) An improved hetero- appear unlikely given that the winter temperatures skedasticity and autocorrelation consistent covariance matrix estimator. Econometrica 60: 953−966 typically experienced are below the thermal toler- Andrews J (2013) How cold Europe was in March 2013. www. ances of M. atlantica discussed above. The M. atlan - accuweather.com/en/weather-blogs/andrews/ how-cold- tica population in the Western English Channel europe-was-in-march-2013/9313909 (Accessed on 15 Jan- likely represents the northern component of a spa- uary 2015) ć ć ć ć tially extensive Northeast Atlantic metapopulation Batisti M, Lu