Ecosystems (2014) 17: 550-566 DOI: 10.1007/s 10021-013-9742-4 ECOSYSTEMS! ) 2014 The A uthor(s). This article is published w ith open access at Springerlink.com Large-Scale Spatial Dynamics of Intertidal Mussel (Mytilus edulis L.) Bed Coverage in the German and Dutch Wadden Sea Eelke O. Folm er,1* Jan D rent,1 Karin Troost,2 Heike Büttger,3 Norbert D ankers,4 Jeroen Jansen,4 M arnix van Stralen,5 Gerald M illat,6 Mare H erlyn,7 and Catharina J. M. P hilippart1 1Department of Marine Ecology, Royal Netherlands Institute for Sea Research (NIOZ), P.O. Rox 59, 1790 AE Den Eurg, Texel, The Netherlands; 2Institute for Marine Resources & Ecosystem Studies, Yerseke, The Netherlands; 3EioConsult SH, Schobüller Str. 36, 25813 Husum, Germany; 4Institute for Marine Resources Er Ecosystem Studies, Texel, The Netherlands; 5MarinX, Elkerzeeseweg 77, 4322NA Scharendij'ke, The Netherlands; 6National Park Administration Wadden Sea Lower Saxony, Virchowstr. 1, 26340 Wil helmshaven, Germany; 7Niedersächsischer Landesbetrieb für Wasserwirtschaft, Küsten- und Naturschutz, An der Mühle 5, D-26548 Norderney, Germany A bstr a c t Intertidal blue m ussel beds are im portant for the functioning and com m unity com position of coastal ecosystems. M odeling spatial dynam ics of intertidal m ussel beds is com plicated because suitable habitat is spatially heterogeneously dis tributed and recruitm ent and loss are hard to predict. To get insight into the m ain determ inants of dispersion, grow th and loss of intertidal m ussel beds, we analyzed spatial distributions and grow th patterns in the G erm an and D utch W ad den Sea. We considered yearly distributions of adult intertidal m ussel beds from 36 connected tidal basins betw een 1999 and 2010 and for the period 1968-1976. We found th at in both periods the highest coverage of tidal flats by m ussel beds occurs in the sheltered basins in the sou th ern W adden Sea. We used a stochastic grow th m odel to investigate the effects of density dependence, w inter tem perature and storm iness on changes in m ussel bed coverage betw een 1999 and 2010. In contrast to expectation, we found no evidence th at cold w inters consistently induced events of synchronous population grow th, no r did we find strong evidence for increased rem oval of adult m ussel beds after storm y w inter seasons. H ow ever, we did find synchronic grow th w ithin groups of proxim ate tidal basins and th at syn chrony betw een distant groups is m ainly low or negative. Because the boundaries betw een syn chronic groups are located n ear river m ouths and in areas lacking suitable m ussel bed habitat, we suggest th at the m etapopulation is u n d er the control of larval dispersal conditions. O ur study dem onstrates the im portance of m oving from simple habitat suitability m odels to m odels th at incorporate m etapopulation processes to R eceived 27 M ay 2013; accepted 27 N ovem ber 2013; published online 7 J a n u a ry 2014 E le ctro n ic s u p p le m e n ta r y m a teria l: The online version of this article (doi:10 .1 0 0 7 /sl0 0 2 1 -0 1 3 -9 7 4 2 -4 ) contains su p p lem en tary m aterial, w h ich is available to a u th o rized users. A u th o r C o n trib u tio n s: EOF, JD, ND, MvS, HB an d CJMP conceived an d designed th e study. C arried o u t in tertid al m ussel bed m o n ito ring and prep ared th e GIS data: KT, ND, JJ, MvS (N etherlands); GM an d MH (Lower Saxony); HB (Schlesw ig-H olstein). EOF com piled th e data sets an d analyzed th e data. EOF w rote th e p ap er to w h ich all co-authors contributed. *Corresponding author; e-mail: E.0.Folm [email protected] gm ail.com 550 Spatial Dynamics of Intertidal M ussel Beds 551 u n d erstan d spatial dynam ics of m ussel beds. The spatio-dynam ic structure revealed in this paper will be in stru m ental for th at purpose. K ey w ords: population synchronization; tivity; supply side ecology; larvae; habitat ity; spatial autocorrelation; spatial dynam ic ecosystem; M ytilus edulis ; Getis tistic. connec suitabil stability; Ord sta I n t r o d u c t io n strategy th at adjusts spaw ning according to envi ronm ental conditions (Gosling 1992b). Larvae and settling individuals are observed in spring, sum m er, and au tu m n (Bayne 1976; Pulfrich 1996; de Vooys 1999; Philippart an d others 2012). Dispersal of the pelagic planktotrophic m ussel larvae is strongly determ ined by w ater currents (Levin 2006; Metaxas an d Saunders 2009; Carson and others 2010; W hite and others 2010) an d the duration of the planktonic larval stage w hich depends on envi ronm ental factors such as food availability, tem perature, and salinity (Gosling 1992b). Blue m ussel larvae have been estim ated to reside in the w ater colum n for approxim ately 3-5 weeks (Gosling 1992b; de Vooys 1999). W hen larvae reach a size of approxim ately 0.2 m m they settle and produce byssus threads to attach to substrata and to each other w hich protects th em from dislodgem ent due to currents an d w aves (Bayne 1964, 1976; De Blok and Tan-M aas 1977; M cGrath and others 1988). Im portant substrata on the intertidal soft sedim ents in the W adden Sea are adult mussels, Pacific oys ters (Crassostrea gigas), cockles (Cerastoderma edule), beds of dead shells, protruding tubes of Lanice conchilega (Callaway 2003), m acroalgae and seagrass (Pulfrich 1996; w a Kangeri and others 2013). Bed form ation results in increased bottom surface roughness w hich reduces hydrodynam ic force (W iddows an d Brinsley 2002; D onker an d others 2012). Intertidal m ussel beds th at have established on soft sedim ents provide substratum for attach m en t an d shelter of postlarvae w hich results in positive feedback betw een m ussel bed presence and local recruitm ent allow ing a bed to persist over long tim e spans (Bayne 1964; M cGrath and others 1988; Gutiérrez and others 2003; Herlyn and others 2008). Foundations of dead shells w hich are built up u n d ern eath m ussel beds m ay also provide sub stratum for postlarvae thereby upholding th e p o sitive feedback even w h e n living m ussels have disappeared (Herlyn an d others 2008). Settlem ent m ay, how ever, also occur outside m ussel beds or shell foundations in years w h e n there are large quantities of settlers. Predation by shrim p and crabs (van der Veer and others 1998; A ndresen and van der M eer 2010) is an im portant loss factor shortly after settlem ent w hile w aves and currents (Nehls Reef-building bivalves such as intertidal blue m ussel (M ytilus edulis) are im portant for the com m u n ity com position and functioning of coastal ecosystems for various reasons. Intertidal m ussel beds provide structural heterogeneity on otherw ise hom ogeneous m udflats and thus provide habitat for other species (Gutiérrez an d others 2003; B uschbaum an d others 2009; van der Zee and others 2012). Mussels m ay com pete w ith other suspension feeders by consum ing large am ounts of ph y to plankton (Kam erm ans 1994; Cadée and Hegem an 2002). C onsequently, secondary p ro duction and biom ass of intertidal m ussel beds is high (Asrnus 1987) m aking th em an im portant food resource for, am ongst others, birds and crabs (Nehls and others 1997). M ussels m ay also prom ote planktonic prim ary production by accelerating n u trie n t cycling (Asrnus an d Asrnus 1991) and lo cal b enthic prim ary production by deposition of n u trien t-rich feces an d pseudofeces. Lastly, m ussel beds m ay locally reduce hydrodynam ic forces w hich influences local sedim ent properties (Widdows and Brinsley 2002). To develop a better u n derstanding of th e roles th at intertidal m ussel beds play in coastal ecosystems, it is im portant to identify th e m ain determ inants of m ussel bed dis tributions an d dynam ics. Insight into spatial dis tributions, h abitat suitability, reproduction, dispersal, settlem ent, and survival are prerequisites to u n d erstan d an d m odel large-scale spatio-tem poral variability of m arine bivalve populations. The purpose of th e present paper is to foster our u n derstanding of long-term spatio-tem poral developm ent of adult intertidal m ussel bed distri butions in th e W adden Sea as a step tow ards the developm ent of spatial nretapopulation m odels of blue mussels. Spatial distributions and dynam ics of intertidal m ussel beds are determ ined by th e com bination of larval supply, settlem ent an d various post-settle rnent loss processes w hich in their tu rn depend on th e interplay of large-scale hydrodynam ic forces and small-scale facilitation (van de Koppei and others 2005, 2012). The blue m ussel is a gonochoric bivalve species w ith a flexible reproductive 552 E. O. Folm er an d others and Thiel 1993; B rinkm an and others 2002; H am m ond an d Griffiths 2004; Herlyn and others 2008; D onker and others 2012) and ice-scouring during cold w inters (Strasser and others 2001) m ay be im portant during all life stages of m ussels and m ussel beds. A dult bivalve population sizes m ay fluctuate vig orously as a consequence of high tem poral variation in larval supply, early stage survival and variation in adult m ortality rate (Thorson 1950; Gosling 1992a; B eukem a and others 1993, 2001, 2010; Caley and others 1996; van der M eer and others 2001; Levin 2006). In th e W adden Sea, Macoma balthica, Cerasto derma edule , Mya arenaria , and Mytilus edulis often show recruitm ent peaks after cold w inters (Beukema and others 2001; Strasser and others 2003). Bivalve recruitm ent peaks following cold w inters have also been observed in other northw estern European seas and estuaries such as the Swedish w est coast (Möller 1986) and the W ash, UK (Young and others 1996). Various explanations have been suggested for this. First, adults produce m ore or larger eggs u n d er cold conditions because lower m aintenance m etabolism is required, thus leaving m ore energy for gonad p ro duction (Honkoop and van der M eer 1998). It should be noted, how ever, th at the experim ent th at H on koop and van der M eer (1998) perform ed w ith M a coma balthica shows th at the tem perature effect on gonad production is rather small. Secondly, bivalve spawning m ay occur earlier in the season after w arm w inters w hich m ay lead to a tem poral m ism atch w ith th e spring bloom of phytoplankton (which, inter alia , depends on light conditions) w hich decreases recruitm ent (Philippart and others 2003). The third line of explanations concerns the tem perature-re lated phenology of mussels and their m ain crustacean predators, shore crab Carcinus maenas and brow n shrim p Crangon crangon (van der Veer and others 1998; A ndresen and van der M eer 2010). Specifically, post-settlenrent survival on intertidal flats is relatively high after cold w inters because of low er densities or later arrival of shrimps and crabs (Beukema 1991, 1992; Young and others 1996; Strasser and G ünther 2001). Reduced densities or later arrival of predators leads to a larger proportion of young bivalves to grow too large to be handled and consum ed by predators (Philippart and others 2003; B eukem a and Dekker 2005). M ortality of adult bivalves m ay also peak during cold w inters because frost is lethal for cockles (Cerastoderma edule) and ice-scouring m ay eradicate intertidal mussel beds (Strasser and others 2001; Büttger and others 2011). The observation th at recruitm ent peaks often follow cold w inters suggests th at w in ter tem p era ture drives large-scale spatial population synchro nization because w in ter tem peratures are correlated over large geographic ranges. Particu larly, B eukem a and others (2001) show th at recruitm ent of th e bivalve species Macoma balthica , Cerastoderma edule , an d M ya arenaria is spatially synchronized betw een three distant locations in the w estern D utch and w estern G erm an W adden Sea betw een 1973 and 1999 (Mytilus was n o t in cluded in th e m onitoring in tw o of the three sites). Synchronous dynam ics on sm aller spatial scales can result from spatial autocorrelation in m ore re gional grow th determ inants. For instance, diverg ing regional trends in food availability, predation, and hydrodynam ic im pact could lead to distinct synchronization patterns betw een regions. Syn chronous dynam ics am ong nearby populations can also result from distance-lim ited dispersal of larvae betw een populations (Bjornstad an d others 1999; R anta and others 2006). The supply of planktonic larvae in a particular region depends on th e p ro x im ity of larvae sources and hydrodynam ic transport tow ard th e region (Levin 2006) and, to a certain extent, also on larval swim m ing behavior (Knights and others 2006). As the sw im m ing capacity of larvae is lim ited relative to w ater currents, the spatial connectivity of th e m ussel populations, and therefore th e degree of population synchrony, will largely depend on regional geonrorphology and currents during the larval stage (Young and others 1996; M cQuaid an d Phillips 2000; Gilg and Hilbish 2003; W atson and others 2010). Tem perature, the spatially diverging grow th determ inants and dispersal will lead to character istic signatures in spatial population covariance. That is, the spatial range over w hich th e determ i nants operate, will show up in the spatial range of population synchronization (Bjornstad an d others 1999). If tem perature is th e dom inant determ inant, it is expected th at synchronization will be observed in th e entire population. If regionally distinct determ inants are im portant th e n synchronization patterns will show autocorrelation on a regional scale. And m ore specifically, if geonrorphology and hydrodynam ic transport are im portant determ i nants of population synchronization, physical b ar riers such as rivers will show up as boundaries of synchronizing populations. Because tem perature and th e regional m echanism s operate sim ulta neously, the tem perature effect m ust dom inate the regional m echanism s to achieve visible synchroni zation at a large spatial scale. Similarly, regional effects m ust be pronounced in order to show up if there are strong tem perature effects. The objective of this study is to obtain insight into the distribution and dynam ics of intertidal Spatial Dynamics of Intertidal M ussel Beds m ussel beds in the W adden Sea by exploring large scale spatio-tem poral trends and scale-dependent population synchronizations. First, w e com pare the m ean spatial distribution of intertidal m ussel beds for th e period 1999-2010 w ith the spatial distri b u tio n of th e period 1968-1976 to get insight into long-term spatial variability and into the determ i nants of h ab itat suitability. Secondly, w e analyze spatial patterns in the synchronization of intertidal m ussel bed dynam ics for th e period 1999-2010. Thirdly, w e use a stochastic grow th m odel to a n a lyze th e impacts of w inter tem perature and storm iness on intertidal m ussel bed dynam ics for th e period 1999-2010 w hile controlling for density dependence. The analyses presented here aim to stim ulate th e developm ent of spatially explicit nretapopulation m odels including habitat suitabil ity, resource dynam ics an d hydrodynam ic con nectivity. Insight into these factors is relevant from a m an ag em en t and conservation point of view w h ere insight into long-term variability an d n a tu ral establishm ent of intertidal m ussel beds is a basic requirem ent. M ethods Study Region and Tidal Basins The W adden Sea (52°57,-5 5 °3 7 ,N, 404 4 '-8 012'E) is a shallow sea located in the southeastern part of the N orth Sea bordering the coastal m ainland of D en m ark, Germ any, and the N etherlands (Figure 1) [For a recent description of the biotic an d abiotic properties of th e W adden Sea see Philippart and Epping (2010)]. It is one of the w orld's largest coherent systems of intertidal sand an d m ud flats (Reise 2005; Reise and others 2010). The area contains coastal w aters, intertidal sandbanks, m udflats, shallow subtidal flats, drainage gullies and deeper inlets and channels. The Elbe, W eser, and Erns are th e m ain rivers discharging into the W adden Sea (Figure 1) w ith average freshw ater run-off of 700, 310, an d 125 nr3 s_1, respectively (Philippart an d Epping 2010). Freshw ater also e n ters th e W adden Sea at an average rate of approx im ately 470 nr3 s_1 through tw o sluices in the dike ("Afsluitdijk” ) in tidal basin 39 (van R aaphorst and de Jonge 2004). Tidal currents and exposure to w aves differ b etw een regions due to differences in tidal range, geonrorphology, fetch, and the occur rence of barrier islands (Figure 1 ). Tidal am plitudes range betw een 1.5 an d 3.0 nr in the N orthern (Ti dal Basins 1-10; henceforth TB denotes "tidal b a sin") and S outhern (TB 23-39) W adden Sea and are greater th a n 3.0 nr in the central part (TB 11- 553 22) (Staneva and others 2009). The m oderate and storm y w inds in th e area are m ainly southw esterly. The area can be divided into tidal basins w hich are natu ral geonrorphological an d hydrodynam ic subunits. A tidal basin is delineated by the m ain land, barrier islands, tidal divides and it connects to the N orth Sea via tidal inlets (Figure 1). The vol um e of w ater th at is being exchanged w ith the N orth Sea through tidal inlets is m uch larger th a n the volum e of w ater th at is displaced betw een neighboring tidal basins (Ridderinkhof and others 1990). W e use tidal basins as units of observation because locations w ithin tidal basins share various morphological, hydrodynam ic, and trophic p ro p erties th at are relevant for the abundance of m ussel beds (van B eusekom and others 2012). F u rth er m ore, the tu rn o v er tim e of a tidal basin— defined by R idderinkhof and others (1990) as the tim e necessary to reduce th e mass present in a basin to a fraction e_1 of the original mass—is relevant for the transport of larvae betw een tidal basins. In partic ular, R idderinkhof and others (1990) show th at the sm aller tidal basins in the D utch W adden Sea have short tu rn o v er times betw een 8 and 10 tidal periods (one tidal period is approxim ately 12.5 h) w hereas larger tidal basins such as th e Erns (Figure 1, TB 30) estuary have tu rn o v er times of 25 tidal periods. Data We used tw o data sets on m ussel bed distributions from tw o periods. The first consists of data collected in 36 tidal basins in the G erm an an d D utch W ad den Sea (Figure 1) for the period 1999-2010. The D anish tidal basins (TB 1-3 and the N orthern part of TB 4) w ere n o t included in this study because in these basins an incom parable survey protocol was applied and surveys took place every 2 years rath er th a n every year. A nnual m apping of m ussel bed contours in Schleswig-Holstein (TB 4-18), Ger m any, Lower Saxony (TB 18-30), G erm any and the N etherlands (TB 30-39) was perform ed on the basis of a com m on definition of a m ussel bed (de Vlas and others 2005) an d according to trilateral TMAG agreem ents (TMAG = Trilateral M onitoring and Assessm ent Group 1997). TB 4 refers to the southern half of th e tidal basin only because the n o rth ern half is located in D enm ark an d n o t su r veyed in a com parable fashion. The contours of visited m ussel beds w ere determ ined by a com bi n ation of aerial surveys, aerial photographs, and by w alking around th e bed w ith a han d -h eld GPS device following th e TMAG protocol (de Vlas and others 2005). In Schleswig-Holstein intertidal m ussel beds w ere m onitored by BioConsult SH 554 E. O. Folm er an d others I 50 1 0 0 km I - M u s s e l b e d c o v e r a g e (% ) Denmark n o d a ta I I < 0.01 I I 0.01 - 0.28 I I 0.28 - 0.59 I I 0.59 - 1.04 North Sea ■ 1.04 - 1.26 ■ 1.26 - 1.85 ■ 1.85 - 2.28 ■ 2.28 - 2.75 ■ 2.75 - 3.35 ■ 3.35 - 6.01 SH n o r th Germany s o u th W e a th e r s ta tio n s ★ e a st LS w est C uxhaven N o rd ern ey Germany NL w est V lie la n d H o o r n T e r s c h e llin g E m den L au w erso o g F igure 1. The 39 tidal basins of th e W adden Sea w ith average intertidal m ussel bed coverage over th e years 1999-2010. Intertidal m ussel bed coverage is defined as th e percentage of tidal flat area th a t is covered by m ussel beds. The intertidal flats in tidal basins 1-3 in D enm ark are colored gray because no com parable m ussel bed data w ere available. The nam es of th e w eath er stations are printed against a yellow background to clearly distinguish th em from th e region nam es. The geographic ex ten t ranges b etw een 52°57'-55°37' N orth an d 4 °4 4'-8°12' Bast (Color figure online). W eser Netherlands since 1998; in Lower Saxony by NLPV and NLWKN since 1994 (except for 1995 an d 1998); in the N etherlands by IMARES and M arinX since 1995. The largest com plete an d hom ogeneous data set th u s results from th e period 1999-2010. Region specific survey details are described below. In Schleswig-Holstein aerial photographs w ere taken an n u ally (except for the years 2006 and 2009) w ith scale of 1:15.000 or 1:25.000. The photographs w ere checked visually and the con tours of ascertained m ussel beds w ere digitized and included in a GIS. Results w ere com bined w ith field surveys. Each year about 40% of th e m ussel bed locations w ere visited. For those locations w here field surveys w ere no t possible, aerial photographs provided th e only source of inform ation. For the beds n o t visited in 2006 an d 2009 it was assum ed th at th e contours w ere th e same as the year before. It is noted, how ever, th at during the field cam paigns in 2006 and 2009 n ew beds and the disap pearance of beds w ere observed. In Lower Saxony contours of m ussel beds w ere determ ined on the basis of black and w hite aerial photographs w ith a scale of 1:15.000. Since 2006 true color photographs w ith a scale of 1:20.000 have been used. Photographs w ere analyzed w ith a stereoscope an d contours of m ussel beds w ere digitized and included in a GIS. For the year 2010 digital orthophotos w ere used and digital m apping of the contours was done on screen. Results w ere validated in the field. In the N etherlands, aerial inspection flights w ere carried out in spring prior to the field survey to assess w h e th e r the beds th at w ere m apped in the previous year w ere still present, w h e th e r there w ere n ew seed beds, and w h e th e r large portions of beds had disappeared. Locations w here changes w ere observed during flights w ere prioritized d u r ing the field surveys. The available tim e was u s u ally no t sufficient to m ap all m ussel beds by m eans of GPS. The percentage of bed locations th at w ere surveyed in the field was betw een 40 an d 95% . For beds n o t visited it was assum ed th at the contour is the sam e as in the preceding year if the aerial survey confirm ed the presence. W hen a bed is visited in a following year the contour in the m issing year is estim ated on the basis of the con tours before and after the missing year. The Spatial Dynamics of Intertidal M ussel Beds polygon vector files describing the contours of the m ussel beds w ere im ported into a GIS. The online supplem entary appendix A provides further details of th e survey protocols and institutes. In all th ree areas, according to th e TMAG p ro tocol, beds qualify as m ussel beds if the percentage cover by mussels is 5% or m ore. In th e period 1999-2010, m ussel beds m ay have contained Pa cific oysters (Crassostrea gigas) w hich w ere first ob served in th e w estern D utch W adden Sea (TB 39) in late 1970s (Fey and others 2010; Troost 2010) and in th e central W adden Sea (TB 26) in 1998 and in th e n o rth ern W adden Sea (TB 4 and 5) in the early 1990s (W ehrm an and others 2000). During th e w in ter 2009/2010 ice-scouring destroyed m ussel beds in Schleswig-Holstein (Büttger and others 2011). To avoid bias caused by rem oval of m ussel beds by ice-scouring, w e excluded from the analysis grow th estim ates from TB 4-18 for the year 2009. The second data set is a com pilation of m ussel bed contours for th e entire W adden Sea based on aerial photographs recorded in 1968, 1973, 1974, 1975, and 1976 (Dijkema and others 1989). The m ethodology used by Dijkema and others (1989) differs from th e m ethodology used to obtain the 1999-2010 data set. In particular, m ussel beds w ere dem arcated on th e basis of stereoscopy of aerial photographs and the contours aro u n d m ussel beds w ere draw n m ore loosely and w ider th a n for the 1999-2010 data set. A lthough com parison of absolute areas betw een the data sets is not possible, it is nevertheless possible to com pare spatial p a t terns in relative abundances. Maps of th e intertidal areas and of land b o u n d aries w ere provided by the Com m on W adden Sea Secretariat (CWSS), W ilhelm shaven. The intertidal area was assum ed to be constant over the period 1968-2010. A vector file representing the tidal basins (Figure 1) was obtained from Kraft and others (2011). All spatial data w ere organized in a GIS. Lam bert A zim uthal Equal Area projection (using th e ETRS89 datum ; EPSG code 3035) was used to obtain tru e area representation over the full study area. To investigate th e effects of w inter tem perature and storm iness on the dynam ics of m ussel bed coverage w e m ade use of air tem perature and w ind m easurem ents from different w eath er stations in the W adden Sea area. The w eath er stations are all located n ear the coastline (Figure 1). Germ an tem p eratu re data w ere obtained from th e Germ an M eteorological Service (w w w .dw d.de) for the sta tions List auf Sylt, C uxhaven, N orderney and Em den. W e used G erm an w ind m easurem ents from 555 the stations List auf Sylt and N orderney only b e cause the storm iness at the stations C uxhaven and Ernden is substantially low er th a n at the other stations because they are located m ore land in wards. D utch tem perature and w ind data w ere obtained from the Royal N etherlands M eteorolog ical Institute (w w w .knrni.nl) for the stations Lauwersoog, H oorn Terschelling, Vlieland and De Kooy. A m easure for w inter severity (T) was ob tained by calculating th e H ellm ann n u m b er w hich is th e absolute value of the sum of the daily average tem peratures below zero degree Celsius for the period 1 Novem ber to 31 M arch. It is noted th at a strong correlation betw een air- an d w ater-tem perature has been observed over a long period of tim e in the w estern D utch W adden Sea (TB 39 in Fig ure 1) suggesting th at air tem perature is a solid indicator for w ater tem perature (van A ken 2008). We defined storm iness as th e n u m b er of days w ith average w ind speed greater th a n 15 m /s (~ 7 Bit) in the period 1 A ugust through 30 June. M ussel bed grow th rates w ere related to th e Heilm an nu m b er of the preceding w in ter (1 Novenrber-31 M arch) and th e storm iness of the following period (1 August-30 April). For instance, grow th betw een 2005 and 2006 (and labelled R2oos) is related to the H eilm an nu m b er over the period 1 N ovem ber 2004 and 31 M arch 2005 and the storm iness betw een 1 A ugust 2005 and 30 April 2006. For each tidal basin w e used the w eath er data from the nearest w eath er station (stations C uxhaven and Ernden are excluded for w ind). The nearest w eath er station is defined as the one w ith the shortest straight line distance from the centroid of the tidal basin. Statistical Models Mussel Bed Area Per Tidal Basin and Spatial Clustering We define m ussel bed area per tidal basin (A¡) as the sum of th e areas of the m ussel bed polygons intersecting w ith the tidal basins. M ussel bed cov erage per tidal basin i (C¡ in %) is th e percentage of total intertidal flat area th at is covered by m ussel beds. For the period 1999-2010 w e calculate yearly sets of coverages and the m ean coverage over the period; for the period 1968-1976 only one set of m ussel bed coverage can be calculated. Ord and Getis (1995) have developed tw o local statistics, Gi an d G\, to evaluate the degree of spatial clustering aro u n d a focal location, i. Statistic G¡ excludes location i front th e cluster w hereas G¡ includes it. The G and G* statistics express the sum of th e values in th e vicinity of a particular location as the proportion of the sum of the values for the entire area. G and G* are positive w h e n the m ean of 556 E. O. Folm er an d others a d u ste r is greater th a n the overall m ean and negative w h e n it is less. Specifically, the statistics identify those clusters of locations w ith values higher or low er in m agnitude th a n m ight occur by chance. G an d G* are similar to the z score in th at th ey indicate how m any standard deviations a v a lue is above or below the m ean. For the m ean, the standard deviation an d testing of the G and G* statistics, see Getis (2009). We use the G* statistic because w e are interested in the degree of cluster ing rath er th a n in the influence of observation i on th e value of j (Getis 2009). We use the G* to identify clusters of tidal basins w ith high and low values of m ean m ussel bed coverage for the periods 1999-2010 an d 1968-1976. Following the notation in Getis (2009) G] is defined as follows: % ¿ W ij(d )C j _ i= l w *c ----- tidal basin i for the period 1999-2010. M ussel bed grow th rates for the years 1999 to 2009 are calcu lated for the 25 tidal basins w ith m ean coverage greater th a n 0.2% . W e tested for density dependence an d the effects of H ellm ann tem perature ( T ) of the preceding w inter an d storm iness (S ) during the following season on grow th rate (R) by m eans of a first-order linear m odel (Royarna 1992). Because w e had no a priori expectations of the functional form, w e also considered n o nlinear relationships betw een the dependent variable R and the explanatory variables coverage (C), T and S by also including their squared term s. We also considered interactions betw een T and C an d betw een S and C to investigate w h e th e r the effects of th e H ellm ann tem perature and storm iness depend on m ussel bed coverage. The first-order linear m odel can be w ritten as: Ri,t = ßo + ß i Q,t + ßiTi,tp + ß^S i'tf + ß4Cft + ßs Tçtp + ßöSjtf + ß i Ci,tT¡,tp + ß &Q,tSi,tf n w * = X! WiiW and s*¡= X! wl for a11j j=1 j= 1 w h ere w ¡j(d ) is the spatial w eight describing the topological relationships betw een tidal basins i and j as a function of the distance (d ) betw een their centroids. It equals 1 if the distance is less th a n 30 1cm and 0 otherw ise (note th at w u= 1). We take th e cut-off distance of 30 1cm to avoid isolation of large tidal basins. A large cut-off distance w ould lead to large num bers of neighbors for the smaller tidal basins. The m axim um distance betw een cen troids of neighboring tidal basins is 28 lent (TB 37 and TB 36) an d th e shortest about 6 lent (TB 16 and TB 15). C and s are th e m ean an d standard devia tion of th e coverages of m ussel beds for the 36 tidal basins. A cluster of large positive values of G¡ indicates a hotspot of tidal basins w hereas a cluster of large negative values indicates a cold spot. Growth, Density Dependence, W inter Temperature and Storminess We calculated relative yearly m ussel bed grow th rate per tidal basin i (R,-1) as the difference in the percentage coverages of m ussel beds betw een suc cessive years divided by th e m ean coverage of tidal basin i in th e period 1999-2010, w h ere Q t is th e m ussel bed coverage in tidal basin i in year t and Q is th e m ean m ussel bed coverage of TLtp denotes the H ellm ann tem perature of the w inter preceding tinte t in tidal basin i (e.g., grow th in year 2005 depends on the tem perature of w inter 2004/2005). S l y denotes the storm iness during the season following tinte t in tidal basin i (for example, grow th in year 2005 depends on th e storm iness of the season 2005/2006). W e used linear-m ixed ef fects m odels (LMM) to account for possible differ ences betw een tidal basins an d years. Particularly, the effects of C, and S w ere allow ed to differ by tidal basin an d year by including th em as random slopes. We did no t m odel T w ith a random effect to keep the nu m b er of variables w ith random effects low to avoid unw ieldy models. The observations w ere w eighted by Q to account for the fact th at the relative changes in coverage are m ore precisely determ ined in the tidal basins w ith large Q. Fol lowing, Bolleer and others (2009) w e used re stricted m axim um likelihood (REML) estim ation. The original distribution of R was leptokurtotic w ith positive and negative tails w hich resulted in heteroscedastic m odel residuals. W e therefore norm alized the distribution of R by using th e pow er transform ation (lRli sgn(R) — 1 ) /A ; A > 0 (Bickel and Doksurn 1981). The value of X was selected on the basis of the distribution of th e residuals. We checked for norm ality and hom ogeneity by visual inspections of plots of residuals against fitted val ues. W e set X = 0.5 w hich yielded satisfactory re sults. W e used th e following tw o-step m odel selection procedure. Step 1 : We started by estim ating th e full m odel w hich included all fixed effects and the Spatial Dynamics of Intertidal M ussel Beds random intercepts and slopes for C and S for both th e levels "tidal basin" an d "y ear". W e rem oved th e random slopes (from bo th levels) in order of increasing standard deviation starting w ith the random slope w ith the lowest standard deviation. The successive m odels w ere com pared on the basis of likelihood ratios. If the p-value for the likelihood ratio test (LRT) was greater th a n 0.05, preference was given to th e sim pler m odel. The random intercepts for the levels "tidal basin" an d "y ear" w ere retained to m eet th e assum ption of hornoscedasticity. Step 2: Given the obtained random effects structure, w e estim ated all plausible nested m odels by setting coefficients of the fixed effects at zero (or rem oving their variables from the m odel). Models including quadratic term s and interactions w ere only considered if the low er order effects w ere also in th e m odel. The estim ated m odels w ere ranked in order of increasing AIC levels (B urnham and A nderson 1998). Synchronization Synchronization is defined as th e correlation b e tw een th e yearly m ussel bed grow th rates of the tidal basins. We consider the correlation in yearly grow th rate rath er th a n the correlation in cover ages because coverages m ay be affected by tem po ral trends (Turchin 2003). For each tidal basin pair, synchronization is calculated using Pearson's cor relation coefficient (r). To avoid possible influence of tidal basins w ith low m ussel bed coverages, the correlation coefficients are calculated betw een tidal basins w ith average coverage greater th a n 0.2% . This resulted in a sym m etric 25 x 25 correlation m atrix. The correlation m atrix for th e entire data set—th at is, th e data set including the tidal basins for w hich th e average coverage is less th a n or equal to 0.2% —is presented in th e online appendix B. We visualized the correlation m atrix by plotting ellipses shaped as contours of a bivariate norm al 1999 -2 0 1 0 557 distribution w ith u n it variance and correlation r. The points of th e contour are given by (x,y) = (cos(0 + £ ), cos(0 —£)) an d d = arccos (r) w here d e [0, 27i] (M urdoch an d Chow 1996). W e used R 2.15.2 (R D evelopm ent Core Team 2009) for data handling, estim ation, and plotting. For spatial data handling and statistics w e used the R packages sp (Pebesma an d Bivand 2005; Bivand and others 2008), rgeos (Bivand and R undel 2011), rgdal (Keitt and others 2011). For m ixed m odeling w e used th e R package lm e4 (Bates an d others 2012). For plotting w e used ellipse (M urdoch and Chow 1996, 2012) and ggplot2 (W ickham 2009). R esults Mean M ussel Bed Coverage and Spatial Stability The m ean m ussel bed coverages strongly varied betw een the tidal basins (Figure 1). The highest m ean coverages w ere found in the tidal basins in the eastern part of th e D utch W adden Sea (TB 3 1 36) an d in the w estern part of Lower Saxony (TB 22-TB 29). In some of the areas m ean coverage ranged betw een 4 and 6% of the intertidal area. The coverages w ere low in the n o rth ern tidal basins in Schleswig-Holstein (TB 4-10) (betw een 0 and 1.5%) w hereas m ussel beds w ere virtually absent in TB 11-18. The coverages w ere also relatively low (< 0 .5 % ) in the w estern D utch W adden Sea (TB 37-39). The spatial variation in m ussel bed coverages presented in Figure 1 is m ore clearly visible in the left panel of Figure 2 w hich shows the distribution of m ussel bed h o t and cold spots (that is, th e G* statistic) for the sam e period. The right panel shows G* for th e 1968-1976 period. In bo th periods the hotspots are situated in the eastern D utch W adden Sea and in w estern Lower Saxony w hile th e cold spots are situated in n o rth ern Schleswig-Holstein 1968 - 1976 5 ^ 3 2 ^ 0 -1 -2 -3 F igure 2. Getis-Ord hotan d cold spot m aps for average intertidal m ussel bed coverages for th e periods 1999-2010 an d 1968-1976 (Color figure online). 558 E. O. Folm er an d others and in th e w estern D utch W adden Sea. The highest G* value occurred in w estern Lower Saxony in the period 1968-1976 and in the eastern D utch W ad den Sea in th e period 1999-2010. Temporal Trends and Synchronization Figure 3 shows for each tidal basin the tren d in m ussel bed coverage over th e period 1999-2010. The yearly fluctuations are high in the tidal basins w h ere th e average coverages are high and m uch low er in th e tidal basins w ith low average cover ages. For exam ple, the coverages fluctuate irregu larly betw een 0 and 8% in the eastern Dutch W adden Sea (TB 31-36) w hereas th e yearly ch an ges in coverage in n o rth ern Schleswig-Holstein (TB 4, 6, 7, 9, and 10) are m ore regular and range b e tw een 0 and 2.5% . Figure 3 also shows th a t tidal basins in each o ther's vicinity show similar fluctu ation patterns. For instance, TB 22-26 in Lower Saxony show a peak in th e year 2000 after w hich th e coverages decrease u n til 2006 after w hich there is a slight an d regular increase. In the eastern Dutch W adden Sea (TB 31, 32, 34, 35, 36), a peak is ob served in 2002, after w hich th e coverages decline, until a second peak is observed in m ost of th e tidal basins in 2006. The grow th in the eastern D utch W adden Sea in 2002 and 2006 is no t observed in the G erm an (TB 4-30) or w estern D utch W adden Sea (TB 37-TB 39). Figure 4 shows th e correlation in m ussel bed grow th betw een all tidal basins w ith average cov erages above 0.2% . The figure shows three blocks of tidal basins w ith synchronic dynam ics. The m ost northerly synchronic block (TB 4-10) is located in n o rth ern Schleswig-Holstein. The second syn chronic block (TB 19-30) is bordered by the river Elbe (TB 18) in th e east and by th e Erns estuary (TB 30) in the west. The dynam ics of th e relatively small TB 20 are no t synchronized w ith the other tidal basins in the block. The third block of syn chronic tidal basins (TB 30-39) is located in the D utch W adden Sea. W ithin the three blocks the correlation betw een tidal basins is highest w ithin the centers of the blocks and decreases w ith dis tance betw een th e tidal basins. A lthough shortdistance synchronization dom inates, long-distance synchronization is also observed betw een the blocks consisting of TB 36, 37, and 39; and TB 19, 20, and 22, respectively. Figure 3. Intertidal mussel bed coverage (%) of the tidal flats for 36 tidal basins for the period 1999-2010. The numbers in the strips of the panels represent the tidal basin number (see Figure 1). Spatial Dynamics of Intertidal M ussel Beds O ^ O ^ - C O I ^ O ) - ! — -I— 559 ) O W ( O t i r ) C D N C O O ) O ^ W ( O t l i ) 0 N O ) -I— C \ I C \ I C \ I C \ I C \ I C \ I C \ I C \ I C \ I C O O O C O C O C O C O C O C O C O C Û C Û C Û C Û C Û C Û C Û C Û C Û C Û C Û C Û C Û C Û C Û C Û C Û C Û C Û C Û C Û C Û C Û C Û C Û I— I— I— I— I— I— I— I— I— I— I— I— I— I— I— I— I— I— I— I— I— I— I— I— I— tb4 /OO0 OOOOOOOOO0 OOOOOOOOOO tb6 ✓ /o/ooooooooo/oos%o%o%oo tb7 o o OOOOOOOOOOOOOOOOOOOOOO TB9 TB10 TB14 TB19 TB20 TB22 TB23 TB24 TB25 TB26 TB27 TB28 TB29 TB30 TB31 TB32 TB33 TB34 TB35 TB36 TB37 TB39 ooo ooooooxoooooooooooooo 0000 00000000000000 x00000 0)0000 o o o o o o o o o o o o o o o o o o o o o o o o o oooooooooooooooooo o o o o o o o 00000000000000000 oooooooo //0 /00/000000000 O O O O O O O O / 0*0000000000000 000X0000/0 //OO0OOOOOOOOO OOOOOOOO0// /OO0OOOOOOOOO 00000000/0// 00/000000000 ooooooooooooo ooooooooooo 00O O 0O O O O O O O O O o o o o o o o o o o 00000000/000/00 OOOOOOOO0 oooooooooooooooo OOOOOOOO /o o o o o o o s o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o / 0/0000 O O 0 OO 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 / 0 0OOO 0 0 0 0 X 0 0 0 0 0 0 0 0 0 0 0 0 0 0 O O O O O O O O OOOOO OOOO OOO0 o o o O O O O OO OO OOOOO OOOO OO0OO o o O O O O O O O O O O O O O O O O O O O O O O O 0 OOOOOOO0OOOOOOOOOOOOOOO0 F igure 4. Ellipse m atrix of th e correlations b etw een yearly intertidal m ussel bed grow th rates of th e tidal basins for w hich th e average coverage w as greater th a n 0.2% . The ellipses are th e contours of a bivariate n o rm al distribution w ith u n it variances an d Pearson's correlation coefficient r (M urdoch an d Chow 1996). In th e case of positive correlation, th e color of th e ellipse is blue. In th e case of negative correlation, th e color is red. The intensity of th e color relates to th e (absolute) size of th e correlation coefficient. Correlations of th e elem ents along th e diagonal are always 1 an d n o t inform ative an d th u s n o t presented. The ellipse m atrix for all tidal basins is presen ted in th e supplem entary appendix B (Color figure online). M ussel Bed Growth Models In th e full m odel of step 1—th at is, the m odel including all fixed effects and random intercepts and ran d o m slopes for coverage and storm iness at th e levels tidal basin and year—th e relative stan dard deviation of the random slope of coverage in level tidal basin was small. Accordingly, the likeli hood ratio test show ed th at deleting this random effect from th e full m odel did not cause a significant reduction in likelihood and thus preference was given to th e simpler m odel (Table 1, m odel 4). This elim ination procedure was continued until only th e intercept at the levels of tidal basin and storm iness at the year level rem ained as random effects. On th e basis of the thus obtained random effects m odel (that is, random intercept for tidal basin level and storm iness at the year level) all nested fixed effects m odels w ere estim ated and ordered by AIC (Table 2). The best m odel only in cludes a negative term for C suggesting th at negative density-dependent processes are acting (—0.160, SE 0.024). The second best m odel in term s of AIC—w hich is considerably w orse th a n the best m odel (AAIC = 4.486)—indicated in addition to a negative effect of C a slight positive b u t highly insignificant effect of storm iness (S). The third best m odel (AAIC = 5.512) also indicated a negative effect of C an d a positive effect of C2 suggesting th at the negative effect of coverage leveled of at high values of C. Models 4 and 5 fit th e data substantially w orse th a n m odels 1-3 (AAIC > 7.859). M odel 4 shows a negative effect of S and a positive effect of S2 in addition to the negative C effect. M odel 5 shows a positive b u t highly insignificant effect of S b u t a negative an d significant interaction betw een C and S. W inter severity (T) does n o t show up as a predictor in the models 1-5. The variance of the random effects at the year level indicated consid erable heterogeneity across years w hile the vari ance across tidal basin is very low. 560 T able 1. E. O. Folm er an d others S e le c tio n of R a n d o m E ffects of th e L in e a r M ix e d M o d e ls of M u s se l B e d G ro w th M odel TB level SD Year level SD 1 Intercept Residual Intercept 0.232 Intercept 0.100 Intercept 0.341 0.958 0.706 0.279 0.922 0.707 0.106 0.296 0.882 0.587 0.129 0.254 0.856 1.514 0.258 2.521 0.844 2 S 3 Residual Intercept 0.153 Intercept C S 4 Residual Intercept C 5 Residual Intercept C S 0.326 0.128 Intercept 0.227 0.103 0.045 Intercept C S C S Residual X2 Df P value 21.421 2 < 0 .0 0 0 2.141 3 0.544 3.375 2 0.185 0.000 3 1.000 C = intertidal mussel bed coverage (% ); S - storminess; TB = tidal basin. In this stage o f the model selection procedure the fixed effects (C, C2, T, I 2, S, S2, C x S, C x T) remain unchanged. The column SD gives the standard deviations fo r the random effects o f the intercept, C and S for the levels TB and Year. The columns y 2, D f and the P value present the relevant outcomes o f the y f tests. The num ber o f observations is 269 after removing the observations that were possibly influenced by ice-scouring; the num ber o f tidal basins is 25 (viz. the ones that fit the criterion o f mean coverage greater than 0.2%) and the number o f years is 11. D is c u s s io n and C o n c l u s io n s We analyzed th e spatial variability of the coverage of m ussel beds in the international W adden Sea on th e basis of data collected in the periods 1968-1976 and 1999-2010. W ithin each period, notable re gional differences in m ussel bed coverages w ere observed. B etw een the periods, how ever, these distribution patterns w ere rem arkably stable in th at th e regions th at w ere densely populated in the period 1968-1976 w ere also densely populated in th e period 1999-2010. The highest coverages occur in th e relatively small an d sheltered tidal basins in th e eastern D utch an d in th e w estern Lower Saxon W adden Sea w hereas m ussel beds w ere virtually absent in tidal basins th a t are n o t sheltered by barrier islands (TB 11-18). The strikingly stable m ussel bed distributions m ay be explained by the regionally variable levels of exposure to hydrody nam ic forces w hich are relatively constant on the scales of decades. A lthough analysis of long-term regional differ ences in m ussel bed coverage provides insight into the relatively invariable determ inants of habitat suitability, analysis of year-to-year variability including spatial synchrony, sheds light on the determ inants of short-term dynam ics. W e expected th at yearly increases in coverage w ould be ex plained by episodic recruitm ent events following cold w inters and th at decreases w ould occur during storm y w in ter seasons. In particular, previous studies perform ed at various locations have show n th at bivalve recruitm ent events are often preceded by cold w inters (Möller 1986; Young and others 1996; B eukem a and others 2001; Strasser and others 2003) w hich has led to th e com m only held view th at w in ter severity is th e m ain driver of b i valve recruitm ent events and population synchro nization in N orth W est Europe. Because w inter tem peratures are strongly correlated in the entire study region (Figure 5) w e hypothesized th at grow th peaks w ould show up at the same tim e th ro u g h o u t th e entire study area. If regional grow th determ inants or larval dispersal w ould dom inate, th e n population synchronization w ould show up at sm aller spatial scales. Our analyses did n o t show th at cold w inters system atically induced increases in m ussel bed coverages. Nor did w e find strong evidence for negative effects of storminess. We did find, how ever, th at synchronic grow th was constricted to confined blocks of tidal basins b o r dered by hydrodynam ic barriers such as rivers and large extents of unsuitable habitats. These results point tow ard th e im portance of considering larval transport an d early stage survival to fu rth er d e velop understanding of the spatial dynam ics of m ussel beds in th e W adden Sea. The im portance of larval transport has also been dem onstrated for coastal m ussel populations in southw est England (Gilg and Hilbish 2003) and th e southeast South Africa (M cQuaid and Phillips 2000) w here coastal Spatial Dynamics of Intertidal M ussel Beds 561 T able 2. Regression Table w ith th e Full and th e Five Best Mixed-Effects Models for M ussel Bed Grow th Rate Selected on th e Basis of AIC Values F u ll M odel Intercept C C2 S S2 T T2 C x S C x T R andom effects Intercept for TB Intercept for year S for year residual AAIC AIC BIC Deviance Log-likelihood -1 .6 1 8 (-0 .3 0 3 ) -0 .2 7 7 (-0 .0 7 8 ) 0.028 (0.009) -0 .1 1 4 (0.169) 0.046 (0.029) -0 .0 0 1 (0.008) -0 .0 0 0 (0.000) -0 .0 4 0 (0.016) -0 .0 0 0 (0.001) M o d el 1 -1 .8 9 4 (0.098) -0 .1 6 0 (0.024) M o d el 2 -1 .9 5 5 (0.229) - 0 .1 5 8 (0.024) M o d el 3 -1 .7 3 0 (0.124) -0 .3 2 2 (0.072) 0.024 (0.009) 0.024 (0.087) M o d el 4 M o d el 5 -1 .7 1 8 (0.223) -0 .1 3 6 (0.024) -2 .0 7 9 (0.234) -0 .0 9 9 (0.034) -0 .2 9 0 (0.142) 0.069 (0.026) 0.111 (0.094) -0 .0 4 1 (0.015) 0.010 0.498 0.078 0.851 0.019 0.427 0.061 0.868 990.216 1040.542 962.216 -4 8 1 .1 0 8 930.489 955.651 916.489 -4 5 8 .2 4 4 0.019 0.471 0.068 0.868 4.486 934.975 963.733 918.975 -4 5 9 .4 8 8 0.012 0.416 0.061 0.864 5.512 936.001 964.758 920.001 -4 6 0 .0 0 0 0.015 0.360 0.060 0.866 7.859 938.348 970.700 920.348 -4 6 0 .1 7 4 0.023 0.470 0.070 0.856 8.188 938.677 971.030 920.677 -4 6 0 .3 3 9 C = mussel bed coverage (% ); S - storminess; TB = tidal basin. Standard errors in parentheses underneath the regression coefficients. The num ber o f observations (269) is the same as in Table 1. circulation is an im portant determ inant of larval dispersal. A lthough our hypothesis and results are consistent w ith findings from other systems, w e do n o t rule o u t alternative explanations. For instance, regional differences in the change of eutrophication (van B eusekom and others 2012) an d carrying capacity or predation rates could lead to regional differences in vital rates w hich could cause the observed synchronization patterns. A nother possi ble cause of diverging dynam ics in the separate blocks m ay be related to local hydrodynam ic dis turbances w hich w ere no t detected by regressing adult m ussel bed dynam ics on storm iness. To dis tinguish betw een th e above-m entioned hypotheses it is im portant to obtain detailed long-term spatial data on com petition and food lim itation, rep ro ductive capacity and output, early stage predation, recruitm ent, hydrodynam ic disturbance and transport. In this context it is also im portant to have better u n derstanding of the positive and negative interactions betw een Pacific oysters and mussels. Particularly, mussels and oysters benefit from each other in that they provide suitable substrate but they suffer from each other due to competition for food resources. The long-term consequences of the establishment of Pacific oyster populations for mussel populations will depend on the responses of both populations to environm ental conditions and on the interactions betw een the populations. The fact th at w e did n o t find a tem perature effect on grow th does n o t im ply th at tem perature m ay n o t have an effect on the long-term dynam ics and sustainm ent of intertidal m ussel beds at all. There are several reasons w h y w e m ight have missed to detect a w in ter tem perature effect. Firstly, our tim e series covered the period 1999-2010 w hich was relatively w arm . B eukem a an d others (2001), w ho did find a tem perature effect, used data for the period 1969-1999 w hich contained several cold w inters. In particular, the severe w inters of 1978/ 1979 an d 1995/1996 w ere followed by extraordi n ary recruitm ent events for different species of 562 E. O. Folm er an d others 150 - F igure 5. W inter severity an d stornriness for the different w ea th er stations along th e W adden Sea coast. W inter severity or th e H ellm ann W e a th e r s ta tio n List a u f Sylt -*■ C u x h a v e n -*■ N o rd e rn e y E m den L a u w e rs o o g -*■ H o o rn T e rsc h e llin g -»■ V lieland -*■ D e K ooy -+■ g 100 CD CD i— B temperature was c 5 50 - o98/99 10.0 99/00 00/01 01/02 02/03 03/04 04/05 05/06 06/07 07/08 08/09 W e a th e r s ta tio n - -*■ List a u f Sylt -*■ N o rd e rn e y L a u w e rs o o g -*■ H o o rn T e rsc h e llin g -*■ V lielan d -*■ D e K ooy 7.5 CD CD CD ■| 09/10 calculated as th e absolute value of th e sum of th e daily average tem peratures th a t w ere below zero degrees Celsius for th e period 1 N ovem ber-31 M arch. Storm iness is defined as th e n u m b e r of days w ith average w in d speed greater th a n 15 nr/s (~ 7 Bft) in th e period 1 A ugust-30 Ju n e (Color figure online). 5.0 - 1— o CO 2.5 - 0.0 - 98/99 99/00 00/01 01/02 02/03 03/04 04/05 05/06 bivalves including M ytilus edulis. The study of M öller (1986) also included th e year 1979 after w hich an eightfold increase in th e num erical d en sity of M ya arenaria was observed. The study of Strasser and others (2003) focused on recruitm ent of several species of bivalves in the w hole W adden Sea following th e cold w in ter of 1995/1996. They found higher th a n average densities of recruits of Cerastoderma edule in 1996 in the entire W adden Sea. For th e species Macoma balthica an d M ya are naria higher th a n average recruitm ent was only observed in th e southern W adden Sea. It is th e re fore possible th at w e w ould have found statistically significant tem p erature effects if our tim e series had included m ore and colder w inters. It is im portant to note, how ever, th at cold w inters are no t a p re requisite for high recruitm ent events as relatively m ild w inters m ay also be followed by high recru itm en t events. For instance, after the rela tively m ild w inters of 1974/1975 and 1990/1991, B eukem a and others (2001) found high densities of recruits for different species of bivalves. A nd in our data set th e m ild w inters of 2000/2001 and 2004/ 2005 w ere followed by substantial increases of 06/07 07/08 08/09 09/10 adult m ussel beds in th e D utch W adden Sea (ob served as peaks in 2002 an d 2006). Because cold w inters are expected to occur less frequently in the future due to clim ate change, the factors u n d erly ing recruitm ent after relatively w arm w inters will becom e increasingly im portant determ inants of bivalve m etapopulations. A nother potential cause of th e absence of a w in ter tem perature effect is th at loss processes operate betw een th e m om ents of recruitm ent an d th e surveys of adult m ussel beds w hich m ay distort the correlation betw een w inter tem perature and m ussel bed grow th to som e ex tent. This implies th at w inter tem perature effects m ight be m ore easily detected for endobenthic species such as Macoma balthica th a n for epibenthic species th at are m ore vulnerable to hydrodynam ic disturbance. Note, how ever, th at w e did observe bivalve-characteristic episodic grow th peaks by considering yearly changes in the coverage of adult m ussel beds (Figure 3). The absence of evidence of a stornriness effect in the grow th m odel does no t im ply th at stornriness does n o t influence m ussel bed dynam ics. The ab sence of a stornriness effect m ay be explained by Spatial Dynamics of Intertidal M ussel Beds th e fact th at w e considered dynam ics of adult m ussel beds w hile especially settlem ent an d new ly settled m ussel beds are sensitive to hydrodynam ic disturbance and stornriness because individual m ussels have n o t yet attached them selves firm ly to th e sedim ent an d each other. As a result estab lishm ent and developm ent of adult m ussel beds tends to occur at relatively sheltered locations w h ere th e im pact of storms on adult beds is rela tively low. For exam ple, Nehls and Thiel (1993)—w ho studied survival of m ussel beds including new ly recruited m ussel beds—found th at only 49 o u t of 94 beds survived their first w inter and th at all of the surviving beds w ere located in sheltered areas. These areas w ere also found to be populated w ith m ussel beds in several decades b e fore and after their investigation. Our paper con cerns th e dynam ics of adult m ussel beds w hich implies th at w e are analyzing th e dynam ics of beds th at already survived the initial young stages and th u s occur in relatively sheltered locations only. The lim itation of our em pirical approach to d e tect stornriness effects could be cleared up by integrating observations of young and adult in te r tidal m ussel beds w ith hydrodynam ic models (Donker and others 2012). A lthough such in te grated m odels provide insight into th e m echanic interactions b etw een hydrodynam ic forces and intertidal m ussel beds, the approach will be difficult to apply to large tem poral and spatial scales for the following tw o m ain reasons. Firstly, the hydrody nam ic conditions of all the relevant loca tions—w hich include tides, w ind and fetch—need to be k n o w n w hich requires very detailed data sets, extensive m onitoring and com plex and com puter intensive m odeling. Secondly, because w aves and currents are atten u ated w h e n they travel over the rough m ussel bed surfaces, they cannot be con sidered in d ep endent variables. Yet, it w ould be useful to have m ore insight into the long-term impacts of waves and currents on the tidal flats in order to identify suitable m ussel bed habitat in term s of hydrodynam ics independently of the observations presented in this paper. In this paper, w e have suggested th at synchronic spatial population dynam ics w ithin blocks of neighboring tidal basins m ay result from distancedependent dispersal of larvae. In particular, larvae th at are produced in one tidal basin m ay be tran s ported to hydrodynam ically connected tidal basins. This suggestion is supported by the fact th at the borders of th e synchronic blocks are located at tidal basins w ith estuaries an d rivers (in w hich w ater currents form a barrier tow ard the next tidal b a sins) an d at regions th at lack suitable habitat. As 563 m entioned above, th e tu rn o v er tim e for th e Erns estuary (Figure 1, TB 30) is substantially longer th a n the tu rn o v er times for the other D utch tidal basins (Ridderinkhof an d others 1990) so th a t the chance for a larva to survive the tim e it takes to pass the estuary and arrive at suitable habitat is relatively low. The chance to survive the passing of the Elbe estuary (TB 18) an d arrive at suitable habitat w ould conceivably be even lower. Our re sults, th at is, th at th e population dynam ics w ithin the blocks separated by th e Erns an d Elbe estuaries are distinct, are in line w ith this observation. H ydrodynam ic transport m odeling, on the basis of tides and w ind, could elucidate possible fates of m ussel larvae and provide insight into the con nectivity of suitable habitat patches w ith in tidal basins. This type of m odeling w ould help to provide insight into the structure of the nretapopulation netw ork and w ould be able to show w hich sub populations produce larvae w ith high survival and settling probability. In addition to w ater m ove m ent, the characteristics of the w ater affecting larval grow th and developm ent (such as food availability, salinity, an d tem perature) should be considered (M etaxas and Saunders 2009). Particu larly, a low er grow th- an d developm ent rate during the larval phase increases th e tim e a larva needs to com plete developm ent w hich increases the risk of being predated by ctenophores and benthic feeders or of being transported offshore (Young and others 1998; Philippart and others 2003; M etaxas and Saunders 2009). H ydrodynam ic connectivity m o d eling and w ater quality m easurem ents in com bi n ation w ith larval requirem ents could provide a fram ew ork for the developm ent of a dispersal m odel describing the corridors betw een suitable habitats from th e perspective of larvae. In th e W adden Sea there is high quality, com parable long-term distribution data available for intertidal m ussel beds b u t n o t for subtidal m ussel beds. Because our quantitative analyses required hom ogeneous data w e have focused on intertidal m ussel beds only. Subtidal m ussel populations may, how ever, influence the dynam ics of intertidal populations by producing larvae th at end up in the intertidal an d vice versa. Hence, a com plete m odel of spatial population dynam ics w ould include su r vival, production, dispersal, and settlem ent of b oth intertidal and subtidal populations. The spatial configuration of suitable habitat and population grow th dynam ics presented in this p a per is relevant from a n atu ra l resource m an ag e m en t perspective. If it is com bined w ith the outcom es of fu rth er research suggested above, it will be possible to pinpoint those locations and 564 E. O. Folm er an d others sub-populations th at are of pivotal im portance for th e resilience of the nretapopulation of m ussel beds and hence require special protection. ACKNOWLEDGMENTS This study was financially supported by the W ad denfonds, province of Noord Holland and the Pro gram tow ards a rich W adden Sea. W e th a n k the following organizations for supporting the in te r tidal m ussel bed surveys and providing data: 1. The Schleswig-Holstein Agency for Coastal Defence, National Park an d M arine C onservation—National Park A uthority, 2. N ational Park A dm inistration W adden Sea Lower Saxony an d The Lower Saxon ian Agency for W ater M anagem ent, Coastal Pro tection and N ature Conservation, 3. IMARES. The A lfred-W egner-Institute (Sylt) kindly provided aerial photographs. W e th a n k D ietm ar Kraft for constructing th e vector file w ith the tidal basins and Carola van Zweeden for GIS database m a n agem ent of IMARES data. We are very grateful to Folieert de Jong for stim ulating an international and com parative approach. W e th an k our col leagues Jaap van der Meer, Ulrich Callies, Theunis Piersma, Sofia Saraiva and David Thieltges for dis cussion and Jaap van der M eer, Jan B eukem a and Peter H erm an for providing challenging and v alu able com m ents on th e m anuscript. Finally, we th an k th e anonym ous review er and th e nonanonym ous review er Dr. M artin Thiel for their critical and useful com m ents. OPEN ACCESS This article is distributed u n d er th e term s of the Creative Com m ons A ttribution License w hich perm its an y use, distribution, and reproduction in any m edium , provided th e original author(s) and th e source are credited. REFERENCES A ndresen H, v a n der M eer J. 2010. B row n shrim p (Crangon crangon, L.) functional response to density of different sized juv en ile bivalves Macoma balthica (L.). J Exp M ar Biol Ecol 390:31-8. Asmus H. 1987. Secondary production of a n intertidal m ussel bed com m unity related to its storage a n d tu rn o v er com part m ents. M ar Ecol Prog Ser 39:251-66. Asmus RM, Asinus H. 1991. M ussel beds: lim iting or prom oting p hytop lan k to n ? J Exp M ar Biol Ecol 148:215-32. Bates D, M aechler M, Bolker B. 2012. Ime4: Linear m ixed-effects m odels using S4 classes. http://CRA N .R-project.org/package= lm e4. B ayne BL. 1964. Prim ary a n d secondary settlem ent in Mytilus edulis L. (m ollusca). J A nim al Ecol 33:513. Bayne BL. 1976. M arine m ussels, their ecology a n d physiology. Cambridge: Cambridge U niversity Press. B eukem a JJ. 1991. The abundance of shore crabs Carcinus mae nas (L.) on a tidal flat in th e W adden Sea after cold a n d m ild w inters. J Exp M ar Biol Ecol 153:97-113. B eukem a JJ. 1992. Dynam ics of ju venile shrim p Crangon crangon in a tidal-flat n ursery of the W adden Sea after m ild a n d cold w inters. M ar Ecol Prog Ser 83:157-65. B eukem a JJ, D ekker R. 2005. Decline of recruitm ent success in cockles a n d o th er bivalves in the W adden Sea: possible role of clim ate change, p redation o n postlarvae a n d fisheries. M ar Ecol Prog Ser 287:149-67. B eukem a JJ, Essink K, M ichaelis H, Zwarts L. 1993. Y ear-to-year variability in the biom ass of m acrobenthic anim als on tidal flats of the W adden Sea: h o w predictable is this food source for birds? J Sea Res 31:319-30. B eukem a JJ, D ekker R. Essink K, M ichaelis H. 2001. S ynchro nized reproductive success of th e m ain bivalve species in the W adden Sea: causes a n d consequences. M ar Ecol Prog Ser 211:143-55. B eukem a JJ, D ekker R, Philippart CJM. 2010. Long-term vari ability in bivalve recruitm ent, m ortality, a n d grow th a n d their contribution to fluctuations in food stocks of shellfish-eating birds. M ar Ecol Prog Ser 414:117-30. Bickel PJ, D oksum KA. 1981. A n analysis of transform ations revisited. J A m Stat Assoc 76:296-311. Bivand RS, R undel C. 2011. rgeos: Interface to G eom etry E n gine-O pen Source (GEOS). http://CRA N .R-project.org/ package=rgeos. Bivand R, Pebesm a EJ, Góm ez-Rubio V. 2008. Applied spatial data analysis w ith R. N ew York: Springer. Bjornstad ON, Inis RA, L am bin X. 1999. Spatial population dynam ics: analyzing patterns a n d processes of population synchrony. Trends Ecol Evol 14:427-32. Bolker BM, Brooks ME, Clark CJ, Geange SW, Poulsen JR. Stevens MHH, W hite JSS. 2009. G eneralized linear m ixed m odels: a practical guide for ecology a n d evolution. Trends Ecol Evol 24:127-35. B rinkm an A, Dankers N, V an Stralen M. 2002. A n analysis of m ussel bed habitats in the D utch W adden Sea. H elgoland M ar Res 56:59-75. B u rn h am KP, A nderson DR. 1998. M odel selection a n d in fer ence: a practical inform ation—theoretic approach. New York: Springer. B uschbaum C. D ittm ann S, Hong J-S, H w ang I-S, Strasser M, Thiel M, Valdivia N, Yoon S-P, Reise K. 2009. M ytilid mussels: global habitat engineers in coastal sedim ents. Helgoland M ar Res 63:47-58. B üttger H, Nehls G, W itte S. 2011. High m ortality of Pacific oysters in a cold w inter in the N orth-Frisian W adden Sea. H elgoland M ar Res 65:525-32. Cadée GC, H egem an J. 2002. Phytoplankton in the M arsdiep at th e end of th e 2 0 th century: 30 years m onitoring biomass, prim ary production, a n d Phaeocystis bloom s. J Sea Res 4 8 :9 7 110. Caley M, Carr M, H ixon M, Hughes T, Jones G, M enge B. 1996. R ecruitm ent a n d the local dynam ics of open m arine p o p u la tions. A n n Rev Ecol Syst 27:477-500. Callaway R. 2003. Long-term effects of im itation polychaete tubes on benthic fauna: th ey anchor Mytilus edulis (L.) banks. J Exp M ar Biol Ecol 283:115-32. Spatial Dynamics of Intertidal M ussel Beds 565 Carson HS, López-D uarte PC, R asm ussen L, W ang D, Levin LA. 2010. Reproductive tim ing alters p opulation connectivity in m arine m etapopulations. Curr Biol 20:1926-31. Kraft D, Folm er EO, M eyerdirks J, Stiehl T. 2011. Data inventory of the tidal basins in the trilateral W adden Sea. Leeuw arden: P rogram m a n aar een Rijke W addenzee. de Blok JW , Tan-M aas M. 1977. F unction of byssus threads in young postlarval Mytilus. N ature 267:558. Levin LA. 2006. R ecent progress in u n derstanding larval dis persal: n e w directions a n d digressions. Integr Com p Biol 46:282-97. de Vlas J, B rinkm an AG, B uschbaum C, Dankers N, H erlyn M, K ristensen PS, M illat G, R u th M, S teenbergen J, W eh rm an n A. 2005. Intertidal Blue M ussel Beds. Trilateral M onitoring a n d A ssessm ent Group, C om m on W adden Sea Secretariat, W ilhelm shaven, G erm any. M cG rath D, King PA, Gosling EM. 1988. Evidence for th e direct settlem ent of Mytilus edulis larvae o n adult m ussel beds. M ar Ecol Prog Ser 47:103-6. de Vooys CGN. 1999. N um bers of larvae a n d prim ary p la n ti grades of the m ussel Mytilus edulis in th e w estern D utch W adden Sea. J Sea Res 41:189-201. M cQ uaid C, Phillips T. 2000. Lim ited w ind-driven dispersal of intertidal m ussel larvae: in situ evidence from the plan k to n a n d th e spread of the invasive species Mytilus galloprovincialis in South Africa. M ar Ecol Prog Ser 201:211-20. Dijkem a KS, van T hienen G, v a n Beek JG. 1989. Habitats of the N etherlands, G erm an a n d D anish W adden Sea. Texel: R e search In stitute for N ature M anagem ent, V eth Foundation, Leiden. M etaxas A, Saunders M . 2009. Q uantifying th e 'B io-' com po n en ts in biophysical m odels of larval transport in m arine b enthic invertebrates: advances a n d pitfalls. Biol Bull 216:257-72. D onker JJA, v a n der Vegt M, Hoekstra P. 2012. W ave forcing over an intertidal m ussel bed. J Sea Res 82:54-66. M öller P. 1986. Physical factors a n d biological interactions reg ulating infauna in shallow boreal areas. M ar Ecol Prog Ser 30:33M 7. Fey F, D ankers N, S teenbergen J, G oudsw aard K. 2010. D evel opm ent a n d distribution of the non-indigenous Pacific oyster (Crassostrea gigas) in th e D utch W adden Sea. A quae In t 18:4559. Getis A. 2009. Spatial A utocorrelation. H andbook of applied spatial analysis, softw are tools, m ethods a n d applications. Heidelberg: Springer, p 255-78. Gilg M, Hilbish T. 2003. The geography of m arine larval dis persal: coupling genetics w ith fine-scale physical oceanogra phy. Ecology 84:2989-98. Gosling EG. 1992a. Ecology of bivalves. In: The m ussel Mytilus: ecology, physiology, genetics, a n d culture. D evelopm ents in aqu acu ltu re a n d fisheries science. A m sterdam : Elsevier, pp 44-86. Gosling EG. 1992b. R eproduction, settlem ent a n d recruitm ent. In: The m ussel Mytilus: ecology, physiology, genetics, and culture. D evelopm ents in aqu acu ltu re a n d fisheries science. A m sterdam : Elsevier, pp 131-168. G utiérrez JL, Jones CG, Strayer DL, Iribarne OO. 2003. M ollusks as ecosystem engineers: th e role of shell p roduction in aquatic habitats. Oikos 101:79-90. H am m ond W, Griffiths CL. 2004. Influence of w ave exposure on S outh African m ussel beds a n d th eir associated infaunal com m unities. M ar Biol 144:547-52. H erlyn M, M illat G, Petersen B. 2008. D ocum entation of sites of intertidal blue m ussel (Mytilus edulis L.) beds of th e low er Saxonian W adden Sea, so u th e rn N orth Sea (as of 2003) and th e role of th eir structure for spatfall settlem ent. Helgoland M ar Res 62:177-88. H onkoop PJ, v a n der M eer J. 1998. Experim entally induced effects of w ater tem p eratu re a n d im m ersion tim e on re p ro ductive o u tp u t of bivalves in th e W adden Sea. J Exp M ar Biol Ecol 220:227-46. K am erm ans P. 1994. Sim ilarity in food source a n d tim ing of feeding in deposit- a n d suspension-feeding bivalves. M ar Ecol Prog Ser 104:63. Keitt TH, B ivand R, Pebesm a E, R ow lingson B (2011) rgdal: bindings for the geospatial data abstraction library, h ttp :// CRAN.R-project.org/package=rgdal. Knights AM, Crowe TP, B urnell G. 2006. M echanism s of larval transport: vertical distribution of bivalve larvae varies w ith tidal conditions. M ar Ecol Prog Ser 326:167-74. M urdoch DJ, C how ED. 1996. A graphical display of large cor relation m atrices. A m Stat 50:178-80. M urdoch D, C how ED. 2012. ellipse: functions for draw ing ellipses a n d ellipse-like confidence regions. http://CRAN.Rproject.org/package=ellipse. Nehls G, Thiel M. 1993. Large-scale distribution patterns of the m ussel Mytilus edulis in the W adden Sea of Schleswig-Holstein: do storm s structure th e ecosystem ? J Sea Res 31:181-7. Nehls G, Hertzler I, Scheiffarth G. 1997. Stable m ussel Mytilus edulis beds in th e W adden Sea: th ey 're ju st for the birds. H elgoländer M eeresu n tersu ch u n g en 51:361-72. Ord JK, Getis A. 1995. Local spatial autocorrelation statistics: distributional issues a n d a n application. Geogr A nal 2 7:286306. Pebesm a EJ, B ivand R. 2005. Classes a n d m ethods for spatial data in R. R News 5:9-13. Philippart CJM, Epping HG. 2010. The W adden Sea: a coastal ecosystem u n d e r continuous change. In: K ennish MJ, Paerl HW, Eds. Coastal lagoons: critical habitats of e nvironm ental change. M arine science book series. Boca Raton: Taylor & Francis. Philippart CJM, v a n A ken HM, B eukem a JJ, Bos OG, Cadée GC, Dekker R. 2003. C lim ate-related changes in recruitm ent of the bivalve Macoma balthica. Lim nol Oceanogr 48:2171-85. Philippart CJM, A m aral A, A sm us R, v a n Bleijswijk J, B rem ner J, Buchholz F, C abanellas-Reboredo M, Catarino D, Cattrijsse A, Charles F, C om tet T, C unha A, D eudero S, D uchêne J-C, Fraschetti S, Gentil F, G ittenberger A, Guizien K, Gonçalves JM, G uarnieri G, Hendriks I, Hussel B, Vieira RP, R eijnen BT, Sam paio I, Serrao E, Pinto IS, Thiebaut E, Viard F, Z uur AF. 2012. Spatial synchronies in the seasonal occurrence of larvae of oysters (Crassostrea gigas) a n d m ussels (Mytilus edulis/gallo provincialis) in E uropean coastal w aters. E stuar Coast Shelf Sei 108:52-63. Pulfrich A. 1996. A ttachm ent a n d settlem ent of post-larval mussels (Mytilus edulis L.) in th e Schleswig-Holstein W adden Sea. J Sea Res 36:239-50. R anta E, K aitaia V, L undberg P. 2006. Ecology of populations. Cambridge: Cambridge U niversity Press. R D evelopm ent Core Team. 2009. R: A Language a n d E n v iro n m en t for Statistical C om puting. 566 E. O. Folm er an d others Reise K. 2005. Coast of change: h abitat loss a n d transform ations in the W adden Sea. Helgoland M ar Res 59:9-21. Reise K, Baptist MJ, Burbridge P, D ankers N, Fischer L, Flem m ing B, Oost AP. 2010. The W adden Sea 2010: a universally outstanding tidal w etland. W ilhelm shaven: CWSS. R idderinkhof H, Z im m erm an JTF, Philippart MB. 1990. Tidal exchange betw een the N orth Sea a n d D utch W adden Sea and m ixing tim e scales of th e tidal basins. J Sea Res 25:331-50. R oyam a T. 1992. A nalytical population dynam ics. London: C hapm an & Haii. Staneva J, Stanev BV, Wolff JO, B adew ien TH, R euter R, Flem m ing B, B artholom , Bolding K. 2009. Hydrodynam ics a n d sedim ent dynam ics in the G erm an Bight. A focus on observations a n d num erical m odelling in th e Bast Frisian W adden Sea. C ont Shelf Res 29:302-19. Strasser M, G ünther C-P. 2001. Larval supply of p redator and prey: tem poral m ism atch b etw ee n crabs a n d bivalves after a severe w inter in th e W adden Sea. J Sea Res 46:57-67. Strasser M, R einw ald T, Reise K. 2001. Differential effects of the severe w in ter of 1995/1996 on th e intertidal bivalves Mytilus edulis. Cerastoderma edule a n d Mya arenaria in the N orthern W adden Sea. Helgoland M ar Res 55:190-7. Strasser M, Dekker R, Essink K, G ü n th er CP, Jaklin S, Kröncke I, M adsen PB, M ichaelis H, Vedel G. 2003. How predictable is h igh bivalve recruitm ent in the W adden Sea after a severe w inter? J Sea Res 49:47-57. T horson G. 1950. Reproductive a n d larval ecology of m arine b otto m invertebrates. Biol Rev 25:1M 5. TMAG— Trilateral M onitoring a n d A ssessm ent Group. 1997. TMAP m anual. The Trilateral M onitoring a n d Assessm ent Program (TMAP). C om m on W adden Sea Secretariat, W il helm shaven, Germ any. Troost K. 2010. Causes a n d effects of a highly successful m arine invasion: case-study of the introduced Pacific oyster Crassostrea gigas in continental NW E uropean estuaries. J Sea Res 64:14565. T urchin P. 2003. Com plex population dynam ics: a theoretical/ em pirical synthesis. Princeton, NJ: Princeton University Press. van A ken HM. 2008. Variability of th e w ater tem perature in the w estern W adden Sea on tidal to centennial tim e scales. J Sea Res 60:227-34. van B eusekom JEE, B uschbaum C, Reise K. 2012. W adden Sea tidal basins a n d the m ediating role of th e N orth Sea in eco logical processes: scaling up of m anagem ent? O cean Coastal M anage 68:69-78. van de Koppel J, Rietkerk M, D ankers N, H erm an PM. 2005. S cale-dependent feedback a n d regular spatial patterns in young m ussel beds. A m Nat 165:66-77. van de Koppei J, B oum a TJ, H erm an PMJ. 2012. The influence of local- a n d landscape-scale processes on spatial self-organi zation in estuarine ecosystem s. J Exp Biol 215:962-7. van der M eer J, B eukem a JJ, D ekker R. 2001. Long-term vari ability in secondary production of a n intertidal bivalve p o p u latio n is prim arily a m atter of recruitm ent variability. J A nim al Ecol 70:159-69. van der V eer HW, Feller RJ, W eber A, W itte J. 1998. Im portance of p redation by crustaceans u p o n bivalve spat in the intertidal zone of th e D utch W adden Sea as revealed by im m unological assays of gut contents. J Exp M ar Biol Ecol 231:139-57. van der Zee B, van der Heide T, D onadi S, Eklöf J, Eriksson B, Olff H, van der V eer H, Piersm a T. 2012. Spatially extended h abitat m odification by intertidal reef-building bivalves has im plications for consum er-resource interactions. Ecosystems 15:664-73. van R aaphorst W, de Jonge VN. 2004. R econstruction of the total N a n d P inputs from the Ijsselm eer into th e w estern W adden Sea b e tw ee n 1935-1998. J Sea Res 51:109-31. w a K angeri AK, Jan sen JM, B arkm an BR, D onker JJA, Joppe DJ, D ankers NMJA. 2013. P erturbation induced changes in substrate use by th e blue m ussel. Mytilus edulis, in sedim entary systems. J Sea Res W atson JR. M itarai S, Siegel DA, Caselle JE, Dong C, M cW il liams JC. 2010. Realized a n d poten tial larval connectivity in th e S o u th ern California Bight. M ar Ecol Progr Ser 401:31-48. W eh rm an n A, H erlyn M, B ungenstock F, H ertw eck G, M illat G. 2000. The distribution gap is closed—first record of n aturally settled Pacific oysters (Crassostrea gigas) in the Bast Frisian W adden Sea, N orth Sea. Senckenbergiana M aritim a 30:15360. W hite C, Selkoe KA, W atson J, Siegel DA, Zacherl DC, T oonen RJ. 2010. O cean currents help explain population genetic structure. Proc R Soc B 277:1685-94. W ickham H. 2009. ggplot2: elegant graphics for data analysis. N ew York: Springer, h ttp ://had.co.nz/ggplot2/book. W iddow s J, Brinsley M. 2002. Im pact of biotic a n d abiotic p ro cesses on sedim ent dynam ics a n d the consequences to the structure a n d functioning of the intertidal zone. J Sea Res 48:143-56. Young B, Bigg G, G rant A. 1996. A statistical study of e n v iro n m en tal influences on bivalve recruitm ent in the W ash, E ng land. M ar Ecol Prog Ser 143:121-9. Young BF, Bigg GR, G rant A, W alker P, B row n J. 1998. A m odelling study of en v ironm ental influences on bivalve se t tlem en t in The W ash, E ngland. M ar Ecol Prog Ser 172:197214.