EIDO Escola Internacional de Doutoramento

TESE DE DOUTORAMENTO: Temporal variability of in the north and northwest Iberian shelf: Understanding plankton dynamics from monitoring time-series.

Variabilidad temporal del plancton en el norte y noreste de la plataforma ibérica: Conociendo la dinámica del plancton a través de series temporales de monitoreo

Lucie Buttay

2018

Mención internacional

CONTENTS

LIST OF FIGURES ...... 1

LIST OF TABLES ...... 1

THESIS ORGANIZATION ...... 1

INTRODUCTION 3

MARINE PLANKTON ...... 5

TEMPORAL STRUCTURE OF BIOLOGICAL COMMUNITIES ...... 9

THE RADIALES MONITORING PROGRAM IN THE NW AND N IBERIAN ATLANTIC ...... 13

TIME-SERIES ANALYSIS ...... 14

THESIS OBJECTIVES ...... 17

CHAPTER 1. SEASONAL AND LONG-TERM VARIABILITY OF MESOZOOPLANKTON ABUNDANCE AND BIOMASS ALONG THE NORTH IBERIAN ATLANTIC SHELF 19

INTRODUCTION ...... 21

MATERIAL AND METHODS ...... 23 Sampling strategy ...... 23 Data analysis ...... 25 RESULTS ...... 26 Patterns of variability of total biomass and abundance ...... 26 Biomass-abundance relationship (average individual weight) ...... 31 Seasonal patterns of SST and Chl-a concentration ...... 33 Zooplankton cycles along the northern Iberian shelf ...... 35 DISCUSSION ...... 38 Modes of variation ...... 38 Long-term trends ...... 41 CONCLUSIONS ...... 43

SUPPLEMENTARY INFORMATION ...... 45

CHAPTER 2. LONG-TERM AND SEASONAL ZOOPLANKTON DYNAMICS IN THE NORTHWEST IBERIAN SHELF AND ITS RELATIONSHIP WITH METEO-CLIMATIC AND HYDROGRAPHIC VARIABILITY. 47

INTRODUCTION ...... 49 METHOD ...... 51 Study area and data collection ...... 51 Numerical analysis ...... 56 Coupling between zooplankton dynamics and environmental variability ...... 58 RESULTS ...... 58 Zooplankton abundance ...... 58

Zooplankton composition ...... 60 Coupling between zooplankton dynamics and environmental variability ...... 64 DISCUSSION ...... 66 Zooplankton seasonal and inter-annual variability ...... 66 Long-term changes ...... 67 Coupling between zooplankton dynamics and environmental variability ...... 68 CONCLUSIONS ...... 71

SUPPLEMENTARY INFORMATION ...... 73

CHAPTER 3. ENVIRONMENTAL MULTI-SCALE EFFECTS ON ZOOPLANKTON INTER-SPECIFIC SYNCHRONY. 75

INTRODUCTION ...... 77

MATERIAL AND METHODS ...... 79 Study area ...... 79 Zooplankton community and environmental variables ...... 79 Data preparation ...... 80 Wavelet analysis ...... 81 RESULTS ...... 83 Dynamics of zooplankton aggregated properties and environmental variables ...... 83 Dynamics of zooplankton taxa and species ...... 86 DISCUSSION ...... 88 Temporal patterns and coupling between aggregated zooplankton properties and environmental variables ...... 89 Inter-specific synchrony: variability and possible drivers ...... 90 Ecological implications ...... 92 CONCLUSIONS ...... 93

ACKNOWLEDGMENTS ...... 93

SUPPLEMENTARY INFORMATION ...... 94

CHAPTER 4. HOW ENVIRONMENTAL FORCING CAN SYNCHRONIZE POPULATION FLUCTUATIONS 99

INTRODUCTION ...... 101

MATERIAL AND METHODS ...... 102 Natural system description ...... 102 Model of competition for essential (=non-substitutable) resources ...... 105 RESULTS ...... 106 Natural system observation ...... 106 Model of competition for essential resources ...... 109 DISCUSSION ...... 112 Temporal structure of the community and its drivers ...... 112

CONCLUSION ...... 116

GENERAL DISCUSSION 117

SCALE MATTERS ...... 122

PLANKTON VARIABILITY AND ITS DRIVERS...... 124 Spatial patterns ...... 124 Changes in abundance or biomass ...... 125 Phenology and community temporal structure...... 128

CONCLUSION 133

APPENDIX 137

REFERENCES 143

ABSTRACT 159

RESUMEN EN ESPAÑOL 163

ACKNOWLEDGMENTS 176

List of Figures

Figure I.1: Size categories of plankton organisms ...... 5 Figure I.2: The biological carbon pump ...... 7 Figure I.3: Compensatory vs synchronous dynamics ...... 10 Figure I.4: Map of RADIALES stations ...... 12 Figure I.: Wavelet decomposition of a sinusoidal signal ...... 16

Figure 1.1: Observed series of zooplankton total biomass and annual pattern ...... 27 Figure 1.2: Annual anomalies of zooplankton total biomass ...... 28 Figure 1.3: Observed series of zooplankton total abundance and annual pattern ...... 29 Figure 1.4: Annual anomalies of zooplankton total abundance ...... 30 Figure 1.5: Relationship between zooplankton Total biomass and abundance ...... 32 Figure 1.6: Annual anomalies of mean individual weight ...... 33 Figure 1.7: Seasonal pattern od Chl-a ...... 34 Figure 1.8: WPS of zooplankton total biomass series ...... 36 Figure 1.9: Cluster build on the WPS similarities ...... 37 Figure SI1.1: Monthly anomaly of sea surface temperature ...... 45

Figure 2.1: Map of the study area ...... 51 Figure 2.2: Total zooplankton abundance and cumulative sum ...... 59 Figure 2.3: GLMM estimated values and standart error ...... 60 Figure 2.4: Seasonal variability of zooplankton abundance...... 61 Figure 2.5: Interannual variability of the 6 main taxonomic groups ...... 63 Figure 2.6: Time-series of principal component PC1 ...... 64 Figure 2.7: Annual time-series of precipitation, upwelling index and Gulf Stream North Wall ...... 65 Figure SI2.1: Seasonal variability of composition ...... 73 Figure SI2.2: Time-series of principal component PC2 ...... 73

Figure 3.1: Wavelet Power Spectrum (WPS) of zooplankton biomass, abundance, Miño outflow and upwelling index ...... 84 Figure 3.2: Wavelet coherence between zooplankton biomass, upwelling Index and Miño outflow ...... 85 Figure 3.3: Annual oscillations of all taxonomic groups and copepod species ...... 87

Figure 3.4: Annual Phase Angle Variance (PAV) and average duration of upwelling events .... 88 Figure SI3.1: Wavelet Power Spectrum of zooplankton biomass and abundance at E3VI ...... 96 Figure SI3.2: Wavelet coherence between zooplankton biomass, upwelling index and Miño outflow (E3VI) ...... 97 Figure SI3.3: Annual oscillations of all taxonomic groups and copepod species (E3VI) ...... 98 Figure SI3.4: Correlation between Phase Angle Variance (PAV) and environmental variables .98 Figure SI3.5: Annual Phase Angle Variance (PAV) (E3VI) and average duration of upwelling events ...... 100 Figure 4.1: Annual oscillations of species and Phase Angle Variance (PAV) ...... 107 Figure 4.2: Wavelet Power Spectrum (WPS) of daily upwelling Index...... 108 Figure 4.3: Covariance between both consumers along a supply amplitude gradient ...... 110 Figure 4.4: Effect of increasing supply amplitude on the dynamics of consumers C1 and C2 ...... 111 Figure 4.5: How resource fluctuation can affect consumer’s dynamics ...... 115 Figure SI4.1: Wavelet Power spectrum of the 12 diatom species...... 117

Figure D.1: The Thesis in a glimpse ...... 121 Figure D.2: Diatom absolute phases for the annual oscillation ...... 129 Figure D.3: Wavelet Power Spectra (WPS) of the main plankton taxa in E2CO ...... 130

List of Tables

Table 1.1: Details of zooplankton sampling at each station ...... 24 Table 1.2: Basic statistics of zooplankton abundance and biomass ...... 28 Table 1.3: Long term trend parameters in abundance and biomass ...... 31

Table 2.1: Taxonomic groups in E1VI and E3VI ...... 54 Table 2.2: List of environmental variables ...... 55 Table 2.3: Model selection. Akaike criterion ...... 57

Table SI3.1: List of observed taxonomic groups and copepod species at E1VI ...... 94 Table SI3.2: List of observed taxonomic groups and copepod species at E1VI ...... 95

Table 4.1: List of the selected diatom species ...... 104 Table 4.2 : Model parameters ...... 106

Table A.1: List of taxonomic groups considered in chapter 2 and 3 ...... 138 Table A.2: List of Copepod species considered in the chapter 3 ...... 139 Table A.3: Diatom species considered in the chapter 4 ...... 140

1

Thesis organization

The four chapters that compound this Thesis have been designed as scientific articles, and as such they conserve the typical structure of this type of publications (Introduction, Material and Methods, Discussion and Conclusion). Thus, as each chapter must be able to stand alone, some repetitions exist, principally within the description of methods. Also, some general or methodological aspects have been too shortly described in the chapters and I have therefore made the most of the Thesis introduction to develop them.

The first chapter “Seasonal and long-term variability of mesozooplankton along the Northern Iberian Atlantic shelf” encompasses the description of the zooplankton variability in the Atlantic coast of the north of Spain and a comparison between locations based on their cycles. This work represents the first exploration I carried out on all the RADIALES data I had access to. However, it arose as Thesis chapter and article much later, while writing the other chapters and realizing the importance of descriptive studies to compare with. This chapter is currently review in Estuarine, Coastal and Shelf Science.

The second chapter “Long-term and seasonal zooplankton dynamics in the northwest Iberian shelf and its relationship with meteo-climatic and hydrographic variability” focuses on the time series of zooplankton off Vigo, where the most noticeable abrupt change in abundance and biomass was observed. In this chapter, I made a particular effort in describing precisely the zooplankton variability, in terms of abundance and taxonomic composition, and quantify the long-term change with classical but robust statistical methods. This chapter was the first work of the Thesis, published in 2015 in Journal of Plankton Resarch.

The work within the third chapter “Environmental multi-scale effects on zooplankton inter-specific synchrony” has been made in collaboration with Dr. Bernard Cazelles from the École Normale Supérieure in Paris (France) with whom I realized a three month stay in autumn 2013. In this chapter, we continued with the exploration of the Vigo time-series, but using a different method (Wavelet analysis) that allowed us to observed new features of the zooplankton variability. In particular, I tracked the changes in synchrony among taxonomic groups and species of the zooplankton community and this work gave rise to a 2 publication in Limnology and Oceanography in 2016 and a serious fascination/obsession for synchrony related subjects.

While the third chapter opened numerous questions about community dynamics, I started collaboration with Dr. David A. Vasseur from his Theoretical Ecology lab of the Yale University in New Haven (CT, USA) to explore the mechanisms that force community to shift from compensatory dynamics to synchrony. This work gave rise to the fourth chapter “How environmental forcing can synchronize population fluctuations?”, in which I present the effect of enhanced amplitude of environmental forcing fluctuations for the temporal structure of the community, combining two complementary approaches: 1) wavelet analysis on the natural community of diatom species to quantify the synchrony and how it change in time, and 2) a theoretical simulation, using a chemostat type model, to understand how competition for fluctuating ressources can shape the temporal structure of the communities. 3

General Introduction

General introduction 5

Marine Plankton

Plankton represents a very diverse group of organisms whose distinctive characteristic is to be drifting along with the currents due to their limited swimming performance. They occur in a myriad of shapes and sizes. Different categories have been built to classify plankton organisms according to their size, feeding habits, life history, etc. Size is often a primary classification criterion of zooplankton organisms (Dussart 1965): femtoplankton contain organisms that measure less than 0.2 µm; from 0.2 up to 2 µm; nanoplankton account for organisms between 2 and 20 µm; microplankton from 20 to 200 µm, mesoplankton from 0.2 to 20 mm, macroplankton organisms measure between 20-200 mm and megaplankton account for organisms over 200 mm (Figure I.1).

Figure I.1 : Size categories of plankton organisms (adapted figure from Sieburth et al., 1978).

Across those size groups we can find a diversity of groups, including viroplankton, , and zooplankton. It is worth noting that size is a master ecophysiological trait (Odum 1956; Hutchinson and MacArthur 1959; Ward et al. 2012) which is related, through allometric relationships, to individual process rates such as growth, fecundity, respiration, etc.

The work presented in this Thesis has been centered in the dynamics of two different groups of plankton: mesozooplankton organisms (Chapter 1 to Chapter 3) and , one of the dominant groups of phytoplankton (Chapter 4). A complete list of the taxon and 6 species considered is given in the Appendix, table 1 to 3. Some additional information about those groups is therefore given in the present introduction.

Mesozooplankton comprises heterotrophic animals living in the water column and measuring from 200 µm up to 2000 µm. They belong to various taxonomic groups such as crustaceans, cnidarians, siphonophors, chaetognaths, gastropods, annelids… The most abundant and studied group being the crustaceans and in particular the . Another important classification that can be made on zooplankton organisms concerns their life history. Organisms that spend their entire life in a planktonic form, such as pteropods, cladocers, copepods, siphonophores, etc. are part of the while the temporary members of the plankton, such as most larval forms of echinoderms and crustaceans, marine worms, some marine snails, most fish, etc. are called meroplankton. They spend only a part of their life cycle (usually the larval stage) in a planktonic form. This specificity in their ecology has important ramifications as different environmental drivers can have different importance depending of their life stage.

The Phytoplankton group encompasses procaryotes organisms such as and eucaryotes such as diatoms, dinoflagelates, cocolithophors and haptophytes as some examples. Phytoplankton organisms live mainly within the photic zone where the light enables them to conduct the photosynthesis (Falkowski 1998). Indeed, as , their growth is also controlled by the availability of inorganic nutrients. Diatoms are often one of the most abundant organisms composing the phytoplankton community and are adapted to high productivity moments (i.e. high-to-moderate turbulence and inorganic nutrient concentrations, Margaleff (1978)). A particular morphologic specificity is that the cell is enclosed within a silica which makes them dependent of the availability of silicate.

Phytoplankton organisms are responsible for approximately half (45%) of the world’s primary production (Field et al. 1998). Then, approximatively 10 to 40% of the primary production is transferred to the the trophic web (Stock et al. 2014) by the secondary producers, being them mainly zooplankton organisms (Banse 1995) depending on the degree of coupling between production and grazing. Another part of the primary production is re-mineralized within the microbial loop (Azam 1993). In addition, and by grazing on phytoplankton, zooplankton organisms generate detritus particles (i.e sloppy feeding, Moller 2005) and faecal pellets (Turner 2002) that contribute to the sinking of General introduction 7 organic matter to the deep ocean. Planktonic organisms have therefore a crucial role in the coupling of global biogeochemical cycles (Figure I.2) and therefore, at the global scale, on climate regulation. In this regard, knowledge about their dynamics is essential to sharp future predictions of the marine environment (Falkowski et al. 2004).

Figure I.2 : The biological carbon pump (credit: U.S. JGOFS)

In a more practical aspect, phytoplankton and zooplankton organisms are also convenient for the study of environmental impacts and population dynamics. They have generally short life span (some days to months), are sensitive to their environment and respond rapidly to environmental variability. Planktonic organisms have therefore been proposed to be good sentinels of environmental changes (Taylor et al. 2002; Perry et al. 2004; Hays et al. 2005). In addition, within our study area the effects of marine resource exploitation of plankton are limited – in comparison with other available time series such 8 as fish populations – and concern mainly the meroplankton species whose adult phases are targeted directly (e.g mainly of mollusks, cephalopods, decapods and fishes), while holoplankton species should only be affected indirectly through changes in some of their predator populations, mainly small pelagic fishes, which are heavily fished in the area.

The diversity of organisms within each size class is surprisingly high and redundancy can occur when accounting for the functional traits of each organism. Indeed plankton is much more diverse than expected for a seemingly homogeneous aquatic environment where competitive exclusion should happen and that constitutes the plankton paradox suggested firstly by Huchinson (1961). Yet the ocean is not as homogeneous as it seems, and plankton organisms are typically heterogeneously distributed in both spatial and temporal dimensions marked by hydrographic and physico-chemical gradients determined by temperature, salinity, pressure, light availability, concentration of nutrients, etc. Moreover, at the scale of the community, this high diversity and redundancy of functional traits may confer more resilience to environmental changes (Loreau 2001).

Plankton also tend to aggregate in swarms (Margaleff 1979) whose density can exceed up to 1000 times the average density (Omori and Hammer 1982). This phenomena is resumed as the so-called patchy distribution of plankton and can be caused by a variety of factors (e.g. temperature and salinity, light intensity, etc.) gathered in ephemeral structures that arise from the movement of water masses and ocean surface mixing. Those patches are rarely mono-specific since foraging behaviour of predators tends to reflect the patchy distribution of their prey (Margalef 1979).

Typical gradients of distribution exist, however. The upper layers of the water column, where light allows primary producers to grow, are generally more densely occupied than the deeper ones. The swimming performance of some zooplankton allows them to engage in vertical migration through the day making the vertical structure of organism’s density to fluctuate. Crustaceans like copepods can make nycthemeral vertical migration of hundreds of meters each day (Ohman1990, Jonasdottit et al 2015, Hansen and Visser 2016).

Typical horizontal zonation also exists. For instance, the coastal environments, which receive continental runoff with the associated nutrient inputs, are generally more productive than the oceanic domain. Besides, currents and surface friction in the shelf and oceanic domains often create hydrographic features such as fronts, eddies and the General introduction 9 emergence of sub-surface waters by coastal upwelling, which may intensify plankton growth and production.

The most conspicuous and well described mode of temporal variability is the seasonal cycle that occurs at the annual scale at mid and boreal latitudes (e.g. Alcalà et al., 2004; Beaugrand et al., 2000; Colebrook, 1979; Levasseur et al., 1984; Mackas et al., 2012; Romagnan et al., 2015). In temperate latitudes, the primary production classically follows a bimodal pattern through the year, with peaks occurring during the periods of transient thermoclines: in spring, after winter mixing, when the thermal stratification of the water column begins, and in autumn, with the deepening of the seasonal thermocline (Harris 1986; Longhurst 1998). Finally, the distribution in space and time of planktonic organisms can also be driven by density dependent processes, such as inter-specific competition; the next section of the introduction will be dedicated to this subject.

Overall, the understanding of the plankton distribution is not a trivial question and it gains interest in the context of ongoing global change, which is expected to impact both the spatial and temporal distribution of planktonic organisms (Parmesan et al. 2003; Edwards et al. 2004; Mackas et al. 2012a).

The work gathered in the present Thesis focused on the temporal aspect of plankton organism distribution along the North and Norwest coast of Spain, in relation with abiotic and biotic factors. For this purpose, I have employed diferent statistical analisis (GLMM, PCA, etc) on the long time series obtained within the framework of the RADIALES monitoring program.

Temporal structure of biological communities

Understanding the rules that control population fluctuations is a fundamental topic in ecology, especially within the context of the ongoing climatic changes that can produce a variety of perturbations (e.g habitat modification, latitudinal displacements in species distribution, change in frequency and amplitude of abiotic forcing) susceptible to lead to a reorganization of ecological communities in a wide range of spatial and temporal scales and at different organization levels. 10

Figure I.3: Two contrasted community dynamics. Compensatory dynamics (a) and Synchronous dynamics (b)

The temporal structure of the community is one of the community characteristics that has been proposed to have strong consequences for community stability (Pimm et al. 1988; Tilman et al. 1998; Inchausti and Halley 2003; Downing et al. 2008; Gonzalez and Loreau 2009; Gouhier et al. 2010a). The wealth of theories to explain community structure and dynamics follow two different basic argumentations. At one side, some theories postulate that fluctuations are regulated by density-dependent processes (MacArthur 1984; May and McLean 2007). As such, interspecific competition leads populations of similar species within a community to fluctuate in a compensatory way (i.e the increase in abundance of one species is accompanied by the decrease of another functionally similar species, Figure I.3) (Tilman et al. 1998; Ives et al. 1999; Klug et al. 2009) because they need some kind of temporal (or spatial) niche differentiation in order to coexist (Loreau and de Mazancourt 2008; Kalyuzhny et al. 2014). The alternative hypotheses suggests that density- independent environmental forcing mechanisms are the major drivers of population fluctuations (Turchin 1999) and this seems to be the mainstream line of thought followed to explain community assembly and dynamics of marine pelagic populations.

Overall, there is a controversy between those two points of view that otherwise should not necessarily viewed as opposite. For example, Houlahan (2007) reported that a compensatory dynamic was found in less than 30% of the animal and plant they tested for, General introduction 11 using a covariance method, and concluded that environmental forcing was the dominant force controlling communities’ structure. A year later, Ranta et al. (2008) added that this vision was not completely accurate since they demonstrated that the method they used was of limited value to detect the presence of interspecific interaction among species within a community. Apparently, an important part of the controversy seems to come from different methodologies that provide different insights on community structure and dynamics.

The detection of synchrony or compensatory dynamics within natural communities requires having a deep understanding of the biological system under study and of the environmental drivers that are operating, which must be based on the analysis of long-term data sets. Because population fluctuations and species interactions are not expected to have constant properties through time (stationary dynamics), appropriate statistical methods have to be employed in order to describe the characteristics of the time series. In addition, population variability can occur at different scales. Surprisingly, most previous studies have avoided those problems and used metrics based on covariance methods (e.g, Gonzalez and Loreau 2009, Jochimsen et al. 2014).

Within this Thesis I employ a wavelet analysis technique, which estimates the amount of variation in a time series attributable to a particular frequency (scale) at a particular instant (time) to quantify the temporal association between different time series of biotic and abiotic variables, being mostly those sampled within the RADIALES monitoring program.

Figure I.4 : Map of the study area showing the position of the four sections and oceanographic stations sampled within the frame of the time-series observing programme RADIALES (http://www.seriestemporales-ieo.com). From east to west and from coastal to more oceanic sites, the stations of the different sections are named as follows: Santander (E2SA, E4SA and E6SA), Gijón (E1GI, E2GI and E3GI), A Coruña (E2CO) and Vigo (E1VI and E3VI). General introduction 13

The radiales monitoring program in the NW and N Iberian Atlantic

The RADIALES monitoring program (http://www.seriestemporales-ieo.com) was implemented in 1988 at an inner-shelf site off A Coruña, and extended to cover other locations along the northern Iberian Atlantic shelf (Figure I.4), with a set of oceanographic stations off Santander and Gijón in the Cantabrian Sea and off Vigo in the southern Galician shelf (Bode et al. 2012b). It covers an important part of Spanish Atlantic coastal and shelf waters, from Cantabria to Galicia, which correspond to the Spanish North-Atlantic Demarcation within the frame of the (Marine Strategy Framework Directive, MSFD, Directive 2008/56/EC). The data used in the present Thesis come from 9 stations distributed along 4 sections covering different domains of the Bay of Biscay and Iberian Coast Ecoregion (ICES areas VIIIa and IXa-north): off Santander and Gijón (Cantabrian Sea) and off A Coruña and Vigo (Galician shelf). Routine sampling within the time series monitoring programme started in the different sections in 1991, 2001, 1988 and 1994 respectively.

The entire study zone is subject to an intermediate latitude regime, but high heterogeneity exists due to the change in coastal orientation and the differential intensity of hydrodynamic processes that occur along the northern Atlantic coast of Spain. The Galician coast, located at the northernmost limit of the Canary Current upwelling system, is characterized by the occurrence of a seasonally varying coastal upwelling. From mid- spring to early autumn, the predominance of northern winds blowing parallel to the coast force sub-surface waters to upwell (Fraga 1981; Blanton et al. 1987). During upwelling events, colder, nutrient enriched waters from deeper layers, inject nutrients into the illuminated surface layers, favouring phytoplankton growth (Wooster and Reid, 1963). In winter, predominant winds roll to the southwest. This situation is accompanied by coastal downwelling and higher precipitation inland which in turn increase river outflow and favor the formation of the western Iberian buoyant plume (WIBP, Peliz et al., 2005). The Iberian Poleward Current (IPC), a slope undercurrent flowing opposite to the general surface circulation pattern, intensifies also during winter months, favoring the formation of sharp across-shelf gradients (Figueiras et al. 2003; Peliz et al. 2005). Along the Cantabrian coast, the main annual input of nutrients to the upper layers is driven by the deep convective mixing occurring in winter (Llope et al. 2007). During summer, when the surface layer is 14 nutrient depleted, the fertilization of the surface layer is related to wind-driven upwelling events prompted by easterlies. In the eastern Cantabrian Sea, it is also noticeable the supply of nutrients due to outflow from rivers (González-Nuevo and Nogueira 2014).

Time-series analysis

The more efficient way to identify patterns of temporal variability and to understand system dynamics is to proceed to local repetitions of observations (time-series) (Greve et al. 2004; Perry et al. 2004). The main objective of the time-series analysis is therefore to employ mathematical models to provide a description of the data variability while a second one would be, once the system is well described, to make predictions for the future. In the present Thesis we have focused on the description of zooplankton dynamics and their relationship with environmental variables. The different methods employed within this Thesis are described in the methods section of each chapter. Here, I seek to explain the specificities of time-series data, to define some vocabulary and to describe some of the applied methods in a more handily way.

Time series usually break one of the basic requirements for most of the classical statistical analysis: observations, adjacent in time, are not independent and often not identically distributed. Consequently, appropriate models, able to handle time correlated data, are needed. There are two approaches that can potentially be complementary: the time domain approach and the frequency domain approach (Shumway and Stoffer 2006a).

Within the time domain approach, the correlation between adjacent points in time is treated as the dependence of the current value on the past value(s). For example, in the present Thesis, GLMs (Generalized Linear Model) with an ARMA(p, q) (autoregressive integrated moving average) structure have been employed, where p represent the number of autoregressive parameters and q the number of weight parameters of the moving average. Hence the time series residuals at a time s are modelled as a function of the p previous residuals, in addition to noise (Zuur et al. 2009).

The frequency domain approach considers that the main characteristics of time series relate to the periodic or harmonic components found in most of the data, which are caused General introduction 15 by a variety of phenomena. In spectral analysis, a separate evaluation of the variance within each periodicity is made. One of the fundamental methods to accomplish this objective is the Fourier decomposition of the time series (Chatfield 1989). Ecological time series typically present non-stationary dynamics; as such, they present changes through time in local average, in the repartition of variance among periodicities, in the relationship with environmental forcing or in a combination of these (Cazelles and Hales 2006). Both types of empirical modeling approach, in the time and in the frequency domains, have been used frequently in ecology and population dynamics (Platt and Denman 1975). They make the assumption, however, that the time series are stationary, failing therefore to describe the transient dynamics that are susceptible to play an important role in the functioning of natural systems. The Wavelet analysis overcomes the problems of non-stationarity in time series by performing a local time-scale decomposition of the signal (Lau et al. 1995; Torrence and Campo 1998). Figure I.5 presents a wavelet decomposition of two signals, that contain the same periodic components, but in one example both components are present throughout all the series, while in the second signal the first component occurs during the first half of the series followed by the second component during the second half. While the Fourier analysis is well suited to quantify periodic components in a time series, it fails to characterize signals whose frequency composition changes through time.

The wavelet coherence is a direct measure of the correlation between the spectrum of two time series, normalized by the spectrum of each signal. It allows to detect transient associations between series (Liu 1994) by giving local information about when the two non- stationary time series are linearly correlated and at which frequency. The temporal association between signals can also vary in time. With complex wavelets function, such as the Morlet wavelet employed in Chapter 1, 3 and 4, it is posible to extract the phases at each scale-time location and therefore to obtain information about the possible delay in the relationship by computing the phase diference. In addition, the Phase Angle Variance (PAV) is a suitable circular statistics to compare the phase spread of multiple time-series (Cazelles and Stone 2003). Within chapters 3 and 4, PAV has been used as an estimation of the synchrony among functional groups and species: the higher is the PAV, the lower is the synchrony because high variance testify of high value dispertion. 16

Figure I.5: Wavelet decomposition of sinusoidal signals. (a) A signal with two periodic component (p1 = 0.25 unit of time (u.t.) and p2 = 1 u.t.) present during the whole series. (b) Fourier spectrum of the signals displayed in a. (c) A signal with two periodic component, but one (p1=0.25 u.t.) is localized on the first half of the series while the other (p2=1 u.t.) is localized on the second half (from ts=5). (d) Fourier spectrum of the signals displayed in c. (e) Wavelet power spectrum of the signal displayed in a ; The colors code for power values are graded from dark blue (low values) to dark red (high values). The black line indicates the cone of influence that delimits the region not influenced by edge effects. (f) Average wavelet power spectrum of the signal in a. (g) Wavelet power spectrum of the signal displayed in b. (h) Average wavelet power spectrum of the signal in b. (Figure extracted from Cazelles et al., 2008)

General introduction 17

Thesis objectives

The main objective of the present Thesis is to explore the patterns of variability of plankton along the northern Atlantic coast of Spain by analyzing the time-series obtained within the RADIALES monitoring program, to investigate the connections between the meteo-climatic and hydrographic environment and plankton fluctuations and finally, to infer, from the plankton population dynamics, how the environment can shape the temporal structure of ecological communities.

Chapter 1 aims to describe the variability in zooplankton aggregated properties (biomass and abundance) in the 9 stations sampled monthly within the framework of the RADIALES monitoring program.

Chapter 2 seeks to explore the changes in zooplankton observed in the Vigo section, presented in Chapter 1, both for the aggregated properties biomass and abundance and taxonomic composition and their relationship with environmental forcing.

Chapter 3 used the same data set than in Chapter 2 but employing a different statistical approach, wavelet analysis, to understand the relationship between zooplankton variability, temporal structure and the fluctuations of environmental forcing.

Chapter 4 seeks to solve some of the ecological questions that emerged in Chapter 3. What is the relationship between environmental forcing and inter-specific synchronicity? For this purpose we combined wavelet analysis of a diatom community and used a model approach to simulate changes in a chemostat system of two species competing for two resources with variable stoichiometry.

19

CHAPTER 1. SEASONAL AND LONG-TERM

VARIABILITY OF MESOZOOPLANKTON ABUNDANCE

AND BIOMASS ALONG THE NORTH IBERIAN

ATLANTIC SHELF

Abstract:

To identify abiotic forcing on zooplankton communities, it is crucial to obtain robust information about their dynamics. Here we used monthly data of zooplankton total biomass and abundance obtained within the monitoring program RADIALES at 9 stations distributed in 4 across-shelf sections along the North-western shelf of Spain. Seasonality accounted for the highest proportion of variability in all the time series. The annual mode was the main component, but semi-annual components were in addition observed in the Cantabrian stations. From the periodicities observed in biomass time-series, three main spatial domains could be distinguished: oceanic, coastal Cantabrian and coastal and shelf Galician stations. In all sections, coastal communities were made up of comparatively smaller organisms. At the decadal scale, the Santander section presented a positive trend in biomass and a negative trend in abundance (significant in the mid-shelf), resulting in an increase in the average individual weight. A Coruña station presented positive and significant trends in biomass and abundance and a decrease in the average individual weight through time. The more conspicuous increase in biomass and abundance was observed in the section off Vigo without changes in individual weight. All the observed trends suggest that local factors may override the large scale effects of ocean warming.

Submitted to Estuarine, Coastal and Shelf Science, authors : Lucie Buttay, Ana Miranda, Gerardo Casas, Antonio Bode, Eneko Aierbe, Bernard Cazelles, Enrique Nogueira, Rafael González-Quirós

Chapter 1 21

INTRODUCTION

Zooplankton represents an important component of the ecosystem, which links primary producers and higher trophic levels and contributes to propagate environmental changes to the whole pelagic system. Zooplankton organisms respond rapidly to environmental forcing or anthropogenic pressures due to their short generation times (Hays et al. 2005), making them good sentinels of environmental changes (Taylor et al. 2002; Perry et al. 2004; Hays et al. 2005). Natural ecosystems are facing increasing threats associated to climate change and other anthropogenic pressures (e.g. fishing and pollution) whose effects can act synergistically to impact marine communities (Möllmann et al. 2008; Planque et al. 2010). Whereas no direct commercial exploitation steers towards zooplankton in Europe, indirect effects of fishing have been already reported (e.g Möllman et al 2008). Similarly, effects on plankton composition due to pollution and the presence of distinct types of contaminants have been encountered (Serranito et al. 2016).

Recent European legislation, such as the EU Water Framework Directive (WFD, 2000/60/EC) and the Marine Strategy Framework Directive (MSFD, 2008/56/EC), focus on protection and restoration of the ecological quality and ecosystem integrity within estuarine, coastal, shelf and offshore systems across Europe. The MSFD explicitly aims to maintain good environmental status (GES) for marine waters, their habitats and resources. The achievement of GES implies that marine resources are used at a sustainable level, ensuring their continuity for future generations. To track the possible departure from the GES, a series of ecosystem indicators, seen as an evaluation and decision tool that allows measuring the state of the ecosystem in regard to a given descriptor, have been defined. Some of the indicators proposed within the MSFD are based on zooplankton, considered at different levels of aggregation, from bulk properties such as total abundance and/or biomass to functional groups or indicator species. There is, therefore, an increasing need for information about zooplankton dynamics to define baseline levels and variability, identify ‘anomalous’ patterns and discern if the observed changes are natural or anthropogenic driven. 22

The more efficient way to identify patterns of temporal variability and understand system dynamics is to proceed to local repetitions of observations (time-series) (Greve et al. 2004; Perry et al. 2004). From this point of view, the RADIALES monitoring program (http://www.seriestemporales-ieo.com) was implemented in Galicia (NW Spain), and extended shortly afterwards to cover other locations along the northern Iberian Atlantic shelf, with a set of oceanographic stations off Santander and Gijón in the Cantabrian Sea (Bode et al., 2012b). It covers an important part of Spanish Atlantic coastal and shelf waters, from Cantabria to Galicia, which correspond to the Spanish North-Atlantic Demarcation within the frame of the MSFD. This area is subject to an intermediate latitude regime but the change in coastal orientation and the differential intensity of hydrodynamic processes make possible the division of the area in different eco-hydrodynamic regions.

The Galician coast, located at the northernmost limit of the Canary Current upwelling system, is characterized by the occurrence of a seasonally varying coastal upwelling. From mid-spring to early autumn, the predominance of northern winds blowing parallel to the coast force the upwelling of sub-surface waters (Fraga 1981; Blanton et al. 1987). In winter, predominant winds roll to the southwest. This situation is accompanied by coastal downwelling and higher precipitation inland which in turn increase river outflow and favour the formation of the western Iberian buoyant plume (WIBP, Peliz et al., 2005). The Iberian Poleward Current (IPC), a slope undercurrent flowing opposite to the general surface circulation pattern, intensifies also during winter months, favouring the formation of sharp across-shelf gradients (Figueiras et al. 2003; Peliz et al. 2005). Along the Cantabrian coast, the main annual input of nutrients to the upper layers is driven by the deep convective mixing occurring in winter (Llope et al. 2007). During summer, when the surface layer is nutrient depleted, the fertilization of the surface layer is related to wind-driven upwelling events prompted by easterlies (Botas et al. 1989). In the eastern Cantabrian Sea, it is also noticeable the supply of nutrients due to outflow from rivers (González-Nuevo and Nogueira 2014).

The purpose of this paper is to review the set of time-series of zooplankton total biomass and total abundance gathered within the RADIALES time-series monitoring program between 1991 and 2010 in 9 oceanographic stations covering the Spanish North Atlantic coastal and shelf waters aiming to: 1) describe the temporal patterns of aggregated Chapter 1 23 zooplankton properties (total abundance and biomass) and their change through time and 2) to assess the spatial patterns across the studied area.

MATERIAL AND METHODS

Sampling strategy

Within the framework of the RADIALES monitoring program, a set of oceanographic stations distributed along the North and Northwest Spanish shelf (Figure I.4) have been sampled monthly for the characterisation of the zooplankton community. The data used in the present work come from 9 stations distributed along 4 sections covering different domains of the Bay of Biscay and Iberian Coast Ecoregion (ICES areas VIIIa and IXa-north): off Santander and Gijón (Cantabrian Sea) and off A Coruña and Vigo (Galician shelf). Routine sampling within the time series monitoring programme started in the different sections in 1991, 2001, 1988 and 1994 respectively. In the present work we used the data until 2011 except for the A Coruña station whose data were available until December 2009. Whereas the methodology has been strictly consistent in time within each section, some differences exist among sections. The more conspicuous methodological difference concerns the sampling protocol for the characterisation of the zooplankton community (Table 1.1).

For instance, while in the Santander section zooplankton sampling has been based on oblique trawling on the upper 50 meters using a Juday-Bogorov net of 250 µm mesh- size, the other sites have been sampled by means of vertical or oblique hauls between the surface and 5 meters above the seafloor (maximum depth 100m) with WP2-type net (Gijón) or a Bongo-type net (Vigo) or Juday-Bogorov net (A Coruña) with 200 µm mesh-size. These differences in the sampling scheme difficult the comparison of biomass or abundance values among sections in absolute terms.

24

Table 1.1 : Details of zooplankton sampling in the time-series monitoring programme RADIALES (http://www.seriestemporales-ieo.com), indicating: name of section; code, location (latitude and longitude), depth at sampling stations and the characteristics of the zooplankton sampling method.

Mesh Haul Latitude Longitude Depth Period Station size type Site name (º) (º) (m) (year) Net type (µm) E2SA 43.500 -3.783 30 06/1991- Juday- Santander E4SA 43.573 -3.783 110 250 Oblique 11/2011 Bogorov E6SA 43.710 -3.783 850 E1GI 43.580 -5.607 31 04/2001- E2GI 43.675 -5.578 108 WP2 200 Vertical 09/2011 Gijón E3GI 43.778 -5.547 160 04/1988- Juday- 200 Oblique A Coruña E2CO 43.422 -8.437 77 12/2009 Bogorov

Vigo E1VI 42.213 -8.850 39 01/1994- Bongo 200 Oblique E3VI 42.142 -8.958 97 10/2011 At each station off Gijón and Vigo, two samples were taken using double nets while at the Santander and A Coruña stations the unique samples were split in two sub-samples with a Motoda box or a Folson splitter (Harris et al., 2000). From one sample, preserved in 4% sodium tetraborate-buffered formaldehyde, sub-samples were taken to count zooplankton organisms under a stereoscopic microscope. At least 400 organisms per sample were counted to estimate total zooplankton abundance per cubic meter (ind·m-3). The second sample was taken to estimate the dry weight biomass as a proxy of total zooplankton biomass per cubic metre. For this purpose, samples were retained in a glass- fibre filter (GF/F) and dried 24h at 60ºC before being weighted and then converted to mg of dry weight per cubic meter (mgDW·m-3). Additionally, we computed the ratio between biomass and abundance in order to quantify the average individual weight (mg·ind-1).

In addition, average surface (i.e. upper 10 m of the water column) temperature from CTD casts (SBE25) (SST, °C), and Chlorophyll-a concentration (Chl-a, g·L-1), from water samples collected with Niskin bottles at ca. 2, 5 and 10 m depth, were analysed at one station of each section (E4SA, E2GI, E2CO and E3VI) to describe the seasonal and along- shelf variability of key habitat variables for zooplankton dynamics. Chl-a concentration was determined from water samples (100 mL) filtered through glass fibre filters (GF/F), Chapter 1 25 extracted in 90% acetone overnight and measured with a spectrofluorometer (Parsons et al., 1984; Neveux and Panouse, 1987).

Data analysis

Seasonality and long-term trends To describe seasonal patterns of zooplankton aggregated properties (abundance and biomass) and Chl-a concentration, we computed, for each month, the median (50th percentile) and the 25th-75th and 5th-95th percentile envelopes to highlight monthly variability. In addition, we computed for each month the mean and the standard deviation of SST. Time-series of annual anomalies of zooplankton total abundance, biomass and average individual weight (biomass-abundance ratio) were computed to depict inter-annual variability. For this purpose, we subtracted the general mean of the time-series to annual means for each location. Trends have been subsequently computed on annual anomalies using Ordinary Least Square (OLS) regression.

Across-shelf zooplankton patterns Spatial variability can be appreciated by comparing seasonal and inter-annual patterns across sites. We examined the across-shelf pattern of the relationship between zooplankton total biomass and abundance in each section. To this aim, taking into account that abundance and biomass are not independent variables, we opted for a type II regression, namely the major axis (MA) regression, in order to describe the linear relationship between biomass and abundance (lmodel2 R package, Legendre, 2014).

Comparison of zooplankton cyclic components across- and along-shelf The differences in zooplankton sampling methods between sections complicate the direct comparison of absolute values of abundance and biomass between sites. To overcome this problem, we focus on the periodic components exhibited by the time-series instead of the actual abundance or biomass values. To describe the periodic components and how they fluctuate in time, we computed the wavelet transformation on each biomass time-series focusing on the period common to all series: from 2001 to 2010. Preliminary, the series were standardized, square root transformed and regularized using the ‘regul’ function from the Pastecs R package (Grosjean and Ibañez 2014) based on the area method (Fox and Brown 1965). Periods greater than 4 years, which can’t be well resolved given the 26 length of the series, were removed before the analysis using a low pass filter (Shumway and Stoffer 2006a). Wavelet transformation performs a local time-scale decomposition of the variance of the signal and allows, therefore, coping with the non-stationary behaviour of the series (Cazelles et al. 2008). We used the Morlet wavelet, a continuous and complex function that enables the extraction of time-dependent amplitude cycles and whose scales are related to frequencies (Ménard et al. 2007; Cazelles et al. 2008). The obtained local Wavelet Power Spectrum (WPS) displays the relative importance of frequencies for each time step. In addition, we also computed the global WPS as the time-average of the local WPS for each frequency component. This summarizes the dominant periodicities of a series (Percival 1995). Significance levels (5%) for WPS were determined through a bootstrapping scheme that used a Hidden Markov Model (Cazelles et al. 2014), being the null hypothesis that the observed time-series patterns were different from those expected by chance alone (Klvana et al. 2004). In addition, we calculated the percentage of variance that corresponded to seasonality in each of the 9 series. Finally, to compare the dynamics between the standardized and normalized series we calculated the dissimilarity between the local WPS of each station based on maximum correlation analysis method (Bretherton et al. 1992; Rouyer et al. 2008). For this purpose, singular value decomposition was performed on the covariance matrix between two local WPS. The distance between each spectrum singular value, obtained in a diagonal matrix, was therefore represented in a dendrogram using the complete linkage method for clustering (Legendre and Legendre 1998).

RESULTS

Patterns of variability of total zooplankton biomass and abundance

Observed values of zooplankton total dry weight in each station were highly variable (Figure 1.1). Part of this variability occurred at seasonal scales, as shown in the monthly quantile values (Figure1.1, right panels). Indeed, in all stations biomass maxima and higher variability was observed from spring to autumn while minima values and lower variability occurred in winter (December to February). There is, however, high variability among stations in the shape of the seasonal pattern. In the Cantabrian sections (Santander and Chapter 1 27

Gijón), all stations presented their maxima abundances in April/May. Additionally, a secondary peak occurred occasionally in September or October.

Figure 1.1: Observed time series of total zooplankton biomass (i.e. dry weight, mgDW·m-3) for the 9 stations (left panel) and seasonal pattern (right panel). The grey dots are the observed values, 5th-95th and 25th-75th percentiles monthly envelope are depicted in light blue and blue respectively and the 50th percentile (median) is the red dotted line.

This bimodality of the seasonal cycle was even more apparent in the station of A Coruña, where nearly equivalent peaks were generally observed around May and September, although the biomass remained relatively high during the summer. In contrast, the spring peak in the Vigo stations rarely occurred and an extended period of high abundance was observed from July to September. The maxima values of biomass recorded in each section were 98.71, 147.28, 80.98 and 251.90 mgDW·m-3 for the sections off Santander, Gijón, A Coruña and Vigo respectively (Table 1.2). It must be recall that absolute values are not strictly comparable due to differences in sampling methods. 28

Table 1.2: Basic statistics of zooplankton time-series. 5th, 50th (median) and 95th percentiles for total zooplankton biomass and abundance.

Biomass (Dry weight, mgDW·m-3) Abundance/1000 (ind·m-3) 5th 50th 95th 5th 50th 95th E2SA 1.82 13.87 58.70 0.18 1.60 6.05 E4SA 2.00 11.63 39.37 0.19 1.08 3.72 E6SA 0.66 7.35 38.25 0.04 0.58 2.22 E1GI 4.61 33.99 96.05 0.94 6.88 17.42 E2GI 3.08 18.34 53.48 0.42 1.99 5.72 E3GI 1.53 17.31 60.14 0.34 1.91 4.80 E2CO 2.97 18.95 63.37 0.28 2.25 7.83 E1VI 2.54 37.42 148.90 0.45 5.09 28.96 E3VI 2.00 29.21 110.50 0.24 2.03 13.85

Figure.1.2: Annual anomaly of biomass (total zooplankton dry weight, mgDW·m-3) for each of the 9 series. Linear regression lines have been added in red, the significant ones are highlighted by the ** symbol. Chapter 1 29

The time-series of biomass (Figure 1.1) exhibited also strong inter-annual variability, which is summarized by the time-series of the annual anomalies of total biomass (Figure 1.2). In the sections located in the Cantabrian Sea (Santander and Gijón), the annual means before 2005 were mainly below the general mean (negative anomalies), while from 2005 onwards there was a dominance of positives anomalies. Indeed, the trends were positive in all stations but significant only for coastal and oceanic stations along the Santander section (E2SA and E6SA, p-values>0.05) with annual rates of change of 0.74 and 0.33 mgDW·m-3·y-1 respectively (Table 1.3).

Figure 1.3: Observed time series of total zooplankton abundance (ind·m-3) for the 9 stations (left panel) and seasonal pattern (right panel). The grey dots are the observed values, 5th-95th and 25th-75th percentiles monthly envelope are depicted in light blue and blue respectively and the 50th percentile (median) is the red dotted line.

In the station of A Coruña, negative and positive annual anomalies alternate every 2- 4 years, but the general trend is positive and significant (p-value<0.05 ; slope= 0.58 mgDW·m-3·y-1). The stations off Vigo presented the more conspicuous long-term increase. Until 2001, all annual means were below the general mean, while from 2002 30 onwards there were only positive anomalies, with the exception of one value in 2004 at E3VI. The trend in both stations was therefore positive and significant (p-value<0.05, for a slope of 3.64 and 2.27 mgDW·m-3·y-1 for E1VI and E3VI respectively).

The time-series of total abundance (Figure 1.3) presented also high variability. Some sporadic peaks of very high abundance (more than 4 times the general mean) occurred more frequently than in the series of total biomass. Yet, the seasonal patterns were similar to those of total biomass: all stations in the Cantabrian Sea and the mid-shelf station of A Coruña presented typically 2 periods of high abundance each year, in spring and autumn, whereas the stations of the Vigo section presented one extended period of high abundance centred in July.

Figure 1.4: Annual anomaly of total zooplankton abundance for each of the 9 series. Linear regression lines have been added in red, the significant ones are highlighted by the ** symbol. Chapter 1 31

Interestingly, in the Cantabrian stations, the long-term variability of abundance shown in Figure 1.4 did not match the positive long-term trends observed for biomass (Figure1.1). Annual anomalies of abundance in Gijon and Santander stations presented mainly slightly negatives trends, which were only statistically significant in the mid-shelf station of the Santander section (E4SA, annual rate o change of -0.05 ind·m-3·y-1). The Galician stations presented positive and significant trends, similarly to those observed for biomass (p-value<0.05, rate of change of 0.14 ind·m-3·y-1 in the A Coruña station and of 0.48 and 0.33 ind·m-3·y-1 for E1VI and E3VI respectively).

Table 1.3 : Long-term trends: Slope and coefficient of determination (R2) of the OLS regression on annual mean of total zooplankton biomass and abundance.

Biomass (Dry weight) Abundance Slope R2 P-Value Slope R2 P-Value E2SA 0.74 0.52 2.10-4 *** -0.01 0.02 0.56 E4SA 0.15 0.06 0.26 -0.05 0.49 6 .10-4 *** E6SA 0.33 0.23 0.03 * 0.002 0.03 0.86 E1GI 0.49 0.04 0.57 -0.04 0.009 0.79 E2GI 1.00 0.36 0.05 -0.06 0.06 0.49 E3GI 1.16 0.12 0.30 0.09 0.16 0.25 E2CO 0.58 0.32 6.10-2 ** 0.14 0.67 3 .10-6 *** E1VI 3.64 0.73 5.10-6 *** 0.48 0.34 0.01 * E3VI 2.27 0.46 2.10-2 ** 0.33 0.28 0.03 *

Biomass-abundance relationship (average individual weight)

The relationship between total biomass and abundance in each site is shown in Figure 1.5. As expected, the relationship was positive and significant in all cases (p-value<0.05). Comparison between sites is made difficult by the use of different sampling methods; however, in each across-shelf section the methods were kept consistent trough time. It is worth noting that the slope of the relationship between biomass and abundance increased with the distance to the coast (sections off Santander, Gijon and Vigo). This indicates that the size structure of the zooplankton community followed a coastal-ocean gradient caused by the relative dominance of small-sized zooplankton in coastal domains relative to those environments with more oceanic influence. 32

The time-series of the annual biomass-abundance ratio allowed us to follow the average individual dry weight (Figure 1.6). In the Santander stations, the annual anomalies were mainly negative before 2005 and positive onwards, showing a statistically significant positive linear trend in coastal and mid-shelf sites (E2SA and E4SA respectively). Along the Gijon section, despite the increase of the biomass-abundance ratio, no significant trends were observed. The station of A Coruña presented an opposite trend to the Cantabrian stations, with a significant negative trend in the ratio biomass-abundance, indicative of a decrease in the average individual weight. In the section of Vigo, no significant trend was observed in any of the sampled stations.

Figure 1.5 Relationship between total zooplankton biomass and total zooplankton abundance. Major axis (red line) and its confidence interval (blue dotted line) have been calculated for each station. Chapter 1 33

Figure 1.6: Annual anomaly of mean individual weight (biomass and abundance ratio) for each of the 9 series. Linear regression lines have been added in red, the significant ones are highlighted by the ** symbol.

Seasonal patterns of SST and Chl-a concentration

The seasonal pattern of Sea Surface Temperature (SST) and chlorophyll-a concentration (Chl-a) at the mid-shelf station of each section (E4SA, E2GI, E2CO and E3VI) are presented in Figure 1.7. In the Cantabrian Sea (E4SA and E2GI), the lowest temperatures were observed in late winter, started to increase in early spring and reached the highest values in summer. Overall, the amplitude of the seasonal cycle is relatively high, ranging from 13 to 21 ºC. In contrast, in the Galician shelf (E2CO and E3VI), SST increased slightly between spring and autumn but remained always below 17 ºC, and thus the amplitude of 34

SST variation throughout the year was much lower. No long-term trends in SST were detected using the data collected at E4SA, E2GI, E2CO and E3VI. An analysis carried out on the NOAA daily Optimum Interpolation Sea Surface Temperature (https://www.ncdc.noaa.gov/oisst) computed from ¼  cells close to the sampled stations revealed that there was a significant increase in SST in the Cantabrian sections (Santander an Gijon) at a rate of 0.04ºC·y-1. In the Galician Stations, the rate was lower (about 0.02ºC·y- 1) and the trend not statistically significant.

Figure 1.7: Seasonal pattern of Chlorophyll-a and temperature in the surface layer (down to 10m depth). The grey dots represent the observed values of Chlorophyll-a, 5th-95th and 25th-75th percentiles monthly envelope are depicted in light green and green respectively and the 50th percentile (median) is the green dotted line. Monthly mean of sea surface temperature (SSTmean) are represented by the red line with their respective confidence interval.

The seasonal pattern of Chl-a also varied significantly between the Cantabrian Sea and Galician shelves. In the mid-shelf off Santander (E4SA), there was a clear peak of Chl-a in April and a secondary one of lower amplitude in October. The Chl-a concentration in the Chapter 1 35 mid-shelf off Gijon (E2GI) presented also a bimodal pattern with the difference that the spring peak is wider, extending from March to May, and the autumn peak start in October but monthly averages remain relatively high until December. On the other hand, in the Galician shelf the average levels of Chl-a concentration were higher than in the Cantabrian Sea, there was no a clear major peak and the average monthly concentration remained relatively high from March to November.

Zooplankton cycles along the northern Iberian shelf

The local Wavelet Power Spectrum (WPS) displays the variability of each biomass time-series analysed in the time and frequency domains (Figure1.8), allowing the distinction of the dominant modes of variation (i.e. period, in the y-axis) and how they changed through time (i-e. time, in the x-axis). All series presented a significant 1-year periodic component in their local and global WPS. This annual component has been continuously significant in E2SA, E4SA, E1GI, E2CO and E1VI, while in E6SA it was only significant until 2004, in E2GI it begun to be significant in 2004 and in E3GI was only significant between 2006 and 2008. Additionally, in E3VI the annual component presented only a short interruption in 2005. 36

Figure 1.8: Wavelet Transformation of all the total zooplankton biomass time series (E2SA, E4SA, E6SA, E1GI, E2GI, E3GI, E2CO, E1VI, E3VI) during the concurrent period 2001-2009. Left panels: Local Wavelet Power Spectrum (WPS); color code for power values is graded from blue (low values) to dark red (high values), and the black line defines the cone of influence below which the information is affected by edge effect. Right panels: the respective global WPS. On both panels, the black dotted lines denote the 5% significance areas determined with a bootstrapping scheme based on HMM (hidden Markov chain model) (Cazelles et al. 2014).

Overall, both local and global WPS revealed that the annual cycle represents the main mode of variation in the Galician stations (E1VI, 65%; E3VI, 58%; E2CO, 57%) and in the coastal stations of the Cantabrian Sea (E2SA, 55%; E4SA, 51%; E1GI, 54%; E2GI, 38%). In the more oceanic stations, seasonality accounted for a much lower proportion of the total variance of the series (E6SA, 24%; E3GI, 18%). In the two stations off Vigo, the annual component was the only significant one, whereas in the other stations there was an intermittent 0.5-year periodic component that was more conspicuous towards the end of Chapter 1 37 the time-series. In E2SA and E4SA, for instance, this half-year mode was significant around 2009 while in E6SA was present in 2005-2006 and then again in 2007-2009. This cycle was also present in the stations off Gijón (around 2006 and 2008 in E2GI and in 2007-2008 in E3GI) and in the station off A Coruña (around 2007-2008 in E2CO). Since the semi-annual component occurred only in some years it scored little power in the global WPS.

The local WPS of each series have been grouped according to their similarity, as showed in the dendrogram (Figure 1.9): the shorter the branch, the higher the similarity between two WPS. The first branch separates coastal stations, characterised by a neat and predominant annual component, from those with a more oceanic character (E3GI and E6SA). Within the first branch of the cluster, we distinguished 2 main groups: one for the Cantabrian coastal stations (E1GI, E2GI and E2SA), and other for the Galician stations and E4SA. Within this later group, the similitude was high between the stations in the Ria de

Vigo (E1VI and E3VI) and also between the mid-shelf stations E2CO and E4SA.

3.5

2.5

1.5

Height

0.5

E1VI E3VI

E2GI E1GI E3GI

E2SA E4SA E6SA E2CO coastal coastal and shelf oceanic Cantabrian Galician Cantabrian stations stations stations

hclust (*, "complete") Figure 1.9: Cluster (k-means on the similarity matrix) build on the local WPS similarities within the cone of influence.

38

DISCUSSION

In the present study, we have analysed time-series of aggregated zooplankton properties, total abundance and biomass, from 9 stations along the Northern Iberian Atlantic shelf to describe their temporal and spatial patterns of variability at different scales. Particular attention was paid to analyse the year-to-year variability in the timing and amplitude of the seasonal signal of zooplankton abundance and biomass. Changes in the phenology of the seasonal signal could be a good indicator of changes in the pelagic ecosystem driven by inter-annual variations in meteo-climatic conditions.

Modes of variation

The wavelet analysis, applied to all the biomass series during their concurrent period 2001-2010, depicted different scales of variability. Seasonality is the main mode of variation, accounting for more than 50% of the total variability of zooplankton biomass in the more costal domains. This contribution is higher in Galicia than in the Cantabrian Sea, particularly in the Ría de Vigo where it reached 65% (E1VI). These results are consistent with previous studies made on shorter periods in some of these series (Valdés et al., 1991; Bode et al., 2009; González-Gil et al., 2015; Buttay et al., 2016). The high contribution of the annual mode to the total variance of zooplankton biomass in the time-series of Galicia (NW Iberian shelf) is due to the modulation exerted by diverse meteo-hydrographic processes occurring in this sub-region and that exhibit strong seasonal dynamics, such as coastal upwelling (Casabela et al. 2014), the influence of the Iberian poleward current (IPC) (Xu et al. 2015) or runoff from the drainage basins (González-Nuevo and Nogueira 2014). Indeed, the influence of these processes is more marked in the southern part of the Galician coast (i.e Rias Baixas, off Vigo) (e.g. Figueiras and Niell, 1987; Nogueira et al. 2000;; Tilstone et al., 2000; Varela et al., 2004) than in the northern Rias Altas (off A Coruña) (Varela and Prego 2003). In accordance with those findings, we have observed that the average seasonal pattern of Chl-a concentration in Galician shelf (E2CO and E3VI) consists in one extended period of high values lasting from March to October, coincident with the upwelling season, thus confirming previous descriptions of the cycle (e.g. Bode et al., 2011). In the Cantabrian coast, seasonality represents also the main mode of variation, accounting Chapter 1 39 between 51 and 55 % of the total variability in the costal sites. The contribution of the annual mode to total variability decreases towards the oceanic edge along the across-shelf section. For instances, in the section off Gijón it represented around 38% in the mid-shelf (E2GI) and only 18 % in the outer shelf (E3GI). Overall, the seasonality described in the present work is obtained from a rather long time-series, the shortest being 10 years long (Gijon section), providing thus consistent information about the patterns and the range of variability. On the other hand, while in the Galician stations and in the coastal site off Santander no significant periodicities other that the annual were found in the total zooplankton biomass series, in most of the Cantabrian series appears also 0.5-year periodicities. This mode of variation is, however, only transitory and significant towards the end of the series. It corresponds to the typical pattern of primary production observed in temperate latitudes, marked by two periods of relatively high abundances coincident with the periods of transient thermoclines in spring and autumn (Winder and Cloern 2010). Indeed, the pattern of Chl-a concentration at the Cantabrian stations (E4SA and E2GI) revealed that even though there are two peaks throughout the year, the autumn peak is weaker and more variable than the spring one. Even the 5th percentile for the month of April is high in comparison with the rest of the year testifying that low values are rare during this period. Zooplankton biomass and abundance also present an important peak in spring in all Cantabrian stations. The autumn peak is, however, less consistent as suggested by the high variability observed in zooplankton abundance and biomass. In fact, the wider 5-95th percentile envelope is found during this period. Previous work carried out in the Cantabrian coast on two abundant copepod species (Calanus helgolandicus and Calanoides carinatus) concurred with our observations and found, in addition of spring maxima in abundances, higher production of eggs associated with the spring phytoplankton bloom (Ceballos and Álvarez-Marqués 2006). Moreover, they revealed that the copepod generations associated with the where the more successful in terms of survival and also that the autumn peak of abundance occurred also occasionally. Those findings are in accordance with the classical functioning of temperate latitude region in which the winter mixing constitute the most effective fertilizing process (Harris, 1986) allowing important phytoplankton blooms that support higher abundance of consumers such as zooplankton organisms. 40

The similarity matrix, build on the pair-wise comparison of the local Wavelet Power Spectra and resumed in the dendrogram, revealed 3 main groups of stations that can be assimilated to distinct eco-hydrodynamic regions, defined as homogeneous geographical domains in terms of their biological characteristics (considered at the level of aggregated properties, communities or species) and their physical structure and dynamics (Scherer et al. 2016): 1) Oceanic locations in the Cantabrian Sea (that may be assimilated to the eco- hydrodynamic typology of ‘offshore stratified waters’); 2) Coastal sites in the Cantabrian Sea (assimilated to ‘regions of freshwater influence’ –ROFIs); and 3) Coastal and shelf locations in Galicia, corresponding to an upwelling ecosystem. Eco-hydrodynamic regions are useful to establish environmental units for management purposes (for instances in relation to WFD and MSFD) to supply reference conditions against which to assess the effect of anthropogenic or natural disturbances. The more oceanic stations, E6SA and E3GI, constitute the first ramification of the cluster. Those stations are indeed the ones whose seasonality accounted for less than a third of the total variability and that presented additional cycles of 0.5 and 2-year periodic components. According to their location near the shelf break, they are more likely to receive oceanic influence. However, previous studies in the Cantabrian Sea revealed that the dynamics of zooplankton at the shelf break is highly governed by the presence of slope water oceanic eddies (Fernández et al., 2004) and fronts (Fernández et al., 1993). Because of the dynamics associated to these processes, samples collected at fixed sampling locations correspond alternatively to neritic and oceanic waters depending of sampling at one or another side of the frontal zone (Albaina and Irigoien 2007). Given the high variability associated to the stations located at the shelf break, it would probably be more difficult in those locations to disentangle the different causes of variability. Despite the logistic difficulties that they would imply, sampling off-shelf, oceanic stations would be preferred when trying to discriminate the effects due to natural variability or anthropogenic disturbances as required within the MSFD. Finally, the coastal and shelf stations were then sorted in two groups: one containing the Cantabrian stations, characterized by the presence of semi-annual periodicities, and another group formed by the Galician stations and the Cantabrian mid-shelf E4SA station. Chapter 1 41

Long-term trends

Seasonality was in average the main mode of variability, but the amplitude and timing of the seasonal cycle varied from year-to-year. Globally, towards the end of the series in the Cantabrian Sea there was an increase in the amplitude of the annual, semi-annual and bi- annual modes of variation. There was an increase in zooplankton biomass in all annual series, although increasing trends were not significant in the shorter series off Gijon and in station E4SA off Santander. The increase in zooplankton biomass in Galicia was already reported in previous analysis of part of these series (Bode et al., 2009; Buttay et al., 2016, 2017).

This result can be surprising because in the present scenario of global warming, primary production is expected to decrease (Sarmiento et al. 1998). This decrease has been observed at the global scale (Gregg et al. 2003) and, intuitively, it would be propagated to zooplankton as consumer of primary production (Chust et al., 2014). At the local scale, however, those patterns are not always observed and if primary production has been shown to decrease in the west of Cape Peñas (off Cudillero, Cantabrian Coast) from 1997 to 2007, an opposite pattern has been encountered off A Coruña, where primary production has increased (Bode et al., 2011). Overall, there are uncertainties concerning the relationship between temperature increase and primary production (Falkowski 1998) and going up on the food web, the sources of uncertainties increase due to the complexity of trophic relationships. In the area off A Coruña, where primary production has been described to increase, there is also evidence that the phytoplankton composition has changed from 1989 to 2008 (Bode et al., 2015). Such changes in phytoplankton compositions are susceptible to influence zooplankton growth and, more generally, the transfer of energy to higher trophic levels (Banse 1995).

Significant positive trends in abundance were found in all the Galician stations and in some of the Cantabrian Sea. It is worth to mention that there were long-term trends in abundance that did not match the long-term evolution in biomass. This is the case in station RSt4 in which abundances decreased significantly while biomass increased, although not significantly. Indeed, when computing the ratio between biomass and abundance, it appears that the mean individual weigh in RSt2 and RSt4 increased significantly. In contrast, it decreased in the station off A Coruña. The increase in mean individual weight represents 42 also a surprising result, as body size reduction has been suggested to be one of the universal responses of organisms facing temperature increase (Gardner et al. 2011). Indeed, in the Cantabrian coast there is some evidence that it already occurred on bacterial communities (Morán et al. 2015). If the decline of body size is a result of trait selection more than phenotypic plasticity alone, the length of our series is insufficient to detect such effects on zooplankton organism whose generation time are much longer than those of bacteria. In addition, in the present work, we used the ratio between total zooplankton biomass and abundance to obtain the mean individual weight but the repartition of biomass by size classes does not necessarily follows a normal distribution and therefore few large organisms are susceptible to impact greatly the mean individual weight values. A proper analysis of zooplankton size spectra may allow a better assessment of those questions.

Overall, the northern Atlantic coast of Spain is subject to heterogeneous oceanographic processes varying in an ample range of spatial/temporal scales, from local/short to regional/decadal, which may blur the effect of temperature increase at the local scale. For example, the more conspicuous change in biomass and abundance took place in the Vigo section, where both aggregated properties showed a significant increase. These series have already been analysed, highlighting the existence of three long-term periods which occurrence was linked to meteo-climatic and hydrographic variability at multiple, interacting temporal and spatial scales including short-term scale of upwelling events, seasonality of upwelling and river outflow and long-term, basin-scale climate indices such as the position of north wall of the Gulf Stream (Buttay et al. 2016, 2017). However, since both biomass and abundance showed parallel increasing trends in the stations of this section, there were not long-term significant changes in the time series of individual weight (biomass:abundance ratio). This fact suggests that despite the observed long-term changes in zooplankton biomass and abundance, the structure of the community remained stable, at least at the level of bulk community properties (i.e. community size spectra).

Chapter 1 43

CONCLUSIONS

Seasonality accounted for the main proportion of zooplankton variability in all the 9 stations of the study area and in addition, a semi-annual component was observed in the Cantabrian shelf, reflecting the occurrence of peaks in spring and occasionally in autumn. In contrast, off Vigo, in the southern part of the Galician shelf, the seasonal cycle presented only one main wider peak from late spring to autumn.

According to the periodicities observed in zooplankton biomass time-series at all scales, three main spatial domains could be distinguished: oceanic (stations: E6SA and E3GI), coastal Cantabrian (E2SA, E1GI and E2GI) and coastal Galician (E1VI, E3VI, E2CO and E4SA). In addition, a significant across-shelf pattern revealed that coastal communities were made up of comparatively smaller organism.

At the decadal scale, while the biomass increased in all the series, diferent patterns were found in abundance and consequently in the organisms average individual weight, suggesting that local factors may override the large scale effects of ocean warming.

Chapter 1 45

Supplementary Information

Figure SI1.1 : Monthly anomalies of Sea Surface Temperature (in blue). The red line corresponds to the long-term trends (the dotted line depict the trends that were not significant). SST data used in the present analysis correspond to the NOAA 1/4° daily Optimum Interpolation Sea Surface Temperature (https://www.ncdc.noaa.gov/oisst) that combine observations from different platforms (satellites, ships, buoys) on a regular global grid. Four locations where selected near each of our stations: Santander (3.875°W ; 43.625°N), Gijon (5.625°W ; 43.625°N), A Coruña (8.375°W ; 43.375°N) and Vigo (8.875°W; 42.125°N) and where averaged per months.

To assess the long term trend in SST, generalized linear models (GLMs) were fitted to the time series of monthly SST anomalies during the period 1991-2011 using the GLM function from the nlme R package (Pinheiro et al. 2012).

푆푆푇퐴푛표푚 = 푡푖푚푒 + 휀 ; 퐴푅푀퐴(푝 = 1, 푞 = 1)

The fixed term corresponded to time and the random part of the model included an autoregressive-moving average (ARMA) model to correct for autocorrelation in the time series (Zuur et al. 2009; Pinheiro et al. 2012).

47

CHAPTER 2. LONG-TERM AND SEASONAL

ZOOPLANKTON DYNAMICS IN THE NORTHWEST

IBERIAN SHELF AND ITS RELATIONSHIP WITH METEO-

CLIMATIC AND HYDROGRAPHIC VARIABILITY.

Abstract: Long-term and seasonal dynamics of zooplankton, in terms of abundance and taxonomic composition, and its relationship with meteo-climatic and hydrographic factors were investigated on the northwest Iberian shelf. Zooplankton samples were collected monthly (1995–2011) at two locations within and off the Ría of Vigo (Stations E1VI and

E3VI, respectively). Total abundance of zooplankton (NZT) varied annually following on average a unimodal cycle, with peak values between late spring and early autumn. In the long term, the time series of NZT exhibited three contrasting periods: A) 1995–2001, characterized by low abundance and low amplitude seasonality, with a stepped increase towards 2001; B) 2001–2006, of high abundance and marked seasonality enclosing the maximum values of the time series; and C) 2006–2010, of intermediate abundance and amplitude of the seasonal cycle. The most common taxonomic groups showed similar long- term and seasonal patterns. Principal component analysis revealed a shift in zooplankton dynamics from 2001 onwards, which affected annual averages and seasonality of all taxa. The observed changes in zooplankton dynamics were concomitant with sustained trends for upwelling intensity (increasing), precipitation (decreasing) and Gulf Stream North Wall position (equatorward) between 2000 and 2005. The results stress the importance of hydrodynamics, driven by meteo-climatic conditions, in the control of the abundance levels of zooplankton at seasonal and long-term interannual scales.

Journal of Plankton Research (2016) 38(1): 106–121. doi:10.1093/plankt/fbv100, authors: Lucie Buttay, Ana Miranda, Gerardo Casas, Rafael González-Quirós and Enrique Nogueira

Chapter 2 49

INTRODUCTION

Zooplankton responses to environmental variability bring about fundamental changes in the general dynamics of marine ecosystems (Banse 1995), causing fluctuation in primary production and other environmental conditions to propagate to higher trophic levels. Zooplankton dynamics are affected by processes occurring over a wide range of spatial and temporal scales. At local and short-term scales, for instance, life history traits of zooplankton, such as growth, fecundity and survival, are directly affected by environmental factors such as temperature and the quantity/quality of nutritional resources (e.g. Gillooly 2000; Ceballos and Ianora 2003; Devreker et al. 2005). Mesoscale physical processes, such as currents, frontal structures, buoyancy-driven plumes or coastal upwelling, govern the dispersion or accumulation of individuals and the availability of nutrients (e.g. Garçon et al. 2001; Gonzalez-Gil et al. 2015). At a global scale, zooplankton dynamics, manifested in terms of biomass, species composition, diversity, size structure or phenology, are significantly correlated with large-scale modes of climate variability such as the North Atlantic Oscillation (NAO), El Niño Southern Oscillation (ENSO) or the Pacific Decadal Oscillation (PDO) (Fromentin and Planque 1996; Mantua et al. 1997; Chiba and Saino 2003; Greene et al. 2003; McGowan et al. 2003; Beaugrand and Reid 2012; Drinkwater et al. 2013). Environmental variability is to a large extent rapidly mirrored in zooplankton population dynamics due to their short life cycles (Hays et al. 2005) and for that reason, these organisms have been suggested as indicators of climatic variability (Taylor et al. 2002; Perry et al. 2004; Hays et al. 2005). These studies have also shown the crucial importance of long-term time series in order to detect and resolve environmental effects on biological systems.

Here, we investigated the links between zooplankton dynamics and environmental variability by analysing a long-term, monthly zooplankton time series collected within and off the Ría of Vigo, the southernmost part of the four coastal embayments that form the Rías Baixas system of Galicia (NW Iberian Peninsula). The Galician Rías are located in the northernmost limit of the Canary Upwelling System, and consequently, coastal upwelling has a major influence on their hydrodynamic behaviour (Blanton et al. 1987; Prego and Fraga 1992; Alvarez-Salgado et al. 1993). The Galician coastal and shelf regions are a highly 50 productive ecosystem (Alvarez-Salgado et al. 1993) in which the exploitation of marine resources and aquaculture represent important commercial activities.

From late spring to early autumn, the intensification of the Azores high-pressure atmospheric cell promotes the predominance of northerly winds along the Western Iberian coast, causing the offshore displacement of surface waters and the upwelling of sub-surface waters of the subtropical branch of the Eastern North Atlantic Central Water (ENACWst) (Fraga 1981; Blanton et al. 1987; Nogueira et al. 1997). Nutrient enrichment is further enhanced within the Rías, partially mixed estuaries with a two-layered positive residual circulation pattern (Bowden 1975). During autumn and winter, the weakening and southward displacement of the Azores high causes the predominance of south and south- westerly winds that promote coastal downwelling episodes and the slowdown or even reversal of the estuarine circulation pattern (Villacieros-Robineau et al. 2013). Besides, the intensification of the Iberian Poleward Current (IPC) during this part of the year, a slope northward flow that transports saltier and warmer (subtropical) waters (Peliz et al. 2005), causes the decrease in across-shelf exchange processes.

Previous zooplankton studies on the Northwest Iberian Shelf have mostly been short term, focusing on the understanding the role of mesoscale and sub-regional processes (e.g. upwelling or the IPC) on zooplankton dynamics over the shelf or within the Rías. These investigations were based on observations at fixed locations, i.e. Eulerian approach (Valdes et al. 1990; Fusté and Gili 1991; Tenore et al. 1995; Blanco-Bercial et al. 2006; Ospina- Alvarez et al. 2010) or on Lagrangian sampling following drifting water masses (Halvorsen et al. 2001; Joint et al. 2001; Batten et al. 2001; Riser et al. 2001; Isla and Anadon 2004).

Seasonal dynamics of zooplankton in the region has been documented using the monthly zooplankton sampling carried out by the Instituto Español de Oceanografía (IEO) within the frame of the time series monitoring programme RADIALES off A Coruña and Vigo. Long-term changes and the role of meteo-climatic and hydrographic drivers at these scales have received, however, less attention. More recently, Bode et al. (2009), using data from the Continuous Plankton Recorder and the RADIALES programme, detected a decreasing trend in zooplankton total biomass and copepod abundance in oceanic realms and an increasing trend in coastal areas, and related this to a decrease in upwelling intensity. Also using CPR data, although covering a wider zone, Nogueira et al. (2012) explored the spatial Chapter 2 51 and temporal patterns of copepod species diversity over the North-East Atlantic. Both studies covered identical periods (1958–2006) and did not consider other zooplankton taxonomic groups.

In the present study, we aimed to: (i) describe long-term and seasonal variability of zooplankton abundance and taxonomic composition in the southernmost part of the Galician upwelling system from 1995 to 2010; and (ii) explore the physical–biological coupling between the observed temporal zooplankton dynamics and environmental meteo-climatic and hydrographic drivers.

METHOD

Study area and data collection

Figure 2.1: Map of the study area showing the position of the two zooplankton sampling stations: Station E1VI within the ría and station E3VI off the mouth.

Zooplankton samples were collected monthly between 1995 and 2010 within the framework of the ongoing time series monitoring programme RADIALES (http://www.seriestemporales-ieo.com) at two oceanographic stations (Figure 2.1): within 52 the Ría of Vigo (Station E1VI, at 42.2138°N, 8.8508°W over the ca. 40 m isobath) and on the mid-shelf off the Ría of Vigo (Station E3VI at 42.1428°N, 8.9588°W over the ca. 100 m isobath). Zooplankton were sampled by means of double-oblique hauls, using a double 40 cm diameter Bongo net with 200 mm mesh size. During the haul, a fixed length of cable is realeased (100 and 50m in E3VI and E1VI respectively), controlling that the angle of the cable during the trawl is maintained at 45°. The depth attained during the haul is recorded with a TD (temperature-depth sensor) and the volume of water filtered is determined with flowmeters allocated at the mouth of each of the nets. The sample from one cod-end of the net was preserved in 4% sodium tetraborate-buffered formaldehyde (ind·m-3). Subsamples were taken until at least 1000 zooplankton organisms per sample were identified to the lowest possible taxonomic level under a stereomicroscope and counts in the subsamples were converted to full-sample number by cubic metre for both total abundance (NZT) and abundance per individual taxa. In this study, the analysis was restricted to the temporal dynamics of the major taxonomic groups of zooplankton (Table 2.1). Sampling and sample processing methods were consistent throughout the time series, and zooplankton identifications were always carried out by the same expert taxonomist. Environmental variables previously suggested to play a significant role on zooplankton dynamics were also considered ( 2.2). The water column temperature standard deviation (ºC), calculated from monthly CTD profiles carried out at the mid-shelf station (E3VI), was used as an index of thermal stratification. Daily values of local precipitation (mm) and atmospheric temperatura (ºC) were obtained from the Vigo airport meteorological station (Spanish “Agencia Estatal de Meteorología”). Daily outflow (m3·s-1) from the nearest large river, the Miño, was provided by the Confederación Hidrográfica Miño-Sil. The upwelling index (m3·s-1·km1) was calculated from geostrophic winds according to the method proposed by Bakun (Bakun 1973) for a point located at 43.88°N, 11.88°W, which is considered representative of the occurrence and intensity of coastal upwelling along the western Galician coast (Lavín et al. 1991). Given its localization, the Galician coast is susceptible to be influenced by the the North Atlantic Oscillation (NAO) (Ottersen et al. 2001), the Eastern Atlantic (EA) pattern, which is a mode of low frequency variability generally associated with temperature and precipitation patterns (Wallace and Gutzler 1981), the anomalies of the North Hemisphere Ocean Temperature (NHOT, Beaugrand and Reid 2012) and the Gulf Stream North wall position (Gulf StreamNWP, Taylor et al. 1980). Chapter 2 53

The study region is also susceptible to be affected by the Atlantic Multidecadal Oscillation which has a 60-year period of fluctuation (Edwards et al. 2013b). From 1994 to 2011, it only corresponded to a continuous increase, and therefore, it has not been included in the environmental data set.

54

Table 2.1 : Taxonomic groups observed in stations E1VI (middle of the ría) and E3VI (mid-shelf off the ría) that are present in at least 50% of the samples. Those groups encountered in less than 50% of the samples were pooled in the “Others” category. For each station, the table reports: taxonomic group name, average relative abundance, the mean, standard deviation and maximum concentration (ind·m-3), and the percentage of samples in which the taxonomic group was present.

% of samples Maximum with Relative Mean Standard concentration concentration Taxon Station abundance concentration deviation observed >0 Copepoda 59,45 4391,09 4889,10 32380,58 100,00 Larvacea 11,11 912,38 1283,68 8057,30 96,71 Cirripedia larvae 7,39 695,43 1622,66 16678,00 92,49 Cladocera 5,86 631,00 1102,06 7308,89 82,63 Echinodermata larvae 3,28 465,39 1502,69 11007,48 64,79 Siphonophorae 3,08 264,67 408,93 2299,77 76,99 Cnidaria 2,01 226,88 408,17 3636,75 79,81 E1VI Others 1,59 194,19 752,08 7251,86 67,60 Bivalvia larvae 1,36 207,90 683,46 8467,33 65,73 Decapoda larvae 1,16 81,64 186,75 2007,78 82,16 Gastropoda larvae 1,12 131,65 269,41 2648,19 78,40 1,05 71,70 139,47 765,28 65,73 Euphausiacea 0,68 48,83 162,10 1800,19 52,11 Polychaeta 0,51 57,36 290,08 4146,39 63,38 Bryozoa larvae 0,35 26,05 55,11 430,23 51,64 Copepoda 71,59 2596,60 3404,39 23173,93 100.00 Larvacea 5,37 216,10 387,99 3472,59 88.27 Cirripedia larvae 4,37 167,67 398,26 2628,00 79.33 Echinodermata larvae 3,12 301,38 1444,08 16919,66 61.45 Siphonophorae 3,06 133,41 275,24 2212,74 79.89 Euphausiacea 2,55 104,88 301,73 2949,00 79.89

Others E3VI 2,37 132,17 359,33 3049,20 79.89 Cladocera 1,99 119,46 308,73 3029,27 53.63 Bivalvia larvae 1,10 121,35 517,88 4517,67 56.42 Decapoda larvae 1,04 22,17 37,72 276,85 73.74 Cnidaria 0,93 55,82 160,25 1659,55 62.56 Gastropoda larvae 0,84 67,14 325,99 4016,93 72.62 Chaetognatha 0,83 36,04 130,97 1567,35 60.89 Foraminifera 0,82 38,28 147,28 1659,55 55.30 Chapter 2 55

Table 2.2 : List of the environmental variables (meteo-climatic and hydrographic) that were considered as potential drivers of zooplankton dynamics

Variable name Units Description Sampling Missing Institution responsible of frequenc monthly data y values acquisition/provision Stratification ºC Standard deviation of water Monthly 29 Instituto Español de column temperature based on Oceanografía

CTD measurement at station E3VI Sea Surface ºC Temperature within the 10 Monthly 29 Temperature superficial meters from CTD measurement at station E3VI River Miño m3.s-1 Miño outflow at the station Daily 0 Confederacion del niño-Sil Outflow 1641 (42.15N 8.19W) (http://saih.chminosil.es/) Upwelling m3·s-1 Average Ekman transport Daily 1 http://www.indicedeaflora Index km-1 miento.ieo.es/ Precipitation mm·day-1 Local precipitation at the Monthly 0 Agencia Estatal de meteorological station in the Meteorología Vigo’s airport (www.aemet.es) Atmospheric ºC Local temperature at the Monthly 0 Temperature meteorological station in the Vigo’s airport Gulf Stream - Anomaly of the position of the Monthly 0 http://web.pml.ac.uk/gulfs north wall Gulf Stream North Wall tream/data.htm position (Gulf StreamNWP) East Atlantic - North–south dipole of anomaly Monthly 0 NOAA, Climate Prediction (EA) pattern centres spanning the North Center Atlantic from east to west http://www.cpc.noaa.gov/ data/teledoc/ea.shtml Winter and - Atmospheric sea level pressure Quarterly 0 University of East Anglia, Summer North anomaly: difference in the Climatic Research Unit: Atlantic normalized sea level pressure http://www.cru.uea.ac.uk/ between Iceland and the Oscillation cru/data/nao/ Azores. NAO indices averaged, (NAOWinter respectively, over December– and February (Winter) or June- NAOSummer) August (Summer) Northern ºC Temperature anomalies relative Monthly 0 https://www.ncdc.noaa.go Hemisphere to the 20th century average v/monitoring- Ocean (1901-2000) references/faq/anomalies. Temperature php (NHOT) Anomalies 56

Numerical analysis

All time series were first explored for the presence of outliers, normality, homoscedasticity and co-linearity among variables following the protocol proposed by Zuur et al.(2010). Daily time series of ancillary climatic, meteorological and hydrographic data (Table 2.2) were monthly averaged. The cumulative sum of the mean distances method (Ibañez et al. 1993) was applied to detect local trends and abrupt changes in the time series. This technique removes from each temporal observation the general mean of the time series and plots the cumulative sum of residuals. The interpretation is based on the linear slope of the cumulative sum (CS) function, since persistent departures from the general mean of the time series cause a persistent change of the slope. The CS function shows a positive (negative) slope in periods for which the mean of the data is higher (lower) than the general mean of the original time series.

To assess the main patterns of variability, including an objective determination of the timing of the long-term periods detected by the CS function and the effect of sampled locations (i.e. within and off the ría), a set of generalized linear models (GLMs) were fitted to the time series of NZT using the GLM function from the nlme R package (Pinheiro et al. 2012). The fixed term included the effect of location (stations E1VI and E3VI), seasonality (month as factor) and long-term periods as identified by the cumulative sum method. The set of alternative models (listed in Table 2.3) always included the effect of month, as seasonality was apparent in the time series of abundance at both locations; testing the set of alternative models excluding the seasonal effect was considered unnecessary. In the model, the interaction between ‘month and period’ accounts for differences in the seasonal cycle between periods. Likewise, the interaction between ‘month and station’ accounts for seasonal differences between locations.

Chapter 2 57

Table 2.3 : Model selection. Akaike criterion for the corresponding fixed parts tested and the residual standard error.

Fixed part AIC Residual standard error: Month 6701.916 1371.8 Month +Period 6678.291 1379.933 Month*Period 6299.262 1247.133 Month +Period+Station 6650.909 1235.462 Month +Period*Station 6467.725 1167.32 Month*Period+Station 6266.811 1247.51 Period*Month+Station*Month+Station*Period 6060.912 1092.942 Period*Month+Station *Period 6241.245 1312.892 Month*Station*Period 5679.144 1091.597 The random part of the model included a parameterization of the variance against the different months and an autoregressive-moving average (ARMA) model to correct for autocorrelation in the time series (Zuur et al. 2009; Pinheiro et al. 2012):

퐴푏푢푛푑푎푛푐푒푇표푡푎푙 = 퐹푖푥푒푑 푝푎푟푡 + 푎푗푏 + 휀 2 푎푗~푁(0, 휎푗 ), 푗 = 1, … ,12 퐴푅푀퐴(푝 = 1, 푞 = 1)

Model selection was based on the Akaike information criterion (AIC), considering the best model the one which presents the lowest AIC value (Burnham and Anderson 2002). Then, the best model structure was used to improve the estimation of the transition date between long-term periods estimated initially by means of the cumulative sum plot. We therefore tested each posible combination of transition dates between periods and used

the AIC to select the best of them. Predicted and standard error of NZT as a function of month, long-term period and location (station) were extracted using the predictSE function from the AICcmodavg R package (Mazerolle 2013).

Plots of the year-to-year variability of the annual cycle for the six major taxonomic groups (in terms of abundance) were plotted using the “interp” function from the Akima R packages (Akima et al. 2013) that allowed linear interpolation of the abundance between months and years. 58

Coupling between zooplankton dynamics and environmental variability

Principal component analysis (PCA) is widely used to identify coherent patterns of temporal variability (e.g. Storch and Zwiers 1999; Hare and Mantua 2000; Mantua 2004; Molinero et al. 2008; Vandromme et al. 2011; García-Comas et al. 2011) by reorganizing the total variance of the data by means of an orthogonal linear transformation. The first component is the one that retains the highest amount of total variance. To assess and compare the main components of temporal variability on plankton and environmental variables, PCAs were done separately on the abundance data set of zooplankton groups (present in more than 50% of the samples) for stations E1VI and E3VI and on the environmental variables (Mantua 2004). Prior to PCA, data were averaged per year after substituting the missing monthly values by the mean of all values for that given month. The resulting annual time series were then standardized due to their different units of measurement and zooplankton data were log-transformed. For physical variables that were highly correlated (r > 0.90), only one variable was included in the analysis to avoid excessive strength of principal components due to redundant information. The selected variables are presented in Supplementary Information, Figure SI2.2.

RESULTS

Zooplankton abundance

Average zooplankton abundance (NZT) between 1995 and 2011 was twice higher within the Ría of Vigo than on the mid-shelf when expressed as a concentration (8421.3 ind·m-3, SD= 9259.2 ind·m-3 and 4101.6 ind·m-3, SD = 5942.9 ind·m-3 at Station E1VI and Station E3VI, respectively). When expressed in surface units, both stations showed similar averaged integrated abundance (3.28x105 and 3.98x105 ind·m-2 for stations E1VI and E3VI, respectively). Anyhow, the time series exhibited large fluctuations around their respective general means, with minimum and maximum monthly values ranging between 85.9 and 41 813.9 ind·m-3 at station E1VI and between 21.9 and 36 878.9 ind·m-3 at E3VI (Figure 2.2). Most of this variability was related to a conspicuous seasonal signal, although largely variable in its amplitude. Winter minima varied within the same order of magnitude across the series, but summer maxima showed large differences. Chapter 2 59

Cumulative sums of NZT (CS, Figure 2.2) identified three long-term periods in the time series, which were approximately coincident at both locations: (i) from the beginning of the series in 1995 until 2001, the CS decreased monotonically, indicating that monthly abundances were always lower than the general mean; (ii) between 2002 and 2006, in contrast, the CS showed an increasing trend (positive slope of the CS), outlining a period with the average abundance above the general mean; (iii) after 2006, the slope of the CS was close to zero, indicating that the average abundance during this period was similar to the general mean. In addition, the saw-toothed aspect of the CS function in the second and third periods captured the higher amplitude of the seasonal cycle.

Figure 2.2: Total zooplankton abundance (grey dots and grey lines) and cumulative sums (CS) of the mean distances (black line) at station E1VI (A) and E3VI (B) between 1995 and 2011.

The best GLM fitted to assess the main patterns of NZT variability included a triple interaction between location, seasonality and long-term period (Table 2.3). The best AIC was found for transitions between long-term periods in January 2001 (from period A to B) and in January 2006 (from period B to C). The parameters estimated by the model (Figure 2.3) confirmed the average seasonal pattern, characterized by a unimodal cycle with 60

minima NZT in winter and maxima NZT in summer. It further depicted that the amplitude of the seasonal signal was lower during period A, higher in B and intermediate in C in both locations. At station E1VI, there was a higher NZT than at E3VI, but this fact has to be taken carefully because the depth of sampling, and hence the volume of water filtered, differed between sites and the vertical distribution of plankton in the water column is heterogeneous.

Figure 2.3: GLMM estimated values and standard error for station E1VI (A) and E3VI (B) for the different long-term periods. Period A: 1995-2001(light grey line), period B: 2001-2006(medium grey line) and period C: 2006-2011 (dark grey).

Zooplankton composition

Taxonomic groups that were present in more than 50% of the samples are listed in Table 2.1 Groups with lower occurrences were pooled in the category “Others”, which included Amphipoda, Ascidiacea larvae, Branchiostomidae larvae, Crustacean larvae, Doliolidae, Isopoda, Mysida, Ostracoda, Gastropoda, Salpidae and Radiolaria at both stations, and, additionally, Foraminifera at station E1VI and Bryozoan larvae and Polychaeta at station E3VI. Details of the seasonal composition of the meroplankton groups are presented in Supplementary Information, Figure SI2.1. Chapter 2 61

Figure 2.4: Seasonal variability of zooplankton abundance composition expressed as the monthly relative abundance of the major zooplankton groups for station E1VI (A) and station E3VI (B). The composition of the meroplankton group is detailed in Supplementary Information (Figure SI2.1).

Copepoda largely dominated the zooplankton community, representing on average around 59% of total abundance within the Ría of Vigo and 72% on the mid-shelf (Table 2.1). Copepoda was the only group present in 100% of the samples at both locations. However, its contribution to NZT varied seasonally (Figure 2.4), rising in December to 91.6 and 79.6% of the total abundance and decreasing in summer to 40.3 and 52.3%, for station E1VI and E3VI respectively. The lower relative contribution of Copepoda during the summer was due to the increase in the relative abundance of other taxonomic groups, namely Larvacea, Cirripedia larvae, Echinodermata larvae, Siphonophorae and Cladocera. These five groups plus Copepoda represented on average 87% of NZT in both stations. The interannual variability of the annual cycle of these six taxonomic groups at each station during the 17- year time series is plotted in Figure 2.5. Abundance increased for all these groups from late spring to early autumn, although the width of the anual period of relatively higher abundance differed between groups. For Cladocera, Echinodermata larvae and Siphonophora, the seasonal period of higher abundances extended for 5–6 months, whereas it appeared longer for Copepoda, Larvacea and Cirripedia (8–9 months). In general 62 terms, for a given taxonomic group, the duration of the period of relatively high seasonal abundance was shorter on the mid-shelf than within the Ría. The clearest case is for Cladocera, for which the difference between both sites was 2 months. The period of maximal abundance in copepods occurred later during period B than during periods A and C. For the long-term dynamics, there was an apparent increase in abundance around 2002 which affected all these six major taxonomic groups at both locations. It is also worth noting that no consistent changes in phenology were detected during the period analysed in any of these groups.

-3 Figure 2.5: Interannual variability of the 6 main zooplankton groups abundance (log10 ind·m ). 64

Coupling between zooplankton dynamics and environmental variability

The first principal component of the data set of physical variables (PC1PHY) retained 27.5% of the total variance and was within the range of previous studies (Hare and Mantua, 2000; Garcı´a-Comas et al., 2011; Vandromme et al., 2011). The time series of PC1PHY scores (Figure 2.6, A) showed low values from 1995 to 2002 (except for 1998), followed by positive values from 2002 to 2010 (except for 2006). The transition between these two periods corresponds to a monotonic increase from 2000 to 2005. The PC2PHY (Supplementary Information, Figure SI2.2) represents 22.2% of the total variance but has no clear pattern of variability.

Figure 2.6: Time series of the principal components derived from PCA. PC1 for physical variables (A); PC1 for log-transformed taxonomic composition in station E1VI (B) and in station E3VI (C). Loadings of the physical variables for PC1(D) (loadings superior to 0.30 coloured in black); loadings of PC1for taxonomic composition in station E1VI (E) and in station E3VI (F).

The physical variables with highest contribution to PC1PHY (|loading| > 0.30) were precipitation and Gulf Stream north wall position (Gulf StreamNWP), with negative loadings, and upwelling index (UI), with a positive loading (Figure 2.6, B). The annual means for these variables are shown in Figure 2.7. Within the high variability of these three series, the transition period depicted in the PC1PHY is observable and has been marked by an arrow.

From 2000 to 2005, both precipitation and Gulf StreamNWP had a decreasing trend, whereas Chapter 2 65

UI showed an increasing trend (Figure 2.7). The first principal components obtained from the zooplankton abundance data sets (i.e. within and off the Ría of Vigo, PC1ZOO_1 and

PC1ZOO_3, respectively) retained a similar percentage of their respective total variances

(around 63%). From 1995 to 2001 for Station E1VI, and to 2000 for Station E3VI, PC1ZOO only presented negative scores, whereas they were positive afterwards (Figure 2.6, C and E). Thus, two clear periods appeared at both sites, although it was relatively more variable at Station E3VI. All taxonomic groups showed positive loadings (Figure 2.6, D and F).

Figure 2.7: Annual time-series of (A) Precipitation (mm·days-1), (B) Upwelling Index (UI, m3·s-1 km-1) and (C) Gulf Stream North Wall Position (Gulf StreamNWP).

The period between 2000 and 2005, which corresponded with an increasing period of the first principal component of the physical variables (PC1PHY), matched the positive period observed for the first principal component of the zooplankton variables (PC1ZOO), more marked for station E1VI (Figure 2.6, A C and E); the correlation between time series of PC1PHY and PC1ZOO scores was 0.56 and 0.36 for Station E1VI and Station E3VI, respectively.

66

DISCUSSION

In this study, we focused on zooplankton dynamics at long-term and seasonal scales within the Ría de Vigo and off its mouth, both representative locations of inner and mid- shelf domains of the Galician sub-region, in the northern limit of the Canary Current Upwelling System (Arístegui et al. 2009). In contrast to most previous zooplankton studies in Galician waters, that focused mainly in short-term processes (e.g. Batten et al., 2001; Blanco-Bercial et al., 2006; Bode et al., 2003, 2005; Fusté and Gili, 1991; Halvorsen et al., 2001; Isla and Anadon, 2004; Riser et al., 2001; Roura et al., 2013; Tenore et al., 1995; Valdes et al., 1990), the analyses of 17 years of monthly time series of zooplankton abundance and composition allowed us to describe the average and year-to-year variability of the seasonal cycle. The analysis provides strong evidence of the occurrence of abrupt changes of zooplankton abundance and community composition that may be linked to meteo-climatic and hydrographic factors.

Zooplankton seasonal and inter-annual variability

Total zooplankton abundance varied seasonally by four orders of magnitude and exhibited, on average, a similar unimodal seasonal cycle at both locations, characterized by maxima from May to September and minima in winter. A similar unimodal pattern of abundance has been observed in the Oregon coastal upwelling system where it is well established that high zooplankton abundance occurs from May to September matching the seasonal upwelling pattern (Peterson and Keister 2003; Hooff and Peterson 2006). The zooplankton community is largely dominated by copepods even if the contribution of other groups is higher at Station E1VI mostly because of the influence of meroplankton groups near coastal sites.

Previous studies on the northern coast of Spain already described the major contribution, in relative terms, of copepods offshore (Valdés and Moral 1998; Blanco- Bercial et al. 2006). A previous study on the Cantabrian Sea shelf reported 68% of copepods at a coastal station and 83% at an offshore station (Valdés and Moral 1998). A recent work carried out in coastal and mid-shelf waters off Vigo at high spatial resolution also reported the increased contribution of meroplankton close to the coast and proposed the ratio Chapter 2 67 between holoplankton and meroplankton as an indicator of coastal influences (Roura et al. 2013). The relative contribution of copepods and of the remaining major taxonomic groups also varied seasonally. Larvacea and meroplankton groups, such as Cirripedia larvae and Echinodermata larvae, were particularly abundant during the summer, decreasing strongly in winter, eventually below detectable limits. A previous study (Valdés et al. 1990) already described the major, but seasonally varying contribution of copepods to the zooplankton community along the whole Galician coast (60% in June 1994 and above 90% in September 1994).

The duration of the annual peak of the five main taxonomic groups differed between locations, resulting in shorter periods of high abundance on the mid-shelf than at the inner- shelf locations. As reported in previous studies in the north Atlantic, zooplankton main groups showed high interannual variability in the timing of the seasonal cycle (Atkinson et al. 2015). Although the copepod group seemed to experience a delay in their period of maximal abundance, the method employed did not allow consistent trends in the timing of other taxonomic groups to be detected.

Here, we summarized the observations of 17 years of monthly sampling providing a more robust description of the seasonal variation of total zooplankton abundance. Long- term observations of zooplankton are necessary to establish reference values of abundance, seasonality and composition. Besides, within the frame of the Marine Strategy Framework Directive (European Commission 2008), these long-term records are necessary to formulate indicators to assess the environmental status of marine ecosystems. The proposed indicators are based on abundance or biomass of the total zooplankton community or of selected taxonomic or functional groups such as the ones analysed in the present work.

Long-term changes

The more conspicuous long-term change observed implied a significant step increase in the average abundance and amplitude of the seasonal signal. It is worth noting that in November 2002, a major oil spill (63 000 tof heavy crude) due to the shipwreck of the oil tanker Prestige, occurred close to the sampled area. Two studies were conducted to evaluate the impact of the accident on the plankton community. The first one (Salas et al. 68

2006) analysed the aliphatic hydrocarbons in zooplankton several months after the oil spill, whereas the second study used the monthly time series of the programme Radiales from 1995 to 2003 in order to compare zooplankton biomass and abundances before and after the accident (Varela et al., 2006). Both studies concluded that the effect of the oil spill on zooplankton was imperceptible. Varela et al. (2006), however, detected an increase in zooplankton after the accident, but they rejected the hypothesis of association between the oil spill and the increase in zooplankton abundance. With the wider perspective given by 8 additional years in the time series and the use of GLM, we were able to identify that the Prestige oil spill occurred after the start of the increasing trend in zooplankton abundance that occurred in 2001. We were also able to detect a second change in 2006 when abundances decreased slightly but remained higher than during the first of the long- term local periods detected. The differences observed in annual average abundance between periods were the consequence of differences in the general mean and in the amplitude of the seasonal signal. Although summer maxima show large differences, they can be the result of comparatively smaller fluctuations of the overwintering stock, which through exponential population growth lead to important changes in population abundance (Colebrook 1985). Maxima abundances are therefore very sensitive to the conditions experienced by the overwintering population.

The abundance of all the main taxonomic groups followed similar long-term patterns as revealed by the PCA. Although all taxonomic groups did not increase simultaneously and responded with different magnitude, they all had positive, similar scores on the first PCs

(PC1ZOO_1 and PC1ZOO_3). This suggests that the drivers behind these long-term changes can affect groups with contrasting ecophysiological traits (e.g. meroplankton, holoplankton or gelatinous plankton), but without causing apparent changes in the composition of the community that may cause a change in the functional dynamics of the plankton food web, as it has been suggested in other cases classified as regime shifts (e.g Beaugrand 2004).

Coupling between zooplankton dynamics and environmental variability

Concomitant with the abrupt shifts in zooplankton abundance (total and per taxonomic group), the physical component variable derived from the PCA of the data set of meteo-climatic and hydrographic variables (PC1PHY) showed minimum values in 2000 Chapter 2 69

followed by a 6-year period of monotonic increase. The correlation between PCAPHY and

PCAZOO is relatively high but not enough to exclude the effect of other variables not included in our data set of physical variables. The correlation was better for Station E1VI than for E3VI, which corresponds to an interface zone between coastal and oceanic domains that shows higher variability (Roura et al. 2013). Previous studies have shown that pelagic populations do not simply follow environmental variability linearly but that their responses to external forcing can be non-linear, such as the amplification of physical forcing through biological interactions (Hsieh et al. 2006). Our results reveal that the abrupt increase in zooplankton abundance is concomitant with a sustained change in the physical environment, which includes downward trends in precipitation, southward displacement of the Gulf Stream north wall position and upward trends for upwelling intensity.

Precipitation on the Galician coast is classically related to the dominance of southwest winds. The resultant river run-off generates the Western Iberian Buoyant Plume (WIBP)

(Peliz et al. 2002) that strongly influences shelf circulation and across-shelf transport processes. During downwelling events also driven by southern winds, there is a confinement of the WIBP towards the coast (Otero et al. 2008) favouring the coastal retention of planktonic organisms. The annual mean precipitation was notably high in 2000 (6.7 mm·day-1), but decreased monotonically from that year until 2005 (3 mm·day-1).

The latitudinal position of the Gulf Stream north wall close to US coast has been also suggested to be a climatic indicator for understanding inter-annual biological changes in

European continental shelf seas (Taylor and Stephens 1980). The association between the displacement of the northern wall of the Gulf Stream and zooplankton abundance has been described in a range of different ecosystems from the Norwegian Sea to the central North

Sea (Taylor 1995a) and seems to drive weak local perturbations of atmospheric circulation patterns (Taylor 1995b). Indeed, coastal upwelling in the area is driven by northerly winds, contrasting to precipitation that tends to occur when southwest winds are dominant. These two variables appear opposite in the first principal component derived from the data set of physical variables. Bode et al. (2009) detected a decrease in upwelling intensity on the NW Iberian shelf from 1968 to 2006, and suggested that this should lead to a reduction in offshore export of planktonic organisms and enhancement of zooplankton retention and therefore abundance. In the Ría de Arousa, 16 nautical miles to the north of the Ría de Vigo, 70

Perez et al. (Pérez et al. 2010) related an increase in coastal phytoplankton and harmful algal blooms with the decreasing trend of upwelling intensity. They hypothesized that this decreasing trend may lead to more stratification and more remineralization of organic matter which in turn may compensate for the decrease in nutrients caused by reduced upwelling. Our observations, corresponding to a different temporal window (1995– 2010), showed a decreasing trend in the annual mean of upwelling in early years until 2002, followed by an increasing trend until 2010. In contrast to Bode et al. (2009), we therefore associated the increase in zooplankton total abundance with a decreasing trend in precipitation and an increasing trend of upwelling from 2002 to 2006.

Upwelling events occur at higher frequencies (days to weeks) and the horizontal displacement of the WIBP has also been described to respond quickly, in the order of hours, to upwelling–downwelling changes (Otero et al. 2008). In our study, as in Bode et al. (Bode et al.

2009b) and Perez et al. (Pérez et al. 2010), upwelling statistics were computed as annual means. We are aware that this simplification may lead to misrepresenting the relationships between upwelling and the dynamics of planktonic populations, which have been shown

(García-Reyes et al. 2014) to respond to drivers at the scale of the order of weeks in the case of zooplankton (Tenore et al. 1995). Therefore, to understand in detail the mechanisms driving changes in zooplankton in the future, it should be appropriate to take into account smaller temporal scales.

In this study, we identified a set of environmental variables susceptible to inducing a significant change in the dynamics of zooplankton that affected average abundance and the amplitude of the seasonal cycle but, apparently, neither its timing nor its composition. Some of these variables are indicators of local-to-regional scale meteo-hydrographic processes, such as precipitation and upwelling intensity, whereas the Gulf StreamNWP is an indicator at the scale of the Atlantic basin. Such global climate indices do not necessarily have a strong link with local-to-regional weather conditions (Stenseth et al. 2003), but they can be good predictors of ecological processes (Hallett et al. 2004; Ménard et al. 2007). Upwelling intensity and precipitation, for example, are affected by these large-scale climatic changes through modification of wind regimes. Other variables, such as the strength of along and across-shelf currents or nutrient concentrations, unfortunately not taken into account in Chapter 2 71 this study, are also dependent on these climatic modes and can have an impact on zooplankton populations.

According to our results, two different mechanisms may have induced the abrupt changes in zooplankton abundance: the first implies a transient perturbation corresponding to 1 year of relatively extreme values. In 2000– 2001, weak upwelling associated with high precipitation should have favoured the retention at the coast of planktonic organisms (downwelling and WIBP reinforcement). The second mechanism implies a constant trend from 2001 to 2006 in several of the environmental variables and notably an increasing trend of upwelling, a southward displacement of the northern wall of the Gulf Stream and a decreasing trend in precipitation. Within this configuration, the amount of upwelled nutrients should increase as well as primary production, as observed in the Ría de Arousa by Perez et al. (Pérez et al. 2010). The end of the monotonic increase in the composite variable of environmental data resulted in the third period with intermediate plankton abundance.

It is worth nothing that in a recent study of zooplankton, carried out in the Western English channel, has highlithed similar long-term periods (1995–2000, 2001–2007, 2008– 2012) (Reygondeau et al. 2015). Although based on different types of data, these authors reach similar conclusions about a progressive long-term modification of the environment which consists mainly in a progressive intensification of the warm period and a decrease in the intensity and depth of thermocline. Those concomitant changes support the idea of a large-scale modification of the environment.

CONCLUSIONS

By analysing a 17-year time series of monthly data, we found zooplankton to show a strong seasonality marked by one main period of high abundance in summer and a change in community structure throughout the year. Copepods were largely dominant in winter, and are mainly accompanied in summer by other groups as Larvacea, Cladocera and larvae of benthic organisms (Cirripedia, Echinodermata, Bivalvia, Decapods and Gasteropods). Copepod dominance also showed spatial differences, with lower relative abundance at the coastal station due to the increased presence of Larvacea, Cladocera and larvae of benthic organisms. 72

We also detected seasonal and long-term changes in the total abundance of zooplankton at the inner- and mid-shelf station that shaped the series in three periods. Zooplankton abundance increased suddenly in 2001 and remained high until 2006, when abundance decreased again but remained higher than the initial values. These changes appear to be related to continuous changes in large climatic modes influencing, for example, the intensity of upwelling and the amount of precipitation in the study zone through changes in wind regime.

Chapter 2 73

Supplementary Information

Figure SI2.1 : Seasonal variability of meroplankton composition expressed as the monthly relative abundance of the major meroplankton groups at station E1VI (A) and E3VI (B).

Figure SI2.2 : Principal components time-series. PC2 for physical variables (A) and its loadings (B).

75

CHAPTER 3. ENVIRONMENTAL MULTI-SCALE

EFFECTS ON ZOOPLANKTON INTER-SPECIFIC

SYNCHRONY.

Abstract

Knowledge on the mechanisms that drive population dynamics and shape community structure is a key issue in ecology. Using wavelet methods, we analyzed 17-yr of monthly time-series of marine zooplankton (taxonomic composition, total abundance, and biomass) and their relationship with environmental factors (upwelling index and river outflow). The main mode of variation in all series was annual and exhibited year-to-year variability. The dynamics of zooplankton aggregated properties showed a strong association with upwelling index and river outflow. The annual oscillation of biomass and abundance increased in 2000 corresponding to the highest amplitudes of environmental forcing. Concomitantly, enhanced synchrony was observed among the main taxonomic groups of zooplankton and among copepod species, the most relevant group in terms of occurrence and abundance. The degree of synchrony appeared to be correlated with the upwelling index and, more closely, with the duration of the upwelling events. The amplified seasonality of the environmental variables from 2000 to 2004, combined with a reduction of off-shore exportation by shortening of upwelling events, favored retention in winter, and primary production in summer. These changes modulated community aggregated properties and affected the stability of the zooplankton community through an increase in inter-specific synchrony allowing the community to shift to another state and likely a reorganization of the community size structure.

Limnology and Oceanography (2017), doi: 10.1002/lno.10501, authors: Lucie Buttay, Bernard Cazelles, Ana Miranda, Gerardo Casas, Enrique Nogueira and Rafael González- Quirós

Chapter 3 77

INTRODUCTION

Knowledge on the processes that drive population dynamics and shape community structure is a key issue in ecology. Much attention has been given to the temporal association of spatially distant populations (Blasius et al. 1999; Cazelles et al. 2001; Engen and Saether 2005; Gouhier et al. 2010a; Fox et al. 2011) in which synchronization, caused by large-scale environmental factors ( i.e. the ‘Moran effect’; Ranta et al. 1997), has been linked to changes in meta-population persistence (Earn et al. 1998; Palmqvist and Lundberg 1998). In contrast, inter-specific synchrony in local communities has received less attention even though it has been proposed to have strong consequences for their stability (Pimm et al. 1988; Tilman et al. 1998; Inchausti and Halley 2003; Downing et al. 2008; Gonzalez and Loreau 2009; Gouhier et al. 2010b). Within a community, populations of similar species are expected to present compensatory dynamics (temporal niche differentiation) in order to coexist through the combined effects of demographic stochasticity and density-dependent processes (Hubbell 2001; Lande et al. 2003; Loreau and de Mazancourt 2008; Kalyuzhny et al. 2014). Intuitively, the balancing replacement of functionally complementary species should maintain the ecological properties at the community level by buffering the effects of disturbances (Gonzalez and Loreau 2009). Those mechanisms can be, however, counterbalanced by environmental forcing (Loreau and de Mazancourt 2008) and thus, compensatory dynamics appear to be rare in natural populations (Houlahan et al. 2007; Vasseur et al. 2014). Empirical work, conducted on freshwater ecosystems, showed that the loss of compensatory dynamics observed within a community after a disturbance can facilitate its shift to a different stable state (Keitt 2008; Jochimsen et al. 2013). More generally, it has been shown that abiotic fluctuations can strongly modify system dynamics instead of just increasing system variance (Chesson 2003). Those findings suggest that community composition data is needed in order to assess the biological effects of environmental disturbances.

Determining the relative strength of biotic and abiotic forces needed to shift between compensatory and synchrony dynamics remains controversial. One of the reasons is the difficulty to identify and model the effect of key environmental drivers (Mutshinda et al. 2009) and the need to use quantitative methods able to cope with the nonstationary nature 78 of the data, given that community properties and environmental forcing response are susceptible to fluctuate in time. Wavelet analysis is a suitable numerical tool for nonstationary time-series. It performs a local time-scale decomposition of the signal (Lau and Weng 1995; Torrence and Compo 1998; Cazelles et al. 2008) and allows to determine the dominant modes of variability and track their change through time (Klvana et al. 2004). It also permits a scale-dependent analysis of the coupling between drivers and response variables (Ménard et al. 2007) and the inspection of the level of synchrony among signals (Keitt et al. 2006; Vasseur et al. 2014).

The functioning of the Galician ecosystem has been intensely studied (e.g. Bode et al. 1998; Nogueira et al. 2000; Cabal et al. 2008; Ospina-Alvarez et al. 2010; Roura et al. 2013). The region, influenced by contrasted seasonal fluctuations in upwelling and precipitation, offers a good case study to infer the relationship between environmental forcing and community dynamics. Additionally, zooplankton has been proposed to be a good sentinel of environmental changes (Taylor et al. 2002; Hays et al. 2005) being therefore an appropriate biological model to study the effects of abiotic factors on population dynamics. For instance, competitive interactions between two marine zooplankton species, giving rise to compensatory dynamics at decadal scale, have been identified in the North Atlantic (Fromentin and Planque 1996). However, investigations focusing on the dynamics of whole zooplankton communities have been based mainly on freshwater ecosystems (Keitt et al. 2006; Keitt 2008; Vasseur et al. 2014), so there is a scarcity of studies dealing with zooplankton dynamics of natural marine populations. A previous analysis of the dataset used in the present work described significant changes on the zooplankton community (Buttay et al. 2016). More precisely, an important increase of the abundance of the whole zooplankton community as well as all the main taxa studied was observed from 2001 onwards. Some meteo-climatic variables were highlighted as possible drivers for such long- term increment, but the methods employed did not allow us to explore in depth neither the coupling between zooplankton abundances and environmental drivers nor the temporal scales involved in such coupling.

In the present study, we pursue three main objectives: (1) to analyze the components of temporal variability in zooplankton aggregated properties and meteo-hydrographic variables, how their periodic components fluctuate in time and their possible association; Chapter 3 79

(2) to quantify the synchrony at the annual scale and its changes through time at two community levels: among major taxonomic groups and among copepod species; and (3) to explore the relationship between synchrony and the multi-scale variability of the environmental drivers in order to infer the processes that shape the structure of the zooplankton community.

MATERIAL AND METHODS

Study area

The northwest Iberian shelf is located in a temperate latitude, at the northern limit of the Canary current upwelling system (Arístegui et al. 2009). As such, the region is characterized by a strong seasonality mainly driven by the alternation, at the annual scale, of wind regimes. From mid-spring to early-autumn, the predominance of northerly winds promotes the upwelling of sub-surface waters, injecting inorganic nutrients to the surface layers that fuel primary production. Upwelling dynamics interact with the circulation within the rías, reinforcing across-shelf export. The rest of the year, southerly and westerly winds predominate, inducing downwelling over the shelf and rainfall over the western Iberian Peninsula. The combination of river runoff and downwelling favor the establishment of the western Iberian buoyant plume (WIBP) (Peliz et al. 2002), which further contributes to shelf retention processes. Nevertheless, a large fraction of upwelling–downwelling and river outflow variability occurs at short-term scales (from days to weeks), also affecting plankton dynamics (Nogueira et al. 2000).

Zooplankton community and environmental variables

Zooplankton was sampled monthly from 1994 to 2011 within the framework of the ongoing RADIALES monitoring program (http://www.seriestemporales-ieo.com), at two stations on the southern part of the Galician sub-region: in the central part of the Ría de Vigo (station E1VI; 42.213°N, 8.850°W; ca. 40 m isobath), and in the adjacent mid-shelf, off the ria's mouth (E3VI; 42.142°N, 8.958°W; ca. 100 m isobath). Samples were taken by means of oblique hauls with a double 40 cm diameter Bongo net of 200 μm mesh size. During the haul, a fixed length of cable is realeased (100 and 50m in E3VI and E1VI 80 respectively), controlling that the angle of the cable during the trawl is maintained at 45°. The depth attained during the haul is recorded with a TD (temperature-depth sensor) and the volume of water filtered is determined with flowmeters allocated at the mouth of each of the nets. Methods for sampling and sample processing were consistent throughout the time series. The sample from one cod-end of the net was preserved in 4% tetraborate- buffered formaldehyde. Subsequently, zooplankton organisms from a subsample (at least 1000 individuals) were identified under a stereoscopic microscope at the lowest possible taxonomic level and counts were converted to full-sample number of individuals per m−3. The sample from the second cod-end of the net was filtered and subsequently dried during 24 h at 60°C in order to estimate the sample biomass as mg dry weight·m−3.

Different levels of zooplankton community assembly were considered: aggregated properties (total zooplankton biomass and total abundance) and abundance of the zooplankton taxonomic groups and of the species of copepods, the most relevant taxa in terms of occurrence and contribution to total zooplankton abundance (Supporting Information Tables SI3.1, SI3.2).

Meteo-hydrographic variables, which are known to play a significant role on ecosystem dynamics in the studied area, were also analyzed (Buttay et al. 2016)). Time series of daily upwelling index (m3·s−1·km−1) were provided by the Instituto Español de Oceanografía. The index was calculated by the Ekman transport equation (Bakun 1973) from geostrophic winds estimated at 43.8°N, 11.8°W, which is considered a representative location for the characterization of wind driven costal upwelling/downwelling (forced by northerly/southerly winds) along the Galician coast (Lavín et al. 1991). Also, the short-term variability of upwelling was estimated by calculating the duration of upwelling events, defined as the number of consecutive days with positive upwelling index. Daily outflow (m3·s−1) from river Miño at the gauge station of Frieira, which covers 86% of the total drainage basin of the river system (17,570 km2) that outflows in the study area, were provided by the Confederación Hidrográfica del Miño-Sil (http://www.chminosil.es/es/).

Data preparation

Upwelling and river outflow data had daily resolution whereas zooplankton sampling was carried out on an approximately monthly basis (32 ± 7 d). Prior to the application of Chapter 3 81 wavelet analysis, the time series of environmental variables were monthly averaged and those of zooplankton abundance were regularized using the “regul” function from the Pastecs R package (Ibanez and Grosjean 2013) based on the area method (Fox and Brown 1965). Low-frequency components having periods greater than 6 yr (corresponding to one- third of the total length of the time-series) could not be well resolved, and so they were removed using a 6-yr low-pass filter (Shumway and Stoffer 2006b). Finally, the monthly time series of zooplankton abundance and Miño outflow were normalized by square root transformation and subsequently all series were standardized to mean zero and unit variance due to their different units and scales of measurement.

Wavelet analysis

For all time-series, we used the Morlet wavelet, a continuous and complex wavelet that enables the extraction of time-dependent amplitude cycles and whose scales are related to frequencies (Ménard et al. 2007; Cazelles et al. 2008). The relative importance of frequencies for each time step may be represented in the time–frequency plane to form the local wavelet power spectrum (WPS) on a 2D plot. Discontinuities, however, exist at the border of the data, and the corresponding affected region growing in extent as the scale increases, is delimited by the cone of influence (Torrence and Compo 1998). Therefore, the information below the cone lacks accuracy and should be interpreted with caution. We also computed the global WPS as the time-average of the local WPS for each frequency component, which provides an unbiased, consistent alternative to the Fourier spectrum in order to summarize the dominant periodicities of the series (Percival 1995) and to calculate the relative contribution of the annual component through the time-series.

The wavelet coherence (WCo) is suited to analyze the transient patterns of co- variation between pairs of signals (Grinsted et al. 2004; Cazelles et al. 2008) by representing, in a time–frequency plane, information on where two nonstationary time series are locally linearly correlated. In the present work, it was applied to compare meteo- hydrographic variables (i.e., upwelling index and river outflow) with total zooplankton biomass and, additionally, their phases for the seasonal mode were extracted and compared (Cazelles et al. 2008). To assess whether the wavelet-based quantities (either for WPS or WCo) were not only due to random processes, we determined the 5% significance 82 level through a bootstrapping scheme that used a hidden Markov model (HMM) (Cazelles and Stone 2003). For this purpose, we tested the null hypothesis, that the observed time- series patterns were different from those expected by chance alone, by generating a surrogate time-series that mimicked the original time-series, thus presenting the same distribution of values and identical short-term autocorrelation structure (Cazelles et al. 2014).

The metrics classically employed to quantify the temporal associations between species are based on covariance quantities (e.g. Loreau and de Mazancourt 2008; Gouhier and Guichard 2014) and did not allow to distinguish between the different scales of variation (Vasseur et al. 2014). In the present work, wavelet decomposition has been applied to the abundance time series of taxonomic groups and copepod species which occurred in more than 30% of the samples: 15 taxonomic groups and 12 copepod species in station E1VI (15 taxonomic groups and 15 copepod species in E3VI). A description of the selected taxonomic groups and copepod species cycles is provided in the Supporting Information Table SI3.1. The oscillations that corresponded to the annual scale were then extracted from all series to compare their characteristics (i.e., amplitude and phase) throughout time.

We quantified the dispersion of the phases in order to characterize the temporal synchrony between the annual signals (Cazelles and Stone 2003; Keitt 2008). To this aim, we computed the angular variance from the extracted phase angles of the annual modes at each time step using the MATLAB Circular Statistics toolbox (Berens 2009): the lower the phase angle variance (PAV), the higher is the synchrony among series. The HMM bootstrap method was also used to determine whether the PAV values observed (i.e., degrees of synchrony) were lower from those expected by chance alone, simply by putting the one- sided 95% confidence interval. Additionally, we explored the relationship between environmental variability and the degree of synchrony by comparing the PAV from a precise date each year (1st July) with the annual averaged environmental variables.

The results presented here correspond to the sampling location within the ría (Sta. E1VI). Results for E3VI (mid-shelf, off the ria's mouth) are shown in the Supporting Information. In general terms, similar zooplankton dynamics were observed in both spatial domains, but between-site differences worth to mention are discussed. The numerical Chapter 3 83 analyses were mainly performed using a MATLAB wavelet package (www.biologie.ens.fr/~cazelles/bernard/Research.html).

RESULTS

Dynamics of zooplankton aggregated properties and environmental variables

The time series of zooplankton aggregated variables (total biomass and abundance) and environmental drivers (upwelling index and Miño outflow) are presented in Figure 3.1. The local WPS of total zooplankton biomass (Fig. 3.1a) and abundance (Fig. 3.1b) revealed the presence of a 1-yr periodic component which, despite its presence throughout the entire 17-yr time series, varied in power from year-to-year and was statistically significant from 2000 onwards. The highest power at the annual scale was recorded between 2002 and 2006 for abundance and between 2005 and 2010 for biomass. Similar patterns were observed in the mid-shelf site (E3VI), with the exception of the year 2005 during which the 1-yr periodic component was absent in biomass and a 2-yr periodic component that was present in the abundance series (Supporting Information Fig.SI3.1). The annual component represented on average 61% of the variance of total zooplankton biomass and 67% of the variance of total zooplankton abundance at station E1VI (at E3VI, it represented 63% and 61% of total zooplankton biomass and abundance, respectively).

River Miño outflow showed one main periodic component (Fig. 3.1c), which was annual and intermittent, being disrupted from 1998 to 2000 and between 2005 and 2009. This component represented on average 63% of the total variance of the monthly time series.

Figure 3.1: Wavelet decomposition of the analyzed time series for station E1VI. (a) Total zooplankton biomass (mg DW·m−3); (b) total zooplankton abundance (ind·m−3); (c) upwelling index (Ekman transport, m3·s−1·km−1); (d) Miño outflow (m3·s−1). Left panels: Regularized time series, the blue dotted line is the removed trend. Central panels: The respective local WPS; color code for power values is graded from blue (low values) to dark red (high values), and the black line defines the cone of influence below which the information is affected by edge effect. Right panels: the respective global WPS. On the central and right panels, the black dotted lines denote the 5% significance areas determined with a bootstrapping scheme based on HMM (Cazelles et al. 2014). Chapter 3 85

In addition, 2- and 3-yr cycles also appeared in the local WPS between 2000 and 2003 due to important river discharges during that period, especially in 2001. The local WPS of the monthly time series of upwelling index (Figure 3.1, d) presented a unique significant 1- yr periodic component that was more marked from 2000 to 2004 and interrupted between 2004 and 2006. This annual component represented on average 53% of the total variance of the monthly time series. Thus, the highest annual amplitudes in the time series of upwelling index and outflow from river Miño occurred between 2000 and early 2004, and were concurrent with the beginning of the period (from 2000 onwards) characterized by enhanced annual cycles of zooplankton aggregated variables.

Figure 3.2: Wavelet Coherence (WCo) between total zooplankton biomass at station E1VI and upwelling intensity (a), and between total zooplankton biomass and Miño outflow (b). On (a) and (b), color code for coherence values is graded from blue (low values) to dark red (high values), the dotted-dashed lines denote the 5% significance areas (determined with a bootstrapping scheme based on HMM, Cazelles et al. 2014) and the black line defines the cone of influence below which the information is affected by edge effect. Panels (c) and (d) show the phase differences for the annual periodic component between upwelling index and total zooplankton biomass and between Miño outflow and total zooplankton biomass respectively. On (c) and (d), the red line corresponds to total zooplankton biomass the blue line represents the environmental variable and the black dotted line represents the phase difference between pairs of signals.

The wavelet coherence (WCo) corroborated the strong association between meteo- hydrographic variability and zooplankton total biomass and abundance. The coherence is particularly consistent at the annual scale (Fig. 3.2 a,b). The comparison of phases for the annual component revealed that total zooplankton biomass (and abundance, not shown) fluctuated in synchrony with upwelling. In contrast, a phase difference of π radians (i.e., 6 86 months) was observed between the annual phases of total zooplankton biomass and outflow from river Miño series (Fig. 3.2 c,d).

Dynamics of zooplankton taxa and copepod species

A description of the significant periodic components present in the time series of zooplankton taxa and copepod species, derived from the inspection of the respective global and local WPS, is given in Supporting Information Table SI3.1. All taxa used in this study presented an annual cycle, consistent in both the local and global WPS. In addition, Cirripedia larvae and Polychaeta (only in station E3VI) presented shorter periodic components of the order of 6 months. They also presented, as was the case for most of the major taxonomic groups of meroplankton (larvae of Bivalvia, Gasteropoda, and Decapoda), significant periods of 2 yr. The time series of copepod species were also dominated by the annual oscillation; oscillations at shorter scales were observed sporadically in Pseudocalanus elongatus and Oithona similis. Multi-annual periodicities were more frequent; cycles of 1.5 yr to 3 yr were observed for the time series of Calanus helgolandicus, Pseudocalanus elongatus, Euterpina acutifrons, Oncaea media, Oithona nana and Oithona similis.

The annual component extracted from each zooplankton taxa and copepod species are shown in Fig. 3.3 a and b, respectively. The amplitude of the seasonal mode of all taxonomic groups’ series increased from 2000 to 2003, and decreased afterwards showing, in 2011, amplitude values that were similar to those found at the beginning of the series (Fig. 3.3a). The amplitude of the seasonal mode of copepod species showed a similar pattern (Fig. 3.3 b).

Chapter 3 87

Figure 3.3: Time series of the seasonal mode of zooplankton abundance. Each blue line represents the seasonal mode extracted from each (a) taxonomic group and (b) copepod species series at station E1VI. The red line in both graphs represents the monthly phase angle variance (PAV) of the seasonal component and the red dotted line represents the one-sided 95% confidence interval (computed with a bootstrapping scheme based on HMM) to show the significance of the low values of the PAV.

The time series of the PAV among zooplankton taxa and among copepod species (red line on Fig. 3.3 a,b, respectively) showed a common general pattern. Both series presented an abrupt decrease after 1999, followed by a period of significantly low angle variance values between 2000 and 2004. During this period, the annual phases were closer, which is indicative of a higher degree of synchrony among zooplankton taxa and copepod species. While the PAV among taxonomic groups was significant through the whole series, for the copepod assembly significance was lost from January 1999 to June 2000 and from February 2005 to June 2006. We further explored the relationship between environmental variability and the degree of synchrony (i.e., annual PAV) among zooplankton community components. All environmental variables tested are described in the Supporting Information Fig. SI3.4. We found no significant relationship between the annual average of upwelling intensity and the PAV among copepod species (r = 0.44; p-value = 0.09) or among 88 taxonomic groups (r = 0.36; p-value = 0.18). A significant relationship was found, however, between the PAV and the annual average duration of upwelling events, defined as the consecutive number of days of upwelling occurrence between downwelling episodes). The annual mean duration of upwelling events (Fig. 3.4) ranged between 7 days (in 2003) to 13 days (in 1995 and 2005), and the shorter the duration of events, the lower the PAV (i.e., the stronger the synchrony) among population abundances. This relationship was statistically significant at both community assembly levels, but higher for copepod species (r = 0.60, p- value = 0.02) than for main taxonomic groups (r = 0.51, p-value = 0.05). The period between 2000 and 2004 corresponded to the highest degree of synchrony and shorter duration of upwelling events. In station E3VI (mid-shelf), despite a concurrent period of low PAV between 2000 and 2004, no significant relationships with upwelling index or with the duration of upwelling events were found.

Figure 3.4: Annual phase angle variance (PAV) computed on the seasonal mode of copepod species abundances (red), taxonomic groups (red dotted line) and annual average duration of upwelling events (black) at station E1VI. (r = 0.60, p-value = 0.02 and r = 0.51, p-value = 0.05 for copepod and taxonomic groups, respectively).

DISCUSSION

In the present work, we have studied the dynamics of a natural marine zooplankton community at different levels of organization and their relationship with fluctuating abiotic factors. To this aim, we applied wavelet methods to 17-yr (1995–2012) monthly time series of community aggregated properties (biomass and abundance), functional groups and species composition and meteo-hydrographic factors (coastal upwelling and river runoff). Long-term data of zooplankton represent a considerable effort in taxonomic identification, Chapter 3 89 and to our knowledge no previous studies have focused on the scale-dependent fluctuations of aggregated properties or interspecific synchrony within a natural, non- manipulated marine zooplankton community.

Temporal patterns and coupling between aggregated zooplankton properties and environmental variables

The principal mode of temporal variation observed in all zooplankton and environmental time series was annual. It represented between 53% and 67% of the total variance of the monthly time series and exhibited in all cases a significant year-to-year variability in amplitude. In both time-series of aggregated zooplankton properties (biomass and abundance), the statistical significance of the annual component started in 2000, pointing to an increase in amplitude at the annual scale. A previous study on the same data set that used classical methods unable to distinguish between scales (Cumulative sums to identify long-term trends in the series combined with Generalized linear model to model the main temporal patterns—long-term, seasonal, and autocorrelation structure), described a stepped increase in the abundance series occurring in 2001, followed by a light decrease in 2006 (Buttay et al. 2016). Similarly, the wavelet transform depicted a decrease in the annual component for the abundance series after 2006. In contrast, the total biomass series continued increasing, suggesting a possible structural change within the community towards larger sized organisms. This pattern was observed in both costal and mid-shelf stations but additional data of organisms' size would, however, be necessary to corroborate those changes.

The relationship between the annual cycles of total biomass (and abundance) of zooplankton and upwelling/downwelling index and river outflow revealed by the wavelet coherence (WCo) confirmed the major importance of these weather-induced factors on the seasonal dynamics of zooplankton in the studied area. It is worth noting that the seasonality of zooplankton aggregated properties arose when the annual amplitude of upwelling and river outflow were maximal. The higher amplitudes of the annual signals of these abiotic factors observed between 2000 and 2004 are indicative of enhanced contrast between the winter and summer seasons, due to changes in wind patterns at regional scales and related to sustained trends of large-scale climatic patterns in the North Atlantic during that period 90

(Buttay et al. 2016). Within this configuration, strong upwelling and dry/sunny weather in summer cause higher primary production, while strong downwelling and precipitation in winter promote the reinforcement of the western Iberian buoyant pluem (WIBP) that, by influencing shelf circulation and across-shelf transport processes, favors the retention of planktonic organisms at the coast (Peliz et al. 2002; Otero et al. 2008). Those changes in the winter conditions may have important ramifications for the plankton community because small variations in the overwintering stock of organisms are magnified by exponential growth and can cause major changes in the summer populations (Colebrook 1979). The shortening of upwelling events, also observed during the same 2000–2004 period, could have favored the retention of organisms at the coast by diminishing the strength of the across-shelf exchanges processes as well (Iles et al. 2012). Indeed, a previous study conducted in the Galician region described a decrease in zooplankton abundances within the oceanic domain and pointed out to the reduction of the offshore exportation of organisms as a possible mechanism (Bode et al. 2009a).

Inter-specific synchrony: variability and possible drivers

In a changing environment, coexisting species should tolerate similarly environmental variability and consequently, natural selection exerted by environmental drivers on functionally similar species can lead to synchronous dynamics, where all species rise and fall together (Rocha et al. 2011). In the Galician coast, which is considered a highly variable environment, it is not surprising that zooplankton populations fluctuate in a relative level of synchrony, presenting high abundances from late spring to early autumn during the upwelling favorable period of the year (Valdés et al. 1990; Buttay et al. 2016). The use of wavelet methods in this study allowed us to detect temporal changes in the degree of synchrony by computing the time evolution of the angular variance among the species annual phases. It is worth noting that concomitantly with the enhanced seasonality in upwelling and river outflow observed from 2000 to 2004 and the substantial increase in zooplankton aggregated properties, higher synchrony was observed among copepod species and taxonomic groups at the annual scale in both sampling stations. It is also notable that the selected taxonomic groups are functionally diverse. Indeed, different feeding habits (predators, filter-feeders…) are represented and while most of the groups Chapter 3 91 are part of the holoplankton, some pertain to the meroplankton (those that are planktonic only for a part of their life cycle: e.g., Cirripedia larvae, Echinodermata larvae). Despite the fact that some different dynamics may occur within the taxonomic groups and blur synchrony detection, similar patterns in synchrony were observed among copepod species and taxonomic groups. The degree of synchrony and the duration of upwelling events appeared to be significantly correlated and higher synchrony was indeed associated with an increase of upwelling frequency. Even though high synchrony was encountered in both stations during the period of shorter upwelling events (2000–2004), the relationship was not statistically significant at the mid-shelf station. We hypothesize that this station, being located at the interface between coastal and oceanic domains (Roura et al. 2013), receives occasionally oceanic influences that may blur the effect of coastal upwelling.

It has been shown that the effects of an environmental driver occurring at a specific scale can be redistributed to other frequencies (Greenman and Benton 2005) although there is no clear evidence of the underlying mechanisms: Did shorter upwelling events prevent compensatory dynamics from occurring by increasing the frequency of disturbances? Or did shorter upwelling events support higher primary production leading to a reduction of interspecific competition and thus the need of temporal differentiation among populations? Additional data on phytoplankton production would have been necessary to answer those questions.

Because community composition data are difficult to obtain on a routine basis (i.e., time consuming and taxonomic expertise required), temporal changes in inter-specific synchrony among zooplankton populations have been described in few studies (Keitt et al. 2006; Keitt 2008; Jochimsen et al. 2013; Vasseur et al. 2014). Two freshwater systems were in particular studied: the Constance lake (Switzerland) affected by a strong decrease in phosphate (Jochimsen et al. 2013) and the artificially acidified Little Rock lake (Wisconsin, USA) (Keitt 2008). In both systems, higher synchrony was encountered as a primary effect after the identified perturbations. Interestingly, in the Keith’s study, the synchrony was driven mainly by winter tolerant species that drastically decreased after the perturbation, while in the present work all populations persisted. Jochimsen et al. (2013) hypothesized that the subsequent regime shift they observed was favored by the loss in compensatory dynamics. Our results may concur with their observations: the decoupling of the two 92 community aggregated properties observed after the period of high synchrony (total zooplankton biomass increased while abundance decreased), gives some indications that structural changes have occurred within the community assembly.

Ecological implications

Downing et al. (2008) lamented that despite the advances in theory, relatively little empirical work existed to determine how populations oscillate within a community. Some studies have focused on the analysis of temporal patterns of plankton aiming to discern the type of population dynamics exhibited by these organisms: compensatory or synchronous dynamics (Vasseur et al. 2005; Houlahan et al. 2007; Keitt 2008). In the present work, we were able to quantify synchrony at the seasonal scale and to describe fluctuations in the degree of synchrony in natural marine communities, providing one of the first evidences that interspecific synchrony can fluctuate in relation to environmental forcing.

Temporal variability among populations (or compensatory dynamics) has been proposed to be an indicator of stability while synchrony has been linked to extinction probability (Pimm et al. 1988; Inchausti and Halley 2003) and especially to a decrease in food-web stability (Gonzalez and Loreau 2009). From the Constance lake observations, Jochimsen et al (2013) also hypothesized that the loss of compensatory dynamics facilitated the community shift to another stable state. Although in the present study we couldn't describe changes in the community food-web, the decrease observed after 2005 in the zooplankton community abundance contrasted with the increase observed in the zooplankton community biomass, giving some indications that the loss of compensatory dynamics may have favored deep changes in the size structure of the community.

In the present work, we have tracked changes in zooplankton populations focusing on their seasonal dynamics. Some recent works suggested that while synchrony is the predominant dynamics at the annual scale, a compensatory dynamics can occur simultaneously at other scales (Keitt et al. 2006; Vasseur and Gaedke 2014; Vasseur et al. 2014). The use of scale-specific methods, such as the wavelet analysis, and the increasing length of the marine zooplankton time series, as such obtained within the RADIALES program, will allow further assessments of the zooplankton community dynamics at multiple scales. Chapter 3 93

CONCLUSIONS

In the present work, we were able to: (1) discern the temporal patterns of zooplankton dynamics at different levels of the community assembly, (2) quantify the degree of synchrony among zooplankton functional groups and species of copepods at the annual scale and its variability trough time, and (3) highlight the predominant role of abiotic, weather-induced factors acting at multiple scales to drive zooplankton fluctuations and shape community dynamics.

We hypothesize that higher amplitude of the seasonal cycle (reinforced seasonality) of abiotic factors, upwelling index and river Miño outflow, as well as the reduction of the duration of upwelling events, drove the increases in zooplankton abundance and biomass observed in 2000. The development of the western Iberain buoyant plume (WIBP) and downwelling enhanced retention in winter, whereas primary production on the shelf in summer was enhanced by frequent but short upwelling events that minimize offshore exportation. These environmental changes consisted in a long-term disturbance (2000– 2004) that modulated not only the community aggregated properties, but also affected the stability of the zooplankton community. Indeed, accentuation of the inter-specific synchrony may have caused the community to step to another state and likely led to a reorganization of the community size structure.

Acknowledgments

L.B. was supported by a FPI Ph.D. Grant from the Instituto Español de Oceanografía. B.C. is partially support by the “Pepiniere interdisciplinaire CNRS-PSL Eco-Evo-Devo.” Zooplankton data were collected and analyzed within the framework of the time series monitoring program RADIALES (http://www.seriestemporales-ieo.com). This study was conducted within the framework of the LOTOFPEL project (CTM2013-16053. Ministerio de Ciencia e Innovación, Spain). The open access publication of this paper was funded by the Principado de Asturias (Spain) research project GRUPIN14-144. We also thank Johnna Holding for revising the English.

94

Supplementary Information

Table SI3.1: List of selected zooplankton taxonomic groups and copepod species observed at station E1VI, and their respective occurrence and periodicities observed in the local and global Wavelet Power Spectrum (WPS).

Taxa Occurrence (%) Local WPS Global WPS Copepoda 100 1 1 Acartia clausi 100 1 1 Paracalanus parvus 97.66 1 1 Oithona plumifera 94.39 1 1 Calanus helgolandicus 92.99 1 and 2.5 1 Pseudocalanus elongatus 92.06 0.5, 1 and 3 1 Temora longicornis 91.12 1 3 Euterpina acutifrons 78.04 1.5 and 3 1 Oncaea media 75.23 1 and 4 1 Oithona nana 55.61 1 and 2.5 1 and 2.5 Centropages chierchiae 40.65 1 1 Oithona similis 38.32 0.8, 1 and 3 1 Larvacea 96.73 1 1 Cirripedia larvae 92.52 0.5 and 1 1 Decapoda larvae 82.24 1 and 2 1 Cladocera 82.24 1 1 Cnidaria 79.44 1 1 Gastropoda 78.5 1 and 2 1 Siphonophorae 76.64 1 1 Bivalvia larvae 65.89 1 and 2 1 Chaetognatha 65.89 1 1 Echinodermata larvae 64.49 1 1 Polychaeta 63.55 1 1 Euphausiacea 52.34 1 1 Bryozoa larvae 51.4 1 1 Foraminifera 43.93 1 1 Chapter 3 95

Table SI3.2: List of selected zooplankton taxonomic groups and copepod species observed at station E3VI, and their respective occurrence and periodicities observed in the local and global Wavelet Power Spectrum (WPS).

Taxa Occurrence (%) Local WPS Global WPS Copepoda 100 1 1 Acartia clausi 98.44 1 1 Calanus helgolandicus 97.92 1 1 Oithona plumifera 96.88 1 and 2 1 Pseudocalanus elongatus 95.31 1 and 2 1 Paracalanus parvus 89.58 1,1.5 and 3 1 Oncaea media 84.38 1 1 Paraeuchaeta hebes 73.44 1 and 2.5 1 Temora longicornis 67.71 1 and 2 1 Calanoides carinatus 62.5 1 1 Centropages chierchiae 56.25 1 and 3 1 Euterpina acutifrons 50 1 and 3 1 3 Calocalanus styliremis 42.71 1 1 Oithona similis 41.15 1, 1.5 and 2 1 Oithona nana 40.1 1 1 Mecynocera clausi 35.94 1 1 Clausocalanus arcuicornis 31.25 1 1 Nannocalanus minor 30.04 1 and 3 1 Larvacea 88.54 1 1 Siphonophorae 80.73 1 1 Cirripedia larvae 79.69 0.5 , 1 and 2 1 Euphausiacea 79.17 1 1 Gastropoda 74.48 1 and 2 1 Decapoda larvae 73.96 1 1 Cnidaria 64.06 1 1 Chaetognatha 61.98 1 1 Echinodermata larvae 61.98 1 1 Bivalvia larvae 58.33 1 and 2 1 Foraminifera 55.73 1 1 Cladocera 53.12 1 1 Polychaeta 47.92 0.5 and 1 1 Bryozoa larvae 40.1 1 and 2 1

96

Figure SI3.1 : Wavelet decomposition of the analyzed time series for station E3VI. (a) Total zooplankton biomass (mg DW·m-3); (b) total zooplankton abundance (ind·m-3). Left panels: Regularized time series, the blue dotted line is the removed trend. Central panels: The respective local Wavelet Power Spectrum (WPS); color code for power values is graded from blue (low values) to dark red (high values), and the black line defines the cone of influence. Right panels: the respective global WPS. On the central and right panels, the black dotted lines denote the 5% significance areas determined with a bootstrapping scheme based on HMM (Cazelles et al. 2014).

Figure SI3.2 : Wavelet Coherence (WCo) between total zooplankton biomass (Station E3VI) and upwelling intensity (a), and between total zooplankton biomass and Miño outflow (b). On (a) and (b), color code for coherence values is graded from blue (low values) to dark red (high values), the dotted-dashed lines denote the 5% significance areas (determined with a bootstrapping scheme based on HMM (Cazelles et al. 2014)) and the black line defines the cone of influence. Phase differences for the annual periodic component between upwelling index and total zooplankton biomass (c) and between Miño outflow and total zooplankton biomass (d). On (c) and (d), the red line corresponds to total zooplankton biomass, the blue line represents the environmental variable and the black dotted line represents the phase difference between pairs of signals.

Chapter 3 97

Figure SI3.3 : Time series of the annual oscillation of zooplankton abundance of (a) taxonomic groups and (b) copepod species at station E3VI. The red line in both graphs represents the phase angle variance of the annual oscillations.

98

(a) (b)

Upwelling duration Upwelling duration Upwelling index Upwelling index

NAOw inter NAOw inter NAOsummer Temperatureatm. Gulf StreamNWP NAOsummer Precipitations Gulf StreamNWP Temperatureatm. Precipitations SST SST Miño outflow Stratification EA EA

Stratification Miño outflow

0.0 0.2 0.4 0.6 0.0 0.2 0.4 0.6

-0.4 -0.2 -0.4 -0.2 correlation correlation

Figure SI3.4 : Correlation between annual average of environmental factors and phase angle variance among (a) copepod species and (b) taxonomic groups. Significant correlation (Pearson p- value>0.05) are highlighted in blue.

Upwelling Index (m3·s-1·km-1): Average Ekman transport (source: http://www.indicedeafloramiento.ieo.es/). Upwelling duration (days): Number of successive days of positive values of upwelling index. Miño outflow (m3·s-1): Miño outflow at the station 1641 (42.15ºN, 8.19ºW) (source: http://saih.chminosil.es/). Stratification (ºC): Standard deviation of water column temperature based on CTD casts at station E3VI. SST (Sea Surface Temperature, ºC): Average temperature within the upper 10m of the water column from CTD casts at station E3VI. Precipitation (mm·days-1): Local precipitation at the meteorological station in the Vigo’s airport (source: www.aemet.es). Temperature atm. (ºC): Local air temperature at the meteorological station in the Vigo’s airport (source: www.aemet.es). Gulf StreamNWP: Anomaly of the position of the Gulf Stream North Wall (source: http://web.pml.ac.uk/gulfstream/data.htm). EA (East Atlantic pattern): North–south dipole of anomaly centres spanning the North Atlantic from east to west (source: NOAA, Climate Prediction Center). NAOWinter and NAOSummer (winter and summer North Atlantic Oscillation): Atmospheric sea level pressure anomaly; difference in the normalized sea level pressure between Iceland and the Azores. NAO indices averaged, respectively, over December–February (winter) or June-August (summer) (source: http://www.cru.uea.ac.uk/cru/data/nao/).

Figure SI3.5: Phase angle variance (PAV) computed on the annual oscillation of copepod species abundances (red), taxonomic groups (red dotted line) and annual average duration of upwelling events (black) at station E3VI. (Pearson r = 0.36 and 0.33 for copepod and taxonomic groups respectively p- values>0.05). Chapter 4 99

CHAPTER 4. HOW ENVIRONMENTAL FORCING CAN

SYNCHRONIZE POPULATION FLUCTUATIONS

Abstract: Identifying the mechanisms that control the temporal structure of ecological communities is key to understand their vulnerability and identify early warnings of critical transitions. The effect of environmental fluctuations on the population dynamics of competing species was studied by combining empirical observations and theoretical simulations. Observations consisted in monthly time series of abundance of 12 dominant diatom species, sampled from 1994 to 2009 in shelf waters of Galicia, in the northern limit of the Canary Current upwelling system. A wavelet analysis of the time series of diatom abundances and upwelling intensity allowed us to analyse how the characteristics of the series (i.e. amplitude and phase of each periodic component) changed in time. We observed a sudden increase of the synchrony between diatom species from 1998 to 2002, concomitant with an increase in the amplitude of upwelling intensity at different scales. To better understand the underlying mechanisms that relate inter-specific synchrony to environmental forcing variability, for instances through nutrient fluctuations linked to upwelling dynamics, we have simulated, with a chemostat type model, the dynamics of two species with fixed stoichiometric requirements competing for two essential resources (such it is the case of nitrogen and silicate for diatoms). Environmental variability was simulated for a set of amplitudes of the fluctuations in one of the resources’ inflow. Our results revealed that, for all the stoichiometric ratios that lead to stable coexistence, an increase in the amplitude of resource supply leads to enhanced synchronicity between competitors. This relationship is, however, not linear, and it seems that inter-specific competition buffer the effect of increasing amplitude until certain thresholds (critical transition or tipping points) at which environmental fluctuations become the main driver of community temporal structure.

In preparation for publication. Authors: Lucie Buttay, David A. Vasseur, Ana Miranda, Gerardo Casas, Rafael González-Quirós and Enrique Nogueira

Chapter 4 101

INTRODUCTION

Understanding the processes that drive population fluctuations and bring sudden abrupt changes within communities is a key topic in ecology, especially in order to sharpen predictions for the future. The effect of abiotic factors in population dynamics does not depend only on average conditions (Parmesan et al. 2003) and in fact, changes in amplitude of environmental oscillations may have stronger consequences for biological systems (Stenseth 2012). For instances, the best known aspect of climate change concerns an increase in temperature, but more subtle and complex changes are also predicted (Houghton et al. 2001; Church and White 2006), such as changes in the frequency and strength of upwelling events in eastern boundary upwelling ecosystems (EBUE) (Bakun 1990; Sydeman et al. 2014; Wang et al. 2015). This has important implications since EBUE present among the highest primary productivity rates in the ocean (Ryther 1969) and sustain between 20-25 % of world fish catches, yet occupying less than 1% of the ocean surface.

One of the features that have been highlighted to have strong consequences for communities’ stability is the temporal association between their populations (i.e. inter- specific synchrony; (Pimm et al. 1988; Tilman et al. 1998; Inchausti and Halley 2003; Gonzalez and Loreau 2009; Gouhier et al. 2010a). Compensatory dynamics arise by the need of temporal niche differentiation for similar species in order to coexist (Hubbell 2001; Lande et al. 2003; Loreau and de Mazancourt 2008; Kalyuzhny et al. 2014). The perpetual replacement of complementary species in time is expected to buffer the effects of abiotic disturbances and therefore, to maintain ecological properties at the community level (Gonzalez and Loreau 2009). However, compensatory dynamics appear to be rare in natural populations and synchrony is described as the main mode of fluctuation (Houlahan et al. 2007; Vasseur et al. 2014). In cyclic systems, the phase locking of the different oscillators, such as the populations that compound the community, has been proposed to be an early warning of critical transition (Scheffer et al. 2009). Indeed, a recent empirical work has shown that the increase in synchrony among plankton populations facilitated the occurrence of a regime shift by debilitating community structure (Jochimsen et al. 2013). Hence, it is crucial to understand the mechanisms that cause populations to fluctuate in 102 synchrony within communities as it may be used as an early warning signal of ecosystem regime shift.

Overall, there is still important controversy about the relative strength of biotic and abiotic forces needed to shift between compensatory and synchronous dynamics. It is difficult to assess those questions in natural systems as it requires a deep understanding on ecosystem functioning (Mutshinda et al. 2009), plus data about community composition and abiotic forcing as well as statistical tools able to cope with the non-stationary nature of ecological systems (Cazelles et al. 2008). In the present work we therefore focused on the relatively well described marine ecosystem of Galicia (Northwest Atlantic coast of Spain), where nutrient availability is highly conditioned by the dynamics of coastal upwelling and where sampling of plankton composition has been carried out monthly since 1994 on a routine basis. In a previous study carried out in this area we observed changes in the temporal association among zooplankton species and hypothesized that it was related to changes in amplitude and frequency of upwelling (Buttay et al. 2017). Here, we step down in the trophic food web and focused on the phytoplankton community with the idea that the relationship between upwelling, and hence the availability of nutrients, and phytoplankton community structure is simpler, allowing us to grasp concurrent changes in abiotic drivers and community temporal structure. In addition, we employed a theoretical approach to explore, through a competition model for essential resources, the underlying mechanisms that can explain the observed dynamics. By combining empirical and theoretical approaches we aim to understand the rules that shape community assembly of competing diatoms species over time and, in particular, to assess the effects of changes in the amplitude of nutrient inputs via coastal upwelling, the main abiotic driver in this particular ecosystem.

MATERIAL AND METHODS

Natural system description

The Galician coast is located at the northern limit of the Canary Current upwelling system (Arístegui et al. 2009), one of the four Eastern Boundary Upwelling Ecosystems (EBUES) in the world ocean. As such, the region is subject to the seasonal occurrence of Chapter 4 103 upwelling events. From mid-spring to early-autumn, driven by the predominance of along- shore equatorward winds, coastal upwelling injects inorganic nutrients to the surface lit layer fueling phytoplankton growth and enhancing primary production. During the rest of the year, south and westerly winds predominate, inducing downwelling over the shelf and rainfall (and associated continental runoff) over the western Iberian Peninsula. The relevance of upwelling dynamics in the structure of the plankton community at multiple time scales, from short-term events and seasonal cycles to long-term variability has been extensively studied in the area (Nogueira et al. 2000; Alvarez et al. 2009; Ospina-Alvarez et al. 2010; Buttay et al. 2017).

Phytoplankton community Marine phytoplankton has been sampled on an approximately monthly basis off A Coruña (Northwest Iberian shelf) (E2CO Station: 43.422°N 8.437°W) since 1994, as part of the hydrographic, biogeochemical and plankton observing programme ‘RADIALES’, a time series monitoring project run by the Instituto Español de Oceanografía (IEO) in the Northwest and North Iberian coastal-shelf (http://www.seriestemporales-ieo.com; Bode et al. 2012). Samples for phytoplankton analysis have been taken by the mean of Niskin bottles and preserved in Lugol’s solution (ca. 0.5% final concentration) for their subsequent identification at the laboratory. Organisms were counted using a Nomarski phase contrast Nikon Diaphot TMD microscope, following the technique described by Uthermöhl (Casas et al. 1997). In the present work we focused on the time series of diatom species that where present in the surface layer (averaged abundance from samples at ca. 2, 5 and 10 m depth), in more than 40 % of the samples. The resultant 12 selected species are listed in the Table 4.1.

Upwelling index Time series of daily upwelling index (m3·s-1·km-1) were provided by the Instituto Español de Oceanografía (http://www.indicedeafloramiento.ieo.es/), calculated by the Ekman transport equation (Bakun 1973) from geostrophic winds estimated at 43°N, 11°W, which is considered a representative location for the characterization of wind driven costal upwelling/downwelling dynamics for the northern Galician shelf (Lavín et al. 1991). Upwelling events are described to have the strongest influence on nutrient availability, being responsible, for instances, of >70% of total dissolved inorganic nitrogen inputs 104

(Villegas-Ríos et al. 2011), although episodic continental inputs from the low flow unpolluted rivers at the head of the rías may occur (Otero et al., 2010; Doval and Madriñán 2016).

Table 4.1: List of selected diatom species observed at station E2CO, and their respective occurrence and contribution of the annual periodic component in the total variance calculated on the Wavelet Power Spectrum (WPS).

Annual variance Species Familly Occurrance contribution Nitzschia longissima Bacillariaceae 0.93 0.40 Pseudo-nitzschia pungens Bacillariaceae 0.81 0.46 Thalassionema nitzschioides Thalassionemataceae 0.71 0.31 Pseudo-nitzschia delicatissima Bacillariaceae 0.63 0.41 transitans Naviculaceae 0.60 0.49 Guinardia delicatula Rhizosoleniaceae 0.58 0.52 Leptocylindrus danicus Leptocylindraceae 0.56 0.68 socialis 0.51 0.61 Detonula pumila Skeletonemaceae 0.43 0.45 Asterionellopsis glacialis Fragilariaceae 0.41 0.39 Rhizosolenia imbricata Rhizosoleniaceae 0.41 0.34 Rhizosolenia setigera Rhizosoleniaceae 0.40 0.52

Wavelet extraction of annual amplitudes and phases Wavelet analysis overcomes the problem of non-stationarity, found in the plankton series, by performing a local time-scale decomposition of the signal (Daubechies 1992; Torrence and Compo 1998; Cazelles et al. 2008). Prior to the application of wavelet analysis, the time series of phytoplankton abundance were regularized using the ‘regul’ function from the Pastecs R package (Ibanez and Grosjean 2013) based on the area method (Fox and Brown 1965). Low-frequency components having periods greater than 5 years (one-third of the time-series length) that could not be well resolved were removed using a 5-year low- pass filter (Shumway and Stoffer 2006b). Finally, the regularized monthly time series of phytoplankton abundance were normalized by square root transformation and subsequently all series were standardized to zero mean and unit variance. We employed the Morlet wavelet, a continuous and complex function that enables the extraction of the amplitudes and phases of the periodic components at each time and frequency combination (Cazelles et al. 2008). The relative importance of frequencies for each time step may be represented in the time–frequency plane to form the local Wavelet Power Spectrum (WPS). The local WPS of each of the 12 diatom species selected (>40% Chapter 4 105 occurrence) are presented in Supplementary Information (Fig. S1). Oscillations at the annual scale were extracted to compare their amplitude and phase throughout the time series in order to characterize the temporal synchrony between the annual signals (Cazelles and Stone 2003). We extracted the phases of the annual oscillation at a precise date of each year (July 1st) and computed the Phase Angle Variance (PAV) using the MATLAB Circular Statistics toolbox (Berens 2009). The level of significance of the PAV is computed with a bootstrapping scheme based on a hidden Markov model (HMM) (Cazelles et al. 2014). The PAV summarizes the dispersion of the phase angle among diatom species, and is thus a measure of the synchronicity (Keitt 2008): the lower the PAV, the higher is the synchronicity.

Model of competition for essential (=non-substitutable) resources

Competition is a common interaction among organisms, and thus a major driver of community structure. To assess the relationship between the amplitude of fluctuation of resources and the degree of synchrony between competing populations we have simulated the dynamics of two consumers that compete for two fluctuating resources. The model of resource competition uses a common differential equation framework and specifies that consumers have a fixed stoichiometric demand for two essential resources (e.g. silicate and nitrogen for diatoms). We followed the model formulation presented in the work by Fox and Vasseur (Fox and Vasseur 2008) and Vasseur and Messinger (2011), computing the simulations with the Simecol R package (Petzoldt and Rinke 2007), a list of the parameter and initial values is given in the table 4.2.

The essential resource abundance Ri (with i =1, 2) depends on the flow rate at which the resource is circulating through a chemostat system (d), the inflow supply concentration

(Si) and the consumption (Cj) by the consumers j.

∂Ri gj = d × (Si − Ri) − ∑j Cj × (1) ∂t yij×Ri

The inflow supply concentration fluctuates at an amplitude ai and a period 2π × ힽ.

t S (t) = a × cos ( ) + 1 (2) i i τ 106

The consumers abundance, Cj (with j =1, 2) suffers density independent losses at per capita rate mj and consumes resource i with functional response gj, parameterized by the per capita resource uptake rates Uj for resource 1 and Uj-1 for resource 2.

∂C j = C × (g − m ) (3) ∂t j j j

gj = min(y1j × Uj × R1 , y2j × (1 − Uj) × R2) (4)

We assume that consumer j has fixed resource requirements and Yij is the yield coefficient that gives the units of consumer j produced from a unit of resource i. On the simulations in which the two consumers coexist, the level of synchronicity has been calculated as the covariance (COV) between time series of consumers’ abundance at equilibrium.

Table 4.2: Model parameters

Parameters a1 Amplitude of the supply of the essential resource 1 Plastic [0:1] a2 Amplitude of the supply of the essential resource 2 0 d flow rate (in and out) 0.1 mj Consumer j mortality (density independent) 0.05 U1 and U2 per capita uptake rates Plastic [0:1] S1 Inflow concentration/ supply concentration t time 1:1000 ힽ Period of the fluctuation 10 y11 0.5 y12 1 y21 1 y22 0.5 Initial values R1 Essential resource abundance 0.1 R2 Ressource2 0.1 C1 Consumer 1 0.1 C2 Consumer 2 0.1

RESULTS

Natural system observation

The 12 species of diatoms that have been selected (i.e. those with >40% occurrences) are distributed in 8 diatoms families (Table 4.1): Bacillariaceae, Thalassionemataceae, Naviculaceae, Rhizosoleniaceae, Leptocylindraceae, Chaetocerotaceae, Skeletonemaceae and Rhizosoleniaceae. The local Wavelet Power Spectrum (WPS) of each diatom time series (Suppl. Info. Figure S1) revealed that seasonality accounted for an important part of the Chapter 4 107 total variability, ranging from 31% (Thalassionema nitzschioides) to 68 % (Leptocylindruc danicus). The annual oscillations of each species varied in amplitude through time (Figure 4.1a). Globally, the higher amplitudes were found from 1998 to 2005, even though some species presented different patterns. This is the case, for instance, of Rhizosolenia setigera that showed its higher amplitude at the beginning of the series (1995-1996). Some differences in the temporal association among those 12 species are also noticeable. To highlight them, we have extracted the phases for the annual oscillation and make abstraction of the amplitude variations (Figure 4.1b). Angular phases locate the observed oscillation, for each time step, on the trigonometric circle, fluctuating therefore between - π and π. From 1998 to 2002, the angular phases get closer than during the rest of the time series, indicating higher synchrony during this period at the annual scale, with only one species, Navicula transitans, maintaining temporal distance with the other species. In order to quantify the degree of synchrony, we have calculated the Phase Angle Variance (PAV) and computed, through a bootstrap approach, the 5% confidence interval below which the synchrony is higher than expected by chance alone. The PAV is almost always below the confidence interval unless in 1994 and between 2005 and 2007. The lower values of PAV (i.e. higher synchrony) were found from 1998 to 2002, when an abrupt decrease in the PAV was observed.

Figure 4.1: Time series of the seasonal mode of the abundance of diatoms present in >40 of the samples. Each color line represents the seasonal mode extracted from the time series of each diatom species. The black line represents the phase angle variance (PAV) of the seasonal component and the black dotted line represents the 95% confidence interval (computed with a bootstrapping scheme based on HMM) to show the significance of the low PAV values. 108

Nutrient inputs in the study area are mainly driven by the dynamics of coastal upwelling, whose intensity may be estimated by means of the Ekman transport equation from the velocity of the components of geostrophic winds. We performed a wavelet transformation on the daily time series of upwelling intensity. The main mode of variability was the annual, which is significant throughout the whole series but fluctuated in amplitude (Figure 4.2). Indeed the highest amplitude was found in 1994-1995 and from 2000 to 2004. Other modes of variation were also present. There was, for example, a 4-year periodic component that is significant during all the series and a 1.5 year component that is significant from 1998 to 2004. An important part of the variability consisted in components whose period is lower than 0.5 (i.e. 6 months) and were present sporadically. Thus, in Figure 4.2c we highlighted the variability for three band-periods: between 14 and 21 days, 1.5 month and 1.5 year. These modes of variation exhibited their maximum amplitude from 1998 to 2002.

Figure 4.2: Wavelet decomposition of daily time series of the upwelling index. a) Local Wavelet Power Spectra (WPS); color code for power values is graded from blue (low values) to dark red (high values), the white lines denote the 5% significance areas determined with a bootstrapping scheme based on HMM (Cazelles et al. 2014), and the black dashed line defines the cone of influence below which the information is affected by edge effect. The magenta dotted line surround the periodicities A, B and C extracted and presented below (in c). b) Global WPS, the black dotted lines denote the 5% significance areas. c) Amplitude of the extracted periodicities for period bands A: 0.04 to 0.06 years (14 to 21 days ), B: 0.14 to 0.15 years (1.5 month) and C: 1.4 to 1.6 years. The vertical blue dotted lines enclose the start of 1998-middle of year 2002, when higher diatom synchrony was found. Chapter 4 109

Model of competition for essential resources

To quantify the synchrony between the two consumers of the model (C1 and C2), we computed the covariance (COV) between both time series at the equilibrium. As the supplies of resources, and consequently the consumer biomasses, were constantly fluctuating, we considered that the equilibrium was reached once the time series of both consumers presented no long-term trend. Indeed, we computed the covariance between model run times 750 and 950, when equilibrium was reached in all simulations. Some combinations of resources uptake U1 and U2 led to the extinction of at least one consumer. Figure 3 shows the covariance between the time series of both consumers within the space of coexistence, for four amplitude values of supply fluctuations: 0.02, 0.4, 0.6 and 0.9.

With a low fluctuation in the amplitude of supply of resource 1 (Figure 4.3a, ai = 0.02), the space of coexistence is dominated by asynchronous pattern with covariance close to -1, unless for the segment in which U1=1-U2 (i.e. where both consumers have the same requirement ratio of resources 1 and 2). Increasing the amplitude of supply of resource R1 results in enhanced synchrony, which spread to the zone surrounding the U1=1-U2 segment

(Figure 4.3b, ai = 0.4) and then to all the space of coexistence (Figure 4.3c and 4.3d, for amplitude values of ai = 0.6 and 0.9 respectively). 110

Figure 4.3: Covariance (COV) between consumer populations (C1 and C2) within the space of coexistence allowed by pairs of values of the per capita uptake rates (U1 and U2) for amplitude values (a1) of resource 1 (R1): (a) 0.02, (b) 0.4, (c) 0.6 ,and (d) 0.9. The crosses in each panel correspond to four U1 and U2 coordinates and more details are presented in the Figure 4.

To have a better understanding of the relationship between the amplitude of the resources supply fluctuation and the covariance between the two competing populations, we selected four pairs of values of the resources uptake rates (U1, U2), whose coordinates in the space of coexistence are highlighted in Figure 4.3a): A (0.24, 0.72), B (0.62, 0.48), C (0.22, 0.18) and D (0.74, 0.40). For those four (U1, U2) coordinates, the covariance increased with the resource supply amplitude in a non-linear fashion (Figure 4.4a). Indeed, the covariance remained unchanged until a tipping point is attained, which is different in each case. For coordinate A, the covariance remained stable and close to -1 until a1 = 0.18, subsequently it increased rapidly. The same situation was found for coordinate B, C and D in which the covariance between consumer populations started to increase for supply amplitudes of 0.32, 0.30 and 0.64 respectively. It is worth noting that in coordinates C and D the increase in covariance is very abrupt. In Figure 4.4b, we presented the dynamics of Chapter 4 111 consumer populations C1 and C2 for the combination of per capita uptake rates of case C (0.22, 0.18) to illustrate the increase in covariance with increasing amplitude of the resources supply (ai of 0.04, 0.3, 0.4 and 0.9). With low amplitude of resource supply (a1), C1 and C2 were not fluctuating in synchrony and the covariance was close to -1. Indeed, the fluctuations in resource 1 are propagated to the consumer 2, through the fluctuations of consumer 1 and its consequent fluctuating consumption of resource 2. As the amplitude of resource supply increase, the resource 1 becomes directly limiting for the consumer 2 and therefore the fluctuations of both consumers get closer.

Figure 4.4 Effect of increasing amplitude in the dynamics of the consumer. (a) Covariance between consumer populations C1 and C2 (COV) as a function of the supply amplitude of Ressouce 1(a1) for the four (U1, U2) coordinates highlighted in figure 3 : A (U1=0.24;U2=0.72), B(U1=0.62;U2=0.48), C(U1=0.22;U2=0.18) and D(U1=0.74;U2=0.40). (b) detail of C1 and C2 fluctuations for the coordinate C(U1=0.22,U2=0.18) and four amplitudes (a1): case 1:0.04, case 2:0.3, case 3: 0.,4 and case 4:0.9. 112

DISCUSSION

We have explored the relationship between the amplitude of environmental variability, community temporal structure and dynamics and inter-specific competition combining empirical analysis and theoretical modeling. We have studied the temporal dynamics of diatom species from a natural marine system where nutrient availability is mainly controlled by coastal upwelling. In addition, we have contrasted those empirical observations with model simulations in which species compete for fluctuating essential resources.

Temporal structure of the community and its drivers

Because it has been linked with their stability, the temporal structure of ecological communities is an important feature. Numerous works have proposed to classify communities on whether they fluctuate in synchrony or in compensatory dynamics (Keitt & Fischer 2006; Houlahan et al. 2007; Vasseur & Gaedre 2007; Downing et al. 2008; Keitt 2008; Gonzalez & Loreau 2009; Jochimsen et al. 2013; Vasseur et al. 2014) and only some of them have discriminated different scales of variability (Vasseur et al 2014, Vasseur and Gaedke 2007). Compensatory dynamics may arise by reciprocal negative interactions among competing species (Loreau and de Mazancourt 2008) but may also be seen in species that have opposite responses to an environmental driver (Ives et al. 1999) or respond to negatively correlated forcing factors (Vasseur et al. 2014). However, the species that can exist in a temporally variable environment are expected to present similar traits and tolerance (Rocha et al. 2011) leading, in a changing environment, to more synchronous dynamics, where all species rise and fall together. As such, it is not surprising that synchronous dynamics is the main mode of variation observed at the annual scale in latitudes where seasonal variations of abiotic and biotic factors control, to a large extent, the abundance of organisms. In the studied area, located at intermediate latitude and strongly influenced by coastal upwelling processes, we have shown that diatom species tend to fluctuate in synchrony, presenting high abundances from April to September (i.e. during the upwelling favorable season). Indeed, the phase angle variance (metric of temporal association, inverse of synchrony) is only sporadically above the confidence Chapter 4 113 interval, even though the temporal association among diatoms varied over time. A previous study on zooplankton dynamics in a southernmost location of the Galician coast (Ría de Vigo) also revealed that the synchronicity was the main mode of variation at the annual scale (Buttay et al 2017). Other studies that have been carried out on plankton communities have also revealed that the temporal structure of the community can change through time and, in particular, they have shown that such changes can occur suddenly (Keitt 2008, Jochimsen 2014, Buttay 2017).

According to present knowledge, mainly based on theoretical studies, different forces are expected to control the community temporal structure. On one hand, abiotic forcing is expected to promote synchrony among species that have similar tolerance range (Rocha 2011). On the other hand, density dependent processes, such as competition or predation, have an antagonist effect by forcing the species population growth to be separate in time in order to coexist (i.e. temporal niche differentiation) (e.g. Vallina et al. 2017). Phytoplankton populations’ growth depends on several factors such as light, temperature and nutrient availability. In the Galician coast, the availability of nutrients is highly controlled by wind-driven upwelling of nutrient-rich sub-surface waters (Fraga 1996, Doval and Madriñon 2016). As such, phytoplankton maximum abundances are mainly found from April to October (see Chapter1). Interspecific competition for nutrients operates when they are limiting, giving rise to the succession of different species through time. Therefore, one can intuitively expect that when nutrients are not limiting, the strength of the competition decreases, allowing the different species of the community to fluctuate in synchrony. From our observations, the period of high synchrony occurring from 1998 to 2002 among diatom species corresponds to a period in which upwelling amplitudes increased at several scales. This is particularly noticeable for the periodic components of 2-3 weeks, 6 weeks and 1.5 year period. We can therefore hypothesize that the higher input of nutrients, mainly of nitrate and silicate for diatoms, linked to enhanced upwelling intensity have decreased the competition among diatom species allowing them to fluctuate in synchrony. Interestingly there is one species (Navicula transitrans) that does not follow the same pattern. It corresponds to a benthic, epiphytic species which present different life history traits (Round et al. 1990) and exhibits higher abundances during winter in the studied area (Ospina- 114

Alvarez et al. 2014) and thus it does not really compete with the other pelagic diatom species.

It is worth noting that synchronous and compensatory dynamics can occur simultaneously but at different timescales (Downing et al 2008 Vasseur and Gaedke) or at different periods (Vasseur and gaedke 2007 Buttay et al 2017, Jochimsen 2014). The sampling frequency of phytoplankton in our study does not allow us to study periodicities below 3 months (such as the 2-3 weeks or 6 weeks periodicities depicted in the upwelling index series). Even though other scales have been explored, synchrony among diatom species has been detected only at the annual scale. In a previous work carried out in the southern part of the Galician coast on zooplankton communities, the upwelling high frequency mode was also indicated to have favored synchrony among zooplankton species at the annual scale (Buttay et al 2017). Indeed, it has been shown that the effects of a perturbation occurring at a specific scale can be redistributed to other frequencies (Greenman and Benton 2005) although there is no clear evidence of the underlying mechanisms.

The results of model simulations are in accordance with our empirical observations as higher amplitude in the fluctuation of one resource supply leads to higher synchrony. In all the space of coexistence, when the amplitude of the resources fluctuation increases, the synchrony increases. However, the relationship between resource supply amplitude and synchrony is not linear. Indeed, covariance remains stable from -1 until a threshold of amplitude at which the covariance starts to increase. The amplitude at which such tipping point is reached, and the slope of the subsequent increase, is different for each stoichiometric requirement combination, but a similar pattern is observed in all cases.

Apart from the cases where both consumers present similar stoichiometric requirements, there is always one consumer more dependent on the fluctuating resource (here, R1) than the other. The temporal association between both consumers presents two modes, before and after the tipping point, summarized in Figure 5. Before the tipping point, the fluctuations in resource 1 (R1) lead population abundance of consumer 1 (C1) to fluctuate. As the population fluctuates, the consumption of the resource 2 (R2), and therefore its concentration, starts to fluctuate, which in turn affects the population size of the second consumer (C2). The fluctuation in R1 is therefore propagated through C1 and Chapter 4 115

R2 before reaching C2 and as this is an indirect, longer path, it creates some lag between the fluctuations of C1 and C2 (i.e. asynchrony). The tipping point is reached when the fluctuations in R1 directly affect both consumers, even if one is more dependent. As both consumers are affected simultaneously, their fluctuation becomes synchronous. While the covariance remains stable before the tipping point, in the vicinity of the tipping point and after, subtle changes in the resource supply amplitude can produce abrupt change in the temporal association of the competitors.

Figure 4.5 How the fluctuation in R1 can affect C2: (a) long way: through the consumption of R2 by C1 or (b) direct way: C2 stoichiometry requirement make it directly limited by R1. Based on Vasseur and Messinger (2011)

Synchrony has been proposed to decrease the stability of the community. There is some empirical evidence showing that the loss of compensatory dynamics makes communities more vulnerable to perturbations and to subsequently suffer a regime shift. Indeed, although Scheffer et al (2009) state that there is little empirical evidence, they proposed that in cyclic systems the phase locking of different oscillators may provide an early warning of critical transition. Since then, some examples have arisen. Jochimsen et al. (2013) showed that the plankton community of the Constance lake (Switzerland), affected by a long-term decrease in phosphate, started to fluctuate in synchrony. They hypothesized that the subsequent regime shift they observed was favored by the loss in compensatory dynamics. Similarly, the zooplankton community of the Ría de Vigo, southern Galician coast, presented an abrupt increase in abundances that started with all the components of the community fluctuating in synchrony (Buttay et al 2017). Yet, regime shift doesn’t occur every time a community presents synchronous dynamics. The A Coruña system studied 116 here does not show any evidences of change in the community, and even if the abundance varied during the synchrony period, they get back to standard values soon afterwards. Overall, the changes in the temporal association among the species that compound the community may be a classical effect of disturbance and indeed some of the effects of environmental changes may go unnoticed at the community level.

CONCLUSION

By using a scale dependent statistical method (i.e Wavelet analysis) on natural marine diatoms time-series we have shown that the temporal structure within a community can fluctuate in time. We identified a period of high synchrony concomitant with increases amplitude of upwelling at different scales of variability. To better understand the relationship between competitors within a fluctuating system and how environmental amplitude can shape the community structure, we employed a theoretical approach. The obtained results go along with our empirical observations and with previous theoretical theories. As such, competition, as a density-dependent process, forces the competitors to fluctuate in compensatory way, until a threshold (tipping point or critical transition) is attained at with environmental forcing becomes the main force that drives population fluctuations. 117

Supplementary Information

Figure SI4.1: Wavelet Power spectrum of a) Nitzschia longissima, b) Pseudo-nitzschia pungens, c) Thalassionema nitzschioides, d) Pseudo-nitzschia delicatissima, e) Navicula transitans, f) Guinardia delicatula, g) Leptocylindrus danicus, h) Chaetoceros socialis, i) Detonula pumila, j) Asterionellopsis glacialis, k) Rhizosolenia imbricata, l) Rhizosolenia setigera. Color code for power values is graded from blue (low values) to dark red (high values), and the black dashrd line defines the cone of influence below which the information is affected by edge effect

119

GENERAL DISCUSSION

General discussion 121

The work carried out in the present Thesis stands almost entirely on the time series of plankton aggregated properties or community composition gathered month after month within the framework of the RADIALES monitoring program. The series are now of considerable length allowing to describe the variability of plankton along the N and NW Iberian shelf, to investigate the possible drivers of plankton variability and to explore some new aspects of ecological communities: their temporal structure (Figure D1).

Figure D.1: Thesis chapters in a glimpse

Overall, the work developed in this Thesis contributes to different aspects: 1) description of zooplankton variability, giving reference values of “normal” zooplankton fluctuations, which would be of interest to recognize natural variability from possible future anthropogenic disturbances; 2) assessment of the drivers of zooplankton variability at several temporal scales; 3) brings new empirical evidence and proposes theoretical mechanisms to explain how communities can shift from compensatory dynamics to synchrony, a subject that has recently become of major interest in ecology since it has been suggested to alter community stability. 122

As all the results obtained have been already discussed within their corresponding chapters, in this general discussion I will focus on the main ideas and findings that emerge out of this Thesis and the related literature.

Scale matters

Much of the literature, and some of the present Thesis chapters, studying the effects of environmental forcing on living organisms, have focused on changes in mean environmental conditions, despite organisms usually experience variation of such conditions over many time scales and not only the mean (Dillon et al. 2016). This may be particularly relevant for planktonic organisms that rarely have a multi-year life-span and so experience the environmental conditions of only a part of the year.

Natural environments are high-dimensional systems that we only observe at a limited range of scales and while in some cases the scale of observations have been determined by the feature of the targeted process, but in other cases it is forced by our perceptual capabilities and/or practical limitations (Levin 1992).

Temporal and spatial dimensions are the most evident ones, but complexity of the organization levels at which biological entities could be analyzed can also be important. To center the discussion on plankton organisms within marine ecosystems, behind the spatial variability we can find processes such as daily foraging migrations, passive drifting and larval dispersal, as well as environmental range. Yet, if all those aspects were not targeted directly in the present work, they can impact the patterns we observe in the temporal dimension. Indeed, within the first chapter, we have shown that by sorting the time-series of zooplankton total biomass according to their periodic components, we were able to establish different spatial domains along the north and northwest Iberian shelf.

Then, within the temporal dimension, different processes can occur at different scales, even antagonist processes occurring simultaneously. To get back to what Levin (1992) called “perceptual capabilities”, there are some temporal scales that we understand easily because we experience them (e.g. day, week, year, generation time, etc.), while other seem more abstract and are probably less documented and studied. One additional dimension that has General discussion 123 been reported in the present Thesis to have importance is the organizational complexity: different levels of the ecosystems can show different patterns. For example, intra- and inter- specific density dependence can provoke variability at the individual or population levels that are not necessarily perceptible at the community level. In addition, environmental fluctuations do not just add extra variation to existing dynamics but can significantly modify the dynamics of the whole system (Chesson, 2003). Indeed, it has been shown that the effects of an environmental perturbation occurring at a specific scale can be redistributed to other frequencies (Greenman and Benton 2005).

It is important to acknowledge that different processes occur at different scales in order to adapt the scale at which the observations are made. With an insufficient sampling strategy we would not only be unable to recognize patterns but also obtain wrong conclusions. One of the more graphic examples to illustrate the problem of insufficient temporal sampling are videos of running cars whose wheels seem to be going in the wrong side just because the video frame frequency is slightly lower than the wheel frequency. Yet behind a monitoring program such as the RADIALES project, that focus on both biotic and abiotic components of the pelagic system, the sampling frequency, decided a priori, corresponds to a trade-off between the scale of the process we should observe and the logistic aspect of doing so in the long term. As such, it is also important to acknowledge the scale of observation in order to know which scale of processes can be tackled.

In the case of the data analysed within this Thesis, the monthly sampling of RADIALES implies that we can only look for processes that occur at periodicities above 3 months. Similarly, the length of the series can give the upper limit of the process scale we can study. For example, some literature suggest an association between biological variability and the Atlantic Multidecadal Oscillation (Edwards et al. 2013a) but the AMO has a period of 60 years and the period sampled within the RADIALES project (about 20 years) corresponds only to a monotonic increase in AMO. 124

Plankton variability and its drivers.

In the present Thesis, different aspects of plankton variability have been observed (Figure D.1). They concern mainly the temporal dimension of plankton variability, but some spatial differences have been encountered by comparing the dynamics of the different RADIALES sampling stations (Chapter 1 and 2). And, while I previously alleged the importance of the scale of observation and scale of processes, I have also been using annual means of plankton and environmental condition variables and/or aggregated variables. As such, within the first chapter of the Thesis, only aggregated variables have been used to describe the zooplankton variability. This approach was chosen as a first exploration of the zooplankton variability in all the sampled stations and also to allow a comparison between the Cantabrian stations, where taxonomic composition information was not available, and the Galician stations. Nevertheless, more detailed analyses were performed in the following chapters. In the second chapter, I used also aggregated variables (total abundance) at the monthly frequency in addition to the abundance of the main taxonomic groups. Finally, in chapters 3 and 4, the taxonomic composition was explicitly considered.

Spatial patterns

Spatial variability is associated with the vertical and horizontal movements of water and, at the scale of the RADIALES sampling area, to the different environmental conditions that are present along the northern Iberian Atlantic coast. Within the first chapter we have observed differences between the series related to coastal-ocean gradients or longitudinal gradients. For example, we observed that the coastal zooplankton community was made, on average, of comparatively smaller individuals in average than the oceanic ones. The more oceanic stations, located at the shelf break, presented also higher variability. This was particularly evident in the Cantabrian stations of Gijon and Santander (E3GI and E6SA) but we also noticed it in the Vigo transect (Chapter 2 and 3) where the outer station (E3VI) presented greater variability than the inner station (E1VI). In addition, the period of high abundance for most of the taxonomic groups considered in the Vigo section was longer in the coastal station (E1VI) than in the oceanic one (E3VI, Chapter 2). We hypothesized that the higher variability encountered at the E3Gi, E6SA General discussion 125 and E3Vi stations arise from their location, where the moving frontal structure makes them fluctuate between coastal and oceanic domains. This highlighted one of the limitations of sampling at fixed stations (Eulerian sampling scheme) and indeed, in order to detect biotic or abiotic forcing, the stations E6SA and E3GI are probably not the most suitable ones. The variability arising from their location at the shelf break may blur subtle environmental effects. One station that will be soon of interest is the more recently (2012) implemented E4GI station, a truly oceanic site located at the north of Cape Peñas at the 4640m depth that would receive little continental influences.

Overall, by comparing the periodicities presented in each time series we identified 3 main groups of stations: the more oceanic Cantabrian stations characterized by high variability, the coastal Cantabrian stations in which the typical seasonal pattern of two annual blooms was found and the Galician ones that, influenced by the seasonal upwelling, presented a long period of relatively high amplitude.

Changes in abundance or biomass

Changes in abundances are often the first identified effects of abiotic forcing. In the northern Atlantic coast of Spain, despite the differences between the Galician and Cantabrian facades, we observed that the total biomass and abundance increased in all stations (Chapter 1). Similarly, an increase in biomass and abundance of zooplankton between 1993 and 2010 was reported in the sampling stations located off Cudillero (Cantabrian coast, González-Gil et al., 2015). The similar positive long-term trends were observed in all the stations located along the Galician and Cantabrian coasts, suggesting that processes occurring at a large spatial scale are driving such long-term dynamics.

From the Cudillero time-series, Gonzalez-Gil et al. (2015) suggested that the positive trend in zooplankton biomass was related to intensification of winter mixing and of coastal upwelling towards the end of the summer. Focussing on a longer period, in the Galician coast, the increase of coastal zooplankton and decrease in phytoplankton since the beginning of the 90s has been suggested to be a consequence of a reduction of the upwelling intensity that occurred from 1965 to 2010 through a limitation of the offshore exportation of planktonic organisms (Bode 2009) and an enhanced stratification (Pérez et al 2010). 126

In contrast, in our temporal window of observation, we did not observe a continuous decrease in upwelling index but a decrease until 2002 in the annual mean followed by an increase until the end of the series. Indeed, in the work by Pérez et al. 2010, the decreasing trend in upwelling intensity also seems to break in 2002. Much of the upwelling variability occurs at short-term scales (i.e. high frequency) and the annual mean of upwelling, as used in chapter 2, or the annual mean of the favourable month, as used in the studies of Bode et al. 2009 and Perez et al 2010, only provide a partial view of the upwelling behaviour as they give, for instance, the same weight to very occasional pulses of strong upwelling and to sustained moderate upwelling events. In general, primary production is favoured by upwelling events that last several days and are followed by a period of relaxation, so the primary production can occur and is retained at the coast (Tenore et al. 1995; Joint et al. 2002).

In the Ría de Vigo time series analysed in chapters 2 and 3, the changes in abundance and biomass appeared abrupt. Indeed, in chapter 2, we observed three contrasted periods at E1VI and E3VI : The first one [1994-2001] of low abundance and biomass, the second one [2001-2006] of high values and the third one [2006-2010] of average values. During the period of high zooplankton abundance, sustained changes in the physical environment were identified. Those changes included a decreasing trend in precipitation and increasing trends for upwelling intensity at the local scale as well as a southward displacement of the Gulf Stream’s north wall position, which is associated with change in the atmospheric circulation pattern and has been proposed as an indicator of changes at the scale of the North Atlantic ocean (Taylor and Stephens 1980; Taylor 1995b). It is worth noting that in the Western English Channel, similar long-term periods (1995–2000, 2001–2007, 2008– 2012) were encountered to describe zooplankton long term variability (Reygondeau et al. 2015). Interestingly, these authors also encountered a progressive long-term modification of the environment which at the local scale consisted in a progressive intensification of the warm period and a decrease in the intensity and depth of the thermocline, which was related to large scale climatic modes such as the North Atlantic Oscillation and lead to abrupt zooplankton dynamics. Such global climate indices

(e.g NAO or Gulf StreamNWP) do not necessarily have a strong link with local-to-regional weather conditions, but they reduce complex space and time variability into simple measures (Stenseth et al. 2003) and have been shown to be good predictors of ecological processes (Planque and Fromentin 1996; Post and Forchhammer 2002; Ménard et al. 2007). General discussion 127

From this 2nd chapter, we can hypothesize that the second period, which shows the maximum values, corresponds to the community response of zooplankton organisms facing a disturbance. Within the third period, of post-disturbance, we noticed that the abundances stepped down but remained higher than during the first period. It is not clear whether the changes observed in Vigo constitute a regime shift or not. First, because we have no evidence (no data) that other components of the ecosystem where affected and second because despite abrupt changes in zooplankton total abundance and biomass occurred, they affected all taxonomic groups and, apparently, the structure of the community presented no significant changes (Chapter 1 and 2). Following up the changes observed in the Vigo stations, in the third chapter we identified that the environmental change also corresponded to an enhanced seasonality of upwelling and river outflow from 2000 to 2004, combined with a shortening of upwelling events that could have possibly favoured retention of organisms during winter (reduction of off-shore exportation) and primary production in summer.

Interestingly, right before the abrupt increase, the zooplankton at different levels of organization (broad taxonomic groups and copepod species) started to fluctuate in synchrony. According to theory and some empirical evidences (Keitt et al. 2006; Keitt 2008; Gonzalez and Loreau 2009; Jochimsen et al. 2013), the synchronization, or loss of compensatory dynamics among-population fluctuations, can be interpreted as the response of a community facing disturbance. Synchronization has also been proposed to reduce the community stability and facilitate the occurrence of regime shifts. It is worth noting that detecting changes in the temporal structure of a community requires both composition data and specific statistical tools. Such changes may therefore constitute an ubiquitous community response to perturbations that may go unnoticed at the level of aggregated community variables or when all the temporal scales of variability are mixed (Mutshinda et al. 2009).

128

Phenology and community temporal structure.

One of the identified major effects of climate change is the significant modification of seasonal timing among very diverse organisms and ecosystems (Durant et al. 2007; Mackas and Beaugrand 2010; Ovaskainen et al. 2013).

In mid and high latitudes, the seasonal influence sets much of the total variability experienced by the planktonic organisms (Longhurst 1998; Sheridan and Landry 2004). Consequently, through all the chapters of the present Thesis, we have seen that plankton variability occurs mainly at the annual scale, similarly to their main drivers such as temperature and nutrient availability. Indeed, species have behavioural and life history strategies that exploit favourable periods of the year for growth and reproduction (those matching with their optimal niche requirements) while minimizing the exposure of critical life stages to unfavourable periods. As such, phenological variability is primarily observed as changes in population size. For example, the different seasonal drivers observed along the Galician and Cantabrian coast have resulted in different seasonal patterns of zooplankton (Chapter 1).

Phenological changes can have strong consequences for the fitness of a population because the recruitment success of a predator is conditioned by the spatial and temporal overlap with its prey (e.g. match-mismach hypothesis; Cushing, 1990). Additionally, the decoupling of phenological relationships can impact all trophic levels. For example, in the North Sea, the extended duration of several phytoplankton species’ blooms has been shown to impact the food web structure of the whole community by causing a complete reorganization of the trophic food web (Edwards et al. 2004).

Seasonal fluctuations of predators are therefore often associated with the seasonal pattern of their prey, but among functionally similar species, compensatory dynamics is the mode of variation expected because the exploitation of similar niches would require at least some temporal differentiation to allow co-existence. While environmental forcing appears to control an important part of the phenological characteristics of populations, biotic forcing, such as density dependent processes, is also susceptible to drive phenological changes. In the 3rd chapter we identified changes in the temporal structure of the community that where related to the functioning of the upwelling that controls both the re-circulation of nutrients and the General discussion 129 off-shore exportation of planktonic organisms. Because nutrient and phytoplankton data were not fully available at the Vigo stations, the mechanism linking shorter upwelling events with increase of zooplankton abundance, biomass and synchrony is not clear. Theoretically, compensatory dynamics is promoted by density-dependent processes such as competition or predation, so it could be expected that a reduction in the strength of density dependent processes would allow more synchrony. In general, there is a balance between the environmental forcing, which can be mainly identified as the synchronizing force, and the density dependent processes that promote compensatory dynamics. As such, we can hypothesize that, in the Vigo stations, the modification of the upwelling event duration promoted more primary production and therefore reduced the competition among zooplankton .

To assess those questions, we focused on a lower trophic level (phytoplankton community) from the A Coruña station (E2CO), where taxonomic composition data were available. Diatom species were the group showing the more noticeable phenological change.

Figure D.2: Phase of diatoms species at the 1st of June. Each line corresponding to one species series.

In Figure D.2 are presented the phases of each diatom species at the 1st of June from 1994 to 2009. This representation allows to follow ‘absolute’ changes in phenology (i.e. in reference to a fixed temporal point) while the synchrony accounts the relative temporal change in relation to the other species. Each species presents high inter-annual variations of almost π 130

(which correspond for the annual oscillation to 6 months). There is also a large variability between species. This dispersion among phases, what we calculated as the Phase Angle Variance (PAV), is inverse to synchrony. Indeed, within the 4th chapter, we have shown that the 12 dominant species of diatoms presented higher synchrony from 1998 to 2002. Interestingly, this period was characterized by higher variance in the high frequency mode of the upwelling, similarly to what we observed in chapter 3.

Figure D.3 : Local wavelet power spectrum of the abundance (ind.m−3) for the time-series of the 6 main taxonomic groups in the A Coruña station E2CO. Color code for power values is graded from blue (low values) to dark red (high values), and the black line defines the cone of influence below which the information is affected by edge effect. The black dotted lines denote the 5% significance areas determined with a bootstrapping scheme based onHidden Markov Model (Cazelles et al. 2014). General discussion 131

By using a mechanistic theoretical model (Chapter 4), we identified synchrony as a possible response to an increase in nutrient fluctuations. Interestingly, the relationship between nutrient fluctuation amplitude and synchrony was not linear. As such, in a similar way to natural systems close to the tipping point, only very subtle changes in nutrient fluctuation amplitude are needed to produce abrupt modifications of species temporal associations.

In chapter 4, we focused on the bottom up mechanisms in which changes of nutrient supply affect diatom dynamics. Yet, there are some indications that top-down control may also have occurred. The increase of diatoms between 1998 and 2002, also observable in Figure D.3, may have resulted in a bottom up configuration and an increase in zooplankton. However, abundances of copepods, decapods and cladocerans were low during this period while chaetognaths abundance increased.

Despite being a simplistic view of the planktonic food web, because some groups (i.e. copepods) encompass organisms with diverse feeding habits, the concurrence of an increase in chaetognaths, a decrease in copepods and an increase in diatoms could reflect a top-down control, even though confirming it would require a more specific study. It is worth noting that a decrease in predation strength is another possible mechanism that would reduce diatom compensatory dynamics (Vasseur and Fox 2009).

To my best knowledge, in the regime shifts described in the literature, synchrony has been quantified and encountered in the Lake Constance regime shift (Jochimsen et al. 2014). However, it has been suggested by Sheffer (2009) that the synchronization of the different ‘oscillators’ of a cyclic ecological system is an early warning of critical transition (i.e. regime shift). We have observed periods of enhanced synchrony among zooplankton species in Vigo and among diatoms species in A Coruña driven by a similar disturbance (increase in short-term variability of upwelling). However, while aggregated abundance of diatoms returned to initial levels of abundances after the disturbance, in the Vigo station, zooplankton biomass and abundance have shifted to another stable state of abundance and biomass. It would be interesting then to explore the structure of both communities to understand why one seems more resilient than the other. A more ambitious task would be to re-examine identified regime shift to test whether synchrony is an ubiquous response to disturbance linked to critical transitions.

133

CONCLUSIONS

135

• In all the RADIALES stations sampled for zooplankton, the zooplankton total biomass increased • Three main regions where were depicted according to the periodicities presented by zooplankton biomass: o Outer shelf, characterized by high variability and larger organisms in average, o Coastal-shelf Cantabrian Sea, presenting both spring and autumns peaks of biomass. o Coastal-shelf of Galicia, characterized by one main peak of maximum abundance centered around summer. • In the long term, the time series of zooplankton total abundance exhibited three contrasting periods in the Ría de Vigo and adjacent shelf: A) 1995–2001, characterized by low abundance and low amplitude seasonality, with a stepped increase towards 2001; B) 2001–2006, of high abundance and marked seasonality enclosing the maximum values of the time series; and C) 2006–2010, of intermediate abundance and amplitude of the seasonal cycle. • The observed changes in zooplankton dynamics were concomitant with sustained trends for upwelling intensity (increasing), precipitation (decreasing) and Gulf Stream North Wall position (equatorward) between 2000 and 2005. • In the Ría de Vigo, the annual oscillation of biomass and abundance increased in 2000 corresponding to the highest amplitudes of environmental forcing. Concomitantly, enhanced synchrony was observed among the main taxonomic groups of zooplankton and among copepod species • The amplified seasonality of the upwelling index and river outflows from 2000 to 2004, combined with a reduction of off-shore exportation by shortening of upwelling events, may have favoured retention in winter, and enhanced primary production in summer. • We observed a sudden increase of the synchrony between diatom species from 1998 to 2002 in the Ría de A Coruña, concomitant with an increase in the amplitude of upwelling intensity at different scales. • By simulating the dynamics of two species with fixed stoichiometric requirements competing for two essential resources we have shown that an increase in the amplitude of resource supply leads in all cases to enhanced synchronicity between competitors. • The relationship between environment fluctuation amplitude and synchrony is not linear because inter-specific competition buffers the effect of increasing amplitude until certain thresholds (tipping point or critical transition) at which environmental fluctuations become the main drivers of community temporal structure.

137

APPENDIX

Table A.1: List of taxonomic groups considered in chapter 2 and 3

Taxonomic_Group Kingdom Phylum Class Order/suborder AphiaID Author Foraminifera Chromista Foraminifera 1410 nd Polychaeta Animalia Annelida Polychaeta 883 Grube, 1850 Maxillopoda (Infraclass = Cirripedia larvae Animalia Arthropoda Cirripedia) 1082 Burmeister, 1834 Maxillopoda (Subclass = Milne Edwards, Copepoda Animalia Arthropoda Copepoda) 1080 1840 Decapoda larvae Animalia Arthropoda Malacostraca Decapoda 1130 Latreille, 1803 Cladocera Animalia Arthropoda Branchiopoda Diplostraca/Cladocera 1076 Latreille, 1829 Bryozoa larvae Animalia Bryozoa 146142 nd Chaetognatha Animalia Chaetognatha 2081 nd Larvacea Animalia Chordata Larvacea/Appendicularia 17446/146421 nd Cnidaria Animalia Cnidaria 1267 Verrill, 1865 Siphonophorae Animalia Cnidaria Hydrozoa Siphonophorae 1371 Eschscholtz, 1829 Echinodermata larvae Animalia Echinodermata 1806 Bruguière, 1791 Gastropoda larvae Animalia Mollusca Gastropoda 101 Cuvier, 1795

Table A.3: List of Copepod species considered in the chapter 3

Species Order/suborder Family AphiaID Author Calanus helgolandicus Calanidae 104466 Claus, 1863 Calanoides carinatus Calanoida Calanidae 104462 Kroyer, 1849 Nannocalanus minor Calanoida Calanidae 104469 Claus, 1863 Calocalanus styliremis Calanoida Calocalanidae 104673 Giesbrecht, 1888 Clausocalanus Calanoida Clausocalanidae 104502 Dana, 1849 arcuicornis Pseudocalanus elongatus Calanoida Clausocalanidae 104515 Boeck, 1865 Centropages chierchiae Calanoida Centropagidae 104494 Giesbrecht, 1889 Paraeuchaeta hebes Calanoida Euchaetidae 104563 Giesbrecht, 1888 Mecynocera clausi Calanoida Paracalanidae 104616 Thompson, 1888 Paracalanus parvus Calanoida Paracalanidae 104685 Claus, 1863 Temora longicornis Calanoida Temoridae 104878 Müller, 1785 Oithona nana Oithonidae 106651 Giesbrecht, 1893 Oithona plumifera Cyclopoida Oithonidae 106652 Baird, 1843 Oithona similis Cyclopoida Oithonidae 106656 Claus, 1863 Euterpina acutifrons Euterpinidae 116162 Dana, 1847 Oncaea media Oncaeidae 128938 Giesbrecht, 1891

Table A.3: Diatom species considered in the chapter 4 (Kingdom: Chromista; Phylum: Ochrophyta ; Class: Bacillariophyceae)

Nombre_referencia Order Family AphiaID Authority (Brébisson in Kützing) Ralfs in Pritchard, Nitzschia longissima Bacillariales Bacillariaceae 149150 1861 Pseudo-nitzschia Bacillariales Bacillariaceae 149153 (P.T. Cleve, 1897) Heiden, 1928 delicatissima Pseudo-nitzschia pungens Bacillariales Bacillariaceae 160528 (Grunow ex P.T. Cleve, 1897) Hasle, 1993 Chaetocerotanae incertae Chaetoceros socialis Chaetocerotaceae 149123 H.S.Lauder, 1864 sedis Asterionellopsis glacialis Fragilariales Fragilariaceae 149139 (Castracane) Round, 1990 Leptocylindrus danicus Leptocylindrales Leptocylindraceae 149106 Cleve, 1889 Navicula transitans Naviculales Naviculaceae 149320 Cleve, 1883 Guinardia delicatula Rhizosoleniales Rhizosoleniaceae 149112 (Cleve) Hasle, 1997 Rhizosolenia imbricata Rhizosoleniales Rhizosoleniaceae 149116 Brightwell, 1858 Rhizosolenia setigera Rhizosoleniales Rhizosoleniaceae 149115 Brightwell, 1858 Detonula pumila Thalassiosirales Skeletonemaceae 149647 (Castracane) Gran, 1900 Thalassionema Thalassionematales Thalassionemataceae 149093 (Grunow) Mereschkowsky, 1902 nitzschioides

143

REFERENCES

Akima, H., A. Gebhardt, T. Petzoldt, and M. Maechler. 2013. akima: Interpolation of irregularly spaced data. R Packag. Albaina, A., and X. Irigoien. 2007. Fine scale zooplankton distribution in the Bay of Biscay in spring 2004. J. Plankton Res. 29: 851–870. doi:10.1093/plankt/fbm064 Alcalà, M. R. D., F. Conversano, F. Corato, and others. 2004. Seasonal patterns in plankton communities in a pluriannual time series at a coastal Mediterranean site ( Gulf of Naples ): an attempt to discern recurrences and trends *. 68: 65–83. Alvarez-Salgado, X. A., G. Rosón, F. F. Pérez, and Y. Pazos. 1993. Hydrographic variability off the Rías Baixas (NW Spain) during the upwelling season. J. Geophys. Res. Ocean. 98: 14447–14455. doi:10.1029/93JC00458 Alvarez, I., N. Ospina-Alvarez, Y. Pazos, and others. 2009. A winter upwelling event in the Northern Galician Rias: Frequency and oceanographic implications. Estuar. Coast. Shelf Sci. 82: 573–582. doi:10.1016/j.ecss.2009.02.023 Arístegui, J., E. D. Barton, X. A. Álvarez-Salgado, and others. 2009. Sub-regional ecosystem variability in the Canary Current upwelling. Prog. Oceanogr. 83: 33–48. doi:10.1016/j.pocean.2009.07.031 Atkinson, A., R. A. Harmer, C. E. Widdicombe, and others. 2015. Questioning the role of phenology shifts and trophic mismatching in a planktonic food web. Prog. Oceanogr. 137: 498–512. doi:10.1016/j.pocean.2015.04.023 Bakun, A. 1973. Coastal upwelling indices, west coast of North America, 1946-71. National Oceanic and Atmospheric Administration. Coast. upwelling indices, west coast North Am. 1–3. Bakun, A. 1990. Global climate change and intensification of coastal ocean upwelling. Science (80-. ). 247: 198–201. doi:10.1126/science.247.4939.198 Banse, K. 1995. Zooplankton : Pivotal role in the control of ocean production. ICES J. Mar. Sci. 52: 265–277. doi:10.1016/1054-3139(95)80043-3 Batten, S. D., E. S. Fileman, and E. Halvorsen. 2001. The contribution of microzooplankton to the diet of mesozooplankton in an upwelling filament off the north west coast of Spain. Prog. Oceanogr. 51: 385–398. doi:10.1016/S0079-6611(01)00076-3 Beaugrand, G. 2004. The North Sea regime shift: Evidence, causes, mechanisms and consequences. Prog. Oceanogr. 60: 245–262. doi:10.1016/j.pocean.2004.02.018 Beaugrand, G., F. Ibañez, and P. C. Reid. 2000. Spatial, seasonal and long-term fluctuations of plankton in relation to hydroclimatic features in the English Channel, Celtic Sea and Bay of Biscay. Mar. Ecol. Prog. Ser. 200: 93–102. Beaugrand, G., and P. C. Reid. 2012. Relationships between North Atlantic salmon, plankton, and hydroclimatic change in the Northeast Atlantic. ICES J. Mar. Sci. 69: 1205–1217. doi:10.1093/icesjms/fst048 Berens, P. 2009. CircStat: A MATLAB toolbox for circular statistics. J. Stat. Softw. 31: 1–21. 144

doi:10.1002/wics.10 Blanco-Bercial, L., F. Álvarez-Marqués, and J. Cabal. 2006. Changes in the mesozooplankton community associated with the hydrography off the northwestern Iberian Peninsula. ICES J. Mar. Sci. 63: 799–810. doi:10.1016/j.icesjms.2006.03.007 Blanton, J. O. O., K. R. Tenore, F. Castillejo, L. P. Atkinson, F. B. Schwing, and A. Lavin. 1987. The relationship of upwelling to mussel production in the rias on the western coast of Spain. J. Mar. Res. 45: 497–511. Blasius, B., a Huppert, and L. Stone. 1999. Complex dynamics and phase synchronization in spatially extended ecological systems. Nature 399: 354–359. doi:10.1038/20676 Bode, A., M. T. Alvarez-Ossorio, J. M. Cabanas, A. Miranda, and M. Varela. 2009a. Recent trends in plankton and upwelling intensity off Galicia (NW Spain). Prog. Oceanogr. 83: 342–350. Bode, A., M. T. Alvarez-ossorio, N. Gonzalez, M. T. Álvarez-Ossorio, and N. González. 1998. Estimations of mesozooplankton biomass in a coastal upwelling area off NW Spain. J. Plankton Res. 20: 1005–1014. doi:10.1093/plankt/20.5.1005 Bode, A., M. T. Álvarez-Ossorio, N. González, J. Lorenzo, C. Rodríguez, M. M. M. Varela, and M. M. M. Varela. 2005. Seasonal variability of plankton blooms in the Ria de Ferrol (NW Spain): II. Plankton abundance, composition and biomass. Estuar. Coast. Shelf Sci. 63: 285–300. doi:10.1016/j.ecss.2004.11.021 Bode, A., M. T. M. T. T. M. T. Alvarez-Ossorio, J. M. M. Cabanas, A. Miranda, and M. Varela. 2009b. Recent trends in plankton and upwelling intensity off Galicia (NW Spain). Prog. Oceanogr. 83: 342–350. doi:10.1016/j.pocean.2009.07.025 Bode, A., M. T. T. Alvarez-Ossorio, S. Barquero, J. Lorenzo, A. Louro, and M. Varela. 2003. Seasonal variations in upwelling and in the grazing impact of copepods on phytoplankton off A Coruña (Galicia, NW Spain). J. Exp. Mar. Bio. Ecol. 297: 85–105. doi:10.1016/S0022-0981(03)00370-8 Bode, A., R. Anadón, J. Lorenzo, M. Llope, L. Mene, X. A. G. Morán, E. Teira, and M. Varela. 2012a. Biomasa y produccíon primaria, p. 199–220. In A. Bode, A. Lavín, and L.V.V.L. A. Lavín [eds.], Cambio climático y oceanográfico en el Atlántico del norte de España. Instituto Español de Oceanografía. Bode, A., L. Lavín, and L. Valdés, eds. 2012b. Cambio climático y oceanográfico en el Atlántico del norte de España.Temas de Oceanografía, 5, Instituto. Botas, J. A., E. Fernández, A. Bode, and R. Anadón. 1989. Water masses off the central Cantabrian coast. Sci. Mar 53: 755–761. Bowden, K. F. 1975. Physical factors: salinity, temperature, circulation and mixing processes, p. 38–70. In E. Olausson and I. Cato [eds.], Chemistry and Biogeochemistry of Estuaries. Bretherton, C., C. Smith, and J. Wallace. 1992. An Intercomparison of method for finding coupled pattern in climate data. J. Clim. 5: 541–559. Burnham, K. P. P., and D. R. R. Anderson. 2002. Model selection and multimodel inference: a practical information-theoretic approach, Springer. 145

Buttay, L., B. Cazelles, A. Miranda, G. Casas, E. Nogueira, and R. González-Quirós. 2017. Environmental multi-scale effects on zooplankton inter-specific synchrony. Limnol. Oceanogr. doi:10.1002/lno.10501 Buttay, L., A. Miranda, G. Casas, G.-Q. Rafael, and E. Nogueira. 2016. Long-term and seasonal zooplankton dynamics in the northwest Iberian shelf and its relationship with meteo-climatic and hydrographic variability. J. Plankton Res. 38: 106–121. doi:10.1093/plankt/fbv100 Cabal, J., G. González-Nuevo, and E. Nogueira. 2008. Mesozooplankton species distribution in the NW and N Iberian shelf during spring 2004: Relationship with frontal structures. Oceanogr. Bay Biscay 72: 282–297. doi:10.1016/j.jmarsys.2007.05.013 Casabela, N., M. Lorenzo, and J. Taboada. 2014. Trends of the Galician upwelling in the context of climate change. J. Sea Res. 93: 23–27. Casas, B., M. Varela, M. Canle, N. Gonza, and A. Bode. 1997. Seasonal Variations of Nutrients , Seston and Phytoplankton , and Upwelling Intensity o ff La Corun ˜ a ( NW Spain ). 767–778. Cazelles, B., S. Bottani, and L. Stone. 2001. Unexpected coherence and conservation. Proc. Biol. Sci. 268: 2595–2602. doi:10.1098/rspb.2001.1843 Cazelles, B., K. Cazelles, and M. Chavez. 2014. Wavelet analysis in ecology and epidemiology : impact of statistical tests. J. R. Soc. Interface 11. doi:10.1098/rsif.2013.0585 Cazelles, B., M. Chavez, D. Berteaux, F. Ménard, J. O. Vik, S. Jenouvrier, and N. C. Stenseth. 2008. Wavelet analysis of ecological time series. Oecologia 156: 287–304. doi:10.1007/s00442-008-0993-2 Cazelles, B., and S. Hales. 2006. Infectious diseases, climate influences, and nonstationarity. PLoS Med. 3: 1212–1213. doi:10.1371/journal.pmed.0030328 Cazelles, B., and L. Stone. 2003. Detection of imperfect population synchrony in an uncertain world. J. Anim. Ecol. 72: 953–968. doi:10.1046/j.1365-2656.2003.00763.x Ceballos, S., and F. Álvarez-Marqués. 2006. Seasonal dynamics of reproductive parameters of the calanoid copepods Calanus helgolandicus and Calanoides carinatus in the Cantabrian Sea (SW Bay of Biscay). Prog. Oceanogr. 70: 1–26. doi:10.1016/j.pocean.2006.03.002 Ceballos, S., and A. Ianora. 2003. Different diatoms induce contrasting effects on the reproductive success of the copepod Temora stylifera. J. Exp. Mar. Bio. Ecol. 294: 189– 202. doi:10.1016/S0022-0981(03)00263-6 Chatfield, J. 1989. The analysis of time series: an introduction., Chapman & Hall, London. Chesson, P. 2003. Understanding the role of environmental variation in population and community dynamics. Theor. Popul. Biol. 64: 253–254. doi:10.1016/j.tpb.2003.06.002 Chiba, S., and T. Saino. 2003. Variation in mesozooplankton community structure in the Japan/East Sea (1991–1999) with possible influence of the ENSO scale climatic variability. Prog. Oceanogr. 57: 317–339. doi:10.1016/S0079-6611(03)00104-6 Church, J. A., and N. J. White. 2006. A 20th century acceleration in global sea-level rise. 146

Geophys. Res. Lett. 33: 94–97. doi:10.1029/2005GL024826 Colebrook, J. M. 1979. Continuous Plankton Records: Seasonal cycles of phytoplankton and copepods in the North Atlantic ocean and the North Sea. Mar. Biol. 51: 23–32. doi:10.1007/BF00389027 Colebrook, J. M. 1985. Continuous plankton records: overwintering and annual fluctuations in the abundance of zooplankton. Mar. Biol. 84: 261–265. doi:10.1007/BF00392495 Daubechies, I. 1992. Ten Lectures on Wavelets,. Devreker, D., S. Souissi, and L. Seuront. 2005. Effects of chlorophyll concentration and temperature variation on the reproduction and survival of Temora longicornis (Copepoda, Calanoida) in the Eastern English Channel. J. Exp. Mar. Bio. Ecol. 318: 145– 162. doi:10.1016/j.jembe.2004.12.011 Dillon, M. E., H. A. Woods, G. Wang, S. B. Fey, D. A. Vasseur, R. S. Telemeco, K. Marshall, and S. Pincebourde. 2016. Life in the frequency domain: The biological impacts of changes in climate variability at multiple time scales. Integr. Comp. Biol. 56: 14–30. doi:10.1093/icb/icw024 Downing, A. L., B. L. Brown, E. M. Perrin, T. H. Keitt, and M. A. Leibold. 2008. Environmental fluctuations induce scale-dependent compensation and increase stability in plankton ecosystems. Ecology 89: 3204–3214. doi:10.1890/07-1652.1 Drinkwater, K. F., A. Belgrano, A. Borja, and others. 2013. The response of marine ecosystems to climate variability associated with the North Atlantic Oscillation J.W. Hurrell, Y. Kushnir, G.O. And, and M. Visbeck [eds.]. … Oscil. Clim. … 134: 211–234. Durant, J. M., D. Ø. Hjermann, G. Ottersen, and N. C. Stenseth. 2007. Climate and the match or mismatch between predator requirements and resource availability. Clim. Res. 33: 271–283. Dussart, B. H. 1965. Les différentes catégories de plancton. Hydrobiologia 26: 72–74. Earn, D. J., P. Rohani, and B. T. Grenfell. 1998. Persistence, chaos and synchrony in ecology and epidemiology. Proc. Biol. Sci. 265: 7–10. doi:10.1098/rspb.1998.0256 Edwards, M., G. Beaugrand, P. Helaouët, J. Alheit, and S. Coombs. 2013a. Marine Ecosystem Response to the Atlantic Multidecadal Oscillation ¨. 8: 1–5. doi:10.1371/journal.pone.0057212 Edwards, M., G. Beaugrand, P. Helaouët, J. Alheit, and S. Coombs. 2013b. Marine Ecosystem Response to the Atlantic Multidecadal Oscillation. PLoS One 8: 1–5. doi:10.1371/journal.pone.0057212 Edwards, M., J. Richardson, and A. J. Richardson. 2004. Impact of climate change on marine pelagic phenology and trophic mismatch. Nature 430: 881–884. Engen, S., and B.-E. Saether. 2005. Generalizations of the Moran effect explaining spatial synchrony in population fluctuations. Am. Nat. 166: 603–12. doi:10.1086/491690 European Commission. 2008. Directive 2000/60/EC of the European Parliament and of the Council of 23 October 2000 establishing a framework for Community action in the field of water policy. L164: 1–72. Falkowski, P. G. 1998. Biogeochemical Controls and Feedbacks on Ocean Primary 147

Production. Science (80-. ). 281: 200–206. doi:10.1126/science.281.5374.200 Falkowski, P. G., M. E. Katz, A. H. Knoll, A. Quigg, J. a Raven, O. Schofield, and F. J. R. Taylor. 2004. The Evolution of Modern Eukaryotic. Science (80-. ). 305: 354–360. doi:10.1126/science.1095964 Field, C. B., J. T. Behrenfeld, Michael J. Randerson, and P. Falkowski. 1998. Primary Production of the Biosphere: Integrating Terrestrial and Oceanic Components. Science (80-. ). 281: 237–240. doi:10.1126/science.281.5374.237 Figueiras, F. G., F. F. Pe, S. Groom, and others. 2003. The Portugal coastal counter current off NW Spain: new insights on its biogeochemical variability. Prog. Oceanogr. 56: 281– 321. doi:10.1016/S0079-6611(03)00007-7 Fox, J. W., D. a. Vasseur, S. Hausch, and J. Roberts. 2011. Phase locking, the Moran effect and distance decay of synchrony: Experimental tests in a model system. Ecol. Lett. 14: 163–168. doi:10.1111/j.1461-0248.2010.01567.x Fox, J. W. W., and D. A. a Vasseur. 2008. Character convergence under competition for nutritionally essential resources. Am. Nat. 172: 667–680. doi:10.1086/591689 Fox, W. T., and J. A. Brown. 1965. The use of time-trend analysis for environmental interpretation of limestones. J. Geol. 73: 510–518. Fraga, F. 1981. Upwelling off the Galician Coast, Northwest Spain. In “Coastal Upwelling.” Am. Geophys. Union Washington, DC 176–182. Fromentin, J.-M., and B. Planque. 1996. Calanus and environment in the eastern North Atlantic . II . Influence of the North Atlantic Oscillation on C . finmarchicus and C . helgolandicus. Mar. Ecol. Prog. Ser. 134: 101–109. Fusté, X., and J. M. Gili. 1991. Distribution pattern of decapod larvae off the north-western Iberian Peninsula coast (NE Atlantic). J. Plankton Res. 13: 217–228. doi:10.1093/plankt/13.1.217 García-Comas, C., L. Stemmann, F. Ibanez, L. Berline, M. G. Mazzocchi, S. Gasparini, M. Picheral, and G. Gorsky. 2011. Zooplankton long-term changes in the NW Mediterranean Sea: Decadal periodicity forced by winter hydrographic conditions related to large-scale atmospheric changes? J. Mar. Syst. 87: 216–226. doi:10.1016/j.jmarsys.2011.04.003 García-Reyes, M., J. L. Largier, and W. J. Sydeman. 2014. Synoptic-scale upwelling indices and predictions of phyto- and zooplankton populations. Prog. Oceanogr. 120: 177– 188. doi:10.1016/j.pocean.2013.08.004 Garçon, V. C., A. Oschlies, S. C. Doney, D. McGillicuddy, and J. Waniek. 2001. The role of mesoscale variability on plankton dynamics in the North Atlantic. Deep Sea Res. Part II Top. Stud. Oceanogr. 48: 2199–2226. doi:10.1016/S0967-0645(00)00183-1 Gardner, J. L., A. Peters, M. R. Kearney, L. Joseph, and R. Heinsohn. 2011. Declining body size: A third universal response to warming? Trends Ecol. Evol. 26: 285–291. doi:10.1016/j.tree.2011.03.005 Gillooly, J. F. 2000. Effect of body size and temperature on generation time in zooplankton. J. Plankton Res. 22: 241–251. doi:10.1093/plankt/22.2.241 148

Gonzalez-Gil, R., F. G. Taboada, J. Hofer, and R. Anadon. 2015. Winter mixing and coastal upwelling drive long-term changes in zooplankton in the Bay of Biscay (1993-2010). J. Plankton Res. 0: 1–15. doi:10.1093/plankt/fbv001 González-Gil, R., F. G. Taboada, J. Höfer, and R. Anadón. 2015. Winter mixing and coastal upwelling drive long-term changes in zooplankton in the Bay of Biscay (1993-2010). J. Plankton Res. 37: 337–351. doi:10.1093/plankt/fbv001 González-Nuevo, G., and E. Nogueira. 2014. Temporal and spatial variability of river plumes in the NW and N Iberian shelf (1987-2007). Cont. Shelf Res. 91: 95–108. doi:10.1016/j.csr.2014.09.005 Gonzalez, A., and M. Loreau. 2009. The Causes and Consequences of Compensatory Dynamics in Ecological Communities. Annu. Rev. Ecol. Evol. Syst. 40: 393–414. doi:10.1146/annurev.ecolsys.39.110707.173349 Gouhier, T. C., and F. Guichard. 2014. Synchrony: Quantifying variability in space and time. Methods Ecol. Evol. 5: 524–533. doi:10.1111/2041-210X.12188 Gouhier, T. C., F. Guichard, and A. Gonzalez. 2010a. Synchrony and stability of food webs in metacommunities. Am Nat 175: E16-34. doi:10.1086/649579 Gouhier, T. C., F. Guichard, and B. A. Menge. 2010b. Ecological processes can synchronize marine population dynamics over continental scales. Proc. Natl. Acad. Sci. U. S. A. 107: 8281–8286. doi:10.1073/pnas.0914588107 Greene, C. ., a. . Pershing, a Conversi, and others. 2003. Trans-Atlantic responses of Calanus finmarchicus populations to basin-scale forcing associated with the North Atlantic Oscillation. Prog. Oceanogr. 58: 301–312. doi:10.1016/j.pocean.2003.08.009 Greenman, J. V., and T. G. Benton. 2005. The impact of environmental fluctuations on structured discrete time population models: Resonance, synchrony and threshold behaviour. Theor. Popul. Biol. 68: 217–235. doi:10.1016/j.tpb.2005.06.007 Gregg, W. W., M. E. Conkright, P. Ginoux, J. E. O’Reilly, and N. W. Casey. 2003. Ocean primary production and climate: Global decadal changes. Geophys. Res. Lett. 30: 10–13. doi:10.1029/2003GL016889 Greve, W., F. Reiners, J. Nast, and S. Hoffmann. 2004. Helgoland Roads meso- and macrozooplankton time-series 1974 to 2004: Lessons from 30 years of single spot, high frequency sampling at the only off-shore island of the North Sea. Helgol. Mar. Res. 58: 274–288. doi:10.1007/s10152-004-0191-5 Grinsted, A., J. Moore, and S. Jevrejeva. 2004. Application of the cross wavelet transform and wavelet coherence to geophysical time series. Nonlinear Process. Geophys. 11: 561–566. doi:10.5194/npg-11-561-2004 Grosjean, P., and F. Ibañez. 2014. pastecs: Package for Analysis of Space-Time Ecological Series. R package version 1.3-18. Hallett, T. B., T. Coulson, J. G. Pilkington, T. H. Clutton-Brock, J. M. Pemberton, and B. T. Grenfell. 2004. Estimation of cell numbers Why large-scale climate indices seem to predict ecological processes better than local weather. Nature 430: 71–75. doi:10.1038/nature02638.1. 149

Halvorsen, E., A. Hirst, S. Batten, K. Tande, and R. Lampitt. 2001. Diet and community grazing by copepods in an upwelled filament off the NW coast of Spain. Prog. Oceanogr. 51: 399–421. doi:10.1016/S0079-6611(01)00077-5 Hare, S. R., and N. J. Mantua. 2000. Empirical evidence for North Pacific regime shifts in 1977 and 1989. Prog. Oceanogr. 47: 103–145. Harris, G. P. 1986. Phytoplankton Ecology: Structure, Function and Fluctuation, Chapman and Hall. Harris, R. P., P. H. Wiebe, J. Lenz, H. R. Skjoldal, and M. Huntley. 2000. Zooplankton Methodology Manual, R.P. Harris, P.H. Wiebe, J. Lenz, H.R. Skjoldal, and M. Huntley [eds.]. Hays, G. C., A. J. Richardson, and C. Robinson. 2005. Climate change and marine plankton. Trends Ecol. Evol. 20: 337–44. doi:10.1016/j.tree.2005.03.004 Hooff, R. C., and W. T. Peterson. 2006. Copepod biodiversity as an indicator of changes in ocean and climate conditions of the northern California current ecosystem. Limnol. Oceanogr. 51: 2607–2620. doi:10.4319/lo.2006.51.6.2607 Houghton, T., Y. Ding, D. Griggs, M. Noguer, P. van der Linden, and D. Xiaosu. 2001. Third Assessment Report of the Intergovernmental Panel on Climate Change IPCC. Houlahan, J. E., D. J. Currie, K. Cottenie, and others. 2007. Compensatory dynamics are rare in natural ecological communities. Proc. Natl. Acad. Sci. U. S. A. 104: 3273–3277. doi:10.1073/pnas.0603798104 Hsieh, C.-H., M. D. Ohman, T. H. E. Linear, and T. Window. 2006. Biological responses to environmental forcing: the linear tracking window hypothesis. Ecology 87: 1932–8. Hubbell, S. P. 2001. The unified Neutral Theory od Biodiversity and Biogeography, Princeton University Press. Hutchinson, G. E., and R. H. MacArthur. 1959. A Theoretical Ecological Model of Size Distributions Among Species of Animals. Am. Nat. 93: 117–125. doi:10.1086/282063 Ibanez, F., and P. Grosjean. 2013. pastecs: Package for Analysis of Space-Time Ecological Series. R Packag. version 1.3-18. Ibañez, F., J. M. Fromentin, and J. Castel. 1993. Application of the Cumulated Function To the Processing of Chronological Data In Oceanography. Comptes Rendus L Acad. Des Sci. Ser. Iii-sciences La Vie-life Sci. 316: 745–748. Iles, A. C., T. C. Gouhier, B. A. Menge, J. S. Stewart, A. J. Haupt, and M. C. Lynch. 2012. Climate-driven trends and ecological implications of event-scale upwelling in the California Current System. Glob. Chang. Biol. 18: 783–796. doi:10.1111/j.1365- 2486.2011.02567.x Inchausti, P., and J. Halley. 2003. On the relation between temporal variability and persistence time in animal populations. J. Anim. Ecol. 72: 899–908. doi:10.1046/j.1365-2656.2003.00767.x Isla, J., and R. Anadon. 2004. Mesozooplankton size-fractionated metabolism and feeding off NW Spain during autumn: effects of a poleward current. ICES J. Mar. Sci. 61: 526– 534. doi:10.1016/j.icesjms.2004.03.014 150

Ives, A. R., K. Gross, and J. L. Klug. 1999. Stability and variability in competitive communities. Science 286: 542–544. doi:10.1126/science.286.5439.542 Jochimsen, M. C., R. Kümmerlin, and D. Straile. 2013. Compensatory dynamics and the stability of phytoplankton biomass during four decades of eutrophication and oligotrophication. Ecol. Lett. 16: 81–89. doi:10.1111/ele.12018 Joint, I., S. B. Groom, R. Wollast, L. Chou, G. H. Tilstone, F. G. Figueiras, M. Loijens, and T. J. Smyth. 2002. The response of phytoplankton production to periodic upwelling and relaxation events at the Iberian shelf break: Estimates by the14C method and by satellite remote sensing. J. Mar. Syst. 32: 219–238. doi:10.1016/S0924- 7963(02)00037-4 Joint, I., M. Inall, R. Torres, F. G. Figueiras, X. a. Álvarez-Salgado, A. P. Rees, and E. Malcolm S. Woodward. 2001. Two Lagrangian experiments in the Iberian upwelling system: Tracking an upwelling event and an off-shore filament. Prog. Oceanogr. 51: 221–248. doi:10.1016/S0079-6611(01)00068-4 Kalyuzhny, M., E. Seri, R. Chocron, C. H. Flather, R. Kadmon, and N. M. Shnerb. 2014. Niche versus Neutrality: A Dynamical Analysis. Am. Nat. 184: 1–5. doi:10.1086/677930 Keitt, T. H. 2008. Coherent ecological dynamics induced by large-scale disturbance. Nature 454: 331–4. doi:10.1038/nature06935 Keitt, T. H., and J. M. Fischer. 2006. Detection of scale-specific community dynamics using wavelets. Ecology 87: 2895–2904. Keitt, T. H., J. M. Fischer, I. Biology, J. A. F. Ischer, T. H. Keitt, and J. M. Fischer. 2006. Detection of scale-specific community dynamics using wavelets. Ecology 87: 2895–904. doi:10.1890/0012-9658(2006) Klug, J. L., J. M. Fischer, A. R. Ives, and B. Dennis. 2009. Compensatory Dynamics in Planktonic Community Responses to pH Perturbations Published by : Ecological Society of America COMPENSATORY DYNAMICS IN PLANKTONIC COMMUNITY RESPONSES TO pH PERTURBATIONS. 81: 387–398. Klvana, I., D. Berteaux, and B. Cazelles. 2004. Porcupine feeding scars and climatic data show ecosystem effects of the solar cycle. Am. Nat. 164: 283–297. doi:10.1086/423431 Lande, R., S. Engen, and B.-E. Saether. 2003. Stochastic population dynamics in ecology and conservation,. Lau, K.-M., and H. Weng. 1995. Climate Signal Detection Using Wavelet Transform: How to Make a Time Series Sing. Bull. Am. Meteorol. Soc. 76: 2391–2402. doi:10.1175/1520- 0477(1995)076<2391:CSDUWT>2.0.CO;2 Lavín, A., G. Díaz del Río, J. M. Cabanas, and G. Casas. 1991. Afloramiento en el noroeste de la peninsula Iberica, Indices de afloramiento para el punto 43N 11W. Periodo 1966- 1989. Inf. Técnicos del Inst. Español Oceanogr. 91: 40. Legendre, P. 2014. lmodel2: Model II Regression. R package version 1.7-2. Legendre, P., and L. Legendre. 1998. Numerical ecology, Elsevier S. Elsevier Science. Levasseur, M., J.-C. Therriault, and L. Legendre. 1984. Hierarchical control of phytoplankton 151

succession by physical factors. Mar. Ecol. Prog. Ser 19: 211–222. Levin, S. A. 1992. The Problem of Pattern and Scale in Ecology: The Robert H. MacArthur Award Lecture Author(s): Simon A. Levin Source: Ecology 73: 1943–1967. doi:doi:10.2307/1941447 Liu, P. 1994. Wavelet spectrum analysis and ocean windwaves, p. 151–166. In K.P. Foufoula- Georgiou E [ed.], Wavelets in geophysics. Academic Press, New York. Llope, M., R. Anadón, J. Á. Sostres, and L. Viesca. 2007. Nutrients dynamics in the southern Bay of Biscay (1993-2003): Winter supply, stoichiometry, long-term trends, and their effects on the phytoplankton community. J. Geophys. Res. Ocean. 112. doi:10.1029/2006JC003573 Longhurst, A. R. 1998. Ecological geography of the sea, Elsevier Inc. Loreau, M. 2001. Biodiversity and Ecosystem Functioning: Current Knowledge and Future Challenges. Science (80-. ). 294: 804–808. doi:10.1126/science.1064088 Loreau, M., and C. de Mazancourt. 2008. Species Synchrony and Its Drivers: Neutral and Nonneutral Community Dynamics in Fluctuating Environments. Am. Nat. 172: E48– E66. doi:10.1086/589746 MacArthur, R. H. 1984. Geographical ecology,. Mackas, D. L., and G. Beaugrand. 2010. Comparisons of zooplankton time series. J. Mar. Syst. 79: 286–304. doi:10.1016/j.jmarsys.2008.11.030 Mackas, D. L., M. D. Galbraith, J. E. Keister, and others. 2012a. Marine Science. 69: 1205– 1217. doi:10.1093/icesjms/fst048 Mackas, D. L. L., W. Greve, M. Edwards, and others. 2012b. Changing zooplankton seasonality in a changing ocean: Comparing time series of zooplankton phenology. Prog. Oceanogr. 97–100: 31–62. doi:10.1016/j.pocean.2011.11.005 Mantua, N. 2004. Methods for detecting regime shifts in large marine ecosystems: A review with approaches applied to North Pacific data. Prog. Oceanogr. 60: 165–182. doi:10.1016/j.pocean.2004.02.016 Mantua, N. J., S. R. Hare, Y. Zhang, J. M. Wallace, and R. C. Francis. 1997. A Pacific interdecadal climate oscillation with impacts on salmon production. Bull. Am. Meteorol. 78: 1069–1079. doi:10.1175/1520-0477(1997)078<1069:apicow>2.0.co;2 Margalef, R. 1978. Life-forms of phytoplankton as survival alternatives in an unstable environment. Oceanol. Acta 1: 493–509. doi:10.1007/BF00202661 May, R., and A. R. McLean. 2007. theoretical ecology, oxford university press. Mazerolle, M. J. 2013. AICcmodavg: Model selection and multimodel inference based on (Q)AIC(c). R Packag. McGowan, J. a, S. J. Bograd, R. J. Lynn, and A. J. Miller. 2003. The biological response to the 1977 regime shift in the California Current. Deep Sea Res. Part II Top. Stud. Oceanogr. 50: 2567–2582. doi:10.1016/S0967-0645(03)00135-8 Ménard, F., F. Marsac, E. Bellier, B. Cazelles, F. Me, and B. Cazelles. 2007. Climatic oscillations and tuna catch rates in the Indian Ocean: A wavelet approach to time series analysis. 152

Fish. Oceanogr. 16: 95–104. doi:10.1111/j.1365-2419.2006.00415.x Molinero, J. C., F. Ibanez, S. Souissi, E. Buecher, S. Dallot, and P. Nival. 2008. Climate control on the long-term anomalous changes of zooplankton communities in the Northwestern Mediterranean. Glob. Chang. Biol. 14: 14–26. doi:10.1111/j.1365- 2486.2007.01469.x Möllmann, C., B. Müller-Karulis, G. Kornilovs, and M. A. St John. 2008. Effects of climate and overfishing on zooplankton dynamics and ecosystem structure: Regime shifts, trophic cascade, and feedback loops in a simple ecosystem. ICES J. Mar. Sci. 65: 302–310. doi:10.1093/icesjms/fsm197 Morán, X. A. G., L. Alonso-Sáez, E. Nogueira, and others. 2015. More, smaller bacteria in response to ocean’s warming? Proc. R. Soc. B Biol. Sci. 282: 20150371. doi:10.1098/rspb.2015.0371 Mutshinda, C. M., R. B. O’Hara, and I. P. Woiwod. 2009. What drives community dynamics? Proc. R. Soc. B Biol. Sci. 276: 2923–2929. doi:10.1098/rspb.2009.0523 Nogueira, E., G. González-Nuevo, and L. Valdés. 2012. The influence of phytoplankton productivity, temperature and environmental stability on the control of copepod diversity in the North East Atlantic. Prog. Oceanogr. 97–100: 92–107. doi:10.1016/j.pocean.2011.11.009 Nogueira, E., F. Ibanez, and F. G. Figueiras. 2000. Effect of meteorological and hydrographic disturbances on the microplankton community structure in the Ria de Vigo (NW Spain). Mar. Ecol. Prog. Ser. 203: 23–45. doi:10.3354/meps203023 Nogueira, E., F. Pérez, and A. Rıos. 1997. Seasonal patterns and long-term trends in an estuarine upwelling ecosystem (Rıa de Vigo, NW Spain). Estuar. Coast. Shelf Sci. 44: 285–300. Odum, H. T. 1956. Primary Production in Flowing Waters1. Limnol. Oceanogr. 1: 102–117. doi:10.4319/lo.1956.1.2.0102 Omori, M., and W. M. Hammer. 1982. Patchy distribution of zooplankton: behaviour, population asssesmendt and sampling problems. Mar. Biol. 72: 193–200. doi:10.1007/BF00396920 Ospina-Alvarez, N., R. Prego, I. Álvarez, and others. 2010. Oceanographical patterns during a summer upwelling–downwelling event in the Northern Galician Rias: Comparison with the whole Ria system (NW of Iberian Peninsula). Cont. Shelf Res. 30: 1362–1372. doi:10.1016/j.csr.2010.04.018 Ospina-Alvarez, N., M. Varela, D. María Dolores, M. Gómez-Gesteira, R. Cervantes-Duarte, and R. Prego. 2014. Outside the paradigm of upwelling rias in NW Iberian Peninsula: Biogeochemical and phytoplankton patterns of a non-upwelling ria. Estuar. Coast. Shelf Sci. 138: 1–16. Otero, P., M. Ruiz-Villarreal, and A. Peliz. 2008. Variability of river plumes off Northwest Iberia in response to wind events. J. Mar. Syst. 72: 238–255. doi:10.1016/j.jmarsys.2007.05.016 Ottersen, G., B. Planque, A. Belgrano, E. Post, P. C. Reid, and N. C. Stenseth. 2001. Ecological effects of the North Atlantic Oscillation. Oecologia 128: 1–14. 153

doi:10.1007/s004420100655 Ovaskainen, O., S. Skorokhodova, M. Yakovleva, and others. 2013. Community-level phenological response to climate change. Glob. Chang. Biol. 3: 33–41. doi:10.1073/pnas.1305533110/ Palmqvist, E., and P. Lundberg. 1998. Population Extinctions in Correlated Environments. Oikos 83: 359–367. doi:10.2307/3546850 Parmesan, C., C. Parmesan, G. Yohe, and G. Yohe. 2003. A globally coherent fingerprint of climate change impacts across natural systems. Nature 421: 37–42. doi:10.1038/nature01286 Parsons, T., Y. Maita, and C. Lalli. 1984. A manual of chemical and Biological methods for seawater analysis., Pergamon Press, Oxford. Peliz, Á., J. Dubert, A. M. P. Santos, P. B. Oliveira, and B. Le Cann. 2005. Winter upper ocean circulation in the Western Iberian Basin - Fronts, Eddies and Poleward Flows: An overview. Deep. Res. Part I Oceanogr. Res. Pap. 52: 621–646. doi:10.1016/j.dsr.2004.11.005 Peliz, Á., T. L. Rosa, A. M. P. M. P. Santos, J. L. Pissarra, ??lvaro Peliz, T. L. Rosa, A. M. P. M. P. Santos, and J. L. Pissarra. 2002. Fronts, jets, and counter-flows in the Western Iberian upwelling system. J. Mar. Syst. 35: 61–77. doi:10.1016/S0924-7963(02)00076-3 Percival, D. P. 1995. On estimation of the wavelet variance. Biometrika 82: 619–631. doi:10.1093/biomet/82.3.619 Pérez, F. F., X. A. Padín, Y. Pazos, M. Gilcoto, M. Cabanas, P. C. Pardo, M. D. Doval, and L. Farina-Busto. 2010. Plankton response to weakening of the Iberian coastal upwelling. Glob. Chang. Biol. 16: 1258–1267. doi:10.1111/j.1365-2486.2009.02125.x Perry, R., H. Batchelder, D. Mackas, S. Chiba, E. Durbin, W. Greve, and H. Verheye. 2004. Identifying global synchronies in marine zooplankton populations: issues and opportunities. ICES J. Mar. Sci. 61: 445–456. doi:10.1016/j.icesjms.2004.03.022 Peterson, W. T., and J. E. Keister. 2003. Interannual variability in copepod community composition at a coastal station in the northern California Current: A multivariate approach. Deep. Res. Part II Top. Stud. Oceanogr. 50: 2499–2517. doi:10.1016/S0967- 0645(03)00130-9 Petzoldt, T., and K. Rinke. 2007. simecol: An Object-Oriented Framework for Ecological Modeling in R. J. Stat. Softw. 22: 1–31. Pimm, S. L., H. L. Jones, and J. Diamond. 1988. On the Risk of Extinction. Am. Nat. 132: 757. doi:10.1086/284889 Pinheiro, J., D. Bates, S. DebRoy, D. Sarkar, and R Development Core Team (2012). 2012. nlme: Linear and Nonlinear Mixed Effects Models. R Packag. version 3.1-105. Planque, B., and J.-M. Fromentin. 1996. Calanus and environment in the eastern North Atlantic. I. Spatial and temporal patterns of C.finmarchicus and C.helgolandicus. Mar. Ecol. Prog. Ser. 134: 11–118. Planque, B., J.-M. Fromentin, P. Cury, K. F. Drinkwater, S. Jennings, R. I. Perry, and S. Kifani. 2010. How does fishing alter marine populations and ecosystems sensitivity to 154

climate? J. Mar. Syst. 79: 403–417. doi:10.1016/j.jmarsys.2008.12.018 Platt, T., and K. Denman. 1975. Spectral analysis in ecology. Annu Rev Ecol Syst 6: 189–2010. Post, E., and M. C. Forchhammer. 2002. Synchronization of animal population dynamics by large-scale climate. Nature 420: 168–171. doi:10.1038/nature01064 Prego, R., and R. Bao. 1997. Upwelling influence on the Galician coast: silicate in shelf water and underlying surface sediments. Cont. Shelf Res. 17: 307–318. Prego, R., and F. Fraga. 1992. A Simple Model to Calculate the Residual Flows in a Spanish Ria. Hydrographic Consequences in the ria o f Vigo. Estuar. Coast. Shelf Sci. 36: 603– 615. Ranta, E., V. Kaitala, J. Lindström, and E. Hell. 1997. The Moran Effect and Synchrony in Population Dynamics. Oikos 78: 136–142. Reygondeau, G., J. C. Molinero, S. Coombs, B. R. MacKenzie, and D. Bonnet. 2015. Progressive changes in the Western English Channel foster a reorganization in the plankton food web. Prog. Oceanogr. 137: 524–532. doi:10.1016/j.pocean.2015.04.025 Riser, C. W., P. Wassmann, K. Olli, and E. Arashkevich. 2001. Production, retention and export of zooplankton faecal pellets on and off the Iberian shelf, north-west Spain. Prog. Oceanogr. 51: 423–441. doi:10.1016/S0079-6611(01)00078-7 Rocha, M. R., U. Gaedke, and D. a. Vasseur. 2011. Functionally similar species have similar dynamics. J. Ecol. 99: 1453–1459. doi:10.1111/j.1365-2745.2011.01893.x Romagnan, J. B., L. Legendre, L. Guidi, and others. 2015. Comprehensive model of annual plankton succession based on the whole-plankton time series approach. PLoS One 10: 1–18. doi:10.1371/journal.pone.0119219 Round, F., D. Mann, and M. Crawford. 1990. The diatoms: Biology and morphology of the genera., Cambridge. Roura, Á., X. A. Álvarez-Salgado, Á. F. González, M. Gregori, G. Rosón, and Á. Guerra. 2013. Short-term meso-scale variability of mesozooplankton communities in a coastal upwelling system (NW Spain). Prog. Oceanogr. 109: 18–32. doi:10.1016/j.pocean.2012.09.003 Rouyer, T., J. Fromentin, N. Stenseth, and B. Cazelles. 2008. Analysing multiple time series and extending significance testing in wavelet analysis. Mar. Ecol. Prog. Ser. 359: 11–23. doi:10.3354/meps07330 Ryther, J. H. 1969. Photosynthesis and Fish Production in the Sea. Science (80-. ). 166: 72– 76. doi:10.1126/science.166.3901.72 Salas, N., L. Ortiz, M. Gilcoto, M. Varela, J. M. Bayona, S. Groom, X. A. Alvarez-Salgado, and J. Albaigés. 2006. Fingerprinting petroleum hydrocarbons in plankton and surface sediments during the spring and early summer blooms in the Galician coast (NW Spain) after the Prestige oil spill. Mar. Environ. Res. 62: 388–413. doi:10.1016/j.marenvres.2006.06.004 Sarmiento, J. L., T. M. C. Hughes, R. J. Stouffer, and S. Manabe. 1998. Simulated response of the ocean carbon cycle to anthropogenic climate warming. Nature 393: 245–249. doi:10.1038/nature00909.1. 155

Scheffer, M., J. Bascompte, W. A. Brock, and others. 2009. Early-warning signals for critical transitions. Nature 461: 53–59. doi:10.1038/nature08227 Scherer, C., R. Gowen, and P. Tett. 2016. Assessing the state of the pelagic habitat: a case study of plankton and its environment in the western Irish Sea. Front. Mar. Sci. 3: 1– 30. doi:doi: 10.3389/fmars.2016.00236 Serranito, B., A. Aubert, L. Stemmann, N. Rossi, and J. L. Jamet. 2016. Proposition of indicators of anthropogenic pressure in the Bay of Toulon ( Mediterranean Sea ) based on zooplankton time-series. Cont. Shelf Res. 1–10. doi:10.1016/j.csr.2016.01.016 Sheridan, C., and M. Landry. 2004. A 9-year increasing trend in mesozooplankton biomass at the Hawaii Ocean Time-series Station ALOHA. ICES J. Mar. Sci. 61: 457–463. doi:10.1016/j.icesjms.2004.03.023 Shumway, R. H. R., and D. Stoffer. 2006a. Time Series Analysis and Its Applications With R Examples, Springer. Shumway, R., and D. Stoffer. 2006b. Time series analysis and its applications,. Sieburth, J. M., V. Smetacek, and J. Lenz. 1978. Pelagic ecosystem structure: Heterotrophic compartments of the plankton and their relationship to plankton size fractions. Limnol. Oceanogr. 23: 1256–1263. Stenseth, N. C. 2012. Ecological Effects of Climate Fluctuations. 1292. doi:10.1126/science.1071281 Stenseth, N. C., G. Ottersen, J. W. Hurrell, A. Mysterud, M. Lima, K.-S. Chan, N. G. Yoccoz, and B. Adlandsvik. 2003. Review article. Studying climate effects on ecology through the use of climate indices: the North Atlantic Oscillation, El Niño Southern Oscillation and beyond. Proc. Biol. Sci. 270: 2087–96. doi:10.1098/rspb.2003.2415 Stock, C. A., J. P. Dunne, and J. G. John. 2014. Global-scale carbon and energy flows through the marine planktonic food web: An analysis with a coupled physical–biological model. Prog. Oceanogr. 120: 1–28. doi:https://doi.org/10.1016/j.pocean.2013.07.001 Storch, H. Von, and F. W. Zwiers. 1999. Statistical Analysis in Climate Research H. Von Storch and F.W. Zwiers [eds.]. J. Am. Stat. Assoc. 95: 1375. doi:10.1017/CBO9780511612336 Sydeman, W. J., M. García-Reyes, D. S. Schoeman, R. R. Rykaczewski, S. A. Thompson, B. A. Black, and S. J. Bograd. 2014. Climate change. Climate change and wind intensification in coastal upwelling ecosystems. Science 345: 77–80. doi:10.1126/science.1251635 Taylor, A. 1995a. North-South shifts of the Gulf Stream and their climatic connection with the abundance of zooplankton in the UK and its surrounding seas. ICES J. Mar. Sci. 52: 711–721. doi:10.1016/1054-3139(95)80084-0 Taylor, A. H., J. I. Allen, and P. a Clark. 2002. Extraction of a weak climatic signal by an ecosystem. Nature 416: 629–32. doi:10.1038/416629a Taylor, A. H., and A. Stephens. 1980. Latitudinal displacements of the Gulf Stream ( 1966 to 1977 ) and their relation to changes in temperature and zooplankton abundance in the NE Atlantic. 145: 0–5. Taylor, A. H., A. Stephens, and J. a Stephens. 1980. Latitudinal displacements of the Gulf Stream ( 1966 to 1977 ) and their relation to changes in temperature and zooplankton 156

abundance in the NE Atlantic. Oceanol.acta. 145: 0–5. Taylor, a H. 1995b. North-south shifts of the Gulf Stream: Ocean-atmosphere interactions in the North Atlantic. Int. J. Climatol. 16: 559–583. doi:10.1002/(SICI)1097- 0088(199605)16:5<559::AID-JOC26>3.0.CO;2-Z Tenore, K. R., M. Alonso-Noval, M. Alvarez-Ossorio, and others. 1995. Fisheries and oceanography off Galicia, NW Spain: Mesoscale spatial and temporal changes in physical processes and resultant patterns of biological productivity. J. Geophys. Res. 100: 10943. doi:10.1029/95JC00529 Tilman, D., C. L. Lehman, C. E. Bristow, D. Tilman, C. L. Lehman, and C. E. Bristow. 1998. Diversity-Stability Relationships : Statistical Inevitability or Ecological Consequence ? Am. Nat. 151: 277–282. doi:10.1086/286118 Torrence, C., and G. P. Compo. 1998. A Practical Guide to Wavelet Analysis. Bull. Am. Meteorol. Soc. 79: 61–78. doi:10.1175/1520- 0477(1998)079<0061:APGTWA>2.0.CO;2 Turchin, P. 1999. Population Regulation: A Synthetic View. Oikos 84: 153. doi:10.2307/3546876 Valdes, J. L., M. R. Roman, M. T. Alvarez-Ossorio, a L. Gauzens, and a Miranda. 1990. Zooplankton composition and distribution off the coast of Galicia,Spain. J. Plankton Res. 12: 629–643. Valdés, L., and M. Moral. 1998. Time-series analysis of copepod diversity and species richness in the southern Bay of Biscay off Santander , Spain , in relation to environmental conditions. ICES J. Mar. Sci. J. du Cons. 55: 783–792. Valdés, L., M. R. Roman, M. T. Alvarez-Ossorio, A. L. Gauzens, and A. Miranda. 1990. Zooplankton composition and distribution off the coast of Galicia, Spain. J. Plankton Res. 12: 629–643. doi:10.1093/plankt/12.3.629 Vallina, S. M., P. Cermeno, S. Dutkiewicz, M. Loreau, and J. M. Montoya. 2017. Phytoplankton functional diversity increases ecosystem productivity and stability. Ecol. Modell. 361: 184–196. Vandromme, P., L. Stemmann, L. Berline, and others. 2011. Inter-annual fluctuations of zooplankton communities in the Bay of Villefranche-sur-mer from 1995 to 2005 (Northern Ligurian Sea, France). Biogeosciences 8: 3143–3158. doi:10.5194/bg-8- 3143-2011 Varela, M., R. Anadón, J. Lorenzo, L. Viesca, L. Mene, and A. Bode. 2012. Fitoplancton, In L.V. A. Bode, A. Lavín [ed.], Cambio climático y oceanográfico en el Atlántico del norte de España. Instituto Español de Oceanografía. Varela, M., and R. Prego. 2003. Hydrography and phytoplankton in an isolated and non- pristine ria area: the A Coruña Harbour (NW Spain). Acta Oecologica 24: 113–124. doi:10.1016/S1146-609X(03)00048-1 Vasseur, D. A., J. W. Fox, A. Gonzalez, and others. 2014. Synchronous dynamics of zooplankton competitors prevail in temperate lake ecosystems. Proc. Biol. Sci. 281: 20140633. doi:10.1098/rspb.2014.0633 157

Vasseur, D. A., and U. Gaedke. 2014. Spectral Analysis Unmasks Synchronous and Compensatory Dynamics in Plankton Communities Published by : Ecological Society of America content in a trusted digital archive . We use information technology and tools to increase productivity and facilitate new. Ecol. Soc. Am. 88: 2058–2071. Vasseur, D. a., U. Gaedke, and K. S. McCann. 2005. A seasonal alternation of coherent and compensatory dynamics occurs in phytoplankton. Oikos 110: 507–514. doi:10.1111/j.0030-1299.2005.14006.x Vasseur, D. A., and U. Gaedre. 2007. Spectral Analysis Unmasks Synchronous and Compensatory Dynamics in Plankton Communities. Ecology 88: 2058–2071. Vasseur, D. a, and J. W. Fox. 2009. Phase-locking and environmental fluctuations generate synchrony in a predator-prey community. Nature 460: 1007–10. doi:10.1038/nature08208 Villacieros-Robineau, N., J. L. Herrera, C. G. Castro, S. Piedracoba, and G. Roson. 2013. Hydrodynamic characterization of the bottom boundary layer in a coastal upwelling system (Ría de Vigo, NW Spain). Cont. Shelf Res. 68: 67–79. doi:10.1016/j.csr.2013.08.017 Villegas-Ríos, D., X. A. Álvarez-Salgado, S. Piedracoba, G. Rosón, U. Labarta, and M. . FernándezReiriz. 2011. Net ecosystem metabolism of a coastal embayment fertilised by upwelling and continental runoff. Cont. Shelf Res. 31: 400–413. Wallace, J. M., and D. S. Gutzler. 1981. Teleconnections in the geopotential height field during the northern hemisphere winter. Mon. Weather Rev. 109: 784–812. Wang, D., T. C. Gouhier, B. A. Menge, and A. R. Ganguly. 2015. Intensification and spatial homogenization of coastal upwelling under climate change. Nature 518: 390–394. doi:10.1038/nature14235 Ward, B. A., S. Dutkiewicz, O. Jahn, and M. J. Follows. 2012. A size-structured food-web model for the global ocean. Limnol. Oceanogr. 57: 1877–1891. doi:10.4319/lo.2012.57.6.1877 Winder, M., and J. E. Cloern. 2010. The annual cycles of phytoplankton biomass. Philos. Trans. R. Soc. Biol. 365: 3215–26. doi:10.1098/rstb.2010.0125 Xu, W., P. Miller, G. Quartly, and R. Pingree. 2015. Seasonality and interannual variability of the European Slope Current from 20 years of altimeter dara comared with in situ measurements. Remote Sens. Environ. 162: 196–207. Zuur, A. F., E. N. E. Ieno, N. J. Walker, A. A. A. Saveliev, and G. M. G. Smith. 2009. Mixed Effects Models and Extensions in Ecology with R, Ebooks Corporation. Zuur, A. F., E. N. Ieno, and C. S. Elphick. 2010. A protocol for data exploration to avoid common statistical problems. Methods Ecol. Evol. 1: 3–14. doi:10.1111/j.2041- 210X.2009.00001.x

159

ABSTRACT

Planktonic organisms play a crucial role in pelagic food webs. Half of the global primary production is due to phytoplankton activity and an important part is transferred to higher trophic levels by zooplankton organisms. Because of their short life-span and sensitivity, planktonic organisms have been proposed to be good sentinels of environmental changes and furthermore constitute an appropriate biological model to study population and community dynamics. The general objectives of this Thesis are: to explore the patterns of variability of plankton along the northern Atlantic coast of Spain by analyzing the time- series obtained within the RADIALES monitoring program (http://www.seriestemporales- ieo.com) (Chapter 1); to investigate the connections between environment variables and plankton fluctuations at multiple temporal scales (Chapters 2 and 3); and finally, to infer how environmental variability and community interactions can shape community assembly of plankton populations (Chapter 4).

The first chapter (“Seasonal and long-term variability of mesozooplankton along the Northern Iberian Atlantic shelf”) describes and compares the patterns of variability at 9 oceanographic stations distributed in 4 across-shelf sections along the North-western shelf of the Iberian Peninsula. According to the periodicities observed in biomass time-series at all scales, three main spatial domains could be distinguished: oceanic (RADIALES stations: E6SA and E3GI), coastal Cantabrian (E2SA, E1GI and E2GI) and coastal-shelf Galician (E1VI, E3VI, E2CO plus the Cantabrian shelf station E4SA). Seasonality accounted for the main proportion of variability in all the stations of the study area. However, in the southern part of the Galician shelf the seasonal cycle presented only one main wider peak from late spring to autumn, while a semi-annual component was observed in the Cantabrian shelf, reflecting the occurrence of peaks in spring and occasionally in autumn. At the decadal scale, the Santander section presented a positive trend in annually averaged biomass and a negative trend in abundance (significant in the mid-shelf), resulting in an increase in the average individual weight that was significant in coastal and shelf sites. A Coruña station presented positive and significant trends in biomass and abundance and a decrease in the average individual weight through time. The observed trends in average individual weight may be indicative of shifts in zooplankton community structure. The more conspicuous increase in 160 biomass and abundance was observed in the section off Vigo but without apparent changes in individual weight.

In the second chapter (“Long-term and seasonal zooplankton dynamics in the northwest Iberian shelf and its relationship with meteo-climatic and hydrographic variability”), attention was given to the previously detected abrupt increase in zooplankton abundance and biomass in the Vigo section. Long-term and seasonal dynamics of zooplankton, in terms of abundance and taxonomic composition, and its relationship with meteo-climatic and hydrographic factors were investigated at two locations, within and off the Ría of Vigo (Station E1VI and E3VI, respectively). Total abundance of zooplankton varied annually following on average a unimodal cycle. In the long term, zooplankton abundance exhibited three contrasting periods: A) 1995–2001, characterized by low abundance and low amplitude seasonality, with a stepped increase towards 2001; B) 2001–2006, of high abundance and marked seasonality enclosing the maximum values of the time series; and C) 2006–2010, of intermediate abundance and amplitude of the seasonal cycle. Principal component analysis revealed that the shift in zooplankton dynamics from 2001 onwards affected annual averages abundances of all zooplankton taxa. This shift was concomitant with sustained trends for upwelling intensity (increasing), precipitation (decreasing) and Gulf Stream North Wall position (equatorward displacement) between 2000 and 2005. The results stress the importance of hydrodynamics, driven by meteo-climatic conditions, in the control of the abundance levels of zooplankton at seasonal and long-term scales.

In order to explore in depth the processes that have driven the abrupt changes in the Vigo stations, the temporal structure of the zooplankton community has been studied, given rise to the third chapter (“Environmental multi-scale effects on zooplankton inter- specific synchrony”). Monthly time-series of zooplankton (taxonomic composition, total abundance, and biomass) and their relationship with upwelling index and river outflow have been analyzed using wavelet methods. The annual oscillation of biomass and abundance increased in 2000 corresponding to the highest amplitudes of environmental forcing. Concomitantly, enhanced synchrony was observed among the main taxonomic groups of zooplankton and among copepod species, the most relevant group in terms of occurrence and abundance. The degree of synchrony appeared to be correlated with the upwelling index and, more closely, with the duration of the upwelling events. The results 161 suggest that amplified seasonality of the environmental variables between 2000 and 2004, combined with a reduction of off-shore exportation by shortening of upwelling events, favoured retention in winter, and primary production in summer. These changes modulated community aggregated properties and affected the stability of the zooplankton community through an increase in inter-specific synchrony, allowing the community to shift to another state and likely a reorganization of the community size structure.

Within the fourth chapter (“How environmental forcing can synchronize population fluctuations”), we explore the effect of environmental fluctuations, such as nutrient availability, on phytoplankton community structure by combining both in-situ observations and model simulations. Observations correspond to the monthly time-series of diatom composition collected off A Coruña (Station E2CO) where the availability of nutrients has been described to rely principally on upwelling dynamics. The level of synchrony, derived from wavelet decomposition of the 12 most abundant species of diatoms, varied in time, and the highest synchonicity (lowest values of Phase Angle Variance –PAV) were observed from 1998 to 2002. During this period, the upwelling index series presented its highest amplitude for various periodic components: 2-3 weeks, 1.5 months and 1.5 years. To better understand the mechanisms and test if changes in the amplitude of nutrient inputs, coupled to upwelling dynamics, can impact the temporal association between competitors, we employed a simple chemostat model in which two species compete for two fluctuating essential resources. Synchrony was estimated between the two populations along a gradient of nutrient input amplitude. For each possible stoichiometric ratio requirement, the synchrony increased together with the amplitude of nutrient input in a non-linear fashion. Indeed, inter-specific competition seems to buffer the effect of nutrient fluctuations until a certain threshold where nutrient supply becomes the main force controlling the temporal association of species.

162

163

RESUMEN EN ESPAÑOL

INTRODUCCIÓN

Los organismos que componen el plancton juegan un papel crucial en las redes tróficas pelágicas. Aproximadamente la mitad de la producción primaria global es debida a la actividad del fitoplancton. Una parte importante de dicha producción primaria (entre el 10 y el 40 %) es transferida a los niveles tróficos superiores por mediación de los productores secundarios, y en particular el zooplancton, mientras que otra parte es remineralizada por el denominado “bucle microbiano”. A través de la formación de detritos y pellets fecales, el zooplancton contribuye además a transferir materia orgánica hacia el océano profundo. Desde un punto de vista pragmático, el plancton es un buen modelo para estudiar los efectos de cambios ambientales en las comunidades pelágicas, debido a su sensibilidad a las variaciones ambientales y por presentar ciclos de vida cortos, por lo que dichos cambios ambientales se revelan rápidamente al nivel de especie o de propiedades de la comunidad.

Los trabajos presentados en la presente Tesis se centran en 2 grupos de plancton: el mesozooplancton (Capitulos 1 a 3) y las diatomeas (Capitulo 4), uno de los grupos principales del fitoplancton. El fitoplancton está compuesto por organismos procariotas (p.ej. cianobacterias), y por organismos eucariotas (diatomeas, dinoflagelados y cocolitofóridos, entre otros). Como foto-autótrofos, su crecimiento poblacional se relaciona con la disponibilidad de nutrientes inorgánicos y luz. Por otra parte, el mesozooplancton engloba organismos heterótrofos que miden entre 200 µm y 2000 µm, incluyendo una gran variedad de grupos taxonómicos tales como los crustáceos, cnidarios, quetognatos y gasterópodos, por ejemplo. Algunos organismos, los correspondientes al meroplancton, son planctónicos únicamente durante una parte de sus ciclos de vida, generalmente la fase larvaria. El holoplancton, en cambio, lo componen los miembros permanentes del plancton.

Desde un punto de vista pragmático, el plancton es un buen modelo para entender los efectos de cambios ambientales en comunidades ecológicas. En efecto, el plancton ha mostrado una gran sensibilidad a las variaciones ambientales y al tener mayoritariamente ciclos de vida cortos, éstos se revelan rápidamente al nivel de especie o de comunidad. Sin 164 embargo, las relaciones entre variaciones ambientales y procesos ecológicos no tienen por qué ser lineales, y existen de hecho varios ejemplos de respuestas abruptas a cambios ambientales graduales. También, procesos denso-dependientes, tales como la competencia interespecífica o la depredación, son susceptibles de producir fenómenos de compensación entre varias especies que difuminan los efectos de las perturbaciones cuando se examinan al nivel de la comunidad.

Los cuatros capítulos de esta Tesis se basan en observaciones realizadas en el marco del programa RADIALES (www.seriestemporales-ieo.com). Éste consiste en un programa de monitoreo que se inició en 1988 con una estación costera en A Coruña y que desde entonces se ha ampliado a toda la costa Atlántica norte de la península Ibérica con estaciones oceanográficas de muestreo localizadas en el sur de Galicia, frente a Vigo, y en el Mar Cantábrico, frente de Gijón y Santander.

Figura 1 : Mapa del área de estudio que mostra la ubicación de las 4 secciones y estaciones oceanográficas muestreadas dentro del marco del programa RADIALES de monitoreo de series temporales (www.seriestemporales-ieo.com). De este a oeste y de costa a océano, se mencionan las diferentes estaciones dentro de cada sección: Santander (E2SA, E4SA y E6SA), Gijón (E1GI,E2GI y E3GI), A Coruña (E2CO) y Vigo (E1VI y E3VI).

Los datos utilizados en la presente Tesis se obtuvieron en 9 estaciones, cuya localización se indica en la Figura 1, y que han sido muestreadas mensualmente para obtener información sobre zooplancton, fitoplancton y otras variables, en particular 165 aquellas relacionadas con la hidrografía. La costa Gallega, localizada en el límite norte del sistema de afloramiento de la corriente de Canarias, se caracteriza por una dominancia de vientos de componente norte y noreste paralelos a la costa que favorecen la ocurrencia de eventos de afloramiento (‘upwelling’) principalmente entre mayo y septiembre. En consecuencia, las aguas frías y ricas en nutrientes que afloran desde las capas sub- superficiales hasta las capas superficiales e iluminadas soportan altos niveles de producción primaria. Durante el invierno se invierte la situación, con una dominancia de vientos del suroeste asociados con procesos de hundimiento (“downwelling”) y elevadas precipitaciones que llegan a aumentar la escorrentía en la cuenca hidrográfica y las descargas de los ríos, lo cual favorece la formación de la denominada “western Iberian buoyant plume”. La corriente Ibérica hacia el polo (también denominada corriente de Navidad o contracorriente costera de Portugal) - una corriente sub-superficial en dirección norte que se intensifica en los meses de invierno y favorece la formación de un fuerte gradiente de densidad a través de la plataforma continental. A lo largo de la costa Cantábrica, la entrada de nutrientes en la capa fótica se debe en gran parte a fenómenos de mezcla profunda que ocurren en invierno. En contraste, durante el verano la fertilización de las aguas superficiales se relaciona con la ocurrencia de eventos de afloramiento o a las descargas fluviales y la escorrentía.

Objetivos

Los objetivos de esta Tesis son: explorar el patrón de variación del plancton a lo largo de la costa Atlántica del Norte de España por medio del análisis de las series temporales obtenidas gracias al programa de monitoreo RADIALES; investigar las conexiones entre fluctuaciones ambientales y del plancton; y finalmente, inferir cómo el ambiente puede moldear la estructura temporal de las comunidades planctónicas.

166

CAPITULO 1 :

Seasonal and long-term variability of mesozooplankton along the Northern Iberian Atlantic shelf.

El primer capítulo tiene como objetivo describir y comparar los patrones de variación de nueve estaciones de muestreo distribuidas en 4 transectos a lo largo de la plataforma Nord-Atlántica de la Península Ibérica. Para ello, se usaron los datos mensuales de abundancia y biomasa total del zooplancton, disponibles en todas las estaciones lo que permite llevar a cabo una comparación de los patrones de variación temporal de zooplancton a lo largo de la costa gallega y cantábrica.

A las series de biomasa se les aplicó un análisis de ondículas (wavelets) que permite descomponer la variabilidad en sus diferentes componentes periódicos (Figura 2). De este modo, se detectó que la estacionalidad explicaba un alto porcentaje de la variabilidad total de la biomasa de zooplancton, correspondiente a más de 50 % de la variabilidad total en las estaciones costeras.

En base a las periodicidades observadas en las series temporales de biomasa de meso-zooplancton, se distinguieron tres dominios espaciales principales: Oceánico, Costero Cantábrico y Costero Gallego (Figura 3). La relevancia de la variabilidad anual fue particularmente importante en las estaciones de Vigo, en las que no se apreciaron otros modos de variación estadísticamente significativos. En contraste, en las estaciones costeras del Mar Cantábrico, además de la componente anual se observaron ciclos semi-anuales en algunos intervalos de las series, lo que se corresponde a la ocurrencia de 2 picos de zooplancton a lo largo del año: uno en primavera y otros, más variable, en otoño. En las estaciones más oceánicas del cantábrico, la componente anual representó menos de 30% de la variabilidad total. 167

Figura 2: Análisis mediante ondículas (wavelets) de todas las series temporales de biomasa total de zooplancton (E2SA, E4SA, E6SA, E1GI, E2GI, E3GI, E2CO, E1VI, E3VI) durante el periodo 2001-2009. Panel izquierdo: Wavelet Power Spectrum: (WPS = espectro de energía de ondiculas) para cada localidad; el código de color va de azul (valores bajos) a rojo oscuro (valores altos). La línea oscura define el cono de influencia por debajo del cual la información se ve afectada por un efecto borde (‘edge effect’). Panel derecho: perfil global de la WPS. En ambos paneles, las líneas punteadas indican el área de significación del 5% determinada por el método de bootstrapping basado en HMM.

A largo plazo, todas las estaciones mostraron una tendencia positiva en biomasa (Figura 4). El aumento más marcado fue en las estaciones de la sección de Vigo. Sin embargo, se apreciaron diferentes patrones de abundancia, en particular en las secciones de Santander y A Coruña, lo que sugiriere posibles cambios en la estructura de la comunidad de zooplancton.

168

3.5

2.5

1.5

Height

0.5

E1VI E3VI

E2GI E1GI E3GI

E2SA E4SA E6SA E2CO coastal coastal and shelf oceanic Cantabrian Galician Cantabrian stations stations stations

hclust (*, "complete") Figura 3: Grupos (‘clusters’) (k-medias en la matriz de similitud) obtenuidos en base a las similitudes de los WPS dentro del cono de influencia.

Figura.4: Anomalía anual de biomasa (peso seco total de zooplancton, mgDW·m-3) para cada una de las 9 series. Las rectas de regresión lineal se han añadido en rojo, indicando con el símbolo ** aquellas tendencias estadísticamente significativas.

169

CAPITULO 2 :

Long-term and seasonal zooplankton dynamics in the northwest Iberian shelf and its relationship with meteo-climatic and hydrographic variability

El segundo capítulo se centró en cuantificar la variabilidad del zooplancton en las estaciones de Vigo, a escala anual y decadal, así como su relación con variables meteo- climáticas e hidrográficas.

A lo largo del tiempo, se distinguieron tres periodos contrastados: A) 1995-2001; B) 2001-2006 y C) 2006-2010. Desde el principio de la serie en 1995 hasta 2001, las abundancias de zooplancton fueron relativamente bajas y la estacionalidad presentó también una amplitud baja. Se observó un incremento brusco en 2001. El siguiente periodo, iniciado ese año, empezó con una elevada abundancia de zooplancton y una estacionalidad muy marcada hasta 2006, cuando se inició el tercer y último periodo. Entre 2006 y 2010 la abundancia total del zooplancton volvió a disminuir, y con ella la amplitud del ciclo estacional, situándose ambas a niveles intermedios entre el primer y segundo periodos.

Figura 5: Valores de abundancia y errores estándar mensuales estimados mediante modelos mixtos lineales generalizados (GLMM) para las estaciones E1VI (A) y E3VI (B) en cada uno de los periodos: A: 1995-2001(línea gris clara), B: 2001-2006 (línea gris oscura) y C: 2006-2011(línea negra).

Todos los principales grupos taxonómicos, mostraron un patrón similar, con abundancias muy elevadas durante el segundo periodo. De hecho, el análisis de componentes principales reveló que todos los taxones de zooplancton incrementaron su abundancia a partir de 2001 (Figura 6), en concordancia con tendencias en las variables 170

ambientales entre 2000 y 2005: incremento en la intensidad del afloramiento, descenso de las precipitaciones y desplazamiento hacia el norte de la posición de la pared norte de la Corriente del Golfo.

Figura 6: Serie temporal de los componentes principales derivados del análisis de componentes principales (PCA). PC1 para variables físicas (A); PC1 para la composición taxonómica (transformación logarítmica) en la estación E1VI (B); y PC1 para la composición taxonómica (transformación logarítmica) en la estación E3VI (C). Pesos de las variables físicas para el PC1 (D) (pesos superiores a 0.30 en color negro); pesos del PC1 para la composición taxonómica en la estación E1VI (E) y la estación E3VI (F).

CAPITULO 3

Environmental multi-scale effects on zooplankton inter-specific synchrony

Para explorar los procesos que han dado lugar los cambios abruptos detectados en Vigo en los capítulos uno y dos, se estudió la estructura temporal de la comunidad del zooplancton, así como las variaciones de los principales forzadores regionales identificados previamente, lo que dio lugar al tercer capítulo. En el presente capitulo, se emplearon métodos de análisis mediante ondículas para poder describir las diferentes escalas que componen la variabilidad de las series de forzadores ambientales (índice de afloramiento y descargas de ríos), de las de propiedades agregadas de zooplancton (biomasa y abundancia

171 total), y de las abundancias de los principales grupos taxonómicos de zooplancton y especies de copépodos. Las oscilaciones anuales en biomasa y abundancia aumentaron en el año 2000, coincidiendo con las mayores amplitudes del índice de afloramiento y descargas fluviales.

Figura 7: Series temporales de las oscilaciones anuales de abundancia de (a) grupos taxonómicos y (b)especies de copépodos en la estación E1VI. La línea roja representa la variancia angular de las fases de la onda anual de cada grupo de series y la línea roja discontinua delimita el Interval de confianza del 95%.

A partir de las fases de las oscilaciones anuales extraídas de las series de abundancia de los grupos taxonómicos y especies de copépodos (Figura 7) se observaron variaciones en el grado de sincronía, siendo ésta especialmente pronunciada entre los años 2000 y 2002. El grado de sincronía se correlacionó con el índice de afloramiento y, especialmente, con la duración de los eventos de afloramiento (Figura 8).

Los resultados sugieren que una estacionalidad amplificada de las variables ambientales entre 2000 y 2004, en combinación con una reducción de la exportación hacia el océano debida a eventos de afloramiento más breves, favoreció la retención en invierno y la producción primaria en verano. 172

Figura 8 Varianza de la fase del ciclo anual calculada sobre las series temporales de abundancia de copépodos (rojo) y de grupos taxonómicos (rojo discontinuo) y del promedio anual de la duración de los eventos de afloramiento (negro).

CAPITULO 4 :

How environmental forcing can synchronize population fluctuations

El cuarto capítulo tiene como objetivo resolver una de las preguntas surgidas del capítulo anterior, en concreto estudiar cuál es la relación entre la amplitud de los forzadores ambientales y la estructura temporal de las comunidades. Para abordar esta pregunta se combinaron observaciones en el mar, para explorar los efectos de las fluctuaciones ambientales en la dinámica temporal y estructura de la comunidad de fitoplancton, con simulaciones procedentes de un modelo mecanicista para entender los mecanismos subyacentes, y en particular, cómo la competencia entre especies puede modular la respuesta de la comunidad.

Figure 9: Series temporales de (a) las oscilaciones anuales de diatomeas y (b) las fases correspondientes a la oscilación anual. Cada color representa una especie. La línea negra representa la varianza angular de las fases anuales (inverso de la sincronía). 173

A partir de la serie temporal de composición mensual de diatomeas en la estación de A Coruña, se estimó la sincronía entre las 12 especies más abundantes (Figura 9). Se observó variabilidad en la estructura temporal de la comunidad, y se destacó un periodo marcado por una alta sincronía entre 1998 y 2002.

El periodo 1998-2002, caracterizado por la mayor sincronía entre especies de diatomeas, también correspondió a la mayor amplitud en el índice de afloramiento para los componentes periódicos de 2 a 3 semanas, 1,5 meses y 1,5 años (Figura 10). Para lograr un mejor entendimiento de los mecanismos, se empleó un modelo simple de tipo quimiostato en el cual dos especies compiten por dos recursos esenciales (dos tipos de nutrientes, como sería el caso de nitrato y silicato para diatomeas) fluctuantes.

Figura 10: Transformada de ondiculas de la serie de índice del afloramiento. (a) Wavelet Power Spectrum, el código de color va de azul (valores bajos) a rojo oscuro (valores altos). La línea oscura define el cono de influencia por debajo del cual la información se ve afectada por un efecto borde (edge effect).(b) perfil global de WPS c) Amplitudes por los periodos A: 0.04 a 0.06 año (14 a 21 dias ), B: 0.14 a 0.15 años (1.5 mes) and C: 1.4 a 1.6 año. Las líneas azul verticales delimitan el periodo 1998-2002 marcado con una alta sincronía entre especies de diatomeas.

Las dos especies difieren en los ratios estequiométricos que necesitan para su crecimiento poblacional de forma que la presión ejercida por la competición no es la misma según los ratios de cada especie. En el presente trabajo se hizo una simulación para cada combinación posible de requerimiento estequiométrico haciendo variar la amplitud en la entrada de un nutriente (Figura 11). 174

Figura 11: Covarianza entre las poblaciones de los dos consumidore C1 y C2 por todas las combinaciones posible de requerimiento estequiometrico U1 y U2 y por unas amplitudes de entrada de nutrientes de (a) 0.02, (b) 0.4, (c) 0.6 y (d) 0.9.

Se observó que la sincronía aumentaba con la amplitud de la entrada de nutrientes en todas las combinaciones de requerimientos estequiométricos, pero de forma no lineal. En efecto, la competencia interespecífica parece amortiguar el efecto de las fluctuaciones en nutrientes hasta un umbral a partir del cual la amplitud de las fluctuaciones se convierte en la principal fuerza que controla la dinámica de las especies.

Discusión y Conclusiones

• En todas las estaciones de RADIALES muestreadas para zooplancton, la biomasa total de zooplancton aumentó. • Se identificaron tres regiones principales en base a las periodicidades presentadas por la biomasa zooplanctónica: o Plataforma exterior, caracterizada por una alta variabilidad y organismos, en promedio, más grandes. o Plataforma costera Cantábrica, presentando picos de biomasa en primavera y otoño. o Plataforma costera de Galicia, caracterizada por un único pico de máxima abundancia, centrado en verano. • A largo plazo, la serie temporal de zooplancton total mostró tres periodos contrastados en la Ría de Vigo y la plataforma adyacente: A) 1995–2001, 175

caracterizado por baja abundancia y estacionalidad de baja amplitud, incrementándose hacia 2001; B) 2001–2006, periodo de alta abundancia y marcada estacionalidad, incluyendo los valores máximos de la serie temporal; y C) 2006–2010, de abundancia y amplitud del ciclo estacional intermedias. Los cambios observados en las dinámicas del zooplancton se correspondieron con tendencias sostenidas en intensidad de los afloramientos (aumento), precipitación (disminución) y la posición del límite norte de la Corriente del Golfo (desplazamiento hacia el ecuador) entre 2000 y 2005. • En la Ría de Vigo, la oscilación anual de biomasa y abundancia aumentó en el 2000, correspondiendo a las mayores amplitudes de los forzadores ambientales. Al mismo tiempo, una mayor sincronía fue observada entre los principales grupos taxonómicos del zooplancton y entre especies de copépodos. • La estacionalidad amplificada del índice de afloramientos y de las descargas fluviales entre los años 2000 y 2004, en combinación con una reducción de la exportación hacia el océano debido a un acortamiento de los episodios de afloramiento, puede haber favorecido la retención en invierno y estimulado la producción primaria en verano. • Estos cambios modularon las propiedades agregadas de la comunidad y afectaron a la estabilidad de la comunidad zooplanctónica por medio de un aumento en la sincronía interespecífica que puede haber facilitado un cambio de estado en la comunidad. • Observamos un incremento repentino en la sincronía entre especies de diatomeas desde 1998 hasta 2002 en la Ría de A Coruña, coincidiendo con un aumento en la amplitud de la intensidad de los afloramientos a diferentes escalas. • Por medio de la simulación de las dinámicas de dos especies con requerimientos estequiométricos fijos compitiendo por dos recursos esenciales, hemos demostrado que un aumento en la amplitud del suministro del recurso conduce en todos los casos a una mayor sincronía entre competidores. • La relación entre la sincronía y la amplitud de las fluctuaciones ambientales es no lineal debido a que la competencia interespecífica amortigua el efecto de un incremento en amplitud hasta llegar a un umbral (punto crítico de transición) a partir del cual las fluctuaciones ambientales se convierten en los principales impulsores de la estructura temporal de la comunidad. 176

ACKNOWLEDGMENTS 177 ¡Gracias!

Y finalmente esta Tesis ha llegado a buen puerto. Eso claro, según mis propios criterios. De hecho, de primeras desearía dar las gracias al tribunal - Beatriz Mouriño, Santiago Hernández-León y Juan Carlos Molinero- por aceptar la tarea de evaluar este trabajo.

Esta Tesis no habría existido sin que se me concediera una beca de formación del personal investigador del IEO. Detrás de la beca me imagino que hay el esfuerzo de varias personas. A todos: ¡gracias! (ironías de la vida, esta beca fue invalidada unos años más tarde… pero como seguimos cobrando pues, sin rencor!).

Por supuesto, van aquí mis dos directores de Tesis, Rafael González-Quirós y Enrique Nogueira. Por haberme elegido para este proyecto de Tesis, por haber puesto en mis manos un set de datos de esta magnitud y valor, por las buenas ideas y las conversaciones científicas siempre animadas y por, habiendo capitulado ante mi insistente voluntad de ir de estancia, ayudarme a afrontar los trámites administrativos que éstas suponían. Muy cargados de otras tareas, han sabido poner atención en mi Tesis sucesivamente, fluctuando como la sardina y la anchoa (de aquí saldría mi entendimiento de la dinámica compensatoria y del peligro que puede conllevar la sincronía).

Los valiosos datos con los que he ‘jugado’ todo este tiempo son el fruto del esfuerzo de mucha gente. No conozco ni a la mitad de ellos, pero desearía agradecer a todos los que han salido a la mar mes tras mes en Vigo, A Coruña, Gijón y Santander, a todos los que han procesado muestras en el laboratorio, a los que han organizado muestras, datos, etc.

En particular, mil graciñas a Ana Miranda, quien ha contestado siempre a mis preguntas, ha revisado con mucha atención mis manuscritos y otras comunicaciones y encontrado siempre las faltas que un ojo común, me refiero a uno no entrenado a la taxonomía, no veía. Por haberme, de hecho, iniciado a la taxonomía del zooplankton, 178 transmitido un poquito del cariño que tienes a estos bichos, y por haberme acogido de tal manera en Vigo.

Un grand merci à Bernard Cazelles, gourou des wavelets, j'ai appris beaucoup pendant les trois mois passés rue d'Ulm ! Merci d'avoir toujours répondu à mes questions et m'avoir donné les coups de pouce précis dont j'avais besoin. De cette thèse, trois chapitres utilisent les ondelettes... Et je pense bien continuer!

Thank you very much to David Vasseur that was not afraid of introducing me to the theoretical world! I build my first theoretical model and enjoyed it! (And I will make more of course!). Thanks to all the lab members, for making me one of the team during those three months, being in Yale has been such an experience. Special thanks to Aalyia for making so special my stay, there started a priceless friendship. Aal and Alyssa, I have enjoyed my time very much at Livingston 23, all the conversations and anthropologic observations of the Yale/US world…

A mis compañeros de Tesis del Centro Oceanográfico de Gijón, por compartir alegrías y penas de la Tesis, por los cursos de surf, las bajadas del sella en canoa, las excursiones, copas, cafés, risas, etc. Aunque han ido marchando uno tras otro, cada uno a su manera ha aportado vidilla al despacho VIP. En orden de salida (de graduación, mejor): Virgi, Juan, Eva, Sofía, Paqui, Leti, Tamara, Nestor. Y también a los que compartieron nuestro despacho durante un tiempito pero le dieron sabor por un buen rato: Marga, Sergi, Iria, Aitor, Sdena, AnaMari, Antonio, Pauline y Lara.

Gracias a todos los compañeros del Centro de Gijón por estos años compartidos: Eva S, Laura, Dani, Pili, Revi, Luis Angel, Esther, Sergio A, Carmen C, Rosa, Roberto, Cristobo, Pilar, Mikel, Renate, Iñaki, Rocío, Felipe, Montse, Venicio, Angel L, Paco, César, Raquel, Angel U, José Maria, Alma, Xelu, Laura, Sara, Ale, Sergio V, Pili, Angela, Paloma, Leti, Jesus, a Itziar por enseñarme con minucia lo que necesitaba del laboratorio, a Alejandro por todas estas conversaciones los últimos meses, A Floren por esos miles de cafés, A Maite por sus ruidos de pájaros y su ánimo infinito, A Eva V por todo incluso lo que no se podría citar aquí (no sé qué sería pero así te quedas pensando!)…

Los inclasificables: Gracias a Amanda por sus fugaces (pero siempre apreciadas!) apariciones, a Alexandra Elbakyan sin la cual sci-hub no existiría, a toda la comunidad de 179

R y Matlab que ya había formulado y contestado en los foros a mis preguntas más absurdas.

A mi tribu gijonesa que ha hecho que estos años sean tan ricos (en todos sentidos) y que siempre me ha subido los ánimos (chicas primero): Pili, Eva, Tere, Maite, Jean- François(e), Eneko y Javi. Por todo lo compartido, y por todo lo que queda por compartir (no pensaríais libraros de nosotros por completo, eh?!): Gracias, Eskerrik asko, Merci, Graciès!!

¡Ahora le toca a Albert! Albert se lo ha leído todo, probablemente desde mi carta de motivación al presentarme a esta beca, hasta el punto final 180 páginas más tarde, pasando por las ideas más absurdas, malas o incluso bonitas, que he tenido en la cabeza durante la Tesis. Ha hecho mucho por esta Tesis bajo la forma de ánimos, de escucha atenta, de bromas, de viajes… It is worth noting that tiene la co-autoría en el capítulo no académico más bonito de esta Tesis : Eric. Eric, él no se ha leído nada, de hecho creo que le ha importado un pepino la Tesis, no tiene suficientes figuras de su interés por el momento. Lo que sí, ha rellenado estos dos últimos años de sonrisas, babas y amor. Mis chicos, Gracias! Ahora pasamos a otra cosa!