Ecology, 95(6), 2014, pp. 1451-1457 © 2014 by the Ecological Society o f Am erica E x p erim en tin g w ith eco sy stem in te ra c tio n n e tw o rk s in search o f th re sh o ld p o ten tials in re a l-w o rld m a rin e ecosystem s S im o n F . T h r u s h , 1’2’9 J u d i E. H e w i t t , 1 S a m a n t h a P a r k e s , 1 A n d r e w M . L o h r e r , 1 C o n r a d P i l d i t c h , 3 S a r a h A . W o o d in ,4 D a v id S. W e t h e y ,4 M a r i a c h i a r a C h i a n t o r e ,5 V a l e n t i n a A s n a g h i,5 S ilv ia D e J u a n ,1 C a s p e r K r a a n , 1 I v a n R o d i l , 1 C a n d i d a S a v a g e , 6’8 a n d C a r l V a n C o l e n 7 Abstract. T hresholds p ro fo u n d ly affect o u r u n d erstan d in g a n d m an ag em en t o f ecosystem dynam ics, b u t we have yet to develop practical techniques to assess th e risk th a t th resholds will be crossed. C om bining ecological know ledge o f critical system interdependencies w ith a largescale experim ent, we tested fo r b reak s in the ecosystem in teractio n netw o rk to identify th resh o ld p o ten tial in real-w orld ecosystem dynam ics. O u r experim ent w ith the bivalves M acom ona liliana a n d A ustrovenus stutchburyi o n m arin e sandflats in N ew Z ealan d d em o n strated th a t reductions in incident sunlight ch anged the in teractio n n etw o rk betw een sedim ent biogeochem ical fluxes, p roductivity, a n d m acro fau n a. By d em o n stratin g loss of positive feedbacks a n d changes in the arch itectu re o f the n etw ork, we provide m echanistic evidence th a t stressors lead to b reak p o in ts in dynam ics, w hich th eo ry predicts predispose a system to a critical tran sitio n . Key words: Austrovenus stutchburyi; bifurcations; bivalves; break points; ecosystem dynamics; M acom ona liliana; M anukau Harbour, New Zealand; marine sandflats; thresholds; turbidity. I n t r o d u c t io n M an y changes in ecological system s are a b ru p t, pro fo u n d ly lim iting o u r u n d erstan d in g o f ecosystem dynam ics, an d o u r ability to predict changes a n d cope w ith the consequences (C arp en ter et al. 2001, Scheffer et al. 2001). C oncern a b o u t crossing regional- a n d globalscale thresholds underscores the need to b etter define the risk o f tip p in g an ecosystem in to ecologically a n d socioeconom ically u n fav o ra b le states (B ro o k et al. 2013). T heories o f how critical tra n sitio n s lead to th resh o ld change in ecological systems have progressed rapidly, b u t em pirical su p p o rt is p atch y a n d often focused o n specific system s, restricting generality (Kéfi et al. 2013). A n um ber o f techniques have been developed to identify the statistical pro p erties o f tim e series or spatial p attern s th a t precede th resholds (D ak o s et al. 2012, Scheffer et al. 2012). These techniques represent a valid w ay o f ran k in g different systems in term s o f their M anuscript received 9 October 2013; revised 6 January 2014; accepted 27 February 2014. Corresponding Editor: P. T. Raimondi. 9 Present address: Institute o f M arine Science. U niversity o f A uckland. P rivate Bag 92019. A u ck lan d 1142 New Zealand. E-mail: sim on.thrush(® auckland.ac,nz p red isp o sitio n to ap p ro ach in g a n d crossing thresholds, b u t fo r real-w orld ecosystem s, this requires significant investm ent in high-frequency m o n ito rin g th a t is gener­ ally u n com m o n (H ew itt a n d T h ru sh 2010, L indegren et al. 2012). C onsequently, alternative a n d com plem entary a p p ro ach es are needed to in form decisions a b o u t the risk o f upcom ing tran sitio n s in ecosystem s a n d change in ecosystem fun ctio n ality an d value. T he only w ay to conclusively detect a th resh o ld is to cross it (C arp en ter et al. 2001). Such a shift m ay result fro m excessive en v ironm ental forcing, the fu nctional extinction or invasion o f species, b u t m ay also occur well below these limits due to changes in the arch itectu re o f in teractio n netw orks th a t connect the biological, chem ­ ical, a n d physical com ponents in the ecosystem (Scheffer et al. 2001). The loss o f positive feedbacks is p articularly im p o rta n t in shifting the system into a fu ndam entally different state (Scheffer et al. 2012). Such shifts in the arch itectu re o f ecosystem in teractio n netw orks have b een detected along en v ironm ental gradients, suggesting the p o ten tial fo r critical tran sitio n s (T h ru sh et al. 2012). W hole-lake experim ental m an ip u latio n s have provided em p irical evidence o f th e im p o rta n c e o f positive feedbacks a n d regim e shifts (C arp en ter 2003). H ow ever, questions rem ain as to w hether we can experim entally m an ip u late a n open an d dynam ic m arin e system to 1451 SlAOcCdiL 1National Institute o f Water and Atmospheric Research, P.O. B ox 11-115, Hillcrest, Hamilton 3251 New Zealand ~Institute o f Marine Science, University o f Auckland, Private Bag 92019, Auckland 1142 New Zealand 3Department o f Biological Sciences, University o f Waikato, Hamilton 3216 New Zealand 4Department o f Biological Sciences, University o f South Carolina, 715 Sumter Street, Columbia, South Carolina 29208 USA 3Universitd degii Studi di Genova (D iS T A V ), Corso Europa 26, Genova 16132 Italy 6Department o f Marine Science, University o f Otago , P.O. B ox 56, Dunedin 9054 New Zealand 1Marine Biology Section, Biology Department, Ghent University, Krijgslaan 281-S8, Ghent 9000 Belgium 8Marine Research Institute, University o f Cape Town, Private Bag, Rondebosch, South Africa 1452 SIM ON F. T H R U SH ET AL. Nutrients Organic content Chlorophyll a Macomona Mud ' content Austrovenus, 13 10 \ Coarse sediment F i g . 1. Conceptual diagram of the potential interaction netw ork linking infaunal bivalves (here m arine species M aco­ mona liliana and Austrovenus stutchburyi in New Zealand) to benthic chlorophyll a and sediment characteristics. (1) Grazing by deposit-feeding M acomona affects m icrophytobenthos (MPB, represented as chlorophyll a), and MPB has a positive relationship with Macomona density. (2) Sediment organic content increases food resources for Macomona. (3) MPB are highly productive and exploit nutrients rem ineralized in sediments. (4) Macomona affect pore water nutrient concentra­ tions directly and (2, 5) indirectly. (6) Austrovenus has a positive effect on MPB. (7) Large Macomona and small Austrovenus interact negatively. (8, 9) Sediment m ud content negatively affects density of both bivalve species; high densities of both species can increase fine-sediment erosion. (10) In coarse sediments, Austrovenus density increases. (11) MPB and m ud content have a positive relationship, (8, 9) m ediated by resident m acrofauna. (12) M ud and sediment organic content are tightly correlated and (13) coarse sediments influence sediment organic content by increasing permeability and decay. (14) Degrading MPB also contributes to sediment organic m atter, which (5) is positively correlated with nutrient concentration in pore water through microbial degradation. See M ethods for further details and references. dem onstrate changes in in teractio n netw orks a n d the loss o f positive feedbacks. W e provide a p a th to address critical tra n sitio n s th a t facilitate a fo rm a l linkage betw een em pirical a n d th eoretical study o f ecosystem dynam ics. A s pred icted by th eo retical studies, o u r experim ent dem on strates th a t stress-induced changes in ecosystem interactio n netw orks can occur in real-w orld ecosystems, and this allows us to determ ine w hether or n o t a specific ecosystem is predisposed to u ndergo tran sitio n s. Such em pirical d e m o n stra tio n helps to g ro u n d theory and provide real-w orld advice to su p p o rt m anagers, policy m akers, an d society a b o u t the limits to resilience o f ecosystem s u n d er change. There is a long h istory o f ecological ex p erim entation in m arine systems, w ith early recognition o f the role o f interactions am ong species a n d distu rb an ce influencing com m unity dynam ics (D ay to n 1971, W o o d in 1978). N ew analytical techniques allow us to assess the strength o f b o th direct and indirect effects, define the architecture o f interaction netw orks, an d assess its validity relative to em pirical d a ta (G race et al. 2012). S tru ctu red eq u atio n Ecology, Vol. 95, No. 6 m odels (SE M s) allow us to in te g rate cau sal an d ex p lo rato ry techniques, beginning w ith the developm ent o f a conceptual m odel su p p o rted by em pirical evidence. O ur conceptual m odel (Fig. 1) is focused on interactions am ong prim ary producers, tw o sediment-dwelling bivalves (the deposit feeder M acom ona liliana an d the suspension feeder Austrovenus stutchburyi), sedim ent type, an d pore w ater nutrien t an d organic m atter pools, justified by empirical research an d know ledge o f ecosystem function­ ing. The p otential for interaction betw een these variables an d analogous species is typical o f shallow, non-eutrophic, coastal an d estuarine sandflat systems. This interaction netw ork has some o f the same elements as those previously used to look for thresholds across environm ental gradients (T hrush et al. 2012). The conceptual m odel represents an ecosystem in terac tio n n etw o rk th a t su p p o rts coastal ecosystem services associated w ith h a b ita t stability, nu trien t recycling, an d productivity (B arbier et al. 2011). In this paper we test for changes in the architecture o f the ecosystem interaction netw orks (Fig. 1) associated w ith decreased light levels, an im p o rta n t consequence o f increased turbidity in coastal ecosystems. W e developed SEM s based on experim ental d ata derived from either shaded or unshaded experim ental plots. Changes in land use often elevate terrestrial sedim ent an d n u trien t loading in m an y estuarine a n d coastal ecosystem s. E levated turbidity is generally recognized because o f its negative effects on seagrass (D u arte 2002), w hereas effects on m icrophytobenthos are often overlooked despite their central role in p rim ary productivity, nutrien t interception, an d basal food resource provisioning for m any benthic invertebrates. M eth o d s SE M starts w ith th e d efinition o f th e p o te n tia l in teractio n n etw o rk (num bered effects discussed here refer to num bers in the Fig. 1 netw ork). Specifically, (1) there is a direct negative effect o f the large depositfeeding M acom ona on m icro p h y to b en th o s (M PB; re p ­ resented as chlorophyll a) due to grazing (Lelieveld et al. 2003). M PB are an im p o rta n t fo o d source fo r M a co ­ m ona, b u t there is also a positive relationship betw een chlorophyll a biom ass a n d M acom ona density (T h ru sh et al. 2006). Sim ilarly, (2) sedim ent organic co n ten t increases the fo o d resources for M acom ona (by analogy to the functionally sim ilar species M acom a balthica [Peterson a n d Skilleter 1994]). (3) M PB are highly p ro d u c tiv e (C a h o o n 1999) a n d rely o n n u trie n ts rem ineralizing in sedim ents to sustain th a t p ro d u ctio n (L o h rer et al. 2004). M acom ona affects p o re w ater n u trie n t co n cen tratio n s directly (4) an d indirectly (2 an d 5) th ro u g h influencing organic m a tte r pools via grazing an d behaviorally induced pressure gradients in p erm e­ able sedim ents (V o lkenborn et al. 2012, W oo din et al. 2012). (6) Austrovenus has a positive effect on ch lo ro ­ phyll a (T h ru sh et al. 2006), alth o u g h this effect is density depen d en t (Sandw ell et al. 2009, L o h rer et al. 2011). (7) N egative in teractio n s betw een large M acom o- T H R E SH O LD PO TEN TIA L IN M A R IN E ECOSYSTEM S lia an d sm all Austrovenus have specifically identified (T h ru sh et al. 1996). (8 an d 9) D ensities o f b o th bivalve species are negatively im pacted by m u d (T h ru sh et al. 2003, 2004), b u t high densities o f b o th species can increase the erosion o f fine sedim ents by b io tu rb a tio n (T h ru sh et al. 1996, Jones et al. 2011). (10) In coarse sedim ents, A ustrovenus density increases (T h ru sh et al. 2008). (11) A positive relationship betw een benthic chlorophyll a a n d sedim ent m uddiness exists (Y an de K oppei et al. 2001), due to fine-sedim ent binding by M B P, a relationship (8 a n d 9) m ed iated by resident m acro fau n a (Y an C olen et al. 2013). (12) M u d an d sedim ent organic co n ten t are tightly co rrelated (Sloth et al. 1995) a n d (13) coarse sedim ents influence sedim ent organic content by increasing perm eability a n d acceler­ ating decay rates (H uettel a n d R usch 2000). D egrading chlorophyll a (14) also contrib u tes to the p o o l o f organic m a tte r in the sedim ent (E hrenhauss et al. 2004), a n d (5) the concen tratio n o f pore w ater n u trien ts is positively correlated w ith th e sedim ent organic co n ten t th ro u g h m icrobial d eg ra d a tio n processes (S loth et al. 1995, G rabow ski et al. 2011). T he experim ent w as co n d u cte d at W iro a Island, M an u k au H arb o u r, N ew Z ealand (37°01.3' S, 174°49.2' E; see Plate 1). The intertidal sandflat is ~ 1 .8 km wide an d has a shallow gradient o f 0.097° (T hrush et al. 1997). T he m acro fau n a l com m unity o f the W iro a Islan d sandflats is dom in ated b o th num erically a n d in term s of biom ass by the tellinid M acom ona liliana ( ~ 100/m2 for individuals w ith shell length > 2 0 m m ), a n d the venerid Austrovenus stutchburyi ( ~ 3 2 /n r for individuals w ith shell length > 10 m m ). T o investigate differences in the architecture o f the ecosystem interaction netw ork, we experim entally m an ip u lated nutrien t concentrations, M a ­ comona densities, a n d light. The experim ent was estab­ lished on 25-29 O ctober 2011 in seven blocks th at encom passed ab o u t 800 X 350 m o f sandflat. E xperim en­ tal plots ( 2 X 2 m ) were situated a b o u t 3 m ap a rt in three row s to encom pass an area o f a b o u t 56 m alongshore an d 14 m dow nshore. The experim ent consisted o f three density treatm en ts o f M acom ona (200, 50, an d 0 individuals/m 2 for individuals w ith shell length > 20 m m ) crossed w ith three light treatm ents (shade, shade control, a n d open) crossed w ith three nutrien t addition treatm en ts (high, m edium , none) a n d one am bient sedim ent treatm ent. This resulted in 28 experim ental treatm ents, replicated once per block. T he central 1 in 2 in each 4-m 2 plo t, excluding one treatm en t o f am bient sedim ent, w as excavated to 18 cm d ep th a n d the sedim ent sieved on 10-m m m esh to extract shell h ash an d large m acro fau n a. Sedim ents were then retu rn ed to the excavated plots an d M acom ona (> 2 0 m m shell length) were spread evenly across the 1 in 2 achieve treatm en t densities. Shade cloth designed to cut ~ 7 0 -8 0 % o f the incident light w as su p p o rted o n a 4 - n r m etal grid (m esh size 150 X 150 m m ) suspended 15-20 cm above th e sedim ent surface. Shade c o n tro ls w ere th e m etal g rid only 1453 covering the plots. Shading o f plots w as used to m im ic changes in light levels due to elevated susp en d ed sedim ent co n cen tratio n s, SSC (Billerbeck et al. 2007). D irect field-based m an ip u latio n o f SSC is very difficult in the m edium to long term . A lth o u g h SSC has m ultiple effects o n the ben th o s (T h ru sh et al. 2004), we focused o n effects m ed iated th ro u g h the M PB . Benthic ch lo ro ­ phyll a is a m easure o f M PB stan d in g stock, n o t productivity. T his is im p o rta n t in the in te rp re ta tio n o f o u r ex p erim en t due to th e p o te n tia l fo r b e d lo a d tra n s p o rt o f M P B u n d er sh ad ed p lo ts, p o ten tially w eakening treatm en t effects. N u trien t ad d itio n consisted o f a d d in g fo u r (high), one (m edium ), o r zero 10 cm long sections o f dialysis tubing filled w ith 15 g (± 0 .1 g) o f a g ran u lar slow-release fertilizer (nu trien t ratios: 18.9 N , 2.1 P, 0.0 K , 4.1 S, 3.0 M g, 7.2 Ca). O u r n u trien t a d d itio n treatm en ts were expected to have a direct stim u lato ry effect o n M PB , because N ew Z ealan d sandflat studies show th a t M PB resp o n d to n u trien t released from the sedim ent as a result o f b io tu rb a tio n (R odil et al. 2011). B efore d estru ctiv e sam p lin g a t th e en d o f the experim ent (ab o u t +100 days, 7-10 F eb ru ary 2012), p o re w ater n u trien ts sam ples were collected from each p lo t using reservoirs sam pling the perm eable sedim ents 5 cm below the sed im en t-w a ter interface. P ore w ater w as pressure-filtered th ro u g h a 2.5-cm W h a tm a n G F /C glass fiber filter in a Swinnex filter h o ld er (E D M illipore, Billerica, M assachusetts, U SA ) a n d stored o n crushed ice in the d a rk fo r tra n sp o rt to the lab o rato ry , w here they were frozen to —20°C pending analysis. A nalysis for dissolved inorganic n u trien ts follow ed sta n d a rd m eth o d s fo r seaw ater using a n A storia-Pacific 300 series segm ented-flow au to an a ly zer (A storia-Pacific, C lackam as, O regon, U SA ) w ith d etection lim its o f <0.1 pm ol/L for N a n d P. T hree cores (2.3 cm diam eter, 2 cm deep) were collected fro m each plo t a n d w ere p o o led fo r chlorophyll a , sedim ent particle size, a n d organic co n ten t analyses. T he rem aining sedim ent w ithin the central 1 m 2 w as excavated to a d ep th o f a b o u t 15 cm a n d sieved on 10m m m esh to extract large m acro fau n a, w hich were then identified a n d sized. Sedim ent fo r chlorophyll analysis w as freeze-dried a n d hom ogenized. A sm all q u a n tity o f sedim ent (—0.1 g) w as soaked in 10 m L o f M g C 0 3-buffered acetone (90%) fo r ~ 2 0 h. Sam ples w ere shaken after 12 h an d again p rio r to centrifugation. Sam ples w ere centrifuged fo r 10 m in a t 3300 rp m before fluorom etric analysis (T u rn er 10-A U , T u rn e r D esigns, Sunnyvale, C alifornia, U SA ). A n acidification step (0.15 m L 0.1N H C L ) w as used to sep arate p h aeo p h y tin from p h o to sy n th etic pigm ents (A ra r a n d C ollins 1997). T o tal organic m a tte r con ten t (as a percentage) w as d eterm ined th ro u g h loss on ignition from hom ogenized dried sedim ents (105°C for 24 h o r un til stable m ass), after com b u stio n fo r 5.5 h at 550°C (D ean 1974). Sedim ents fo r particle size analysis w ere digested in 10%) h y drogen peroxide to rem ove organic m atter, until b ubbling ceased before using the SlAOcCdiL June 2014 1454 SIM ON F. T H R U SH ET AL. Ecology. Vol. 95. No. 6 P l a t e 1. Outplanting Macomona in one of the experimental plots on the sandflat (W iroa Island. New Zealand). Photo credit: Rachel J. Harris. sta n d ard operating p rocedure fo r m arine sedim ents o n a M alvern m astersizer-S (300 F R lens; M alv ern In s tru ­ m ents, M alvern, U K ) to determ ine grain size fractions in the range o f 0.05-2000 pm. SEM s were developed using M -P lus softw are (M u th én an d M u th én 2007), based o n Fig. 1, to co n tra st the netw orks a p p aren t in the shaded an d u n sh ad ed exper­ im ental treatm ents. W e developed best-fit m odels for shaded a n d unsh ad ed treatm en ts a n d co m p ared these w ith the results from applying the best shaded n etw ork to the unsh ad ed d a ta a n d vice versa, in o rd er to ensure th a t there were real differences in the netw orks. D a ta tra n sfo rm a tio n s (loge) to im p ro v e th e lin earity o f responses, or n orm ality o f errors, w ere utilized for bivalve abundances, coarse sand, a n d m ud. The initial SE M (Fig. 1) w as exam ined fo r respecification by rem oval o f direct p athw ays based o n p a ram eter tests o f significance, while m ain tain in g nonsignificance tests fo r goodness o f fit (x2) a n d the ro o t m ean square erro r o f ap p ro x im atio n , increasing values o f the co m parative fit index (above 0.95) (Vile et al. 2006) a n d decreasing A k aik e’s in fo rm atio n criterio n (see A ppendix). The sensitivity o f the m odel to new direct p athw ays w as also investigated, w ith the decision on w hether o r n o t to include new links being based o n p ara m e te r significance tests an d n o decreases in goodness o f fit (G race et al. 2 012). R esu lts C onsistent w ith using SE M as a to o l to integrate causal a n d ex p lo rato ry techniques (G race et al. 2012), we have n o t focused on the m o re trad itio n al analyses o f the experim ental d ata. Prelim inary analysis d id n o t ind icate any experim ental a rtefacts asso ciated w ith sh ad in g (see A p p en d ix ). T re a tm e n t differences in M acom ona density were m ain tain ed in the experim ent, a lth o u g h slightly higher densities in shaded plots were a p p a re n t, while Austrovenus decreased. L ow er pore w ater N O x co n cen tratio n s w ere fo u n d in higher M a co ­ mona density a n d zero n u trien t a d d itio n treatm ents. P o re w ater dissolved reactive p h o sp h o ru s co n cen tra­ tio n s w ere significantly a n d n egatively re la te d to M acom ona density, b u t n o t n u trie n t a d d itio n . N o significant differences in benthic chlorophyll a concen­ tra tio n o r p o re w ater am m onia co n cen tratio n s were detected. A lth o u g h m ean benthic chlorophyll a concen­ tra tio n s in the shaded plots were low, high variability lim ited the d etec tio n o f significant differences (see A ppendix). S h ad in g d rastica lly ch an g ed th e n a tu re o f the in teractio n netw o rk (Fig. 2). In the u n sh ad ed tre a t­ m ents, there w ere a to ta l o f 13 interactions; alth o u g h three w ere w eak (P > 0.15 fo r estim ates), they were req u ired fo r m odel stability. A positive feedback loop w as identified betw een Austrovenus, benthic chlorophyll June 2014 T H R E SH O LD PO TEN TIA L IN M A R IN E ECOSYSTEM S Unshaded Shaded Nutrients 0.02 H Nutrients Chlorophyll a 0.24 Chlorophyll a 0.24 ., Organic ^ c o n te n t 0.19 Organic content 0.07 Macomona 0.07 , Macomona Austrovenus 0.30 ▼ 1455 Mud 0.30 Coarse sediment Austrovenus 0.18 ' A Mud 0.10 Coarse ■' sediment a, a n d m ud. A positive coarse sed im en t-o rg an ic m a tte r link fed into this positive feedback loop. A n o th e r w eak positive feedback loop involving M acom ona releasing n u trien ts to su p p o rt chlorophyll standing stock fed into the stronger feedback loop. In co n trast, in the shaded plots, only seven connec­ tions w ere m ade a n d there w ere n o positive feedbacks. T he positive feedback betw een Austrovenus, benthic chlorophyll a, an d m u d w as b ro k en a t the link betw een benthic chlorophyll a a n d m u d , w hich w as now negative, an d the positive effect o f M acom ona o n A ustrovenus was also lost. A dditionally, the negative effect o f M acom ona o n benthic chlorophyll a w as now significant a n d the positive effect o f chlorophyll a o n M acom ona h a d been lost. A lth o u g h b o th bivalves play a central role in the in teractio n n etw o rk u n d er am b ien t light, w hen the sedim ent is shaded, th eir role, a n d in p artic u la r the positive feedbacks, are n o t ap p aren t. U n d e r shaded conditions, we see m ostly simple u n idirectional relatio n ­ ships betw een biological a n d physical variables, w ith no role fo r n u trien t additions. D is c u s s io n O u r em pirical evidence o f a b reak p o in t is consistent w ith theoretical predictions o f a system predisposed to a th resh o ld shift, w hich w o u ld occur in response to changes in sedim ent loading a n d eu tro p h icatio n . These stresso r-in d u ced changes im ply im p o rta n t shifts in ecosystem interactions th a t determ ine organic m a tte r deg rad atio n , n u trien t recycling, an d pro d u ctiv ity th a t drive the functional capacity o f co astal soft sediments. T he reconfiguration o f the in teractio n n etw ork, as a result o f shading, occurred in a b o u t 100 days an d resulted in the loss o f positive feedbacks b o th directly by linking large M acom ona w ith M P B a n d indirectly m ed iated th ro u g h large Austrovenus (Fig. 2). In a d d i­ tio n , the n u trien t ad d itio n s m ed iated the in teractio n betw een M acom ona a n d M P B only in th e u n sh ad ed sedim ents. A lth o u g h m ean M PB w as low er in shaded tre a tm e n ts, h ig h v ariab ility p reclu d ed d etec tin g a significant difference. A n im p o rta n t fa c to r on this sandflat is the b ed lo ad tra n sp o rt o f surficial sedim ents a n d associated M P B (T u rn e r et al. 1997). T his process p rovides a m echanism to resupply M PB to shaded sedim ents a n d a d d variability to o u r d ata . N evertheless, o u r experim ent w as able to detect significant changes in the ecosystem in teractio n netw ork. A t the scale o f o u r experim ent, the m echanism s o f change induced by large M acom ona are behaviors (b io tu rb atio n , feeding, gili cleaning, gut voiding) th a t affect pore w ater oxygen co n cen tratio n s a n d n u trien t cycling (Y o lk en b o rn et al. 2012). M ed iatio n o f the effects o f n u trien t a d d itio n s by large bivalves only in u n sh ad ed plots highlights the p o ten tial fo r cum ulative effects associated w ith in ­ creased tu rb id ity an d eutrop h icatio n . SEM s revealed th a t shading the sedim ent decoupled the ecosystem role o f M acom ona in sedim ent b iogeo­ chem istry. The range o f d a ta encom passed in m odel b uilding is im p o rta n t in generalizing the SEM . A lthough th e differences in en v iro n m en tal cond itio n s en co m ­ p assed in this experim ent w ere subtle, the in teractio n n etw o rk m odels developed fro m this experim ent were sim ilar to those previously developed from broadscale surveys o f large b io tu rb a tin g fau n a, sedim ent grain size, o rg an ic ch ara c teristics, a n d b en th ic c h lo ro p h y ll a (T h ru sh et al. 2012). T he previous analysis revealed shifts in n etw o rk arch itectu re across a benthic ch lo ro ­ phyll a th resh o ld o f 11.6 pg/g chlorophyll a. D espite the sm aller range o f d ata in o u r m o re recent experim ent, a sim ilar th re sh o ld w as d etec ted, im plying th a t the SlAOcCdiL F ig . 2. Changes in ecosystem interaction networks associated with experimental shading, showing R 2 values for each factor. Red arrow s indicate positive effects, black lines indicate negative. Solid lines denote param eter estimates with P < 0.05; dotted lines with P values 0 .0 5 -0 .1 5 ; dashed lines with P > 0 .1 5 . but the relationship needs to be retained in the model for stability. Line thickness increases with the size of the param eter estimate, standardized by the relevant range standardization procedure (Grace et al. 2012); actual values are presented in the Appendix. 1456 SIM ON F. T H R U SH ET AL. netw orks are reasonably stable a n d consistent. F u n d a ­ m ental shifts associated w ith a loss o f positive feedbacks an d restrictions in the im p o rtan ce o f indirect effects were ap p a re n t in b o th this ex p erim en tal stu d y a n d the previous broadscale analysis. P roviding em pirical in fo r­ m atio n to define how a set o f ecosystem com ponents are connected in an in teractio n n etw o rk helps to elucidate ecosystem fu nctio n . M o reo v er, if specific types o f co n n ectio n s a re p resen t, b u t can be c h an g ed by particu lar stressors, this provides a useful m eans for predicting the risk o f critical tra n sitio n related to a specific type an d m agnitudes o f stress. It also allows insight into ecosystem fu n ctio n after the tran sitio n . A lth o u g h there is often a gap betw een th eo ry an d practice, we have been able to experim entally shift the architecture o f an in teractio n netw ork, p roviding m ore im p etu s fo r scientists a n d m an a g e rs to use these indicators o f critical tran sitio n s. O u r results show th a t w ell-designed an d scaled m anip u lativ e field experim ents can contribute to identifying w arnings o f increased risk fo r upco m in g ecosystem tra n sitio n s by testin g fo r changes in in teractio n netw orks. F o r exam ple, specific experim ents m an ip u latin g the level o f shading could be designed an d im plem ented in different locations to define key b reak p oints fo r resource m anagem ent. Shifts in functional perform ance associated w ith shading m ay n o t be g rad u al if in teractio n netw orks fu ndam entally shift. In o u r experim ent, we see changes associated w ith interactions th a t affect key attrib u te s o f the system ranging from p rim ary pro d u ctiv ity a n d the stabilization o f fine sedim ents to th e ability o f the system to recycle key nutrients. This indicates th a t the m ain ten an ce o f com m unity functio n plays critical roles in the resilience o f sandflats to elevated suspended-sedim ent co n c en tra ­ tions an d eutrophicatio n . C hange associated w ith non-p o in t-so u rce lan d use, sedim ents, an d n u trien ts are noto rio u sly difficult to m anage. H ow ever policies th a t generate inertia a n d are blin d to o th er surprises are d angerous a n d ultim ately m ay be very costly in term s o f the loss o f ecosystem services an d associated societal values (H o ra n et al. 2011). This is w hy th e risks o f th resh o ld shifts are im p o rtan t. If n o t pro p erly accounted for, o u r futures m ay be m ark ed by sudden, unexpected reductions in ecosystem functionality a n d value. I f there is a credible risk th a t a p articu la r system can change in a n o n lin ear w ay associated w ith subtle b u t cum ulative stress, then m anagem ent dow n to a p redeterm ined lim it will be highly risky. O u r a p p ro a c h o f investigating the arch i­ tecture o f ecosystem in teractio n netw orks, a n d how they are changed by stressors, provides real-w orld em pirical evidence to su p p o rt m an y o f the theo retical develop­ m ents in regim e shifts a n d resilience. This will allow us to m o re tan g ib ly u n d e rs ta n d th e im p licatio n s o f thresholds in ecosystem s th a t affect o u r daily lives. A lth o u g h we can n o t yet predict w hen b reak p o ints in d ynam ics w ill occur, o u r a p p ro a c h do es identify m echanistic links a n d the key connections betw een Ecology. Vol. 95. No. 6 ecosystem elem ents th a t will facilitate interventions. Such interventions are im p o rta n t in averting regime shifts, especially w here the drivers o f change are difficult to ad ju st o r generate huge legacy effects (Biggs et al. 2009). A cknow ledgm ents We thank the Auckland Council for permission to conduct the experiment (CN37056:FN12876) and Peter Robinson (Grounds and Wildlife H azard M anagem ent Planner for Auckland Airport) and Aviation Security for access to the study site. We thank Katie Cartner, Iain M acDonald. Barry Greenfield. Sarah Hailes. Michael Townsend. Hazel Needham. Carolyn Lundquist, and others involved with the fieldwork and laboratory analysis. This work was funded by M BIE CO 1X1005 and N IW A under the Coasts and Oceans Research Programme 3 (2012/13 SCI). C. K raan was supported by the M arsden Fund. Royal Society of New Zealand (NIW1102), and a Marie-Curie International Outgoing Fellowship (FP7-PEOPLE-2011-IOF). C. Van Colen acknowledges a postdoctoral fellowship provided by the Flemish Fund for Scientific Research (FWO-1.2.380.11.N.00). L it e r a t u r e C it e d A rar. E.. and G. Collins. 1997. M ethod 445.0 in vitro determ ination of chlorophyll a and pheophytin a. Marine and freshwater algae by fluorescence. Revision 1.2 U.S Environm ental Protection Agency. Cincinnati. Ohio. USA. Barbier. E. B.. S. D. Hacker. C. Kennedy. E. W. Koch. A. C. Stier, and B. R. Silliman. 2011. The value o f estuarine and coastal ecosystem services. Ecological M onographs 81:169193. Biggs. R.. S. R. Carpenter, and W. A. Brock. 2009. Turning back from the brink: Detecting an impending regime shift in time to avert it. Proceedings o f the N ational Academy of Sciences USA 106:826-831. Billerbeck. M. H.. H. Roy. K. Bosselmann. and M. Huettel. 2007. Benthic photosynthesis in submerged W adden Sea intertidal flats. Estuarine and C oastal Shelf Science 71:704716. Brook. B. W .. E. C. Ellis. M. P. Perring. A. W. M ackay. and L. Blomqvist. 2013. Does the terrestrial biosphere have plane­ tary tipping points? Trends in Ecology and Evolution 28: 396-401. C ahoon. L. B. 1999. The role of benthic microalgae in neritic ecosystems. Oceanography and M arine Biology 37:47-86. Carpenter. S. R. 2003. Regime shifts in lake ecosystems: pattern and variation. Excellence in Ecology Series N um ber 15. Ecology Institute. Oldendorf/Luhe, Germany. Carpenter. S.. B. W alker. J. M. Anderies, and N. Abei. 2001. From m etaphor to measurement: Resilience o f what to what? Ecosystems 4:765-781. Dakos. V.. et al. 2012. M ethods for detecting early warnings of critical transitions in time series illustrated using simulated ecological data. PLoS One 7:e41010. D ayton. P. K. 1971. Com petition, disturbance and community organization: The provision and subsequent utilization of space in a rocky intertidal community. Ecological M ono­ graphs 41:351-389. D ean. W. 1974. D eterm ination of carbonate and organic m atter in calcareous sediments and sedimentary rocks by loss on ignition: C om parison w ith other m ethods. Journal of Sedimentary Petrology 44:242-248. D uarte. C. M. 2002. The future o f seagrass meadows. Environm ental Conservation 29:192-206. Ehrenhauss. S.. U. W itte. F. Janssen, and M. Huettel. 2004. Decom position of diatom s and nutrient dynamics in perm e­ able N orth Sea sediments. Continental Shelf Research 24: 721-737. T H R E SH O LD PO TEN TIA L IN M A R IN E ECOSYSTEM S Grabowski, R. C., I. G. D roppo, and G. W harton. 2011. Erodibility of cohesive sediment: the im portance of sediment properties. E arth Science Reviews 105:101-120. Grace. J. B.. D. R. Schoolmaster. G. R. Guntenspergen, A. M. Little. B. R. Mitchell. K. M. Miller, and E. W. Schweiger. 2012. Guidelines for a graph-theoretic im plem entation of structural equation modeling. Ecosphere 3:art73. Hewitt. J. E.. and S. F. Thrush. 2010. Empirical evidence of an approaching alternate state produced by intrinsic community dynamics, climatic variability and m anagem ent actions. M arine Ecology Progress Series 413:267-276. H oran. R. D.. E. P. Fenichel. K. L. S. D rury, and D. M. Lodge. 2011. M anaging ecological thresholds in coupled environ­ m ental-hum an systems. Proceedings o f the N ational A cad­ emy of Sciences USA 108:7333-7338. Huettel. M .. and A. Rusch. 2000. T ransport and degradation of p h ytoplankton in perm eable sediment. Lim nology and Oceanography 45:534-549. Jones. H. F. E.. C. A. Pilditch. D. A. Bruesewitz. and A. M. Lohrer. 2011. Sedimentary environm ent influences the effect of an infaunal suspension feeding bivalve on estuarine ecosystem function. PLoS O N E 6:e27065. Kéfi. S.. V. D akos. M. Scheffer. E. H. V an Nes. and M. Rietkerk. 2013. Early warning signals also precede noncatastrophic transitions. Oikos 122:641-648. Lelieveld. S. D.. C. A. Pilditch. and M. O. Greene. 2003. V ariation in sediment stability and relation to indicators of m icrobial abundance in the O kura Estuary. New Zealand. Estuarine Coastal and Shelf Science 57:123-136. Lindegren, M.. V. Dakos. J. P. Gröger, A. G ârdm ark, G. Kornilovs. S. A. O tto, and C. M öllmann. 2012. Early detection of ecosystem regime shifts: a multiple m ethod evaluation for m anagem ent application. PLoS O N E 7: e38410. Lohrer. A. M.. J. E. Hewitt. S. F. Hailes. S. F. Thrush. M. Ahrens, and J. Halliday. 2011. C ontam ination on sandflats and the decoupling of linked ecological functions. Austral Ecology 36:378-388. Lohrer. A. M .. S. F. Thrush, and M. M. Gibbs. 2004. Bioturbators enhance ecosystem function through complex biogeochemical interactions. N ature 431:1092-1095. M uthén. L. K .. and B. O. M uthén. 2007. M plus user's guide. Sixth edition. M uthein and M uthein, Los Angeles. Califor­ nia. USA. Peterson. C. H.. and G. A. Skilleter. 1994. C ontrol of foraging behaviour of individuals within an ecosystem context: The clam M acoma balthica. flow environm ents and siphoncropping fishes. Oecologia 100:256-267. Rodil. I. F.. A. M. Lohrer. L. D. Chiaroni. J. E. Hewitt, and S. F. Thrush. 2011. D isturbance of sandflats by thin deposits of terrigenous sediment: consequences for prim ary produc­ tion and nutrient release. Ecological Applications 21:416426. Sandwell. D. R.. C. A. Pilditch. and A. M. Lohrer. 2009. Density dependent effects o f an infaunal suspension-feeding bivalve (Austrovenus stutchburyi) on sandflat nutrient fluxes and m icrophytobenthic productivity. Journal o f Experimen­ tal M arine Biology and Ecology 373:16-25. Scheffer. M.. S. Carpenter. J. A. Foley. C. Folke. and B. W alker. 2001. Catastrophic shifts in ecosystems. N ature 413: 591-596. Scheffer. M ., et al. 2012. Anticipating critical transitions. Science 338:344-348. 1457 Sloth. N. P.. T. H. Blackburn. L. S. Hansen. N. RisgaardPetersen, and B. A. Lomstein. 1995. N itrogen cycling in sediments with different organic loading. M arine Ecological Progress Series 116:163-170. Thrush. S. F.. J. E. Hewitt. V. Cummings. J. I. Ellis. C. H atton. A. Lohrer. and A. N orkko. 2004. M uddy waters: elevating sediment input to coastal and estuarine habitats. Frontiers in Ecology and the Environm ent 2:299-306. Thrush. S. F.. J. E. Hewitt. M. Gibbs. C. Lundquist, and A. N orkko. 2006. Functional role of large organism s in intertidal communities: Com munity effects and ecosystem function. Ecosystems 9:1029-1040. Thrush. S. F.. J. E. Hewitt. C. W. Hickey, and S. Kelly. 2008. Multiple stressor effects identified from species abundance distributions: Interactions between urban contam inants and species habitat relationships. Journal of Experimental M arine Biology and Ecology 366:160-168. Thrush. S. F.. J. E. Hewitt, and A. M. Lohrer. 2012. Interaction networks in coastal soft-sediments highlight the potential for change in ecological resilience. Ecological Applications 22: 1213-1223. Thrush. S. F.. J. E. Hewitt. A. N orkko. P. E. Nicholls. G. A. Funnell, and J. I. Ellis. 2003. H abitat change in estuaries: predicting broad-scale responses of intertidal m acrofauna to sediment m ud content. M arine Ecology Progress Series 263: 101 - 1 12 . Thrush. S. F.. J. E. Hewitt. R. D. Pridmore. and V. J. Cummings. 1996. Adult/juvenile interactions o f infaunal bivalves: contrasting outcomes in different habitats. M arine Ecology Progress Series 132:83-92. Thrush. S. F., et al. 1997. The sandflat habitat: Scaling from experiments to conclusions. Journal of Experimental M arine Biology and Ecology 216:1-10. Turner. S. J.. J. G rant. R. D. Pridm ore. J. E. Hewitt. M. R. W ilkinson. T. M. Hume, and D. Morrisey. 1997. Bedload and water-column transport and colonization processes by mobile post-settlement benthic macrofauna: Does infaunal density m atter? Journal of Experimental M arine Biology and Ecology 216:51-76. V an Colen. C.. S. F. Thrush. M. Vincx. and T. Ysebaert. 2013. Conditional responses of benthic communities to interference from an intertidal bivalve. PLoS ONE 8:e65861. V an de Koppei. J.. P. M. J. Herm an. P. Thoolen. and C. H. R. Heip. 2001. D o alternate stable states occur in natural ecosystems? Evidence from a tidal flat. Ecology 82:34493461. Vile. D.. B. Shipley, and E. G am ier. 2006. A structural equation model to integrate changes in functional strategies during old-field succession. Ecology 87:504-517. Volkenborn. N.. C. Meile. L. Polerecky. C. Pilditch. A. N orkko. J. N orkko. J. E. Hewitt. S. F. Thrush. D. S. Wethey. and S. A. W oodin. 2012. Interm ittent bioirrigation and oxygen dynamics in permeable sediments: A n experi­ mental and modeling study o f three tellinid bivalves. Journal of M arine Research 76:794-823. W oodin. S. A. 1978. Refuges, disturbance and community structure: a m arine soft-bottom example. Ecology 59:274284. W oodin. S. A.. D. S. W ethey. J. E. Hewitt, and S. F. Thrush. 2012. Small scale terrestrial clay deposits on intertidal sandflats: behavioral changes and productivity reduction. Journal of Experimental M arine Biology and Ecology 413: 184-191. S u p p le m e n ta l M a t e r i a l Appendix Description o f experimental effects and potential artifacts and base statistical tables structured equation modeling (Ecological Archives E095-127-A1). SlAOcCdiL June 2014