DOI 10.1007/s00442-008-9-z POPULATION ECOLOGY - ORIGINAL PAPER Beetle nd plnt density s cues inititing dispersl in two species of dult predceous diving eetles Donld A. Yee Stcy Tylor Steven M. Vmosi Received: 5 Decemer 2007 / Accepted: 17 Novemer 2008 Springer-Verlg 2008 Communicted y Joel Trexler. Electronic supplementry mteril The online version of this rticle (doi:10.1007/s00442-008-9-z) contins supplementry mteril, which is ville to uthorized users. D. A. Yee S. Tylor S. M. Vmosi Deprtment of Biologicl Sciences, University of Clgry, Clgry, AB T2N 1N4, Cnd Present Address: D. A. Yee (&) Deprtment of Biologicl Sciences, University of Southern Mississippi, Httiesurg, MS 39460, USA e-mil: donld.yee@usm.edu Astrct Dispersl cn inxuence popultion dynmics, species distriutions, nd community ssemly, ut few studies hve ttempted to determine the fctors tht Vect dispersl of insects in nturl popultions. Consequently, little is known out how proximte fctors Vect the dispersl ehvior of individuls or popultions, or how n orgnism s ehvior my chnge in light of such fctors. Adult predceous diving eetles re ctive dispersers nd re importnt predtors in isolted qutic hitts. We conducted interrelted studies to determine how severl fctors Vected dispersl in two common pond-inhiting species in southern Alert, Cnd: Grphoderus occidentlis nd Rhntus sericns. SpeciWclly, we (1) experimentlly tested the evect of plnt nd eetle densities on dispersl proilities in ponds; (2) surveyed ponds nd determined the reltionships mong eetle densities nd plnt densities nd wter depth; nd (3) conducted lortory trils to determine how eetle ehvior chnged in response to vrition in plnt densities, conspeciwc densities, food, nd wter depth. Our Weld experiment determined tht oth species exhiited density dependence, with higher eetle densities leding to higher dispersl proilities. Low plnt density lso ppered to increse eetle dispersl. Consistent with our experimentl results, densities of R. sericns in ponds were signiwcntly relted to plnt density nd vried lso with wter depth; G. occidentlis densities did not vry with either fctor. In the lortory, ehvior vried with plnt density only for R. sericns, which swm t low density ut were sedentry t high density. Both species responded to depth, with high eetle densities eliciting eetles to spend more time in deeper wter. The presence of food cused opposite responses for G. occidentlis etween experiments. Behviorl chnges in response to ptch-level heterogeneity likely inxuence dispersl in nturl popultions nd re expected to e importnt for oserved ptterns of individuls in nture. Keywords Complexity Density dependence Dytiscide Metpopultion Predtor Introduction Movement y orgnisms mong hitt ptches hs importnt consequences for the understnding of popultion dynmics, popultion genetics, species distriutions, nd community ssemly nd stility (Hnski nd Gilpin 1997; Hnski 1999; Bilton et l. 2001; Bowler nd Benton 2005). Theoreticl models tht incorporte dispersl often tret movement s Wxed component of the popultion nd ssume tht dispersing individuls re insensitive to withinor etween-ptch vrition (Bowler nd Benton 2005). This ssumption usully is invlid, s host of proximte fctors hve een shown to initite dispersl out of the current ptch, including negtive species interctions (Rosenzweig
1991; Binckley nd Resetrits 2005), food vilility (Hnski et l. 2002; Oh nd Tkgi 2005), nd density of conspeciwcs (Otronen nd Hnski 1983; Denno nd Peterson 1995; Herzig 1995). Although there re some nswers to the question, Wht fctors cuse n orgnism to leve ptch?, little dt exist tht quntify the ehviorl mechnisms tht my chnge in response to fctors inititing dispersl (Stinner et l. 1983). Individul decisions to disperse in or out of ptch often re relted to vrition in underlying iotic or iotic ptch chrcteristics (Rosenzweig 1991; Blustein 1999; Binckley nd Resetrits 2005; Bowler nd Benton 2005; Binckley nd Resetrits 2007). For exmple, density-dependent dispersl my e cused y either of two ehviorl mechnisms: response to decresing resources or n voidnce of conspeciwcs (Bowler nd Benton 2005). The underlying mechnism my inxuence ptterns oserved in the strength of density-dependent dispersl cross ptches (Herzig 1995). For instnce, if individuls disperse in response to decresing resources, density-dependent dispersl my e wek when resource levels re high, or strong when fctors such s incresing hitt complexity mke those resources hrder to locte. Conversely, if individuls disperse to void conspeciwcs, the strength of density-dependent dispersl my not vry in response to resource levels, nd my e wek in complex hitts where conspeciwcs re less likely to interct. The identiwction of how ehvior Vects dispersl decisions my provide ridge to developing generl ptterns of dispersl, predicting movement within nd etween hitts, nd led to etter understnding of metpopultion nd metcommunity dynmics. Indeed, mking ccurte predictions out dispersl cross lndscpes will require knowledge out how dispersl ehvior chnges with vrition in ptch structure (Goodwin nd Fhrig 2002). Freshwter qutic orgnisms tht hve complex life cycle fce diycult chllenge, s dults must disperse through potentilly inhospitle terrestril mtrix in order to rech qutic hitts contining food nd oviposition sites. The fctors inititing dispersl out of hitts nd into the surrounding lndscpe re lrgely unknown (Rundle et l. 2002; Petersen et l. 2004). Predceous diving eetles (Coleopter: Dytiscide) re reltively lrge (4,000 species), widely distriuted fmily of qutic predtors tht occur in vriety of temporry nd permnent wters (Lrson et l. 2000). Adult eetles re primrily qutic, ut they disperse ctively cross the terrestril mtrix (Lrson et l. 2000). The importnce of this group to food wes is s predtor on smll qutic invertertes nd vertertes (Johnson nd Jkinovich 1970; Deding 1988; Hicks 1994; Johnsson nd Nilsson 1992; Lrson et l. 2000; Adity et l. 2006). Even though dytiscids often re loclly undnt, hve high txonomic diversity, nd re importnt memers of the food wes tht they occupy, surprisingly little is known out dult dispersl or the fctors responsile for their movement mong hitts (Vmosi nd Vmosi 2007; Vmosi et l. 2007). Some species of dytiscids re eril colonizers of new qutic hitts (Firchild et l. 2000; Wilcox 2001) nd re good Xiers cple of moving severl kilometers (Lundkvist et l. 2002; Schäfer et l. 2006). Dytiscid Xight usully peks ner sunset nd my e Vected positively y ir tempertures (Nilsson 1997; Lundkvist et l. 2002). For long distnce Xights, eetles hve een shown to locte wter through the cue of polrized ultrviolet light (Nilsson 1997). Movement to novel hitts hs een studied (Southwood 1962; Wilcox 2001), lthough there is lck of informtion on the fctors tht my initite dispersl out of ptch (Lundkvist et l. 2002) or within ptch. Lundkvist et l. (2002) found tht ptterns of dult eetles mong diverent lndscpes ws potentilly consequence of hitt selection, ut the exct environmentl fctors responsile for these choices remins unknown. Our gol ws to exmine fctors inxuencing the dispersl in the Weld for two dytiscid species, Grphoderus occidentlis (Horn) nd Rhntus sericns (Shrp), tht re undnt t our Weld site. Rhntus sericns is 8.6 10.8 mm in length nd thought to e good colonizer of isolted qutic hitts (Lrson et l. 2000). Grphoderus occidentlis is lrger (11.1 14.4 mm), nd oth species re typiclly ctive during the dylight hours nd found in open sunlit ponds (Lrson et l. 2000). Once out of the wter, these species disperse ctively y Xight fter short cclimtion period (D.A.Y., personl oservtion). Currently, no dt exist on the fctors tht initite dispersl out of the qutic environment for these species, nor is there dt linking eetle ehvior with vrition in the qutic environment. We conducted interrelted studies to exmine how diverent fctors my Vect eetle dispersl (Weld experiment), if these fctors could explin ptterns of eetle undnce in nturl ponds (Weld survey), nd how eetle ehvior chnged with these fctors (lortory experiments). We Wrst performed Weld mnipultive experiment in nturl ponds wherein we vried qutic plnt nd eetle densities nd monitored eetle dispersl proilities. Becuse ponds in our study re vry in plnt nd eetle densities mong nd within ponds (D.A.Y., unpulished dt), we hypothesized tht these fctors would Vect eetle dispersl proilities. SpeciWclly, nd in the sence of prior studies on dispersl for these species, we predicted tht reduced plnt density or incresed eetle density would e ssocited with higher dispersl proilities. Furthermore, we expected dispersl proilities to diver etween species, with the lrger species (G. occidentlis) eing more sensitive to chnges in plnt nd eetle densities. We
then conducted Weld surveys tht exmined nturl ssocitions etween eetles nd plnt density nd wter depth in nturl ponds. We predicted tht if plnt or eetle densities were importnt in inititing dispersl in the Weld experiment, then our Weld surveys would revel undnce ptterns consistent with the experimentl results. For exmple, if we found tht eetles hd higher dispersl proilities in low plnt density, then we would predict low plnt density res of nturl ponds to hve low densities of eetles compred to high plnt density res. Finlly, to understnd if eetle ehvior ws importnt in explining Weld ptterns, we conducted lortory experiments wherein individul eetle ctivities nd positions in wter were mesured with respect to vrition in plnt nd conspeciwc densities, wter depth, nd food presence. We predicted tht ehviorl responses to the mnipulted fctors would give us insight into dispersl propensity in the Weld experiments nd ptterns generted from the Weld surveys. Thus, if low plnt density initited dispersl in Weld experiments nd Weld surveys reveled low undnces of eetles in low plnt density res of ponds, we would predict tht eetles might e less likely to inhit low plnt density res of lortory continers when overed choice. We ssumed tht the initition of eetle dispersl would e consequence of the iotic (conspeciwc density, plnt density) nd iotic (depth) environment within ponds, nd would e explined, in prt, y how individul eetles respond ehviorlly to such fctors. Mterils nd methods Study site nd orgnisms The site used for Weld studies ws locted pproximtely 20 km north of Strthmore, Alert, Cnd (51 18 17 N, 113 28 15 W), in priries used nnully for cttle grzing. This fenced site (1.6 km 2 ), known s the Roseud complex, contined pproximtely 35 ponds tht vried in size nd permnence. The dominnt qutic vegettion in temporry ponds consisted of ded nd live individuls of Nuttll s slt medow grss [Puccinelli nuttllin (JA Schultes) AS Hitchc.]. Ponds used were not connected y wter ut were ssumed to e close enough to permit eetles to Xy mong ponds. Field experiment Five semi-permnent or temporry ponds, ligned in north south xis, were selected (Tle S1). Ponds hd mximum depths from 20 to 33 cm. We mesured severl wter relted vriles (turidity, color, nd chlorine s surrogte of slinity) on the 1st dy of experiments ecuse wter chemistry could Vect dispersl. Mesurements were conducted on 150 ml smple of wter collected from ech pond on 12 July 2007 using TC-3000i colorimetric meter (LMotte, Chestertown, Mrylnd). In generl, there ws little vrition mong pond wter in chlorine, lthough wter from pond C hd higher turidity nd ws drker (Tle S1). Field dispersl continers (herefter, skets) were used to test the importnce of eetle density nd qutic plnt density on eetle dispersl propensity. Bskets (Fyllen, IKEA) were cylindricl (42 cm dimeter opening) nd consisted of 1 mm mesh over wire frmes. Bskets were lid horizontlly (47 cm long) in ech pond t 5 10 m intervls with the opening fcing towrd djcent ponds. Ech sket ws positioned in the wter so tht the wter ws deepest t the ck (12 14 cm) nd sloped up towrd the opening, leving 3 5 cm of the open end out of the wter llowing for sumerged wter surfce re of pproximtely 0.19 m 2. This design would hve forced eetles to exit the wter in order to leve the sket. Although we could not gurntee tht ll eetles tht left skets would disperse y Xight, eetle dispersing from skets would minimlly hve to crwl out of the wter, which would e the sme tctic used y eetles to move from pond onto lnd efore dispersing y ir (D.A.Y., personl oservtion). The wter volume within ech sket ws pproximtely 12 l. We plced 10 15 cm dimeter rock t the deepest prt in ech sket to weigh it down. Into ll skets we dded 1,250 ml pond wter contining undnt food for dult eetles (e.g., smll dytiscid lrve, Dphni sp., dmselxy lrve, hydrophilid lrve, corixids, chironomids, mosquito lrve). We excluded non-trget dult eetles, including hydrophilids, nd potentil predtors (e.g., lrger dytiscid dults or lrve, drgonxy lrve). Food ws collected from djcent ponds nd homogenized efore eing dded to skets. Tretments consisted of three densities of eetle (2, 4, or 8) nd two levels of plnt density (low = 10.0 g, high = 100.0 g wet mss) for totl of six skets in ech of the Wve ponds (n = 30). The design ws replicted twice on two non-consecutive dys (G. occidentlis: 12 nd 16 July; R. sericns: 13 nd 17 July) giving totl of ten replictes per tretment comintion per species (one replicte of ech tretment comintion in ech of Wve ponds over 2 dys). Only nine replictes exist from pond C for G. occidentlis s skets were distured during 12 July. The lowest eetle density ws chosen fter smpling 12 nturl ponds nine times ech using 0.16 m 2 smpler on 7 nd 27 June. Men densities were 1.29 1.39 nd 1.36 1.48 individuls of G. occidentlis nd R. sericns, respectively, per smple (7 27 June). Plnts consisted of live nd ded Puccinelli nuttllin collected from ponds. Plnts used in skets nturlly Xoted throughout the wter column, nd high densities of plnts completely oscured the ottom of
ech sket. Adult eetles were collected using qutic D-nets from surrounding non-experimentl ponds. Individul eetles were dded rndomly to skets during the erly fternoon on ech dy; 24 h lter, remining eetles were counted nd relesed. Between dys, skets were restocked with food nd repositioned if necessry to ssure consistency in wter depth. We ssumed tht skets were infrequently colonized, s eetles were never in excess of strting densities nd in only one instnce did n individul of non-trget species pper. Becuse eetles leving or remining in sket is inry response vrile, we modeled the proility of dispersl using inomil regression. This is conceptully similr to nlysis of vrince with proportion of eetles dispersing s the dependent vrile. However, inomil regression ssumes n underlying inomil, rther thn norml, proility distriution (Quinn nd Keough 2002). In stndrd inomil regression (logistic regression) the response vrile is inry (disperse vs remin) nd is consequence of the tretment fctors (eetle density, plnt density, species). The Wnl Wtted regression model descries the proportion of eetles dispersing. An underlying ssumption of regression is independence of oservtions, ut ecuse individul eetles co-occur within given sket, they were sujected to similr locl conditions nd cnnot e treted s independent dt points. Therefore, we used modiwed version of inomil regression, et-inomil regression (Venles nd Ripley 2003), which ccounts for individuls eing clustered within skets. The numer of eetles leving per sket (y) is ssumed to e drwn from inomil distriution, y» inomil(n, λ), where λ is the men proility of eetle dispersing nd n is the numer of individuls in given sket. The men proility of dispersing, λ, is function of the tretment fctors. Mthemticlly, this mens we llow the men proility of dispersing, λ, to vry mong the skets following et distriution. High vriility in λ, known s overdispersion, indictes tht vriility mong skets exceeds the vriility mong individuls within sket. The Wtted model hs one dditionl prmeter, φ, which mesures the degree of correltion within cluster, nd provides correction to the mount of vriility explined y the model (Venles nd Ripley 2003). A vlue of φ = 0 would indicte tht vriility etween nd within skets re comprle, nd individul eetles respond independently. The signiwcnce of min evects nd interctions were determined using three-wy nlyses of devince (ANO- DEV) with pond s lock term. ANODEV is generl liner model (GLM) nlogue of nlysis of vrince (ANOVA) tht permits speciwction of the non-norml error distriutions (Quinn nd Keough 2002) tht re likely to occur when ssuming n underlying inomil distriution. Chnges in model devince were clculted y including n evect lst in model tht lredy included the other explntory vriles, exclusive of higher order interctions. The signiwcnce of ech evect ws ssessed y chi-squre tests of likelihood rtio sttistics (G 2 ) in type III nlyses using the pproprite degrees of freedom. Men seprtion ws conducted on signiwcnt evects using n djustment for multiple comprisons. Anlyses were implemented using pckge od for nlysis of overdispersed dt in the sttisticl pckge R (http://www.r-project.org). Field survey To understnd if eetle undnces corresponded with fctors mnipulted in the Weld experiment (e.g., plnt density), we smpled six ponds during 16 18 July for nturl eetle densities nd ssocitions with reltive plnt densities nd wter depth. Two of the ponds surveyed were lso used in experiments (i.e., A, E, Tle S1), lthough surveys occurred fter the completion of dispersl experiments. Ponds rnged from 248 to 2,800 m 2. In ech pond we took 100 smples. An evort ws mde to smple cross vritions in depth nd plnt density within ech pond. A rigid plstic ottomless tu (49 33 cm opening, 0.16 m 2 ) ws crefully ut quickly plced into the wter column until it met the pond ottom. The density of plnt mteril covering the pond ottom within the tu ws qulittively estimted (1 = 0 20%, 2 = 21 40%, 3 = 41 60%, 4 = 61 80%, 5 = 81 100% cover). Depth (cm) ws recorded using meter stick hlfwy long the longest side of the tu. Six successive sweeps (three wy nd three towrd the resercher) within the tu were mde using n qutic D-net. Cptured dult R. sericns nd G. occidentlis were enumerted nd then trnsferred to holding in so s not to Vect susequent smples. In four of the ponds, three dditionl sweeps were mde for Wve smples to determine the evectiveness of our technique in cpturing dults. We tested for diverences in log 10 + 1 undnce of ech species mong ponds using mixed model nlysis of covrince (ANCOVA) with plnt density s the single fctor, nd depth s the covrite. Becuse we wished to understnd the reltionship etween plnt density nd depth nd eetle densities, nd were not interested in ponds per se, pond ws used s rndom lock term in ll nlyses. This nlysis ws run using SAS (PROC MIXED, SAS 2004). Lortory experiments To understnd how eetles ehved in response to fctors studied in the Weld tht my initite dispersl (e.g., eetle density, plnt density, depth), we performed two seprte concurrent lortory experiments. In one experiment, we
held depth constnt nd vried plnt density (herefter plnt density experiment), nd in nother experiment we used vrile depth environment ut held plnt density constnt (herefter depth experiment). In ech experiment we mesured eetle positions nd ctivities in the sence nd presence of food nd cross three eetle densities. Adult R. sericns nd G. occidentlis were collected from ponds t Roseud in July fter experiments were completed. Puccinelli nuttllin nd severl uckets of Wltered pond wter lso were collected. Pond wter ws used in ll experiments to provide ckground of wter-orne chemicl cues. Beetles were kept in species-speciwc colonies in wlk-in climte chmer set t 23 C on 12:12 light:drk cycle for 1 week prior to the eginning of the experiments nd given d liitum ccess to frozen chironomid lrve. Grss ws stored moist t 5 C. One dy prior to the eginning of experimentl trils, suycient numers of G. occidentlis nd R. serricns were isolted nd strved in individul 125 ml continers. We mrked 24 individuls of ech species using pper tgs (1 2 mm) glued to the elytr using cynurte glue (for methods, see Freilich 1989). Equl numers were coded with either white or lue tg. These tgged individuls were monitored in trils (see elow). For the plnt density experiment, four 40 l quriums were Wlled with mixture of 2.4 l pond wter nd 21.6 l deionized wter in climte chmer set t 23 C. For the high density section, top contining nine groups of P. nuttllin, ech composed of ten strnds, ws mde to Wt onehlf of ech qurium. For the other hlf, the top contined only three groups of grss, ech composed of two grss stems (low density section). For the depth experiment, n dditionl four 40 l quriums were plced t n ngle to produce grdient of depths rnging from 12 to 24 cm. Lines long the front of ech glss pnel were mrked to indicte depth. Ech qurium ws Wlled with wter s in the plnt density quri. Along the entire ck hlf of ech qurium we sumerged Wve regulrly spced groups of P. nuttllin, ech composed of Wve strnds. At the strt of the oservtions in oth experiments, eetles were dded to ech of the quri in densities of 2, 4, or 8 individuls. Two of the eetles in ech density were tgged (one of ech color s detiled ove). The ehviors of the tgged eetles were recorded using instntneous scn census every 2 min for 30 min y single oserver. During scns, the loction nd ctivity of ech tgged eetle ws recorded. Loctions included plnt (eetle touching grss) or spce (eetle not in contct with the grss). Activities included sttionry or swimming. In the plnt density experiment, we recorded the loction (high or low density) of individuls during ech scn. For the depth experiment, we recorded the depth (nerest cm) for ech tgged individul t ech scn. After the initil oservtion period, 1.25 g thwed frozen chironomid lrve ws dded to the center of ech qurium nd oservtions of the two tgged eetles were repeted s efore. Between trils, quriums were thoroughly clened nd the wter nd eetles were replced. For ech experiment, we replicted ech eetle density for ech species eight times (n = 24 experimentl units per species). Becuse of logistic constrints we could not collect dt on ll replictes of ech eetle density in ech experiment in one dy, so we rn experiments over the course of 2 dys (four replictes per dy), with replictes of ll possile tretment comintions (eetle density y plnt density nd eetle density y depth) for ech species conducted on ech dy. Behvior of the two mrked eetles in ech qurium ws verged prior to nlysis. Behvior dt ws rcsinesqure root trnsformed in order to homogenize vrinces nd to meet the ssumption of normlity. To reduce the numer of totl vriles nd to otin uncorrelted descriptors of ehvior, we used principl component nlysis (PCA) on these trnsformed proportions (Yee et l. 2004). Principl components (PCs) with eigenvlues >1.0 were retined for further nlysis (Quinn nd Keough 2002). For ech experiment (plnt density, depth) nd ech species, seprtely, we tested min evects nd the interction of eetle density nd food using multivrite nlysis of vrince (MANOVA) with retined PCs s dependent vriles. For the depth experiment, we lso used men depth of ech eetle s dependent vrile long with retined PCs. SigniWcnt MANOVA evects were interpreted using stndrdized cnonicl coeycients (SSC) tht quntify the mgnitude of the contriutions of the individul PCs in producing signiwcnt multivrite diverences (Scheiner 2001). Anlyses were implemented with the princomp nd mnov pckges in R. Results Field experiment Under ll model itertions, vlues of φ were signiwcntly diverent from zero (P < 0.001), suggesting tht vriility etween nd within skets ws not comprle, nd tht individul eetles cted in non-independent fshion. The min evect of eetle density ws signiwcnt (Tle 1), with more individuls of oth species dispersing t the highest versus the lowest density (Fig. 1). There lso ws mrginlly signiwcnt evect of plnt density on dispersl, with high plnt density promoting lower dispersl proility thn low plnt density (Tle 1, Fig. 1). In no cse ws species signiwcnt, indicting tht the dispersl proilities for G. occidentlis nd R. sericns were the sme under ech fctor.
Tle 1 Results from nlysis of devince on dispersl in Grphoderus occidentlis nd Rhntus sericns in ponds in Alert, Cnd Fctor df G 2 P Pond 4, 102 2.264 0.687 Species (S) 1 0.643 0.423 Plnt density (P) 1 3.332 0.068 Beetle density (D) 2 5.190 0.023 S D 2 1.611 0.204 S P 1 0.039 0.844 D P 2 0.952 0.329 S P D 2 0.035 0.852 Devince vlues Proportion dispersing 0.60 0.50 0.40 0.30 0.20 0.10 0.00 0.60 0.55 0.50 0.45 0.40 0.35 0.30 0.25 Fig. 1 Proportion of Grphoderus occidentlis nd Rhntus sericns eetles dispersing (men 1 SE) from skets in ponds in pririe north of Strthmore, Alert, Cnd in July 2007 in response to eetle densities, nd qutic plnt density (high = 100 g, low = 10 g wet mss Puccinelli nuttllin). SigniWcnce is sed on et inomil regression. The sme lower cse letters shred y mens indicte no signiwcnt diverences fter correcting for multiple comprisons Field survey 2 4 8 Beetle density (no. per sket) Our initil six sweeps ccounted for 91 nd 79% of individuls of G. occidentlis nd R. sericns, respectively, sed on dt from the dditionl sweeps tken in four of the ponds. Thus, even though we did not necessrily cpture ll eetles within tus, we encountered few eetles eyond the Low High Aqutic plnt density Wrst six sweeps. Mximum eetles per smple of G. occidentlis nd R. sericns were 6 nd 14, with mens of 0.60 0.04 G. occidentlis nd 1.06 0.07 R. sericns. For G. occidentlis undnce, plnt density (F 4,585 = 1.09, P = 0.362), depth (F 1,585 = 0.42, P =0.518), nd the interction of plnt density nd depth (F 4,585 = 0.71, P = 0.587) were not signiwcnt. Rhntus sericns undnce in smples vried signiwcntly with depth (F 1,585 = 24.68, P < 0.001), depth y plnt density (F 4,585 = 5.09, P = 0.005), ut ws only mrginlly signiwcnt with plnt density lone (F 4,585 =1.99, P = 0.095). For the interction etween depth nd plnt density, undnce did not vry cross depths in the lowest plnt density (0 20% plnt cover, F 1,38 =1.00, P = 0.333, R 2 = 0.02), ut undnce did increse with depth within the remining levels of plnt density (21 40%, F 1,38 = 4.16, P = 0.049, R 2 = 0.08; 41 60%, F 1,96 = 43.62, P < 0.001, R 2 = 0.31; 61 80%, F 1,91 = 69.29, P < 0.001, R 2 = 0.43: 81 100%, F 1,331 = 131.50, P < 0.001, R 2 =0.28). Lortory experiments For the plnt density experiment, results from PCA for oth species were similr, in tht PCA reduced the two ctivities nd four positions to two PC xes, which summrized etween 92.4 nd 98.1% of the vrition in these mesures (Tle 2). For oth species, we interpreted PC1 s seprting individuls swimming in open wter (lrge positive PC scores) from those tht were sttionry on plnts (lrge negtive PC scores) (Tle 2). Individuls with high positive scores on PC2 were ssocited with the low plnt density portions of the qurium, wheres individuls with lrge negtive scores occupied the high plnt density res (Tle 2). Bsed on MANOVA, we detected signiwcnt evects of food nd mrginlly signiwcnt evect of density for G. occidentlis in the plnt density experiment (Tle 3). In oth cses, PC1 contriuted most to the signiwcnt diverences etween tretment levels (Tle 3). Beetles in low densities spent time sttionry on plnts, ut swimming incresed s eetles were dded (Fig. 2). Without food, dult G. occidentlis lso remined sttionry, ut the ddition of food cused eetles to spend signiwcntly more time swimming in the open wter (Fig. 2). The ddition of food ppered to initite serching ehvior. Behvior of dult R. sericns ws Vected only y food, with PC1 nd PC2 contriuting to the signiwcnt min evect (Tle 3). In the sence of food, dults were sttionry on plnts in the high plnt density portion of the qurium, wheres food led to n increse in swimming within the low plnt density re (Fig. 3). For the depth experiment, PCA produced diverent results for ech species (Tle 2). For G. occidentlis, PCA reduced the two ctivities nd two positions to single PC
Tle 2 Results from principle component nlysis (PCA) for Grphoderus occidentlis (GO) nd Rhntus sericns (RS) for lortory experiments testing the evect of depth nd plnt density on eetle ehviors (ctivities nd positions) Species Experiment PC Eigenvlue Prop. vrince Cum. vrince Behvior GO Plnt density PC1 4.658 0.776 0.776 +(Swimming, Spce), (Sttionry, Plnt) PC2 1.227 0.205 0.981 +(Low), (High) Depth PC1 3.419 0.855 0.855 +(Swimming, Spce), (Sttionry, Plnt) RS Plnt density PC1 4.521 0.754 0.754 +(Swimming, Spce), (Sttionry, Plnt) PC2 1.023 0.170 0.924 +(Low), (High) Depth PC1 2.753 0.688 0.688 +(Plnt), (Spce) PC2 1.246 0.311 0.999 +(Sttionry), (Swimming) The eigenvlue nd proportionte (Prop.) vrince explined y ech PC, s well s the cumultive (Cum.) vrince, re given. Behviors with loding vlues 40 re presented Tle 3 Results of two-wy MANOVA (food nd eetle density) for Grphoderus occidentlis (GO) nd Rhntus sericns (RS) in trils testing the evect of plnt density on ehvior, nd ehvior nd depth ssocitions, respectively PC xes for ehviors re lso listed in this tle. Stndrdized cnonicl coeycients (SCC) indicte the mgnitude of the contriution of ech PCA fctor in producing signiwcnt multivrite diverences Species Source df Pilli s trce P-vlue SSC GO Plnt density PC1 PC2 Density (D) 4, 36 0.409 0.076 1.33 0.33 Food (F) 2, 17 0.341 0.028 1.34 0.09 D F 4, 36 0.147 0.589 1.35 0.10 Depth PC1 Depth Density (D) 4, 36 0.920 0.001 0.68 1.40 Food (F) 2, 17 0.363 0.022 1.36 0.17 D F 4, 36 0.328 0.157 0.58 1.42 RS Plnt density PC1 PC2 Density (D) 4, 36 0.288 0.218 0.90 0.89 Food (F) 2, 17 0.401 0.013 1.06 0.70 D F 4, 36 0.219 0.368 0.73 0.67 Depth PC1 PC2 Depth Density (D) 6, 34 0.830 0.004 0.13 0.11 1.83 Food (F) 3, 16 0.218 0.255 0.75 0.33 1.10 D F 6, 34 0.400 0.241 0.08 0.72 1.20 xis, which summrized 85.5% of the vrition in ehviors (Tle 2). This PC delineted individuls swimming in the open wter (lrge positive scores) from individuls who were sttionry on plnts (lrge negtive scores) (Tle 2). For R. sericns, two PC xes were produced (99.9% vrince explined), with PC1 diverentiting individuls who were t the plnt (positive scores) versus those tht were in the open wter (negtive scores), nd PC2 diverentiting individuls who were sttionry (positive scores) versus those tht swm (negtive scores) (Tle 2). MANOVA yielded signiwcnt density nd food min evects, ut no interction for G. occidentlis (Tle 3). The SSCs for eetle density for oth PC1 nd depth were lrge nd opposite, indicting tht oth were importnt for explining the evects of eetle density (Tle 3). The lowest numer of dults occupied shllow depths nd performed mixture of eing sttionry on plnts nd swimming in the open, moderte numers displyed trend towrd more swimming ut ehviors occurred in wter tht ws signiwcntly deeper, nd high numers cused dults to move to signiwcntly deeper wter nd remin sttionry on plnts (Fig. 4). In the sence of food, individuls ctively swm in spce, wheres dding food cused individuls to remin sttionry on plnts (Fig. 4). When food ws present, we oserved tht dults spent considerle time eting while ssocited with grss stems. For R. sericns, only eetle density signiwcntly Vected dult ehvior, with depth hving the lrgest SSC (Tle 3). As densities incresed, individuls spent more time in deeper wter (Fig. 5). Discussion Dispersl hs importnt consequences tht cn resonte throughout n ecosystem, from Vecting popultion persistence
Swimming, Spce PC1 1.50 1.00 0.50 0.00 Sttionry, Plnt -0.50-1.00-1.50 2 4 8 G. occidentlis density per qurium No food Food tretment Food Fig. 2 Men ( 1 SE) ehviorl response for G. occidentlis cross diverent densities (2, 4, 8 eetles), nd food tretments (food, no food). Activities (swimming, sttionry) nd positions (on plnt, in spce) most closely ssocited with lrge positive or lrge negtive PC scores re indicted prllel to the y-xis. The sme letters shred y mens indicte no signiwcnt diverences fter correcting for multiple comprisons sed on multivrite contrsts Swimming, Spce PC2 Sttionry, Plnt 0.80 0.60 0.40 0.20 0.00-0.20-0.40-0.60-0.80 1.00 Food 0.50 Low plnt density 0.00 PC1 No food -0.50-1.00 High plnt density Fig. 3 Bivrite mens ( 1 SE) for qutic plnt density ssocitions, PC1 (high = 90 strnds, low = 6 strnds of Puccinelli nuttllin) vs ctivities nd positions, PC2 (swimming or sttionry ctivities nd on plnt or in spce positions) tht mde the gretest contriutions to multivrite diverences etween the two food tretments (no food, food) for Rhntus sericns. Mens re signiwcntly diverent sed on multivrite contrsts nd dynmics, to community ssemly nd evolutionry processes, ut fctors tht Vect the dispersl of individuls out of ptch re rrely empiriclly evluted (Stinner et l. 1983; Dok et l. 1992; Bowler nd Benton 2005). Our Weld nd lortory experiments nd oservtions indicted tht movement of Grphoderus occidentlis nd Rhntus sericns, either out of or within pond, cn e Vected y ttriutes of the qutic environment s well s density of conspeciwcs. In the Weld, density dependence ws importnt for oth species, nd eetles in low plnt density dispersed t rte comprle to those in the highest eetle densities (Fig. 1). Thus, our hypothesis tht plnt nd eetle density would initite eetle dispersl ws supported. Bsed on Weld experiments, we predicted tht these fctors would lso help to explin ptterns of undnce in nturl ponds. Plnt density ppered to hve relevnce only for R. sericns, s more eetles were found in deeper wter cross higher levels of plnt density in our Weld surveys. We lso predicted tht ehviorl responses would give us insight into dispersl propensity generted from the Weld. Trils did indicte tht, when given choice, R. sericns ws more likely to ssocite with high plnt density res (Tle 3), result consistent with oth the Weld experiment nd Weld survey. Although oth species responded in similr fshion to plnt density in the Weld experiment (Fig. 1), undnce of G. occidentlis did not vry with plnt density sed on Weld smpling, nor did this species show ssocitions with either level of plnt density in the lortory experiment (Tle 3). Overll, there only ws one diverence in ehvior etween species in the lortory depth experiment, with R. sericns ssociting with plnts more often (79.0 0.03%) compred to G. occidentlis (68.0 0.05%). The fct tht G. occidentlis did not spend s much time ssociting with plnts my indicte n inherent ehviorl diverence etween these species tht my Vect the ptterns of undnce we oserved in the Weld surveys. This fct does not explin why G. occidentlis dispersed t higher rte when plnt densities were low in the Weld experiment. One possiility is tht G. occidentlis ws responding to something else tht ws correlted to plnt density in the Weld experiment tht ws sent in the Weld surveys or in the lortory experiment. Given tht R. sericns were more undnt in our Weld surveys (n = 633 vs n = 357 for G. occidentlis), it lso is possile tht R. sericns is more sensitive to conspeciwc density, nd this my explin the strong hitt ssocites tht my mitigte density evects in our Weld survey nd lortory experiments. Finlly, lrve of G. occidentlis re ctive swimmers tht use open wter hunting tctics, wheres lrvl R. sericns use plnts s perch to hunt (D.A.Y., mnuscript sumitted).
Swimming, Spce 1.5 1.0 0.5 2 4 PC1 Sttionry, Plnt 0.0-0.5-1.0-1.5 0.0 5.0 10.0 15.0 20.0 Depth (cm) 8 No food Food Food tretment Fig. 4 Behviorl response for G. occidentlis. Bivrite mens ( 1 SE) for PC1 of diverent eetle densities nd wter depth, which mde the gretest contriutions to multivrite diverences mong density tretments (2, 4, or 8 individuls), nd men ( 1 SE) response to food tretments. Activities nd positions most closely ssocited with lrge positive or lrge negtive PC scores re indicted prllel to the xes Depth (cm) 16. 0 14. 0 12. 0 10. 0 8.0 6.0 4.0 2.0 0.0 2 4 8 R.sericns density per qurium Fig. 5 Men ( 1 SE) depth of R. sericns with vrition in density (2, 4, or 8 individuls per qurium). The sme lower cse letters shred y mens indicte no signiwcnt diverences fter correcting for multiple comprisons sed on multivrite contrsts Such lrvl ehviors my trnslte to diverences in dult preferences s well, lthough this topic is lrgely unexplored for dytiscids. Although it ppers tht there exist diverences etween species in response to this plnt density lone, future work will need to focus on more detiled response surfce of eetle densities nd levels of plnt density in order to fully determine the role of this fctor in Vecting dispersl in nturl popultions. Other fctors thought to initite dytiscid dispersl include hitt drying (Velsco nd Milln 1998), wter temperture (Smith 1973; Velsco nd Milln 1998), nd reproduction (Lundkvist et l. 2002). In only one study hs density dependence een recognized, lthough t densities much greter thn were used here (24 eetles per 1 l, Smith 1973). In our study, it is unlikely tht hitt drying ws cue for dispersl, s wter levels in ponds did not pprecily diver cross the experiment. In ddition, the study tht identiwed drying s cue concluded tht wter levels <1 cm were necessry for promoting dispersl (Velsco nd Milln 1998); wter depth in our skets ws mintined well c ove this threshold (12 14 cm). High wter tempertures (>32 C) hve een implicted s cue for dytiscids in desert res (Velsco nd Milln 1998), lthough our ponds never reched such extremes. Becuse we did not priori select equl sex rtios for skets, it remins possile tht skewed rtios my hve Vected dispersl. It lso hs een suggested tht the physiologicl sttus of eetle could Vect dispersl rte (Velsco nd Milln 1998), lthough the fct tht individul eetles were rndomly ssigned to skets mkes diverences in physiologicl stte n unlikely explntion for the ptterns we oserved. In generl, our dt from the Weld experiment support the notion tht eetles use the density of conspeciwcs nd plnt density s cues to initite dispersl, nd ehviorl ssys indicte tht these cues my ct in dditive wys to Vect ehvior. It ws evident from our Weld experiment tht eetles of oth species were more likely to sty in skets with high plnt density (Fig. 1), nd there ws evidence tht nturl popultions of R. sericns were higher in res of ponds with high plnt density. Incresing plnt density my medite density-dependent evects y diminishing conspeciwc encounters. Alterntively, high plnt density cn provide sfer hitt from predtors, ut my interfere with n orgnism s ility to disperse or Wnd food (Goodwin nd Fhrig 2002). Our Weld survey nd experiments do not llow us to identify the mechnism tht explins the response to plnt density, lthough the results from our ehviorl ssys suggest tht R. sericns my chnge its use of high versus low plnt density hitts in response to food level (Fig. 3). The Weld survey reveled tht densities of R. sericns increse with incresing plnt density nd wter depth, result tht ws prtilly corroorted y ehviorl trils, where eetles in the highest densities occupied the deepest prt of the quri. Although densities of G. occidentlis did not vry with ny fctor in the Weld survey, high densities did initite higher dispersl in the Weld experiment
(Fig. 1) nd ehviorl trils concluded tht higher densities ltered eetle ehvior nd depth ssocitions (Fig. 4). When we exmined the coeycient of vrition (CV) for ech species in the depth experiment, we noted tht the CV of depth decresed signiwcntly with incresing density (G. occidentlis: F 1,47 =48.18, P < 0.001, R 2 = 0.490; R. sericns: F 1,47 = 48.96, P < 0.001, R 2 = 0.517). This suggests tht low densities of eetles use wider rnge of depths, ut incresing eetle densities led to more limited hitt use. Such movement to deeper wter with incresing densities my help to llevite density dependent evects (e.g., contct with conspeciwcs) s more volume is ville in those res. Decresed movement with incresing density, s seen in the depth experiment for G. occidentlis (Fig. 4), lso my e the mechnism tht reduces conspeciwc encounters. Contrry to our Wndings, dytiscids hve een shown to ssocite more with shllow pond mrgins thn with deeper zones (Julino 1991; Firchild et l. 2003), lthough it must e noted tht even our deepest smples were generlly less thn 30 cm deep. The use of shllow wter hitts my e to void deeper wter predtors (Julino 1991; Lrson et l. 2000), or to sty closer to the wter surfce for rething (Julino 1991). We suggest tht n ssocition etween eetles nd shllow wter lso my e wy to expedite dispersl out of pond. Concentrtions of eetles in the shllow mrgins re lso thought to led to ssemlges of intercting species (Firchild et l. 2003), lthough evidence for competitive interctions mong dytiscids hs not een found (Julino nd Lwton 1990). The Weld surveys for R. sericns showed tht this species ggregtes in deeper wter, lthough it must e noted tht depth nd our qulittive mesure of plnt density in nturl ponds were negtively correlted (r = 0.301, P < 0.001), indicting tht shllow res hd high plnt densities nd deeper wter ws lrgely plnt free. Anlysis indicted tht oth depth nd plnt density were importnt for explining R. sericns densities in ponds, which mens tht these two fctors were cting in concert. The inxuence of food vilility on dispersl hs not een tested in the Weld for dytiscids, ut this fctor hs een shown to medite dispersl in other lrge qutic inverterte predtors (Oh nd Tkgi 2005). In our ehvior ssys, we oserved tht oth species rected to food vilility, lthough the ehviorl responses were species nd context dependent. For instnce, G. occidentlis were more ctive in the presence of food when depth ws constnt, ut were minly inctive when food ws provided in the vrile depth experiment. We ttriute this diverence to feeding versus serching ehviors. When depth ws constnt ut the environment ws vrile (i.e., low vs high plnt density), individuls spent signiwcnt mount of time serching for food. When depth ws vrile ut plnt density ws constnt nd reltively low, dults quickly cptured prey nd then emrked on outs of motionless eting. It is uncler why vrition in depth should elicit such ehviorl chnge, ut it does seem plusile tht n interction etween depth nd plnt density ws involved. There is indirect evidence tht food density cn help to structure eetle communities t the ptch level (Nilsson nd Svensson 1995), ut further work is needed to identify how strong of n inxuence this fctor is in inititing dispersl. Dytiscid dispersl hs some importnt rmiwctions for eetle communities. Resetrits (2001) hs suggested tht individul movement mong ponds my ssist in homogenizing vrition in eetle communities if such movement is rndom, or it my crete distinct ssemlges if dults respond to underlying environmentl diverences mong hitts (see lso Vmosi nd Vmosi 2007). This ltter type of movement involves selection of hitts thought to promote the highest potentil Wtness (Fretwell nd Lucs 1970; Binckley nd Resetrits 2005). From our work it ppers tht eetle movement nd loction within ponds re non-rndom, which my help to explin the nested ptterns of eetle ssemlges noted y others (Nilsson nd Svensson 1995). Other fctors not studied here could lso inxuence eetle communities in the Weld. For instnce, the presence of predtors my structure popultions nd communities of eetles y ltering dispersl ptterns mong prey (Binckley nd Resetrits 2005). Although predtors or other fctors my shpe communities through coloniztion, such fctors re likely importnt for movement of orgnisms out of ptch s well. We used plnt iomss in our study ecuse positive reltionships etween qutic plnt iomss or density nd inverterte undnce nd diversity hve een identiwed in severl freshwter qutic systems (Heck nd Wetstone 1977; Crlisle nd Hwkins 1998). There hs een some dete out the dewnition of hitt complexity, which hs hindered rod generliztion for the ultimte evect of complexity (McCoy nd Bell 1991). We rgue tht the results from this study on the evect of plnt density on eetle dispersl will id in the understnding of the roder topic of hitt complexity in qutic systems. Our dt indicte tht dispersl in predceous diving eetles is not constnt trit, ut is likely Vected y the structure nd chrcteristics of hitt ptches (Goodwin nd Fhrig 2002). Vrition nd undnce of ptches within ponds is lso thought to exert strong inxuence on locl eetle communities (Firchild et l. 2003), nd there is growing ody of evidence showing tht hitt selection is principl driver of popultion ptterns in qutic hitts (Blustein 1999; Resetrits 2001; Binckley nd Resetrits 2005). By linking the results from Weld experiments, Weld surveys, nd lortory ehviorl trils, we hve mde strt to understnding how eetles rect to
iotic or iotic fctors within ptch, nd how those fctors my initite dispersl out of ptch. Acknowledgments We thnk B. Wohlfhrt for ssistnce in the Weld, S.H. Yee for dvice nd ssistnce on sttisticl nlyses, S.J. Mcculey, J. Jnnot, G.W. Firchild, nd n nonymous reviewer for helpful suggestions on n erlier version of this mnuscript, nd the Reinhrdt Fmily nd J. Brunen (Ducks Unlimited, Alert, Cnd) for proving ccess to the Weld site. This project ws supported y funds from n Alert Ingenuity New Fculty Awrd (#20060362) nd n NSERC Discovery Grnt (RGP283114) to S.M.V. References Adity G, Ash A, Sh GK (2006) Predtory ctivity of Rhntus sikkimensis nd lrve of Toxorhynchites splendens on mosquito lrve in Drjeeling, Indi. J Vector Borne Dis 43:66 72 Bilton DT, Freelnd JR, Okmur B (2001) Dispersl in freshwter invertertes. Annu Rev Ecol Syst 32:159 181 Binckley CA, Resetrits WJ Jr (2005) Hitt selection determines undnce, richness nd species composition of eetles in qutic communities. Biol Lett 1:370 374 Binckley CA, Resetrits WJ Jr (2007) EVects of forest cnopy on hitt selection in treefrogs nd qutic insects: implictions for communities nd metcommunities. Oecologi 153:951 958 Blustein L (1999) Oviposition site selection in response to risk of predtion: evidence from qutic hitts nd consequences for popultion dynmics nd community structure. In: Wsser SP (ed) Evolutionry theory nd processes: modern perspectives. Kluwer, Netherlnds, pp 441 456 Bowler DE, Benton TG (2005) Cuses nd consequences of niml dispersl strtegies: relting individul ehvior to sptil dynmics. Biol Rev 80:205 225 Crlisle DM, Hwkins CP (1998) Reltionships etween inverterte ssemlge structure, two trout species, nd hitt structure in Uth mountin lkes. J North Am Benthol Soc 17:286 300 Deding J (1988) Gut content nlysis of diving eetles (Coleopter: Dytiscide). Nt Jutl 22:17 184 Denno RF, Peterson MA (1995) Density-dependnt dispersl nd its consequences for popultion dynmics. In: Cppuccino N, Price PW (eds) Popultion dynmics: new pproches nd synthesis. Acdemic, Sn Diego, pp 113 130 Dok DR, Mrino PC, Kreiv PM (1992) Sptil scle medites the inxuence of hitt frgmenttion on dispersl success: implictions for conservtion. Theor Popul Biol 41:315 336 Firchild GW, Fulds AM, Mtt JF (2000) Beetle ssemlges in ponds: evects of hitt nd site ge. Freshw Biol 44:523 534 Firchild GW, Cruz J, Fulds AM, Short AEZ, Mtt JF (2003) Microhitt nd lndscpe inxuences on qutic eetle ssemlges in cluster of temporry nd permnent ponds. J North Am Benthol Soc 22:224 240 Freilich JE (1989) A method for tgging individul enthic mcroinvertertes. J North Am Benthol Soc 8:351 354 Fretwell SD, Lucs HL (1970) On territoril ehvior nd other fctors inxuencing hitt distriution in irds. I. Theoreticl development. Act Biotheor 19:16 36 Goodwin BJ, Fhrig L (2002) EVects of lndscpe structure on the movement ehviour of specilized goldenrod eetle, Tirihd orelis. Cn J Zool 80:24 35 Hnski I (1999) Metpopultion ecology. Oxford University Press, Oxford Hnski I, Gilpin M (1997) Metpopultion iology: ecology, genetics nd evolution. Acdemic, Sn Diego Hnski I, Breuker CJ, Schöps K, SetchWeld R, Neiminen M (2002) Popultion history nd life history inxuence the migrtion rte of femle Glnville fritillry utterxies. Oikos 98:87 97 Heck KLJ, Wetstone GS (1977) Hitt complexity nd inverterte interctions in vegetted qutic hitts. J Biogeogr 4:135 142 Herzig AL (1995) EVects of popultion density on long-distnce dispersl in the goldenrod eetle Trirhd virgt. Ecology 76:2044 2054 Hicks B (1994) Foregut contents of dult Ilyius Erichson (Dytiscide: Coleopter) from Newfoundlnd. Coleopt Bull 48:199 200 Johnsson A, Nilsson AN (1992) Dytiscus ltissimus nd D. circumcinctus (Coleopter, Dytiscide) lrve s predtors on three csemking cddis lrve. Hydroiologi 248:201 213 Johnson GH, Jkinovich W Jr (1970) Feeding ehvior of the predceous diving eetle Cyister Wmrioltus Wmrioltus (Sy). Bio- Science 20:111 Julino SA (1991) Chnges in structure nd composition of n ssemlge of Hydroporus species (Coleopter: Dytiscide) long ph grdient. Freshw Biol 25:376 378 Julino SA, Lwton JH (1990) The reltionship etween competition nd morphology. II. Experiments on co-occurring dytiscid eetles. J Anim Ecol 59:831 848 Lrson DJ, Alrie Y, Roughley RE (2000) Predceous diving eetles (Coleopter: Dytiscide) of the Nerctic region. NCR Reserch Press, Ottw Lundkvist E, Lndin J, Krlsson F (2002) Dispersing diving eetles (Dytiscide) in griculturl nd urn lndscpes in south-estern Sweden. Ann Zool Fenn 39:109 McCoy ED, Bell SS (1991) Hitt structure: the evolution nd diversiwction of complex topic. In: Bell SS, McCoy ED, Mushinsky HR (eds) Hitt structure. The physicl rrngement of ojects in spce. Chpmn nd Hll, London, pp 3 27 Nilsson AN (1997) On Xying Hydroporus nd the ttrction of H. incognitus to red cr roofs. Ltissimus 9:12 16 Nilsson AN, Svensson BW (1995) Assemlges of dytiscid predtors nd culicid prey in reltion to environmentl fctors in nturl nd cler cut orel swmp forest pools. Hydroiologi 308:183 196 Oh S, Tkgi H (2005) Food shortge Vects Xight migrtion of the gint wter ug Lethocerus deyrolli in the prewintering seson. Limnology 6:85 90 Otronen M, Hnski I (1983) Movement ptterns in Spheridium: diverences etween species, sex, nd feeding nd reeding individuls. J Anim Ecol 52:663 680 Petersen I, Msters Z, Hildrew AG, Ormerod SJ (2004) Dispersl of dult qutic insects in ctchments of divering lnd use. J Appl Ecol 41:934 950 Quinn GP, Keough MJ (2002) Experimentl design nd dt nlysis for iologists. Cmridge University Press, Cmridge Resetrits WJ Jr (2001) Coloniztion under thret of predtion: voidnce of Wsh y n qutic eetle, Tropisternus lterlis (Coleopter: Hydrophilide). Oecologi 129:155 160 Rosenzweig ML (1991) Hitt selection nd popultion interctions: the serch for mechnisms. Am Nt 137:S5 S28 Rundle SD, Foggo A, Choiseul V, Bilton DT (2002) Are distriution ptterns linked to dispersl mechnism? An investigtion using pond inverterte ssemlges. Freshw Biol 47:1571 1581 SAS (2004) SAS/STAT 9.1 user s guide. SAS Institute Inc., Cry, NC Schäfer ML, Lundkvist E, Lndin J, Persson TZ, Lundström JO (2006) InXuence of lndscpe structure on mosquitoes (Dipter: Culicide) nd dytiscids (Coleopter: Dytiscide) t Wve sptil scles in Swedish wetlnds. Wetlnds 26:57 68 Scheiner SM (2001) MANOVA. Multiple response vriles nd multi species interctions. In: Scheiner SM, Gurevitch J (eds) Design nd nlysis of ecologicl experiments, vol 2. Oxford University Press, Oxford, pp 99 133.