lable at ScienceDirect

Estuarine, Coastal and Shelf Science 162 (2015) 98e109

Contents lists avai

Estuarine, Coastal and Shelf Science

journal homepage: www.elsevier .com/locate/ecss

Phytoplankton blooms in estuarine and coastal waters: Seasonalpatterns and key species

Jacob Carstensen a, *, Riina Klais b, James E. Cloern c

a Dept. of Bioscience, Aarhus University, Frederiksborgvej 399, DK-4000 Roskilde, Denmarkb Institute of Ecology and Earth Sciences, Tartu University, Lai 40, 51005 Tartu, Estoniac United States Geological Survey, MS496, 345 Middlefield Road, Menlo Park, CA 94025, USA

a r t i c l e i n f o

Article history:Received 3 November 2014Accepted 2 May 2015Available online 12 May 2015

Keywords:community compositiondiatomsdinoflagellatesenvironmental factorsmeta-analysisphycology

* Corresponding author.E-mail address: jac@bios.au.dk (J. Carstensen).

http://dx.doi.org/10.1016/j.ecss.2015.05.0050272-7714/© 2015 Elsevier Ltd. All rights reserved.

a b s t r a c t

Phytoplankton blooms are dynamic phenomena of great importance to the functioning of estuarine andcoastal ecosystems. We analysed a unique (large) collection of phytoplankton monitoring data covering86 coastal sites distributed over eight regions in North America and Europe, with the aim of investigatingcommon patterns in the seasonal timing and species composition of the blooms. The spring bloom wasthe most common seasonal pattern across all regions, typically occurring early (FebruaryeMarch) atlower latitudes and later (AprileMay) at higher latitudes. Bloom frequency, defined as the probability ofunusually high biomass, ranged from 5 to 35% between sites and followed no consistent patterns acrossgradients of latitude, temperature, salinity, water depth, stratification, tidal amplitude or nutrient con-centrations. Blooms were mostly dominated by a single species, typically diatoms (58% of the blooms)and dinoflagellates (19%). Diatom-dominated spring blooms were a common feature in most systems,although dinoflagellate spring blooms were also observed in the Baltic Sea. Blooms dominated bychlorophytes and cyanobacteria were only common in low salinity waters and occurred mostly at highertemperatures. Key bloom species across the eight regions included the diatoms Cerataulina pelagica andDactyliosolen fragilissimus and dinoflagellates Heterocapsa triquetra and Prorocentrum cordatum. Otherfrequent bloom-forming taxa were diatom genera Chaetoceros, Coscinodiscus, Skeletonema, and Tha-lassiosira. Our meta-analysis shows that these 86 estuarine-coastal sites function as diatom-producingsystems, the timing of that production varies widely, and that bloom frequency is not associated withenvironmental factors measured in monitoring programs. We end with a perspective on the limitationsof conclusions derived from meta-analyses of phytoplankton time series, and the grand challengesremaining to understand the wide range of bloom patterns and processes that select species as bloomdominants in coastal waters.

© 2015 Elsevier Ltd. All rights reserved.

1. Introduction

Phytoplankton biomass, primary production and communitycomposition are all highly dynamic at the land-sea interface wherediverse human actions and climate variability intersect to drivecomplex patterns of change over time (Cloern and Jassby, 2008). Animportant pattern is the occurrence of seasonal or episodic burstsof biomass accumulation as blooms, and research in recent decadeshas identified processes that trigger blooms at the land-sea inter-face, including: pulsed inputs of nutrients from river inflow (Peierls

et al., 2012; Hall et al., 2013), coastal upwelling (Brown andOzretich, 2009), atmospheric deposition (Paerl, 1997), wind-induced entrainment of bottom water (Iverson et al., 1974;Carstensen et al., 2005), and neap-spring variability of tidal mix-ing and stratification (Cloern, 1996); seasonal winds that enhancewater retention in bays (Yin, 2003); heat waves that set up thermalstratification (Cloern et al., 2005); increasing retention time influshed systems (Odebrecht et al., 2015); release of benthic grazingpressure (Carstensen et al., 2007; Cloern et al., 2007; Petersen et al.,2008); and seasonal changes in temperature and solar radiation(Shikata et al., 2008). Phytoplankton blooms have ecological andbiogeochemical significance because much of the annual primaryproduction in estuarine-coastal ecosystems occurs during theseevents when photosynthesis exceeds system respiration (Caffrey

J. Carstensen et al. / Estuarine, Coastal and Shelf Science 162 (2015) 98e109 99

et al., 1998). That production is the energy supply that fuels pro-duction in food webs supporting fisheries (Houde and Rutherford,1993), aquaculture harvest (Bacher et al., 1998), system respira-tion (Hopkinson et al., 2005), and microbial processes that makeestuaries biogeochemical hot spots (Cloern et al., 2014). Recentcomparisons of chlorophyll a time series across a range ofestuarine-coastal ecosystem types reveal a surprising diversity ofseasonal biomass patterns, that these patterns differ from those inthe open ocean (Cloern and Jassby, 2010), and they can changeabruptly (Winder and Cloern, 2010).

Comparison of chlorophyll a time series across sites has beenuseful for discovering how the patterns of phytoplankton biomassvariability are shaped by features that distinguish estuarine-coastalecosystems from the open ocean e nutrient enrichment, tidalmixing, freshwater inflow, shallow depth and tight benthic-pelagiccoupling, sharp vertical and horizontal gradients (Cloern, 1996).Progress has been slower in solving the much more challengingproblem of understanding how these and other processes selectthose phytoplankton species that grow fast enough to developblooms. Our general conceptual understanding recognizes oneseasonal pattern that starts with a spring bloom dominated bylarge, fast-growing diatoms, followed by a number of summerblooms comprised of diatoms, flagellates, and dinoflagellates, andautumn blooms dominated by diatoms and dinoflagellates (Tettet al., 1986; Mallin et al., 1991). However, there are many de-viations from this classical pattern. Blooms in San Francisco Bay aredominated by diatoms throughout the year (Cloern and Dufford,2005), and blooms of nitrogen-fixing cyanobacteria typicallydevelop in low salinity waters during summer (Jurgensone et al.,2011) when dissolved inorganic nitrogen is depleted from thesurface layer and temperatures are high (Paerl and Huisman, 2009).Dinoflagellates dominate spring blooms in parts of the Baltic Sea(Klais et al., 2011). The highly variable physical environment andnutrient regime in estuaries and coastal waters promote differentstrategies at different times (Margalef, 1978), and bloom species areoften selected among those present at suitable inoculum levelsprior to the bloom (Smayda and Reynolds, 2001). Given thecomplexity of the problem, we have not yet identified consistentseasonal patterns of bloom occurrence by individual species orspecies groups in coastal waters.

Understanding bloom dynamics at the species level has beenelusive partly because we are not making sufficient effort to studylife-cycle processes such as sexual reproduction (Sarno et al., 2010),germination of resting stages (Shikata et al., 2008), allelopathy andmutualism between species (Smayda, 1997; Smayda and Reynolds,2001). Second, the information contained in the many empiricalrecords of phytoplankton community variability has not beensynthesized to search for common patterns of bloom occurrenceand composition.We extend the approach of comparing time seriesacross sites to explore patterns of variability in phytoplanktoncommunities and, particularly, species that develop blooms in es-tuaries, bays, and shallow coastal waters. To do this we assembledphytoplankton time series from 86 estuarine and coastal sites, andthen probed this compilation to explore four basic ecologicalquestions:

Q1. Which species and higher taxa dominate phytoplanktonblooms in shallow, nutrient-enriched coastal waters?Q2. Are there characteristic seasonal patterns of bloomoccurrence?Q3. Are blooms dominated by a common set of phytoplanktongroups or species?Q4: Does bloom frequency vary consistently along gradients ofhabitat attributes such as salinity, temperature, light availability,nutrients, or mixing?

Answers to these fundamental questions are essential forexpanding our still-limited knowledge of the natural history ofphytoplankton species succession and blooms.

2. Methods

We used long-termmonitoring data from a diverse set of marineecosystems in North America and Northwestern Europe to identifyblooms as observations of unusually high phytoplankton biomass.Differences in the frequencies and phytoplankton taxonomiccomposition of these blooms were examined across 86 coastal sitesranging from estuaries and lagoons typically affected by land runoffto embayments and nearshore coastal systems (Fig. 1). Forsimplicity we refer to these as estuarine-coastal sites, recognisingtheir differences in landscape and hydroclimatic settings. Thesesites encompass a broad range of salinity, temperature, nutrientconcentrations, tidal mixing, stratification patterns, water depthand transparency, providing a unique opportunity to explorephytoplankton bloom patterns across habitat gradients character-istic of the land-sea continuum. For some analyses we groupedphytoplankton data from the 86 sites into 8 geographic regions(Fig. 1), largely based on latitude, salinity, tidal amplitude andstratification patterns.

2.1. Data sources

Time series (minimum 5 years) of phytoplankton species countsand water quality data (all surface data) were collected fromdifferent national and regional monitoring programs (Table 1). Inaddition to salinity, temperature and Secchi depth, water sampleswere analysed for nutrient and chlorophyll a concentrations usingstandard measurements within the different monitoring programs.The taxonomical composition and biomass of the phytoplanktoncommunity was assumed to be analysed by standard techniques(inverted microscope; Uterm€ohl, 1958) in Lugol's-fixed samples.The taxonomic resolution varied among and even within moni-toring programs due to differences in identification expertise of themicroscopist and the level of taxonomic aggregation (e.g. speci-mens identified to genus level only). We assume that the mostcommon bloom-forming species are well recognised throughoutthe diverse data sets and that differences in taxonomic resolutionare most problematic for the less common species that are notaddressed in this study. The taxonomy used in all data sets wasstandardized according to the World Register of Marine Species(http://www.marinespecies.org/) to enable comparison of bloomspecies across sites (Olli et al., 2015). We recognise limitations tothe taxonomy obtained by microscopy (Jakobsen et al., 2015) andthat the taxonomical identification in some cases includes crypticspecies that include multiple species, which we address below.

In all monitoring programs phytoplankton specimens wereidentified to a standard taxonomical level (mostly at the species orgenus level) and size class. Results of microscopic analyses werereported as either biovolume (NRE and SFB; Table 1) or carbonbiomass (all other data sets) of each species using differentcompendia for translating counts (details were not provided withthe data). If carbon biomass was not reported we estimated it foreach species usingmeasured biovolumes and conversion factors fordiatoms (0.11 pg C mm�3; Strathmann, 1967) and non-diatoms(0.13 pg C mm�3; Edler, 1979). More accurate scaling equationscould not be employed because cell volumes were not reported. Foreach sample, we calculated the total phytoplankton carbon biomassby aggregating biomass of all autotrophic and mixotrophic species.We excluded the mixotrophic ciliateMesodinium rubrum because itwas not consistently identified in all monitoring programs, and we

Fig. 1. Location of the 86 coastal sites distributed among regions (specified by different colours): A) San Francisco Bay (SFB), B) Neuse River Estuary (NRE), C) Chesapeake Bay (CB),D) Rhine-Meuse-Scheldt delta (RMS), E) Wadden Sea (WS), F) Danish Strait and Estuaries (DSE), G) Baltic Sea (BS), and H) Gulf of Bothnia (GB). Note that one Wadden Sea station isplotted in F). Scaling varies between charts. Sites are located in North America (AeC) and in Northwestern Europe (DeH). (For interpretation of the references to colour in this figurelegend, the reader is referred to the web version of this article.)

J. Carstensen et al. / Estuarine, Coastal and Shelf Science 162 (2015) 98e109100

also excluded from analysis all samples having fewer than fivetaxonomical units reported.

2.2. Bloom definition

The time series of total phytoplankton carbon biomass fromeach coastal site was separated into bloom and non-bloom obser-vations, using a modified algorithm of Carstensen et al. (2007). Inthe present study we employed a periodic spline function (gam(bs ¼ “cp”) in R-package mgcv) as opposed to a harmonic functionin Carstensen et al. (2007) to describe the seasonal variation in totalcarbon biomass. The periodic spline allowed for greater flexibilityin describing the seasonal variation. The algorithm for identifyingblooms is illustrated in Supplemental information Fig. S1.

The algorithm rests on the assumption that the total carbonbiomass of non-bloom observations was normally distributed witha seasonal mean described by a periodic spline function. Bloomswere identified as significant deviations above this pattern. Thealgorithm was initialised by setting all observations to the non-bloom population, and the periodic spline for the seasonal varia-tion was estimated. Occurrences of total phytoplankton carbonbiomass exceeding the 99th percentile of the prediction interval ofthe periodic spline were defined as blooms. The periodic spline was

Table 1Data sets used in the present study listing time span, number of stations and phytoplankstations were not monitored throughout the entire period. The data were grouped into eigGB ¼ Gulf of Bothnia, NRE ¼ Neuse River Estuary, RMS ¼ Rhine-Meuse-Scheldt, SFB ¼ S

Monitoring program Period # of stations # of sam

Chesapeake Bay 1984e2009 21 7197Denmark 1979e2013 18 7249Finland 1966e2010 11 4542Germany, regional 1999e2012 1 1002Latvia 1976e2012 4 905Netherlands 1990e2011 14 5645North Carolina 2000e2013 4 624San Francisco Bay 1992e2013 4 727Sweden 1983e2012 9 1571

estimated again on the remaining non-bloom observations and the99th percentile of the prediction interval was used to single outadditional bloom observations. This re-estimation of the periodicspline was continued until all non-bloom observations were belowthe 99th percentile of the prediction interval, and all bloom ob-servations exceeded this upper limit.

For all phytoplankton samples, including those collected duringbloom and non-bloom situations, the species (or taxonomical levelreported) contributing the most to total carbon biomass wasidentified and its proportion of the total biomass was estimated.Even though the taxonomic units employed could be at the genuslevel or higher, we refer to the taxonomical unit with the highestcarbon biomass as the “dominating species”.

2.3. Statistical analyses to answer four fundamental questions

We examined seasonal patterns of bloom frequency across theeight regions by calculating the monthly probabilities of a phyto-plankton sample being identified as a bloom. The monthly proba-bilities of bloom occurrence at each site were calculated using ageneralised linear model (GLM) with bloom versus non-bloom as abinomial response variable and sampling month as an explanatoryfactor. Similarly, monthly means of the biomass proportion of the

ton surface samples included, as well as the source of the data sets. Note that all theht regions: BS¼ Baltic Sea, CB¼ Chesapeake Bay, DSE¼ Danish Straits and Estuaries,an Francisco Bay, WS ¼ Wadden Sea.

ples Provider Region

Chesapeake Bay Monitoring Program CBAarhus University, DCE DSE, WSFinnish Environment Institute BS, GBNiedersachsen, NLPV WSInst. of Aquatic Ecology, Univ. of Latvia BSRijkswaterstaat RMS, WSNC Dept. of Environ. & Natural Res. NREUS Geological Survey, Menlo Park, CA SFBSwedish Meteorological Hydrological Inst. BS, DSE, GB

J. Carstensen et al. / Estuarine, Coastal and Shelf Science 162 (2015) 98e109 101

dominating species were calculated from both bloom and non-bloom samples to examine how much a dominant speciescontributed to the overall biomass in both cases. We described theregion-specific seasonality in bloom frequency by averaging thesemonthly probabilities over all sites within regions to address Q2(seasonal patterns).

To address Q3 (common bloom composition) we identified thetaxonomical group of bloom-dominating species across the rangesof salinity, temperature and biomass proportion of the bloom-dominating species represented by the 86 sites. We also identi-fied the most common bloom-dominating species (those found in>1% of all bloom samples) for each region and across the entire dataset, and we compared the frequencies that these species dominatedbiomass in bloom and non-bloom samples. We further analysedoccurrences of dominant bloom species commonly found across allsites (only those identified to the species level) in relation tosalinity and temperature.

We investigated differences in bloom frequency across sites bycalculating site-specific mean bloom frequencies as the average ofthe site-specific monthly bloom frequencies. We identified differ-ences in the site-specific bloom frequencies between regions usinga one-way ANOVA. For comparison, we calculated site-specificannual means of salinity, temperature, total nitrogen (TN) and to-tal phosphorus (TP) using a GLM approach accounting for differ-ences in sampling frequencies between years and months (seeCarstensen et al., 2006 for details). Time series of water qualitydata were generally longer than the phytoplankton time series, butwe only analysedwater quality data fromyears with phytoplanktonsamples. TN and TP were log-transformed prior to applying theGLM and the site means resulting from the GLMwere subsequentlyback-transformed to their original scale. The frequency of stratifi-cation was calculated for each site as the relative number of sam-ples having a density difference between bottom and surfacesamples larger than 0.5 g L�1. Individual sites were categorized as:‘stratified’ if the frequency of stratification by this definition washigher than 80%; ‘intermittent’ if that frequency was between 20and 80%; or ‘mixed’ if fewer than 20% of density profiles werestratified. Some data sets only had surface measurements and theirstratification patterns were based on information from the litera-ture and data providers. To address Q4 we looked for relationshipsbetween the 86 site-specific mean bloom frequencies and envi-ronmental factors (latitude, water depth, tidal amplitude, stratifi-cation pattern, and overall mean salinity, temperature, TN and TP)with a General AdditiveModel (GAM) using a spline functionwherethe degree of smoothing was determined by generalised crossvalidation.

3. Results and discussion

3.1. Phytoplankton blooms as essential features of estuarine-coastalecosystems

While there is no universally accepted definition of what con-stitutes a bloom, the notion of a substantial deviation above back-ground phytoplankton biomass is common to all definitions. Of the29,462 carbon-biomass measurements compiled for this analysis,7368 (25%) were identified as blooms defined as deviations fromstandard seasonal patterns at each site (Fig. 2). This result is robustevidence that blooms are common in estuarine-coastal waters,although their raw frequency (number of bloom samples relative tototal number of samples) varied across sites, ranging from 6% ofobservations in Laholmsbukten (transition area between the NorthSea and Baltic Sea) to 43% of observations in South San FranciscoBay.

The composition of bloom communities is important becausethe phytoplankton include a diversity of taxonomical groups havinga wide range of cell size, motility, nutritional requirements, lifehistory, biochemical compositions, and food quality for consumers(Litchman and Klausmeier, 2008). Diatoms play central roles insilica cycling, some cyanobacteria species in nitrogen cycling, allspecies in the cycling of carbon, nitrogen, phosphorus and oxygen;large cells are directly accessible to metazoans and their energycontent is transferred efficiently in food webs supporting fisheries;diatoms, dinoflagellates and cryptophytes are highly nutritious forconsumers because they are enriched in essential fatty acids;blooms of some diatom, dinoflagellate, cyanobacteria and flagellatespecies are harmful and degrade habitat and water quality. There-fore, the biogeochemical and ecological responses to blooms varydepending on which phytoplankton species are selected by envi-ronmental conditions that promote biomass accumulation (Cloernand Dufford, 2005).

3.2. Bloom-dominant species

Our data compilation shows that blooms in estuarine-coastalecosystems are events of biomass production by micro-phytoplankton (cell size > 20 mm) and, in particular, diatoms.Diatoms were the dominant biomass component in 58% of bloomsamples, and dinoflagellates were the dominant biomasscomponent in 19% of blooms (Table 2, Fig. 3). In this senseestuarine-coastal ecosystems function as analogs to diatom- anddinoflagellate-producing coastal upwelling systems (Lassiteret al., 2006; Silva et al., 2009), except the source of new nutri-ents is land runoff as well as nutrients below the photic zone thatare mixed up by wind and tidal mixing rather than the deepocean. Nanoplankton were minor components of phytoplanktonblooms, including cryptophytes that were commonly present butrarely observed as bloom-dominating species (<2% of all blooms).For most samples cryptophytes contributed a modest proportionof the total carbon biomass (Fig. 3C). Cyanobacteria and chlor-ophytes dominated biomass in only 7 and 3% of bloom samples,respectively. In the remaining 12% of blooms, species from otheralgal groups dominated. The most frequent genus in this groupwas Phaeocystis spp., found to dominate 2.5% of all blooms;however, it dominated only in the Rhine-Meuse-Scheldt (RMS)and Wadden Sea (WS) regions where it dominated 9e14% ofblooms (Table S1).

Diatoms are the taxonomic group most often associated withphytoplankton blooms, seconded by dinoflagellates (Sarthou et al.,2005; Carstensen et al., 2007), and supported also by this analysis.This is a robust pattern, built from multi-year sampling recordsacross awide diversity of coastal habitats. The success of diatoms asa group derives from their capacity to grow rapidly in turbulenthigh-nutrient environments (Maranon et al., 2012). Diatom cellsare capable of fast growth due to rapid nitrogen (especially NO3�)uptake (Lomas and Glibert, 2000). However, in contrast to smallernano- and picophytoplankton diatoms require high nutrient con-centrations for optimal growth because of their small surface tovolume ratios (Sarthou et al., 2005). Estuarine-coastal waters aregenerally nutrient rich: across the 86 study sites, median values ofDIN (10.4 mM), DIP (0.38 mM), and silicate (14.7 mM) were highenough to support fast diatom growth for large parts of the year.Diatoms are also well adapted to varying light levels and physicalstress characteristic of shallow coastal systems, especially duringspring blooms (Lomas and Glibert, 2000). Diatoms, owing to theirhigh species diversity (Armbrust, 2009), appear to be well adaptedto the habitat gradients of estuaries because they were dominantbloom components across the entire temperature ranges and mostof the salinity ranges represented by the eight regions (see below).

Fig. 2. Partitioning samples of total phytoplankton carbon biomass into bloom and non-bloom observations for eight different coastal sites; one example from each region. Darksolid line shows the estimated mean seasonal carbon biomass for non-bloom observations and the grey solid line is the 99th percentile of the non-bloom prediction interval,separating non-bloom and bloom observations. The eight coastal sites are displayed in a north-south latitudinal sequence. A few bloom observations were outside of the plottingrange and not shown. Note the difference in biomass scale between sites.

J. Carstensen et al. / Estuarine, Coastal and Shelf Science 162 (2015) 98e109102

Dinoflagellates were the secondmost frequent taxonomic groupdominating blooms. Dinoflagellate-dominated spring blooms werenot common in all regions, but they regularly occurred in the BalticSea (BS), varying inter-annually in importance along with diatom-dominated blooms (Klais et al., 2011). The success of di-noflagellates in the Baltic Sea spring phytoplankton communityremains poorly understood, as they are inferior competitors due totheir low growth rates and nutrient uptake capacities undernutrient-replete spring conditions (Spilling and Markager, 2008).Chemical suppression of competitors is suggested as the possiblemechanism promoting cold water bloom dinoflagellates, such asScrippsiella hangoei, to outcompete the regular spring bloom di-atoms, including Skeletonema costatum s.l. and Thalassiosira baltica(Suikkanen et al., 2011). Like the common bloom-forming diatomsthe common bloom-forming dinoflagellates Peridiniella catenataand S. hangoei also form benthic resting stages (cysts) that seedblooms in the Baltic Sea (Kremp, 2001).

Pico- and nanophytoplankton taxa generally did not developblooms in the estuarine-coastal waters we considered. The

exceptions were chlorophyte blooms that developed in low-salinityupper reaches of Scheldt Estuary of RMS region, a highly turbidestuary where light intensity is insufficient to support positive netproduction (Gazeau et al., 2005). This suggests that these chlor-ophyte blooms might develop upstream in freshwater rather thanwithin the estuary. All other cases of small-celled species domi-nated blooms (e.g. cyanobacteria in the BS, Phaeocystis spp. in theRMS and WS) consisted of colonial forms. The generation time ofgrazers often determines the outcome of size competition (Kiørboe,1993). In the absence of grazers, especially the fast-reproducingmicrozooplankton including heterotrophic dinoflagellates, micro-phytoplankton would always be outcompeted by pico- and nano-plankton, regardless of nutrient concentrations.

3.3. Timing of blooms

Our analyses of bloom timing revealed that: (a) spring bloomswere common in all regions, (b) timing of the spring blooms variedacross regions, and (c) blooms developed any time of year, but the

Table 2Listing of the most common dominant bloom species (>1% of the bloom observations). For each dominant bloom species the number of bloom observations and the number ofphytoplankton observations (as totals and as percentages) are listed, where it contributed most to the carbon biomass in the sample. Regions, where the species dominated atleast one sample, are also listed (abbreviations, see Table 1).

Taxonomical group and dominatingbloom genus or species

# of blooms dominated # of samples dominated Regions with observeddominant bloom species

Chlorophytes- Chlorophyceae 127 1.72% 201 0.68% CB, DSE, RMS, WS

Cryptophytes- Cryptophyceae 94 1.28% 1827 6.20% CB, DSE, RMS, WS

Cyanobacteria- Planktothrix agardhii 147 1.99% 318 1.08% BS, GB, RMS, WS

Diatoms- Skeletonema costatum s.l. 673 9.13% 2065 7.01% BS, CB, DSE, RMS, SFB, WS- Cerataulina pelagica 406 5.51% 798 2.71% CB, DSE, RMS, WS- Coscinodiscus spp. 274 3.72% 1379 4.68% BS, CB, DSE, NRE, RMS, SFB, WS- Odontella sinensis 208 2.82% 787 2.67% DSE, RMS, WS- Dactyliosolen fragilissimus 207 2.81% 430 1.46% CB, DSE, RMS, WS- Achnanthes taeniata 144 1.95% 219 0.74% BS, DSE, GB- Thalassiosira spp. 144 1.95% 565 1.92% BS, CB, DSE, RMS, WS- Chaetoceros spp. 90 1.22% 378 1.28% BS, CB, DSE, GB, RMS, WS- Pseudo-nitzschia delicatissima 84 1.14% 156 0.53% DSE, WS- Thalassiosira baltica 84 1.14% 192 0.65% BS- Cyclotella spp. 83 1.13% 309 1.05% CB, DSE, RMS- Rhizosolenia setigera 77 1.05% 271 0.92% CB, DSE, RMS, SFB, WS- Odontella regia 76 1.03% 297 1.01% RMS. WS- Proboscia alata 74 1.00% 180 0.61% CB, DSE

Dinoflagellates- Peridiniella catenata 361 4.90% 699 2.37% BS, DSE, GB- Scrippsiella hangoei 160 2.17% 218 0.74% BS- Heterocapsa triquetra 109 1.48% 374 1.27% BS, CB, DSE, NRE, RMS, SFB- Glenodinium spp. 87 1.18% 174 0.59% BS, CB- Neoceratium tripos 80 1.08% 549 1.86% DSE- Gymnodinium spp. 76 1.03% 474 1.61% BS, CB, DSE, WS- Prorocentrum cordatum 75 1.02% 262 0.89% BS, CB, DSE, NRE, RMS, SFB

Other groups- Phaeocystis spp. 184 2.50% 298 1.01% RMS, WS- Unidentified 127 1.72% 902 3.06% DSE, GB, RMS, WS

J. Carstensen et al. / Estuarine, Coastal and Shelf Science 162 (2015) 98e109 103

occurrence of other seasonal blooms varied between regions(Fig. 2). Spring blooms occurred with a probability of 30e60%across regions. They typically developed during April and May inthe Gulf of Bothnia (GB) and Baltic Sea (BS) regions (Fig. 4A,B), butduring March in the Danish Straits and Estuaries (DSE) region (in33% of the samples, Fig. 4C). The spring bloom was even morepronounced in the WS, with 50% of the April observations cat-egorised as blooms (Fig. 4D). Bloom patterns in the RMS regionwere similar to the WS, although with highest bloom frequency(49%) in May (Fig. 4E). Bloom frequency in Chesapeake Bay (CB)peaked (30e35%) in MarcheApril (Fig. 4F). Blooms were commonin San Francisco Bay (SFB) between February and April (Fig. 4G),and they developed in the Neuse River Estuary (NRE) during thefirst three months of the year, but at lower frequency (~30%) thanspring-bloom frequency in other regions (Fig. 4H).

The timing of spring blooms followed latitudinal gradients inthe annual solar radiation cycle. In the high-latitude GB and BSregions low incident irradiance from December to March sup-presses phytoplankton growth and the bloom seasons were com-pressed (Fig. 2A,B). The bloom season began earlier in the DSEregion and even earlier in CB and NRE. However, the bloom seasonbegan later (AprileMay) in the RMS and WS regions despite theirlatitudes below DSE. Coastal sites in these two regions havestronger tidal mixing and are more turbid (see below), whichprobably delayed the onset of spring blooms. Thus, although thetiming of spring blooms is generally governed by latitude, it can bemodulated by local physical processes such as sediment suspensionby tidal currents.

Beyond the common occurrences of blooms during spring, therewas a wide range across regions in the occurrences of blooms

during other seasons. Summer and autumn blooms developed inDSE, the WS and the RMS regions. Autumn blooms developed inSFB, winter blooms were common in CB, and summer bloomsdeveloped in the NRE. The scatterplots of bloom occurrences inregions DSE (Fig. 2C) and WS (Fig. 2D) suggest that blooms in theseregions might be characterized more accurately as episodic ratherthan seasonal components of phytoplankton variability. Thus, ouranalyses reveal that there is no one canonical pattern of bloomoccurrence. This result is not surprising, given the many processesthat regulate algal growth, mortality and transport and theirdistinct seasonal patterns and frequencies of variability inestuarine-coastal waters (Cloern and Jassby, 2008). For example,river inflow is both a source of nutrients to promote blooms and aflushing process to remove phytoplankton biomass, so bloomsdevelop at intermediate flows that optimize the balance betweenthese processes (Peierls et al., 2012). Pulses of riverine inflowprovide both nutrients and freshwater as a source of buoyancy tostratify estuaries, so pulse events promote large (and harmful)dinoflagellate blooms in the NRE (Hall et al., 2013). Seasonalchanges in wind direction alter the retention time of water incoastal bays and promote development of red-tide dinoflagellatesin Hong Kong coastal waters during the winter monsoon season(Yin, 2003). Seasonal thermal stratification of the BS leads tonutrient depletion in surface waters and low bloom frequencyduring summer (Fig. 4B), but short-term events of thermal strati-fication during heat waves can promote summer or autumn bloomsin nutrient-rich estuaries such as SFB (Cloern et al., 2005). Someblooms are initiated in shallow coastal waters when light pene-trates deep enough to germinate diatom resting stages in bottomsediments (Shikata et al., 2008) or when storms transport

Fig. 3. Functional groups represented in observed blooms versus salinity, temperatureand the carbon biomass proportion of the dominating species. Bloom observationswere divided into groups based on ambient salinity and temperature rounded to thenearest integer and groups of carbon biomass proportion rounded to the nearest 5%.

J. Carstensen et al. / Estuarine, Coastal and Shelf Science 162 (2015) 98e109104

dinoflagellate cysts from sediments into the water column (Kremp,2001). In tidal systems the neap-spring cycle induces biweeklypatterns of mixing and stratification (an increase in nutrients bytidal mixing during spring tides and stratification during neaptides), with blooms developing on neap tides and dissipating onspring tides (Cloern,1996). Tidal currents in SFB have a semi-annualcomponent with largest spring tides around the solstices, sointense mixing suppresses stratification and blooms during sum-mer and winter (Fig. 4G). The characteristic winterespring diatombloom disappeared from Narragansett Bay after decades of warm-ing (Nixon et al., 2009), and the autumn blooms in SFB (Fig. 4G)appeared for the first time in 1999 after a shift in climate forcing ofthe North Pacific Ocean (Cloern et al., 2007). Our synthesis of manydata sets confirms that beyond the common occurrences of springblooms there is no consistent seasonal pattern of bloom timing inthese estuarine-coastal ecosystems.

3.4. Key species

A small number of common species developed blooms in bothNorth American and European waters: diatoms Skeletonema cos-tatum s.l. (representing several species, see below), Cerataulina

pelagica, Dactyliosolen fragilissimus, and dinoflagellates Heterocapsatriquetra and Prorocentrum cordatum. Each of these species devel-oped blooms within different ranges of temperature and salinity.Bloomsdominated by S. costatum s.l. weremost commonat salinitiesbetween 14 and 28 and at low temperatures (Fig. 5A,B), and thisdiatomdidnotdominateblooms in the lowsalinityGBor in thewarmNRE regions. C. pelagica blooms were most common at salinitiesbetween 8 and 20 and temperatures between 5 and 12 �C (Fig. 5A,B)and rarely occurred inwaterswith salinity less than 8, e.g. the coastalsites in the GB and BS. It did not develop blooms in the high-salinitySFB or in the warm waters of NRE. Blooms dominated byD. fragilissimusoccurred in theCB, DSE, RMS andWS regions (Table 2)and most commonly in the salinity range 10e27 and when temper-ature exceeded 12 �C. Blooms of H. triquetra and P. cordatum weremost common in the salinity range 6e15 (Fig. 5C), and P. cordatumblooms generally occurred at higher temperatures than H. triquetrablooms (Fig. 5D). Both dinoflagellate species formed blooms in allregions except the low-salinity regions of GB and BS (Table 2).

The most frequent and widespread bloom forming “species”,Skeletonema costatum s.l., is perceived as a fast-growing and highlyadaptable diatom that thrives in coastal waters throughout theworld. Inmany areas, species identified as S. costatum s.l. are amongthe most important contributors to phytoplankton blooms. Theapparent cosmopolitan distribution of S. costatum s.l. in coastalareas worldwide can now be attributed partly to its inclusion ofcryptic species. Recent molecular and morphological analysesrevealed that this “species” is a complex comprising at least eightmorphologically similar species in addition to S. costatum (Sarnoet al., 2005). Therefore, the ubiquity of S. costatum s.l. bloomsmight be explained by the genetic and physiological variabilitywithin this morphospecies, reflecting a limitation in our capacity tounderstand population dynamics and biogeography of phyto-plankton species from microscopy (Jakobsen et al., 2015).

The other two diatom species with intercontinental distributionand high frequency of bloom formation e Cerataulina pelagica andDactyliosolen fragilissimus e are also chain forming centric diatomshaving a form similar to Skeletonema costatum s.l., although thesespecies are larger (Harrison et al., 2015). Together, S. costatum s.l.,C. pelagica and D. fragilissimus dominated 17% of all blooms iden-tified in our data compilation, i.e. almost every third diatom bloom.Other frequent and globally distributed bloom-forming diatomswere chain-forming taxa (Thalassiosira spp., Chaetoceros spp.) or, ifsolitary, very large cells (Coscinodiscus spp.). Thus, while there is nocanonical seasonal pattern of bloom formation in estuarine-coastalwaters, there are life forms e chains of smaller celled or solitarylarge celled diatoms e that are highly adapted to the opportunitiesfor fast population growth in shallow, turbulent, turbid andnutrient-enriched coastal ecosystems.

Dinoflagellates Heterocapsa triquetra and Prorocentrum corda-tum dominated blooms in all regions except the low-salinity GB andthe mesotidal WS. In the BS the non-toxic H. triquetra formsextensive blooms during summer, when thermal stratification isstrong, and the blooms disappear in late August or early September(Olli, 2004). This pattern was also observed in the DSE, whereas inthe CB and NRE, H. triquetra typically dominated blooms in win-terespring when freshwater discharge establishes strong salinitystratification. P. cordatum blooms occur in systems affected by largefreshwater discharges, which are rich in nutrients and dissolvedorganic matter (Grzebyk and Berland, 1996). However, P. cordatumblooms were most common in late summer and autumn, whenfreshwater discharges are normally at minimum, suggesting thatP. cordatum blooms could be based on organic nutrients (Heil et al.,2005). Preferred water temperatures (20e24 �C) and salinities(10e16) reported by Velikova and Larsen (1999) are consistent withour results (Fig. 5).

Fig. 4. Seasonal distribution of bloom frequency (blue line) and the proportion of the total carbon biomass for the dominant species (right axis), separated into blooms (red line)and non-blooms (grey line), averaged over stations within each of the eight regions (nstat ¼ number of stations). The eight regions are ordered according to decreasing latitude. (Forinterpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

J. Carstensen et al. / Estuarine, Coastal and Shelf Science 162 (2015) 98e109 105

3.5. Blooms as responses to environmental variability

The 86 sites included in our data compilation represent a widerange of physical and chemical conditions (Table S2), providing arobust empirical record to search for general patterns of associationbetween environmental factors and bloom dynamics. Among thesefactors we considered: nutrient concentrations that set the po-tential magnitude for bloom development; latitude that representsgradients of climate (solar radiation, temperature) cycles; waterdepth that influences light availability and strength of benthic-pelagic coupling; salinity as a habitat attribute and indicator ofriver influence; tidal amplitude as an indicator of mixing energy;and density stratification that isolates phytoplankton from benthicconsumers and dampsmixing to establish vertical gradients of lightand nutrients. First we looked for associations between these fac-tors and frequency of bloom occurrence. Although ANOVA revealedno significant differences in bloom frequency between the eightregions (F7,78 ¼ 1.13; p ¼ 0.3553), differences among sites were

large ranging from 5% bloom frequency at Laholmsbukten (DSEregion) to 35% bloom frequency at Veerse Meer (RMS region). OurGAM analysis showed that no significant component of this vari-ability could be explained by any of the environmental factors weconsidered or their interactions, despite the large ranges of latitude,depth, tidal amplitude, stratification, and annual mean tempera-ture, salinity, TN and TP (Fig. 6).

This result suggests that there may be no general relationshipsbetween physical-chemical factors and variability of bloom occur-rence across the wide diversity of coastal ecosystems. These re-lationships do exist within individual ecosystems, but they appearto be site specific and they change over time. For example, nutrientsupply sets the potential for phytoplankton production, but therealisation of that potential varies greatly across estuarine-coastalecosystems. Some, such as Chesapeake Bay, appear to respondstrongly to changes in nutrient supply whereas others, such as SanFrancisco Bay, have shown resistance to nutrient enrichment(Cloern, 2001). This implies that the efficiency with which nitrogen

Fig. 5. Proportion of blooms dominated by selected diatom (top) and dinoflagellates (bottom) species versus salinity (A, C) and temperature (B, D). Bloom observations were dividedinto groups for salinity and temperature rounded to the nearest integer. The selected dominant bloom species were chosen to have a taxonomical resolution at the species level andhave a broad regional presence among the most common dominating bloom species in Table 2.

J. Carstensen et al. / Estuarine, Coastal and Shelf Science 162 (2015) 98e109106

and phosphorus are converted into phytoplankton biomass, andthe potential for bloom development, varies across sites. Moreover,grand experiments of nutrient reduction have shown that this ef-ficiency changes over time (Carstensen et al., 2004, 2007). Perhapsthen we should not be surprised by the absence of general re-lationships between nutrient loading, or any other environmentalfactor, and bloom dynamics indexed as frequency of occurrence.Instead, phytoplankton production dynamics appear to be regu-lated by site-specific, idiosyncratic and time-varying combinationsof all the factors that regulate the balance between production,consumption and transport (Cloern et al., 2014) instead of globalrelationships that operate uniformly across all ecosystems.

On the other hand, we did find general patterns in the taxo-nomic composition of blooms and, in particular, variability ofbloom communities along the gradients of salinity and tempera-ture. Diatoms developed blooms across the full ranges of temper-ature and salinity sampled (Fig. 3A,B), although some nichepartitioning was evident among the most frequently blooming di-atoms e Skeletonema costatum s.l. prevailing at lowest, Cerataulinapelagica at intermediate and Dactyliosolen fragilissimus at highertemperatures (Fig. 5B). At the group level, chlorophytes and cya-nobacteria developed blooms only at low salinities in systems withstrong freshwater input, and dinoflagellate blooms occurred at lowsalinities where they take advantage of their vertical migrationability to swim down and take up nutrients (Fig. 3A). Dinoflagellatebloomsweremore frequent at both high and low temperatures, butless so in intermediate temperatures (Fig. 3B). The dominance ofdinoflagellates at low temperatures is due to cold water di-noflagellates Peridiniella catenata and Scrippsiella hangoei thatfrequently dominated the BS spring blooms, where they not onlyoutcompeted, but even exceeded the typical diatom spring bloombiomass (Klais et al., 2011). These two species showed highestfrequency of bloom dominance among dinoflagellates, but theyonly occurred in the Baltic Sea (Table 2). Lists of the top bloomdominants (Table 3) separate the regions with classical diatom anddinoflagellate dominated communities from coastal areas thatfrequently experienced blooms formed by other algal groups e e.g.the chlorophytes in NRE and RMS, and haptophyte Phaeocystis spp.in WS. Chlorophytes tend to respond positively to elevated fresh-water input events (Paerl et al., 2014), which is typical of both re-gions where chlorophytes dominated blooms. Thus, although there

is great variability in environmental conditions, phytoplanktonbiomass and bloom frequency across the diversity of ecosystemtypes, one general rule does emerge from our synthesis: estuarine-coastal waters support blooms that are primarily composed of highdiatom and dinoflagellate biomass, and blooms of chlorophytes andcyanobacteria are restricted to low-salinity habitats.

4. Perspectives

Meta-analyses can be powerful approaches for identifyinggeneral patterns of community dynamics through synthesis ofmany site-based studies. For example, a meta-analysis spanning203 terrestrial species identified trends of advanced timing of an-imal migrations and plant flowering associated with globalwarming (Parmesan, 2007). This discovery was grounded in directobservations of plants and animals identified to the species level atmany sites. We cannot observe phytoplankton directly in theirenvironment, many specimens can not be identified to species levelby microscopy, and species identified by their morphology can becomplexes of an unknown number of cryptic species (Amato et al.,2007). Moreover, there is a mismatch in the time scales of phyto-plankton biomass variability (days-weeks) and sampling (weeks-months), so bloom events are not well documented and resolved.Finally, there is significant variability among laboratories in themethods used to sample, preserve and count, identify and measurephytoplankton cell volume, the taxonomic nomenclature used andresolution reported. These all place substantial constraints on thereliability of conclusions that can be drawn from meta-analyses ofphytoplankton time series. However, these constraints notwith-standing, several key results emerged from our meta-analysis ofphytoplankton community variability in estuarine-coastal sites.

First, phytoplankton blooms across these sites did not follow acommon seasonal pattern; events of high biomass were mostcommon in spring, but they occurred at any time during the year.This result appears robust because it corroborates another meta-analysis showing that the timing of annual chlorophyll a peaksvaries across and within estuarine-coastal ecosystems (Cloern andJassby, 2008). The within-ecosystem variability includes shiftingseasonal patterns over time, including loss of the winterespringdiatom from Narragansett Bay (Borkman and Smayda, 2009),earlier onset of cyanobacteria blooms in the Baltic Sea (Kahru and

Fig. 6. Bloom frequency versus environmental conditions (Table S2): A) Latitude, B) temperature, C) water depth, D) stratification pattern, E) salinity, F) tidal amplitude, G) totalnitrogen (TN) and H) total phosphorus (TP). The eight regions are shown in different colours using the abbreviations from Table 1. The estimated generalized additive models (GAM)are shown as solid lines with statistics inserted (n ¼ 86, except for TN and TP where n ¼ 81).

Table 3The five most frequent dominant bloom species (taxonomical unit) in each region with the proportion of bloom observations dominated. The full list of bloom dominantspecies for all regions is given in Table S1.

Baltic Sea Chesapeake Bay Danish straits and estuaries Gulf of Bothnia

Peridiniella catenata (19.8%) Cerataulina pelagica (18.8%) Skeletonema costatum s.l. (21.8%) Diatoma tenuis (23.2%)Scrippsiella hangoei (9.34%) Coscinodiscus spp. (11.4%) Dactyliosolen fragilissimus (6.60%) Peridiniella catenata (14.5%)Planktothrix agardhii (8.11%) Skeletonema costatum s.l. (5.64%) Cerataulina pelagica (4.57%) Aphanizomenon spp. (7.97%)Achnanthes taeniata (8.11%) Cyclotella spp. (5.24%) Neoceratium tripos (3.97%) Synechococcus spp. (7.97%)Skeletonema costatum s.l. (5.78%) Dactyliosolen fragilissimus (4.31%) Pseudo-nitzschia delicatissima (3.92%) Chaetoceros spp. (5.80%)

Neuse river estuary Rhine-Meuse-Scheldt delta San Francisco Bay Wadden Sea

Coscinodiscus spp. (33.0%) Odontella sinensis (13.7%) Thalassiosira punctigera (16.3%) Phaeocystis spp. (13.8%)Heterocapsa triquetra (15.1%) Chlorophyceae (11.7%) Ditylum brightwellii (7.76%) Odontella sinensis (10.7%)Trachelomonas spp. (14.2%) Unidentified (10.85%) Coscinodiscus oculus-iridis (7.35%) Odontella regia (10.0%)Gymnodinium instriatum (8.49%) Phaeocystis spp. (8.97%) Thalassiosira rotula (7.35%) Mediopyxis helysia (8.03%)Thalassiosira nordenskioeldii (5.66%) Thalassiosira spp. (5.43%) Skeletonema costatum s.l. (6.53%) Rhizosolenia imbricata (5.35%)

J. Carstensen et al. / Estuarine, Coastal and Shelf Science 162 (2015) 98e109 107

J. Carstensen et al. / Estuarine, Coastal and Shelf Science 162 (2015) 98e109108

Elmgren, 2014), and new occurrences of autumn blooms in SanFrancisco Bay (Cloern et al., 2007). One characteristic of estuarine-coastal ecosystems appears to be their diverse and variable sea-sonal patterns of bloom occurrence.

Second, blooms at these sites were characteristically dominatedby microplankton (cell size > 20 mm) e a distinct contrast frombiomass dominance by picoplankton in oligotrophic regions of theworld ocean. This predominance of larger algal forms is consistentwith other meta-analyses showing that the microplanktoncomponent of phytoplankton communities increases in proportionto resources and biomass (Mara~n�on et al., 2012). Microplankton areconsumed directly by metazoans so their predominance impliesefficient energy transfer from producers to consumers in thesehigh-biomass, high-nutrient systems.

Third, phytoplankton blooms were most commonly dominatedby diatoms, and this was true across the full ranges of salinity andtemperature. Therefore, diatoms as a group appear to be highlyadapted to exploit opportunities for biomass growth along theland-ocean continuum. Our meta-analysis suggests that estuarine-coastal waters function as diatom-producing systems. This hasimportant ecological and biogeochemical implications because ofthe large amounts of high food value of diatoms to consumers andthe unique role played by diatoms in the cycling of silica as it iscarried from rivers to oceans.

Finally, our meta-analysis illustrates the elusiveness of globalrules defining how phytoplankton respond to environmental vari-ability in coastal areas. This implies that bloom dynamics areregulated by site-specific variability of those factors and/or pro-cesses not captured in monitoring programs such as species in-teractions and life cycle events. A grand challenge remains tosynthesize observations from place-based studies into a generalmodel that explains the processes underlying phytoplankton pat-terns in estuarine-coastal waters. Keys to success will be enhancedefforts to: measure processes of species interaction; applyemerging technologies such as imaging flow cytometry and mo-lecular tools to better resolve biomass variability and taxonomiccomposition of phytoplankton communities at higher temporalresolution; and compare time series from tropical and subtropicalecosystems that are under-sampled and may not follow the pat-terns of temperate estuaries, such as diatom-dominance (Cotoviczet al., 2015).

Acknowledgements

We thank all the institutions that kindly provided their moni-toring data for this study (see Table 1). This study is a contributionto the Scientific Committee on Ocean Research (SCOR) workinggroup 137 and has been supported by the WATERS project (fundedby Swedish Environmental Protection Agency, grant agreement10/179), the DEVOTES project (funded by the EU 7th frameworkprogram, grant agreement 308392), and the COCOA project (fundedby the BONUS program for Baltic Sea research, grant agreement2112932-1).

Appendix A. Supplementary data

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.ecss.2015.05.005.

References

Amato, A., Kooistra, W.H.C.F., Ghiron, J.H.L., Mann, D.G., Proschold, T., Montresor, M.,2007. Reproductive isolation among sympatric cryptic species in marine di-atoms. Protist 158, 193e207.

Armbrust, E.V., 2009. The life of diatoms in the world's oceans. Nature 459,185e192.

Bacher, C., Duarte, P., Ferreira, J.G., Heral, M., Raillard, O., 1998. Assessment andcomparison of the Marennes-Ol�eron Bay (France) and Carlingford Lough(Ireland) carrying capacity with ecosystem models. Aquat. Ecol. 31, 379e394.

Borkman, D.G., Smayda, T., 2009. Multidecadal (1959e1997) changes in Skel-etonema abundance and seasonal bloom patterns in Narragansett Bay, RhodeIsland, USA. J. Sea Res. 61, 84e94.

Brown, C.A., Ozretich, R.J., 2009. Coupling between the coastal ocean and YaquinaBay, Oregon: importance of oceanic inputs relative to other nitrogen sources.Estuar. Coasts 32, 219e237.

Caffrey, J., Cloern, J.E., Grenz, C., 1998. Changes in production and respiration duringa spring phytoplankton bloom in San Francisco Bay, California, USA: implica-tions for net ecosystem metabolism. Mar. Ecol. Prog. Ser. 172, 1e12.

Carstensen, J., Conley, D.J., Henriksen, P., 2004. Frequency, composition, and causesof summer phytoplankton blooms in a shallow coastal ecosystem, the Kattegat.Limnol. Oceanogr. 49, 190e201.

Carstensen, J., Frohn, L.M., Hasager, C.B., Gustafsson, B.G., 2005. Summer algalblooms in a coastal ecosystem: the role of atmospheric deposition versusentrainment fluxes. Estuar. Coast. Shelf Sci. 62, 595e608.

Carstensen, J., Conley, D.J., Andersen, J.H., Ærtebjerg, G., 2006. Coastal eutrophica-tion and trend reversal: a Danish case study. Limnol. Oceanogr. 51, 398e408.

Carstensen, J., Henriksen, P., Heiskanen, A.-S., 2007. Summer algal blooms inshallow estuaries: definition, mechanisms, and link to eutrophication. Limnol.Oceanogr. 52, 370e384.

Cloern, J.E., 1996. Phytoplankton bloom dynamics in coastal ecosystems: a reviewwith some general lessons from sustained investigation of San Francisco Bay,California. Rev. Geophys. 34, 127e168.

Cloern, J.E., 2001. Our evolving conceptual model of the coastal eutrophicationproblem. Mar. Ecol. Prog. Ser. 210, 223e253.

Cloern, J.E., Dufford, R., 2005. Phytoplankton community ecology: principles appliedin San Francisco Bay. Mar. Ecol. Prog. Ser. 285, 1e28.

Cloern, J.E., Schraga, T., Lopez, C., Knowles, N., Labiosa, R., Dugdale, R., 2005. Climateanomalies generate an exceptional dinoflagellate bloom in San Francisco Bay.Geophys. Res. Lett. 32 (14), 1e5.

Cloern, J.E., Jassby, A.D., Thompson, J.K., Hieb, K.A., 2007. A cold phase of the EastPacific triggers new phytoplankton blooms in San Francisco Bay. Proc. Natl.Acad. Sci. 104, 18561e18565.

Cloern, J.E., Jassby, A., 2008. Complex seasonal patterns of primary producers at theland-sea interface. Ecol. Lett. 11, 1294e1303.

Cloern, J.E., Jassby, A., 2010. Patterns and scales of phytoplankton variability inestuarine-coastal ecosystems. Estuar. Coasts 33, 230e241.

Cloern, J.E., Foster, S.Q., Kleckner, A.E., 2014. Phytoplankton primary production inthe world's estuarine-coastal ecosystems. Biogeosciences 11, 2477e2501.

Cotovicz, L.C.J., Knoppers, B.A., Brandini, N., Costa Santos, S.J., Abril, G., 2015. A largeCO2 sink enhanced by eutrophication in a tropical coastal embayment (Gua-nabara Bay, Rio de Janeiro, Brazil). Biogeosci. Discuss. 12, 4671e4720.

Edler, L., 1979. Recommendations on methods to marine biological studies in theBaltic Sea. Phytoplankton and chlorophyll. Baltic Mar. Biol. Publ. 5, 38.

Gazeau, F., Gattuso, J.-P., Middelburg, J., Schiettecatte, L.-S., Frankignoulle, M.,Borges, A., 2005. Planktonic and whole system metabolism in a nutrient-richestuary (the Scheldt estuary). Estuar. Coasts 28, 868e883.

Grzebyk, D., Berland, B., 1996. Influence of temperature, salinity and irradiance ongrowth of Prorocentrum minimum (Dinophyceae) from the Mediterranean Sea.J. Plankton Res. 18, 1837e1849.

Hall, N.S., Paerl, H.W., Peierls, B.L., Whipple, A.C., Rossignol, K.C., 2013. Effects ofclimatic variability on phytoplankton community structure and bloom devel-opment in the eutrophic, microtidal, New River Estuary, North Carolina, USA.Estuar. Coast. Shelf Sci. 117, 70e82.

Harrison, P., Jakobsen, H.H., Zingone, A., 2015. Cell biovolumes of marine phyto-plankton from global data sets. Estuar. Coast. Shelf Sci. (this issue).

Heil, C.A., Glibert, P.M., Fan, C., 2005. Prorocentrum minimum (Pavillard) Schiller: areview of a harmful algal bloom species of growing worldwide importance.Harmful Algae 4, 449e470.

Hopkinson, C.J., Smith, E.M., del Giorgio, P.A., Williams, P.J.l.B., 2005. Estuarinerespiration: an overview of benthic, pelagic and whole system respiration. In:del Giorgio, P.A., Williams, P.J.l.B. (Eds.), Respiration in Aquatic Ecosystems.Oxford University Press.

Houde, E.D., Rutherford, E.S., 1993. Recent trends in estuarine fisheries e pre-dictions of fish production and yield. Estuaries 16, 161e176.

Iverson, R.L., Curl Jr., H.C., O'Connors, H.B.J., Kirk, D., Zakar, K., 1974. Summerphytoplankton blooms in Auke Bay, Alaska, driven by wind mixing of the watercolumn. Limnol. Oceanogr. 19, 271e278.

Jakobsen, H.H., Carstensen, J., Harrison, P.J., Zingone, A., 2015. Estimating time seriesphytoplankton biomass: species identifications and comparing volume to car-bon scaling ratios. Estuar. Coast. Shelf Sci. (this issue).

Jurgensone, I., Carstensen, J., Ikauniece, A., Kalveka, B., 2011. Long-term changes andcontrolling factors of phytoplankton community in the Gulf of Riga (Baltic Sea).Estuar. Coasts 34, 1205e1219.

Kahru, M., Elmgren, R., 2014. Multidecadal time series of satellite-detected ac-cumulations of cyanobacteria in the Baltic Sea. Biogeosciences 11,3619e3633.

Kiørboe, T., 1993. Turbulence, phytoplankton cell size and the structure of pelagicfood webs. Adv. Mar. Biol. 29, 1e72.

Klais, R., Tamminen, T., Kremp, A., Spilling, K., Olli, K., 2011. Decadal-scale changes ofdinoflagellates and diatoms in the anomalous Baltic Sea spring bloom. PLoSONE 6, e21567.

J. Carstensen et al. / Estuarine, Coastal and Shelf Science 162 (2015) 98e109 109

Kremp, A., 2001. Effects of cyst resuspension on germination and seeding of twobloom forming dinoflagellates in the Baltic Sea. Mar. Ecol. Prog. Ser. 216, 57e66.

Lassiter, A.M., Wilkerson, F.P., Dugdale, R.C., Hogue, V.E., 2006. Phytoplankton as-semblages in the CoOP-WEST coastal upwelling area. Deep Sea Res. Part II 53,3063e3077.

Litchman, E., Klausmeier, C.A., 2008. Trait-based community ecology of phyto-plankton. Annu. Rev. Ecol. Evol. Syst. 39, 615e639.

Lomas, M.W., Glibert, P.M., 2000. Comparison of nitrate uptake, storage, andreduction in marine diatoms and flagellates. J. Phycol. 36, 903e913.

Mallin, M.A., Paerl, H.W., Rudek, J., 1991. Seasonal phytoplankton composition,productivity and biomass in the Neuse River estuary, North Carolina. Estuar.Coast. Shelf Sci. 32, 609e623.

Mara~n�on, E., Cerme~no, P., Latasa, M., Tadonl�ek�e, R.D., 2012. Temperature, resources,and phytoplankton size structure in the ocean. Limnol. Oceanogr. 57,1266e1278.

Margalef, R., 1978. Life-forms of phytoplankton as survival alternatives in an un-stable environment. Oceanol. Acta 1, 493e509.

Nixon, S.W., Fulweiler, R.W., Buckley, B.A., Granger, S.L., Nowicki, B.L., Henry, K.M.,2009. The impact of changing climate on phenology, productivity, and benthic-pelagic coupling in Narragansett Bay. Estuar. Coast. Shelf Sci. 82, 1e18.

Odebrecht, C., Abreu, P.C., Carstensen, J., 2015. Retention time as controlling factorof short-term chlorophyll a variability in a shallow microtidal subtropical es-tuary. Estuar. Coast. Shelf Sci. (this issue).

Olli, K., 2004. Temporary cyst formation of Heterocapsa triquetra (Dinophyceae) innatural populations. Mar. Biol. 145, 1e8.

Olli, K., Paerl, H.W., Klais, R., 2015. Diversity of coastal phytoplankton assemblages ecross ecosystem comparison. Estuar. Coast. Shelf Sci. (this issue).

Paerl, H.W., 1997. Coastal eutrophication and harmful algal blooms: Importance ofatmospheric deposition and groundwater as “new” nitrogen and other nutrientsources. Limnol. Oceanogr. 42, 1154e1165.

Paerl, H.W., Huisman, J., 2009. Climate change: a catalyst for global expansion ofharmful cyanobacterial blooms. Environ. Microbiol. Rep. 1, 27e37.

Paerl, H.W., Hall, N.S., Peierls, B.J., Rossignol, K.L., Joyner, A.R., 2014. Hydrologicvariability and its control of phytoplankton community structure and functionin two shallow, coastal, lagoonal ecosystems: the Neuse and New River Estu-aries, North Carolina, USA. Estuar. Coasts 37 (suppl. 1), S31eS45.

Parmesan, C., 2007. Influences of species, latitudes and methodologies on estimatesof phenological response to global warming. Glob. Change Biol. 13, 1860e1872.

Peierls, B.L., Hall, N.S., Paerl, H.W., 2012. Non-monotonic responses of phyto-plankton biomass accumulation to hydrologic variability: a comparison of twocoastal plain North Carolina estuaries. Estuar. Coasts 1e17.

Petersen, J.K., Hansen, J.W., Laursen, M.B., Clausen, P., Carstensen, J., Conley, D.J.,2008. Regime shift in a coastal marine ecosystem. Ecol. Appl. 18, 497e510.

Sarno,D., Zingone,A.,Montresor,M., 2010.Amassive andsimultaneous sexeventof twoPseudo-nitzschia species. Deep Sea Res. Part II Top. Stud. Oceanogr. 57, 248e255.

Sarno, D., Kooistra, W.H.C.F., Medlin, L.K., Percopo, I., Zingone, A., 2005. Diversity inthe genus Skeletonema (Bacillariophyceae). II. An assessment of the taxonomy ofS. costatum-like species with the description of four new species. J. Phycol. 41,151e176.

Sarthou, G., Timmermanns, K.R., Blain, S., Tr�eguer, P., 2005. Growth physiology andfate of diatoms in the ocean: a review. J. Sea Res. 53, 25e42.

Shikata, T., Nagasoe, S., Matsubara, T., Yoshikawa, S., Yamasaki, Y., Shimasaki, Y.,Oshima, Y., Jenkinson, I.R., Honjo, T., 2008. Factors influencing the initiation ofblooms of the raphidophyte Heterosigma akashiwo and the diatom Skeletonemacostatum in a port in Japan. Limnol. Oceanogr. 53, 2503e2518.

Silva, A., Palma, S., Oliveira, P.B., Moita, M.T., 2009. Composition and interannualvariability of phytoplankton in a coastal upwelling region (Lisbon Bay,Portugal). J. Sea Res. 62, 238e249.

Smayda, T.J., 1997. Harmful algal blooms: their ecophysiology and general relevanceto phytoplankton blooms in the sea. Limnol. Oceanogr. 42, 1137e1153.

Smayda, T.J., Reynolds, C.S., 2001. Community assembly in marine phytoplankton:application of recent models to harmful dinoflagellate blooms. J. Plankton Res.23, 447e461.

Spilling, K., Markager, S., 2008. Ecophysiological growth characteristics andmodelling of the onset of the spring bloom in the Baltic Sea. J. Mar. Syst. 73,323e337.

Strathmann, R.R., 1967. Estimating the organic carbon content of phytoplanktonfrom cell volume or plasma volume. Limnol. Oceanogr. 12, 411e418.

Suikkanen, S., Hakanen, P., Spilling, K., Kremp, A., 2011. Allelopathic effects of BalticSea spring bloom dinoflagellates on co-occurring phytoplankton. Mar. Ecol.Prog. Ser. 439, 45e55.

Tett, P., Gowen, R., Grantham, B., Jones, K., Miller, B.S., 1986. The phytoplanktonecology of the Firth of Clyde sealochs Striven and Fyne. Proc. R. Soc. Edinb. Sect.B 90, 223e238.

Uterm€ohl, H., 1958. Zur Vervollkommung der quantitativen phytoplankton-meth-odik. Int. Ver. Theor. Angew. Limnol. Verh. 17, 47e49.

Velikova, V., Larsen, J., 1999. The Prorocentrum cordatum/Prorocentrum minimumtaxonomic problem. GRANA 8 (3e4), 108e112.

Winder, M., Cloern, J.E., 2010. The annual cycles of phytoplankton biomass. Philos.Trans. R. Soc. B Biol Sci. 365, 3215e3226.

Yin, K., 2003. Influence of monsoons and oceanographic processes on red tides inHong Kong waters. Mar. Ecol. Prog. Ser. 262, 27e41.

本文献由“学霸图书馆-文献云下载”收集自网络，仅供学习交流使用。

学霸图书馆（www.xuebalib.com）是一个“整合众多图书馆数据库资源，

提供一站式文献检索和下载服务”的24 小时在线不限IP

图书馆。

图书馆致力于便利、促进学习与科研，提供最强文献下载服务。

图书馆导航：

图书馆首页 文献云下载 图书馆入口 外文数据库大全 疑难文献辅助工具