Jurnal Kece Plankton!

8/10/2019 Jurnal Kece Plankton!

1/12

Threshold-driven shifts in two copepod species: Testing ecological theory withobservational data

Ulrike Scharfenberger, Aldoushy Mahdy, and Rita Adrian *

Leibniz-Institute of Freshwater Ecology and Inland Fisheries, Berlin, Germany

AbstractWe used an observed abrupt shift in the dominance pattern of two coexisting copepod species in Muggelsee, a

shallow eutrophic lake in Germany, to investigate mechanisms leading to this shift, by embedding our findingsinto the framework of intraguild predation theory and theoretical scenarios of threshold-driven regime shifts. Weproposed that the abrupt increase in Cyclops kolensis , changing its status from a rare to the dominant species asavailable algal prey declined in the lake, was due to its superior exploitative competition for commonly consumedalgal prey. However, C. kolensis was only able to thrive in the emerging low food niche when abundances of competing larger Cyclops vicinus , a predator of C. kolensis juveniles, fell below a critical threshold. This isconsistent with the state threshold scenario of regime shift theory, for which a response variable exhibits anabrupt shift, here C. kolensis , after the driver ( C. vicinus ) crosses a threshold. We confirmed the nonlinearrelationship between the two copepod species by excluding potentially matching abrupt changes in other abioticand biotic driving variables, and successfully classifying C. kolensis abundance probability on the basis of C.

vicinus abundances using logistic generalized linear modeling. C. vicinus decline followed the driver thresholdscenario of regime shift theory, whereby an abrupt change in a driver (cryptophytes) causes a sudden shift in aresponse variable ( C. vicinus ). We illustrated how observational data on plankton communities match predictionsderived from ecological theory.

In freshwater ecosystems, relative community composi-tion of zooplankton ensembles constantly fluctuates overthe years. This is because the communities are embedded ina highly dynamic and complex system governed by a hugerange of extrinsic and intrinsic forces, where the latter oftenmediate the effects of the former (Sommer et al. 2012).Thus, changing conditions can favor different species,depending on their ecological niches. The resulting changesin abundance can thereby be gradual or abrupt withelapsing time. Abrupt changes constitute regime shifts andmay be the result of different mechanisms. The regime shiftmight arise from an equally sudden change in a drivingvariable, or it may be a drastic answer to a gradualchanging driving variable surpassing a threshold. Therebythe drastic response may or may not be the result of a jumpbetween alternative stable states (Andersen et al. 2009;Scheffer 2009; Bestelmeyer et al. 2011). In recent years,drastic responses in ecosystem dynamics have gainedincreasing attention in ecology because of their seemingunpredictability and potentially large effect on an ecosys-tem. Hence, if a regime shift occurs, it is desirable to first

differentiate whether potential drivers exhibit matchingsudden changes or if their change is gradual. If the latter isthe case, important insight into ecological systems can begained by quantifying potential thresholds in those drivers.

Extrinsic changes driven by climate and trophic states,which many freshwater systems have been subjected to overthe past decades (Mooij et al. 2005; Adrian et al. 2009),have a large effect either at the single species level(Gyllstrom et al. 2005; Seebens et al. 2008) or on planktoncommunities as a whole (Adrian et al. 2006; Elliot and May2008; Wagner and Adrian 2009). General concepts

governing the intrinsic dynamics of a community includecompetition, predation, and intraguild predation (IGP); thelatter being a special case describing the killing and eatingof species that use similar, often limiting, resources and arethus potential competitors (Polis et al. 1989). In contrastto linear food chains, simple IGP models, comprising apredator, a consumer, and a resource, predict theconsumers competitive exclusion by an omnivorouspredator in highly productive environments, while coexis-tence becomes possible at intermediate productivities.Potentially two alternative stable states can occur at thisintermediate productivity level (Holt and Polis 1997;Mylius et al. 2001; Verdy and Amarasekare 2010). Aprecondition for these dynamics is the consumers superi-ority in exploitative competition for the common resourceand the predators significant fitness benefit from theconsumption of the consumer. However, Holt and Huxel(2007) showed that both prerequisites can be relaxed, if alternative resources are added to the simple IGP food-webmodel.

There is an apparent gap between theoretical frame-

works in terms of driving forces of regime shifts and theirapplicability to the abrupt shifts observed in nature, i.e.,testing ecological theory with observational data (but seeScheffer et al. 2001; Scheffer and Carpenter 2003; Dakoset al. 2008). Our present study contributes to filling this gapby exploring the driving forces for a sudden shift in therelative abundances of two coexisting freshwater cyclopoidcopepod species. Cyclops vicinus gradually declined inabundance, whereas Cyclops kolensis populations abruptlyincreased between 1980 and 2010 in the Muggelsee (Berlin,Germany). This lake has experienced an improvement in itstrophic state (Kohler et al. 2005). At the same time watertemperatures have been increasing over the past three* Corresponding author: [email protected]

Limnol. Oceanogr., 58(2), 2013, 741752E 2013, by the Association for the Sciences of Limnology and Oceanography, Inc.doi:10.4319/lo.2013.58.2.0741

741


2/12

decades (Adrian et al. 2006; Wilhelm et al. 2006; Huberet al. 2008). Hence, we were able to test the combinedeffects of changes in the availability of prey and enhancedwater temperature in the context of IGP interactionsbetween these two copepod species, since they share thesame prey spectrum and prey on each others juvenilestages. We then integrated the observed interlinked systemdynamics into scenarios of regime shift theory as proposedby Andersen et al. (2009).

C. kolensis is a winter species of northern European lakes(Rivier 1996) known for its strict life cycle, with a diapausein summer, and a pelagic existence in winter until earlyspring. C. vicinus is a common species in eutrophic lakes. Inless productive lakes, it usually develops two populationmaxima in spring and autumn, interrupted by a diapauseduring summer (Santer and Lampert 1995). In highlyeutrophic lakes, it tends to be present all year round in thepelagic zone (Hansen and Jeppesen 1992; Hansen 1996).Both species are omnivorous (Toth and Zankai 1985; Tothet al. 1987; Adrian 1991) and have similar diets including juvenile stages of their own genus (Adrian 1991; Krylov etal. 1996). However, the larger species C. vicinus has higheringestion rates for both algal and zooplankton prey(Adrian 1991; Krylov et al. 1996) and is known for itshigh food demand (Santer and Lampert 1995; Hopp andMaier 2005). This is in line with observed decliningpopulations of C. vicinus in lakes undergoing reoligotro-phication (Seebens et al. 2008), or their prominence inhighly eutrophic lakes (Adrian and Deneke 1996; Adrian1997; Maier 1998).

On the basis of long-term observational records (1980 to2010) we explored the nature of the driving forcesunderlying the abrupt increase of C. kolensis in terms of (1) intraguild competition for prey between C. vicinus andC. kolensis , (2) threshold-driven release of C. kolensis frompredation pressure by the larger C. vicinus, and (3) the roleof warming trends. We postulate that C. vicinus , due to itshigher food demands, was negatively affected by theobserved reduction in the trophic state of the lake, whichresulted in less available herbivorous prey. By contrast, C.kolensis , with its lower resource demands, profited in twoways: through release from competition for shared foodresources, and through release from predation pressure onits juvenile stages, i.e., by the larger but decliningpopulation of C. vicinus . We first quantified criticalthresholds for prey availability, separating high and lowfood niches, and second C. vicinus densities, separating

high vs. low abundance probabilities of C. kolensis .Our observed data are in line with the IGP model (Poliset al. 1989) such that the two species are able to coexistunder conditions of intermediate productivity. As for thenature of the observed shifts in driver and responsevariables, our data exhibited an intriguing match withscenarios proposed by regime shift theory (Andersen et al.2009).

Methods

Study site The Muggelsee classifies as a shallow,polymictic, and eutrophic lake, situated in the temperate

climate zone at the transition from maritime to continentalclimate, southeast of suburban Berlin (52 u 269N, 13 u 399E).The lake covers an area of 7.3 km 2 , has a mean depth of 4.9 m, and a maximum depth of 7.9 m. The trophic statehas changed from hypereutrophic to eutrophic, with atransitional period in the mid-1990s (Kohler et al. 2005;Huber et al. 2008). In addition, the Muggelsee has beensubjected to seasonal warming trends in water temperaturein the range of 1 u C to 2.4 u C over the last three decades(Adrian et al. 2006). For a comprehensive limnologicalcharacterization of the lake see Driescher et al. (1993).

Data collection and data processing Between 1980 and2010 surface water temperature was recorded daily between08:00 h and 09:00 h at a depth of around 0.5 m. At times of ice cover duration, surface temperatures were assumed tobe 0 u C (Gerten and Adrian 2000). Monthly means werecalculated from the daily water temperatures. Fourremaining gaps, with a maximal length of 1 month, werelinearly interpolated. Subsequently, September to Maymeans were calculated, when the time window of thepelagic phases of C. vicinus and C. kolensis populationsoverlapped. Comparing this water temperature seriesagainst local air temperatures showed the expectedrelationship (data not shown). Unless otherwise stated, allstatistical analyses were performed from September to Mayof the following year, depicted here as yearly means.

Sampling of phytoplankton and zooplankton wascarried out weekly and in winter months every 2 weeks,using standard limnological techniques. A detailed descrip-tion of sampling and sample processing is given inDriescher et al. (1993) and in Gerten and Adrian (2000).To avoid overestimating abundances due to missing valuesand differences in sampling intervals between wintermonths (fortnightly) and the growing season (weekly), wefirst calculated monthly means and linearly interpolatedremaining missing values on the monthly scale. In total, 15gaps in the zooplankton time series were interpolated, alloccurring between 1980 and 1987. For phytoplankton, onlyone missing value had to be interpolated. From these timeseries, unless stated otherwise, means from September untilMay were calculated for each winterspring period.

Abundances of C. kolensis and C. vicinus, the twodominant species during winterspring, refer to adultfemales and males. Before gaining dominance in 1995, C.kolensis was a rare species, resulting in random effects andzero inflation in the first half of the data set. Cyclops

strenuus, also present between September and May, was notconsidered in this study, since its long-term dynamics layoutside the influential time window for C. kolensis .

We tested for abiotic (water temperature) and biotic(prey availability) forces, as well as intraguild predatoryinteractions between C. kolensis and C. vicinus as potentialdrivers for the observed strong shift in their relativeabundance, by using statistical modeling and statisticallybased classification. We considered total algal biomass,cryptophyte biomass (important food source for nauplii[Santer and van de Bosch 1994; Hansen and Santer 1995;Santer and Lampert 1995]), and total rotifer abundances aspotential prey for the two studied cyclopoid copepod

742 Scharfenberger et al.


3/12

species, both known to be omnivorous ( see Introduction). Inthe observed phosphate and nitrogen ranges in Muggelsee,both total algal and cryptophyte biomass responded linearlyto declining nutrients in the lake (data not shown). Beforeanalysis, all variables were standardized by mean centeringand standard deviation scaling. All data manipulation,analysis, and graphical visualization were conducted usingthe R language environment for statistical computing (RCore Team 2012) and associated library extensions. Al-though predation by fish is an important factor in top-downcontrol of zooplankton communities, this was, due to thelack of fish data, not considered in this study.

Statistical analyses Our analytical approach is closelyrelated to the road map proposed by Bestelmeyer et al.(2011) (Fig. 1). The analysis of monotonic and abrupttrends in response and potential driving variables werefollowed by correlation and regression analysis betweenresponse and driving variables. The analysis was thencompleted by investigating the response mechanisms, i.e.,food niche separation and the quantification of thresholds.

Analysis of abrupt and monotonic trends The observedabrupt change over time in C. kolensis abundances couldhave arisen via three different mechanisms: it could havebeen (1) the result of a linearly mediated response to amatching abrupt change in one of the driving variables; (2)a drastic response to a driving variable surpassing athreshold; or (3) the result of a jump between alternativestable states, caused by either an external driver passing abifurcation point or a perturbation of the state variables(Andersen et al. 2009; Scheffer 2009). To distinguishbetween the first two possibilities, we tested all potential

drivers for monotonic and abrupt trends. A differentiationbetween the latter two mechanisms was not possible withthe available data. Regime shift is not a well-defined term.So, here, we use the term for single-system levels in thesense defined by Andersen et al. (2009).

Monotonic trends To access monotonic linear andnonlinear trends for the entire time interval we used theprocedure described in Yue et al. (2002). This iterative

procedure is based on the nonparametric MannKendalltrend test (MKT) and the nonparametric robust TheilSenestimator (TSE) for slopes. The approach of Yue et al.(2002) has the advantage of accounting for the potentialmutual influence between autocorrelation (AR(1)) andlinear trends in time series data. To evaluate the monotonictrends, as well as the TSE, the R package zyp (Bronaughand Werner 2009) was used.

Abrupt trends Before the break-point analysis, allvariables were standardized and significant linear trendswere removed using TSE. For C. kolensis and rotiferabundances, no linear trend components needed to beremoved. Chronological clustering was conducted usingnonparametric univariate regression trees with time as theonly explanatory variable (Borcard et al. 2011). For break-point analyses, we used the R packages mvpart (Death2012) and rpart (Therneau et al. 2012), which follow theclassification and regression trees of Breiman et al. (1984).During model building, the groups are formed such that thesums of squares about the group means are minimized,without changing the chronological order. This leads todetecting changes over time in the mean of the responsevariable. The resulting tree estimates the response variablethrough the group means found for the leaf nodes, i.e.,through a piecewise constant function, which we refer tohere as the break-point model. To avoid overfragmenta-tion, only groups with at least 15 data points were splitfurther. Additionally, the tree size was controlled by usingthe so-called 1 SE rule for selecting the final tree model.

Relation of C. kolensis and C. vicinu s to abiotic and biotic drivers To assess differences in the monotonicrelation of C. kolensis and C. vicinus to the abiotic andbiotic drivers, as well as to each other, we used thenonparametric Spearman correlation coefficient. The rela-tionships of C. vicinus could be analyzed using the entiretime series, whereas for C. kolensis only data from 1995onward were included, due to rare species random effectsand zero inflation for the first half of the data set. Since allvariables exhibited strong linear or nonlinear long-termdynamics, there is on the one hand the problem of detectingspurious correlation, whereas on the other hand rigorousnonlinear detrending might weaken potential signalsbeyond detection, since biological signals tend to havesmall signal-to-noise ratios. Furthermore, common trendsmost likely carry biological meaning. We therefore

performed the analysis for both detrended and trend-affected data. We obtained nonlinear detrended data as theresiduals from a locally weighted regression smoother(LOESS) of polynomial degree 2 against time, with awindow width that spanned 95 % of the data. Detrending of water temperature was performed on log( x) transformeddata. Detrending of all other variables was performed onlog( x+ 1) transformed data. Furthermore, since all signalsfor C. vicinus using detrended data became rather weak, weadditionally analyzed C. vicinus peak abundances in theform of April to May means.

Sequential regression was used to model the relationshipbetween C. vicinus , water temperature, and food variables.

Fig. 1. Flow chart of statistical analyses performed inthis study.

Threshold-driven shifts in two copepods 743


4/12

Sequential regression was chosen to deal with thecomparatively high colinearity between the explanatoryvariables, which makes interpretation of regression coeffi-cients difficult since they are no longer partial. Further-more, weak biological signals are often not detected at alldue to the inflation of variance of the regression parameters(Zuur et al. 2010). For sequential regression, one needs toestablish a hierarchy for the explanatory variables. Suchhierarchy can then be used to decorrelate the variables, i.e.,make them orthogonal through subtraction of commonvariation from the less important variables (Graham 2003;Dormann et al. 2012). On the basis of biological reasoning,we established the following hierarchy: cryptophyte bio-mass (bottleneck for nauplii development), total algalbiomass, rotifer abundances (omnivorous feeding ecologyof C. vicinus and C. kolensis ), and water temperature. Toassess the robustness of our results, we exchangedcryptophyte biomass with total algae biomass in subse-quent analyses, and tested for water temperature leadingthe hierarchy. For this analysis only C. vicinus peakabundances (April to May means) were considered. Tominimize the influence of extreme values, we performed alogarithmic transformation to all the variables beforeanalysis. If the preliminary single variable generalizedadditive models (GAMs) suggested nonlinear relationships,we included higher-order terms of the correspondingvariable in the model. GAMs were calculated using the Rpackage mgcv (Wood 2012).

Violation of independence was accounted for by anautoregressive model for the residuals, and violation of homoscedasticity was corrected by an exponential variancestructure for the residuals within the framework of generalized least-square modeling (R package nlme[Pinheiro et al. 2012]). Model selection was based on theimportance and significance of the regression coefficients,as well as on ANOVA comparison and the Akaikeinformation criterion. If necessary, the hierarchical vari-ables were updated accordingly.

Food niche separation and quantification of thresholds High discriminative power for the relative dominancestatus of C. kolensis and C. vicinus in one of the drivingvariables is an indicator for a niche apportionment betweenthe two species with regard to this variable. Furthermore,the decision border between the different states quantifiesrespective thresholds. We therefore calculated a binaryresponse variable, where a $ 50% share of C. kolensis on

the overall abundance of C. kolensis and C. vicinus wascoded as 1 and 0 otherwise. This new response variable wasused in a classifier based on a logistic generalized linearmodel and a receiver operating characteristic (ROC). Tooptimize the classifier the sum of sensitivity and specificitywas maximized. Whereas sensitivity measures the proba-bility of the classifier detecting a . 50% share of C. kolensis ,when it is really the case, specificity measures theprobability of detecting a , 50% share when this is trulythe case. The point of maximization (the sum of sensitivityand specificity) was found as the minimal ROC distance.The cutoff value connected with this maximum was takenas the decision border of the classifier, and with that as the

threshold value. The area under the curve (AUC) of theROC is used to access the general ability of the model todiscriminate between the binary states of the responsevariable. The AUC takes values between 0.5 and 1, where avalue of 0.5 equals pure guessing and 1 is perfectdiscrimination. For evaluating the ROC, the R packageDiagnosisMed (Brasil 2010) was used.

Quantifying thresholds for intraguild competition and predation To find the threshold at which C. kolensis wasable to overcome the competition and predation of C.vicinus, by degree we varied the cutoff value for the definitionof high and low C. kolensis abundance. Low abundance wasthen coded as 0, high as 1. For each coding scheme, anoptimal classifier was calculated as described above, and theoptimal thresholds for the dependent and independentvariables were again found on the basis of the AUC.

Results

Copepod long-term trends and phenology The cyclopoidcopepod winter community in the Muggelsee was domi-nated by C. kolensis and C. vicinus . Comparing the resultsfrom trend and regression tree analyses revealed thatwhereas C. vicinus abruptly decreased in abundance in 1988and 1993, its overall dynamics were dominated by asignificant gradual decline over the entire study period(Figs. 2a, 3a; Table 1). By contrast, C. kolensis underwentone abrupt abundance increase in 1995 (Figs. 2a, 3b).Between 1980 and 1995before this abrupt change C.kolensis comprised, on average, 2.5 % (0% , 12% , and 2 % inautumn, winter, and spring, respectively) of the cyclopoidcopepod winter community, and became by far the mostdominant species in 1995 to 2010 (6 % , 86% , and 57 % inautumn, winter, and spring, respectively; Fig. 2a).

C. kolensis developed its pelagic population betweenSeptember and May (Fig. 2c). The phenology of C. kolensisremained unchanged over the entire investigation period. C.vicinus was usually present throughout the year, withpopulationpeaks in spring and autumn. A 1 month phenologyshift toward later in autumn was observed for C. vicinus in thesecond investigation period (1995 to 2010; Fig. 2b).

Trends in prey availability and water temperature Totalalgal and cryptophyte biomass declined abruptly in 1991and 1993, dropping by 46 % and 31 % , respectively, from the5 yr period before the abrupt change to the following one

(Fig. 3d,e). Rotifers exhibited an abrupt decline in 1993,where they lost 40 % in mean abundance (5 yr comparisonas above), but increased again in 2002 (Fig. 3f). Rotiferabundances never decreased below 245 individuals (ind.)L2 1 during the entire study period (Fig. 4e,f). In additionto the abrupt changes, significant monotonic trends forwater temperature, as well as cryptophyte and total algalbiomass, were found. Water temperature, with an averagerise of 0.9 u C over the entire study period, was the onlyvariable with an increasing trend (Fig. 3c; Table 1).

Abiotic and biotic drivers Following the Spearmancorrelation coefficient, we found a significant relation



5/12

between C. kolensis and C. vicinus abundances as well asrotifer abundances (Table 2). Spearman correlation coeffi-cients between key drivers and C. vicinus using trend-affected data were high for total algal biomass, cryptophytebiomass, and water temperature. We found no strongrelationships between C. vicinus abundance and detrendedtime series apart from a moderate association withcryptophytes (Table 2). However, the importance of

cryptophyte biomass and water temperature for C. vicinusabundances, as well as the direction of the relationship, wasfurther confirmed in the sequential regression analysis. Forboth detrended and trend-affected data, cryptophytebiomass was significantly positively related to C. vicinusabundance, irrespective of whether cryptophyte biomass ortotal algal biomass led the sequential regression hierarchy.This result was also robust against including a linear trendon top of the hierarchy. If water temperature led thehierarchy, cryptophyte biomass and water temperaturewere both significant for trend-affected data, and againrobust against explicit inclusion of a linear trend into themodel. However, in the case of detrended data, only watertemperature was significant when leading the hierarchy. Asa primary abiotic driver, water temperature did not accountfor . 50% C. kolensis dominance (Table 3). This wasconfirmed by the low power of the logistic regression-basedclassifier to distinguish between C. kolensis or C. vicinusdominance, on the basis of water temperature (Table 3).

Niche separation for high abundance probability of C. kolensis We found a clear segmentation for high vs.

Fig. 2. Copepod long-term trends and phenology. (a) September to May mean abundances of Cyclops kolensis (open circles) andCyclops vicinus (closed circles). (b) Phenology of C. vicinus and (c) C. kolensis , depicted as monthly mean abundances for the periods 1980to 1994 (open circles) and 1995 to 2010 (closed circles). Mean abundances for each period were standardized with the maximal abundanceof the respective species and period.

Table 1. Trend of key drivers and response variables. The

analysis follows the iterative procedure suggested by Yue et al.(2002). Estimated slopes of the trends and their correspondingconfidence intervals (CI) are based on the nonparametric robustTSE. Significant levels ( p-value) of the monotonic trends arebased on the nonparametric MKT test.

Slope CI p value

Water temperature ( u C) 0.03 2 0.01 to 0.06 0.04Total algal biomass (mg L 2 1 ) 2 0.05 2 0.09 to 2 0.03 0.03Cryptophytes (mg L 2 1 ) 2 0.09 2 0.12 to 2 0.06 0.00Rotifers (ind. L 2 1 ) 2 0.02 2 0.05 to 0.02 0.46Cyclops vicinus (ind. L 2 1 ) 2 0.09 2 0.12 to 2 0.06 0.00Cyclops kolensis (ind. L 2 1 ) 0.05 0.01 to 0.09 0.16



6/12

low abundances of C. vicinus and C. kolensis for total algaland cryptophyte biomass (Fig. 4a,b,c,d). C. kolensisoccupied the low food niche, whereas C. vicinus wasdominant when herbivorous food availability was high.The threshold partitioning the two food niches lay at 6.2 mgL 2 1 for total algal mass and at 0.6 mg L 2 1 for cryptophytebiomass (Table 3). With only a few exceptions, boththresholds were surpassed from 1994 onward. A foodniche separation was not obvious for rotifer prey, whichwere present in abundances of . 245 ind. L 2 1 throughoutthe period of investigation (Fig. 4e,f).

Thresholds for C. vicinus competitive and predationsuperiority Inspecting the direct relationship between C.kolensis and C. vicinus abundance, a threshold of 0.7 ind. L 2 1 (AUC 5 0.9) of C. vicinus separated highfrom low abundance probability of C. kolensis (Fig. 5).

Discussion

Abrupt changes in ecosystem responses to climate andenvironmental change can have potentially large effects onecosystem functioning. However, the detailed nature of

Fig. 3. Trends in prey availability and water temperature. Abrupt and linear trend dynamics in yearly means. Solid lines depictrobust regression models combining the largest abrupt changes in the mean from regression tree analysis for each variable, together withsignificant trends from monotonic trend analysis if these exist for the respective variable. (a) Cyclops vicinus , (d) total algal biomass, and(e) cryptophyte biomass were dominated by both abrupt changes and gradual linear trends in the mean. (b) Cyclops kolensis and (f)rotifers were purely break-point driven, whereas (c) water temperature was solely marked by a linear trend. Data were standardized withthe maximum value of the respective variables.



7/12

change is difficult to predict and data-driven proof of model outputs is rare (Carpenter et al. 2011). Weinvestigated mechanisms behind the abrupt increase of anomnivorous copepod species ( C. kolensis ), changing from arare to the dominant species, while coexisting C. vicinusconcurrently decreased in abundance. These two species arelinked in a consumer-structured IGP food web such thattheir coexistence, sharing the same resources, is more likely

under intermediate productivity conditions in their lakehabitat. In our model system, C. kolensis occupies theposition of the consumer with lower prey requirements,whereas C. vicinus acts as its predator. This particularposition of C. kolensis enabled it to profit from theimproved trophic state and subsequent gradual decline of C. vicinus , releasing it from predation on its larval stagesand competing for commonly shared algal prey, while not

Fig. 4. Food niche separation for high abundance probability of Cyclops kolensis. C. kolensis (open circles) and Cyclops vicinus(closed circles) abundance for the periods before 1980 to 1994 (left), and after 1995 to 2010 (right) the shift in C. kolensis abundance withrespect to (a, b) total algal biomass, (c, d) cryptophytes, and (e, f) rotifers. Dashed lines depict the thresholds estimated by theclassification summarized in Table 3.



8/12

being negatively affected by the decline in prey availabilityitself. Our analyses shed light on the mechanisms under-lying the abrupt change in C. kolensis abundance in thecontext of the IGP hypothesis, and we discuss the nature of the driving forces in the context of regime shift theory.

Gradual decline in C. vicinus The decline of C. vicinusclosely mirrored the decline in total algal and cryptophytebiomass (Fig. 3). The significant and robust relationshipbetween C. vicinus and cryptophytes, confirmed by bothcoinciding break points and regression analysis, highlightsthem as the main driver, followed by total algal biomass.These findings agree with results from grazing experimentswhere algae constituted 66 % of the daily ingested dry massof C. vicinus (Adrian and Frost 1992). It is also known thatC. vicinus nauplii rely on phytoflagellates (cryptophytes),with limiting concentrations being between 0.2 and 0.5 mgC L 2 1 (Santer and van de Bosch 1994; Hansen and Santer1995). Assuming a conversion factor between cryptophytefresh weight and carbon of 0.2 pg C mm2 3 (Rocha andDuncan 1985), cryptophytes were most likely close to

limiting concentrations for C. vicinus nauplii over the entirestudy period, but particularly in the second period (from1980 to 1994 and 1995 to 2010, cryptophyte carbondeclined on average from 0.2 to 0.1 mg C L 2 1 ). Sincenauplii are considered the bottleneck in a copepod lifecycle, an insufficient supply of cryptophytes has a largeeffect on their ability to develop into adults (Santer and vande Bosch 1994; Hansen and Santer 1995; Santer andLampert 1995). Furthermore, the high herbivorous food

demand of C. vicinus is also reflected by its presence, andoften dominance, in eutrophic lakes (Hansen and Jeppesen1992; Santer and Lampert 1995; Adrian 1997) aided by itsdisproportional ability to profit from high prey availabilitythrough shorter development times (Hansen and Santer1995).

On the basis of the break-point component in the long-term dynamics of C. vicinus and cryptophytes, therelationship between them matches the driver thresholdscenario hypothesized in regime shift theory (Andersenet al. 2009). In this regime shift scenario, an abrupt changein a driving force (here cryptophytes) causes an equallysudden change in a response variable ( C. vicinus in ourcase), such that the thresholds are only present in the timeseries (Fig. 6).

Abrupt increase in C. kolensis The only significantdirect connection between C. kolensis and any of its preywas found for rotifer abundance. The strong negativecorrelation most likely arose from a top-down effect of C.kolensis on its prey, marking rotifers as an important food

source. However, C. kolensis abundance increased at a time

Table 2. Monotonic relationships between Cyclops kolensis or Cyclops vicinus and abiotic and biotic key drivers. The analysis wasbased on Spearmans rank correlation coefficient. Spearmanss rho is given first, followed by the respective p-value in brackets. Thecoefficients for C. kolensis were based on data ranging from 1995 to 2010, whereas those for C. vicinus were based on data from 1980 to 2010.

C. kolensis* C. kolensis** C. vicinus* C. vicinus** C. vicinus***

Water temperature 0.16(0.54) 0.32(0.22) 2 0.4(0.02) 2 0.26(0.16) 2 0.19(0.32)Total algal biomass 2 0.14(0.62) 2 0.18(0.51) 0.63(0) 0.16(0.39) 0.05(0.77)Cryptophytes 2 0.23(0.39) 0.06(0.84) 0.72(0) 0.13(0.5) 0.26(0.17)Rotifers 2 0.36(0.17) 2 0.63(0.01) 0.1(0.6) 2 0.04(0.83) 0.1(0.61)C. vicinus 0.50(0.05) 0.34(0.20)

* Results from trend-affected data; ** results from nonlinear detrended data; *** results from April to May means of nonlinear detrended data.

Table 3. Classification analysis on the basis of abiotic andbiotic key drivers for Cyclops kolensis dominance ( $ 50% ) relativeto Cyclops vicinus . The values of the classifiers decision border aregiven, which can be interpreted as the threshold for a 50 %

dominance of C. kolensis . The area under the curve (AUC) and itscorresponding confidence intervals (CI) give a measure for thegoodness of fit of the classification with respect to each driver.Whereas an AUC of 0.5 equals pure guessing, an AUC of 1 equalsa perfect classification. High discriminative power in one of the driving variables is an indicator for a niche apportionmentbetween the two copepods with regard to this variable.

Threshold AUC CI

Water temperature 7.02 u C 0.67 0.46 to 0.88Total algal biomass 6.15 mg L 2 1 0.95 0.88 to 1.02Cryptophytes 0.57 mg L 2 1 0.94 0.86 to 1.03Rotifers 419.91 ind. L 2 1 0.65 0.44 to 0.85

Fig. 5. Thresholds for intraguild competition and predationsuperiority. Dashed lines correspond to thresholds found by theoptimal classifier to separate high and low Cyclops kolensisabundance (horizontal line) on the basis of the abundance of Cyclops vicinus (vertical line). Closed circles correspond to correctclassification results (TP 5 true positives, TN 5 true negatives),open circles to false ones (FP 5 false positive, FN 5 falsenegative) on the basis of the estimated thresholds.



9/12

Fig. 6. Driver threshold and state threshold regime shift scenarios. Analyses on the basis of fig. 1 in Andersen et al. 2009. (a, b, c)The driver threshold scenario refers to the existence of a threshold in the driver (cryptophytes), which is linearly mediated to a response(Cyclops vicinus ). (a) The regime shift in cryptophyte biomass in 1993 was immediately followed by (b) a regime shift in C. vicinusabundances, resulting in (c) a linear driver (cryptophytes) response ( C. vicinus ) scatter plot. The solid lines in driver and response variables(panels a and b) stress the break-point components in the dynamics of the two variables. (d, e, f) The state threshold scenario refers to theexistence of a threshold in the response ( Cyclops kolensis ), but not in the driver ( C. vicinus ). (e) C. kolensis abundances underwent a regimeshift (d) after C. vicinus crossed a threshold, such that an abrupt change appears in the time seriesof C. kolensis and (f) a threshold appears inthe scatter plot between the two variables. The line in panel d solely depicts the linear trend component of C. vicinus abundance, since onlythis part of the dynamics is relevant in this regime shift scenario.



10/12

of general decrease in all food categories. Thus, the specieswas unlikely to have been affected by deprivation of prey.Rather, C. kolensis indirectly benefited from algal preydecline, due to the decline in C. vicinus releasing it fromcompetition. The opposing effects of herbivorous fooddecrease on the two species are explained by apportionmentof the food niches. Whereas C. vicinus was dominant underplentiful algal and cryptophyte biomass, C. kolensis seemedparticularly able to thrive under low food conditions, tooccupy the low food niche (Fig. 4). These findings are in linewith lower algal prey ingestion rates for C. kolensis (5.5 mgdm ind. 2 1 d2 1 ) than those observed for C. vicinus (7.3 mg dmind. 2 1 d2 1 ) when offered the same natural lake planktonmixture of algal and invertebrate prey (Adrian and Frost1992). Ingestion rates of invertebrates were within the samerange of 4.4 and 4.2 mg dm ind. 2 1 d2 1 for both species(Adrian and Frost 1992). This is consistent with the lack of any evidence, from either correlation or regression analysis,or food niche partitioning, that the two species might havebeen limited by carnivorous prey (Tables 2, 3; Fig. 4).

The food niche separation marks C. kolensis as thesuperior exploitative competitor, and explains why it wasunaffected by the general decline in algal biomass, much inline with the IGP hypothesis. This is supported by the factthat the larger C. vicinus exhibited higher ingestion rates foralgal prey than C. kolensis when offering the same preyassemblages (Adrian and Frost 1992). Additionally, thestrong relation between C. kolensis and rotifer abundances(Table 2) underlines slight differences in the diet of C.kolensis when compared with C. vicinus . From IGP modelsembedded in a richer food web, the existence of suchalternative food resources is known to relax the assumptionof the consumers superiority in exploitative competitionand still enables coexistence at intermediate productivity(Holt and Huxel 2007). However, this does not necessarilyaccount for the abrupt nature of C. kolensis success.Rather, the dynamics governing its interaction in the IGPfood web must be highly sensitive to a critical abundancethreshold, matching the state threshold scenario proposedby Andersen et al. 2009 (Fig. 6). In this scenario theresponse variable abruptly shifts (here C. kolensis ) after thedriver crosses a threshold (here C. vicinus ), putting thefocus on the monotonic trend component in C. vicinus long-term dynamics (Fig. 3). This is in line with the highdiscriminative power of the logistic classifier using C.vicinus abundance to distinguish between high and lowC. kolensis presence (Fig. 5). The critical threshold of

C. vicinus density that enables C. kolensis to overcome theinhibiting effects of competition for prey and predationpressure in our system was 0.7 C. vicinus L 2 1 .

The shift in C. kolensis abundance could have also beencaused by a transition between alternative stable states,since IGP theory predicts the existence of such states for aconsumer with a superior ability to exploit competition atintermediate productivity levels (Holt and Polis 1997;Mylius et al. 2001; Verdy and Amarasekare 2010).However, we were not able to differentiate the statethreshold further into an explicit transition betweenalternative stable states, or driver-state hysteresis scenarioas coined by Andersen et al. 2009.

Our view of C. vicinus being the main predator of C.kolensis juveniles, although both copepod species prey onnauplii (Adrian 1991; Krylov et al. 1996), stems from thephenology of both species. The abundance peaks of adult,reproducing C. kolensis occurred in March, whereas onaverage those of C. vicinus were in April (Fig. 2b,c),coinciding with the new batch of C. kolensis offspring,which are highly susceptible to predation. The phenologyshift of the second population peak of C. vicinus fromSeptember to October most likely benefited C. kolensis(Fig. 2). The 1 month delay in maximum C. vicinuspopulation might as well have diminished predation onemerging C. kolensis copepodites in September. Within amonth late copepodites easily develop into the adult stage,which are less susceptible to predation by adult C. vicinus ,giving the C. kolensis population a head start.

Water temperature, although an important driver of copepod life cycles and phenology (Adrian et al. 1999;Gerten and Adrian 2002; Seebens et al. 2008), correlatedweakly with C. kolensis abundance, and it was not possibleto adequately distinguish between high and low C. kolensisabundance on the basis of water temperature. Thisexcluded water temperature as a major direct driver of the dominance shift in C. kolensis . Nevertheless, C. kolensismight have indirectly benefited from warmer watertemperatures, not only due to the earlier spring phyto-plankton bloom (Gerten and Adrian 2000), giving itsnauplii an earlier resource, but also temperature-drivenshorter developmental times (Hansen and Jeppersen 1992).Furthermore, water temperature might have played anoth-er indirect role through its negative relationship to C.vicinus abundance (Table 2), possibly further promoting itsdecline. The negative relationship between water tempera-ture and C. vicinus is not easily explainable, since bothcopepod species should profit in terms of shorter genera-tion times due to higher water temperatures.

To conclude, we have presented an example of anintriguing match between observational long-term recordsand ecological food web and regime shift theory. The twocoexisting copepod species are linked in an IGP food web,with C. vicinus as the predator and C. kolensis as theconsumer. The latter proved superior in exploitativecompetition through its ability to occupy a low food nicheat times when the lakes trophic state had improved(Kohler et al. 2005; Huber et al. 2008). Consistent withour observed results, theory predicts that coexistence insuch a food web is only possible at intermediate levels of

resource productivity, in combination with the consumerbeing better at exploiting the resource (Polis et al. 1989;Holt and Polis 1997) or has additional resources (Holt andHuxel 2007). Our results were also in line with results fromdeterministic models (Mylius et al. 2001; Verdy andAmarasekare 2010), as well as laboratory experiments(Morin 1999; Diehl and Feiel 2000) investigating the effectof food enrichment within IGP food webs. As for thenature of the complex interaction between driving forcesand copepod regime shifts, we were able to distinguishbetween true threshold-driven dynamics underlying theabrupt increase in C. kolensis and a mere, linearly mediateddecline in C. vicinus , responding to a matching abrupt



11/12

change in a driving variable (Fig. 6). The most likelyinitiator for the abrupt increase of C. kolensis in 1995 wasthe monotonic decline of C. vicinus crossing a criticalthreshold. Once C. vicinus declined to a critical abundanceof 0.7 ind. L 2 1 , C. kolensis was able to thrive in the lowfood niche that had already opened up between 1991 (totalalgal mass) and 1993 (cryptophytes).

Overall, we have presented a successful case of howobserved changes in nature on the basis of long-termobservational data match results from ecosystem theory. Incombination, they provide powerful tools in understandingthe complex interactions in ecosystems and help to generatehypotheses for a better understanding of the remainingunexplained variability. Since modeling is a prerequisite toproject responses, e.g., toward global environmentalchanges, testing model outputs with long-term observa-tional data or experimental studies is desirable. We havepresented a plausible case study on how threshold-drivenregime shifts at single system levels in a lake have the

potential to change relative species composition withinplankton communities.

AcknowledgmentsWe thank the staff of the Leibniz-Institute of Freshwater

Ecology and Inland Fisheries for sampling and technical support.We acknowledge Judith Dannowski, Maria Gonzalez, Fred Jopp,Danijela Markovic, William Rizk, Jan-Hendrik Schleimer, SilkeSchmidt, Mike Vanni, and Carola Wagner for their fruitful andinspiring discussions related to this topic. We are also grateful tothe two anonymous referees for their valuable comments on themanuscript. Funding was provided by the EU FP-7 Theme 6project Adaptive strategies to mitigate the impacts of climatechange on European freshwater ecosystems (REFRESH, con-tract: 244121). A. M. was supported by a Ph.D. fellowship from

the Mission Office of the Egyptian Government.

ReferencesADRIAN , R. 1991. The feeding behaviour of Cyclops kolensis and

C. vicinus (Crustacea, Copepoda). Verh. Int. Ver. Limnol. 2 4:28522863.

. 1997. Calanoidcyclopoid interactions: Evidence from an11-year field study in a eutrophic lake. Freshw. Biol. 38:315325 , doi: 10.1046/j.1365-2427.1997.00215.x

, AND OTHERS . 2009. Lakes as sentinels of climate change.Limnol. Oceanogr 54: 22832297 , doi:10.4319/lo.2009.54.6_ part_2.2283

, AND R. D ENEKE . 1996. Possible impact of mild winters on

zooplanktonsuccession in eutrophic lakes of the AtlanticEuropeanarea. Freshw. Biol. 36: 757770 , doi: 10.1046/j.1365-2427.1996.00126.x

, AND T. M. F ROST . 1992. Comparative feeding ecology of Tropocyclops prasinus mexicanus (Copepoda, Cyclopoida). J.Plankton Res. 14: 13691382 , doi: 10.1093/plankt/14.10.1369

, N. W ALZ , T. H INTZE , S. H OEG , AND R. R USCHE . 1999.Effects of ice duration on plankton succession during springin a shallow polymictic lake. Freshw. Biol. 41: 621634 ,doi: 10.1046/j.1365-2427.1999.00411.x

, S. W ILHELM , AND D. G ERTEN . 2006. Life-history traits of lake plankton species may govern their phenological responseto climate warming. Glob. Change Biol. 12: 652661 ,doi: 10.1111/j.1365-2486.2006.01125.x

ANDERSEN , T., J. C ARSTENSEN , E. H ERNA NDES -G ARCIA, AND C. M.D UARTE . 2009. Ecological thresholds and regime shifts:Approaches to identification. Trends Ecol. Evol. 24: 4957doi: 10.1016/j.tree.2008.07.014

BESTELMEYER , B. T., AND OTHERS . 2011. Analysis of abrupttransitions in ecological systems. Ecosphere 2: art129doi: 10.1890/ES11-00216.1

BORCARD , D. , F. G ILLET , AND P. L EGENDRE . 2011. Numericalecology with R, 1st ed. Springer.

BRASIL , P. 2010. DiagnosisMed: Diagnostic test accuracy evalu-ation for medical professionals. R package version 0.2.3[accessed 06 July 2012]. Available from http://CRAN.R-project.org/package 5 DiagnosisMed

BREIMAN , L., J. H. F RIEDMAN , C. J. S TONE , AND R. A. O LSHEN1984. Classification and regression trees, new ed. CRC Press.

BRONAUGH , D., AND WERNER , A., FOR THE PACIFIC CLIMATEIMPACTS CONSORTIUM . 2009. zyp: Zhang + Yue-Pilon trendspackage. R package version 0.9-1 [accessed 06 July 2012].Available from http://CRAN.R-project.org/package 5 zyp

C ARPENTER , S. R., AND OTHERS . 2011. Early warnings of regimeshifts: A whole-ecosystem experiment. Science 332: 10791082 , doi: 10.1126/science.1203672

D AKOS , V., M. S CHEFFER , E. H. VAN NES , V. B ROVKIN , V.PETOUKHOV , AND H. H ELD . 2008. Slowing down as an earlywarning signal for abrupt climate change. Proc. Natl. Acad.Sci. USA 105: 1430814312 , doi: 10.1073/pnas.0802430105

D EATH , G. 2012. mvpart: Multivariate partitioning. R packageversion 1.5-0 [accessed 06 July 2012]. Available from http://CRAN.R-project.org/package 5 mvpart

D IEHL , S., AND M. F EIEL . 2000. Effects of enrichment on three-level food chains with omnivory. Am. Nat. 155: 200218doi: 10.1086/303319

D ORMANN , C. F., AND OTHERS . 2012. Collinearity: A review of methodsto deal with it anda simulation studyevaluating their performance.Ecography 35: nono , doi: 10.1111/j.1600-0587.2012.07348.x

D RIESCHER , E., H. B EHRENDT , G . S CHELLENBERGER , AND RSTELLMACHER . 1993. Lake Muggelsee and its environment natural conditions and anthropogenic impacts. Int. Rev. ges.Hydrobiol. 78: 327343 , doi: 10.1002/iroh.19930780303

E LLIOT , J. A., AND L. M AY . 2008. The sensitivity of phytoplanktonin Loch Leven (U.K.) to changes in nutrient load and watertemperature. Freshw. Biol. 53: 3241.

G ERTEN , D., AND R. A DRIAN . 2000. Climate-driven changes inspring plankton dynamics and the sensitivity of shallowpolymictic lakes to the North Atlantic Oscillation. Limnol.Oceanogr. 45: 10581066 , doi: 10.4319/lo.2000.45.5.1058

, AND . 2002. Species-specific changes in thephenology and peak abundance of freshwater copepods inresponse to warm summers. Freshw. Biol. 47: 21632173doi: 10.1046/j.1365-2427.2002.00970.x

G RAHAM , M. H. 2003. Confronting multicollinearity in ecologicalmultiple regression. Ecology 84: 28092815 , doi: 10.1890/02-3114

G YLLSTRO M , M., AND OTHERS . 2005. The role of climate in shapingzooplankton communities of shallow lakes. Limnol. Ocean-ogr. 50: 20082021 , doi: 10.4319/lo.2005.50.6.2008

H ANSEN , A.-M. 1996. Variable life history of a cyclopoid copepod:The role of food availability. Hydrobiologia 320: 223227doi: 10.1007/BF00016823

, AND E. JEPPESEN . 1992. Life cycle of Cyclops vicinus inrelation to food availability, predation,diapause andtemperature.J. Plankton Res. 14: 591605 , doi: 10.1093/plankt/14.4.591

, AND B. SANTER . 1995. The influence of food resource onthe development, survival and reproduction of the twocyclopoid copepods: Cyclops vicinus and Mesocyclops leuck-arti. J. Plankton Res. 17: 631646 , doi: 10.1093/plankt/17.3.631

http://dx.doi.org/10.1046%2Fj.1365-2427.1997.00215.xhttp://dx.doi.org/10.1046%2Fj.1365-2427.1997.00215.xhttp://dx.doi.org/10.4319%2Flo.2009.54.6_part_2.2283http://dx.doi.org/10.4319%2Flo.2009.54.6_part_2.2283http://dx.doi.org/10.4319%2Flo.2009.54.6_part_2.2283http://dx.doi.org/10.1046%2Fj.1365-2427.1996.00126.xhttp://dx.doi.org/10.1046%2Fj.1365-2427.1996.00126.xhttp://dx.doi.org/10.1046%2Fj.1365-2427.1996.00126.xhttp://dx.doi.org/10.1093%2Fplankt%2F14.10.1369http://dx.doi.org/10.1093%2Fplankt%2F14.10.1369http://dx.doi.org/10.1046%2Fj.1365-2427.1999.00411.xhttp://dx.doi.org/10.1046%2Fj.1365-2427.1999.00411.xhttp://dx.doi.org/10.1046%2Fj.1365-2427.1999.00411.xhttp://dx.doi.org/10.1111%2Fj.1365-2486.2006.01125.xhttp://dx.doi.org/10.1111%2Fj.1365-2486.2006.01125.xhttp://dx.doi.org/10.1111%2Fj.1365-2486.2006.01125.xhttp://dx.doi.org/10.1016%2Fj.tree.2008.07.014http://dx.doi.org/10.1016%2Fj.tree.2008.07.014http://dx.doi.org/10.1890%2FES11-00216.1http://dx.doi.org/10.1890%2FES11-00216.1http://dx.doi.org/10.1126%2Fscience.1203672http://dx.doi.org/10.1126%2Fscience.1203672http://dx.doi.org/10.1073%2Fpnas.0802430105http://dx.doi.org/10.1073%2Fpnas.0802430105http://dx.doi.org/10.1086%2F303319http://dx.doi.org/10.1086%2F303319http://dx.doi.org/10.1111%2Fj.1600-0587.2012.07348.xhttp://dx.doi.org/10.1111%2Fj.1600-0587.2012.07348.xhttp://dx.doi.org/10.1002%2Firoh.19930780303http://dx.doi.org/10.1002%2Firoh.19930780303http://dx.doi.org/10.4319%2Flo.2000.45.5.1058http://dx.doi.org/10.4319%2Flo.2000.45.5.1058http://dx.doi.org/10.1046%2Fj.1365-2427.2002.00970.xhttp://dx.doi.org/10.1046%2Fj.1365-2427.2002.00970.xhttp://dx.doi.org/10.1890%2F02-3114http://dx.doi.org/10.1890%2F02-3114http://dx.doi.org/10.4319%2Flo.2005.50.6.2008http://dx.doi.org/10.4319%2Flo.2005.50.6.2008http://dx.doi.org/10.1007%2FBF00016823http://dx.doi.org/10.1007%2FBF00016823http://dx.doi.org/10.1093%2Fplankt%2F14.4.591http://dx.doi.org/10.1093%2Fplankt%2F14.4.591http://dx.doi.org/10.1093%2Fplankt%2F17.3.631http://dx.doi.org/10.1093%2Fplankt%2F17.3.631http://dx.doi.org/10.1093%2Fplankt%2F17.3.631http://dx.doi.org/10.1093%2Fplankt%2F17.3.631http://dx.doi.org/10.1093%2Fplankt%2F17.3.631http://dx.doi.org/10.1093%2Fplankt%2F14.4.591http://dx.doi.org/10.1007%2FBF00016823http://dx.doi.org/10.4319%2Flo.2005.50.6.2008http://dx.doi.org/10.1890%2F02-3114http://dx.doi.org/10.1046%2Fj.1365-2427.2002.00970.xhttp://dx.doi.org/10.4319%2Flo.2000.45.5.1058http://dx.doi.org/10.1002%2Firoh.19930780303http://dx.doi.org/10.1111%2Fj.1600-0587.2012.07348.xhttp://dx.doi.org/10.1086%2F303319http://dx.doi.org/10.1073%2Fpnas.0802430105http://dx.doi.org/10.1126%2Fscience.1203672http://dx.doi.org/10.1890%2FES11-00216.1http://dx.doi.org/10.1016%2Fj.tree.2008.07.014http://dx.doi.org/10.1111%2Fj.1365-2486.2006.01125.xhttp://dx.doi.org/10.1111%2Fj.1365-2486.2006.01125.xhttp://dx.doi.org/10.1046%2Fj.1365-2427.1999.00411.xhttp://dx.doi.org/10.1046%2Fj.1365-2427.1999.00411.xhttp://dx.doi.org/10.1093%2Fplankt%2F14.10.1369http://dx.doi.org/10.1046%2Fj.1365-2427.1996.00126.xhttp://dx.doi.org/10.1046%2Fj.1365-2427.1996.00126.xhttp://dx.doi.org/10.4319%2Flo.2009.54.6_part_2.2283http://dx.doi.org/10.4319%2Flo.2009.54.6_part_2.2283http://dx.doi.org/10.1046%2Fj.1365-2427.1997.00215.x


12/12

H OLT , R. D., AND G. R. H UXEL . 2007. Alternative prey and thedynamics of intraguild predation: Theoretical perspectives.Ecology 88: 27062712 , doi: 10.1890/06-1525.1

, AND G. A. P OLIS . 1997. A theoretical framework forintraguild predation. Am. Nat. 149: 745764 , doi: 10.1086/286018

HOPP

, U., AND

G. MAIER

. 2005. Survival and development of fivespecies of cyclopoid copepods in relation to food supply:Experiments with algal food in a flow-through system.Freshw. Biol. 50: 14541463 , doi : 10.1111/j.1365-2427.2005.01417.x

H UBER , V., R. A DRIAN , AND D. G ERTEN . 2008. Phytoplanktonresponse to climate warming modified by trophic state.Limnol. Oceanogr. 53: 113, doi: 10.4319/lo.2008.53.1.0001

K O HLER , J., S. H ILT , R. A DRIAN , A. N ICKLISCH , H. P. K OZERSKI ,AND N. W ALZ . 2005. Long-term response of a shallow,moderately flushed lake to reduced external phosphorus andnitrogen loading. Freshw. Biol. 50: 16391650 , doi: 10.1111/ j.1365-2427.2005.01430.x

K RYLOV , P. I., V. R. A LEKSEEV , AND O. A. F RENKEL . 1996. Feedingand digestive activity of cyclopoid copepods in activediapause. Hydrobiologia 320: 7179, doi: 10.1007/BF00016806

M AIER , G. 1998. Differential success of cyclopoid copepods in thepelagic zone of eutrophic lakes. J. Mar. Syst. 15: 135138 ,doi: 10.1016/S0924-7963(97)00064-X

M OOIJ , W. M., AND OTHERS . 2005. The impact of climate changeon lakes in the Netherlands: A review. Aquat. Ecol. 39:381400 , doi: 10.1007/s10452-005-9008-0

M ORIN , P. 1999. Productivity, intraguild predation, and popula-tion dynamics in experimental food webs. Ecology 80:752760 , doi: 10.1890/0012-9658(1999)080[0752:PIPAPD]2.0.CO;2

M YLIUS , S. D., K. K LUMPERS , A. M. DE ROOS , AND L. P ERSSON .2001. Impact of intraguild predation and stage structure onsimple communities along a productivity gradient. Am. Nat.158: 259276 , doi: 10.1086/321321

P INHEIRO , J . ,D.B ATES , S . D EBR OY , D . S ARKAR , THE R D EVELOPMENTCORE TEAM . 2012. nlme: Linear and nonlinear mixed effectsmodels. R package version 3.1-103 [accessed 06 July 2012].Available from http://CRAN.R-project.org/package 5 nlme

POLIS , G. A., C. A. M YERS , AND R. D. H OLT . 1989. The ecologyand evolution of intraguild predation: Potential competitorsthat eat each other. Annu. Rev. Ecol. Syst. 20: 297330 ,doi: 10.1146/annurev.es.20.110189.001501

R C ORE TEAM . 2012. R: A language and environment forstatistical computing. R Foundation for Statistical Comput-ing, Vienna, Austria [accessed 06 July 2012]. Available fromhttp://www.R-project.org/

R IVIER , I. C. 1996. Ecology of diapausing copepodids of Cyclopskolensis Lill. in reservoirs of the Upper Volga. Hydrobiologia320: 235241 , doi: 10.1007/BF00016825

R OCHA , O., AND A. D UNCAN . 1985. The relationship between cellcarbon and cell volume in freshwater algal species used inzooplanktonic studies. J. Plankton Res. 7: 279294 , doi: 10.1093/plankt/7.2.279

SANTER , B., AND W. L AMPERT . 1995. Summer diapause incyclopoid copepods: Adaptive response to a food bottleneck?J. Anim. Ecol. 64: 600613 , doi: 10.2307/5803

, AND F. VAN DE BOSCH . 1994. Herbivorous nutrition of Cyclops vicinus : The effect of pure algal diet on feeding,development, reproduction and life cycle. J. Plankton Res. 16:171195 , doi: 10.1093/plankt/16.2.171

SCHEFFER , M. 2009. Critical transitions in nature and society, 1sted. Princeton Univ. Press.

, AND S. CARPENTER . 2003. Catastrophic regime shifts inecosystems: Linking theory to observation. Trends Ecol. Evol.18: 648656.

, , J. A. F OLEY , C. F OLKE , AND B. WALKER . 2001.Catastrophic shifts in ecosystems. Nature 413: 591596 ,doi: 10.1038/35098000

SEEBENS , H., U. E INSLE , AND D. S TRAILE . 2008. Copepod life cycleadaptations and success in response to phytoplankton springbloom phenology. Glob. Change Biol. 15: 13941404 ,doi: 10.1111/j.1365-2486.2008.01806.x

SOMMER , U., AND OTHERS . 2012. Beyond the PEG-model:Mechanisms driving plankton succession. Ann. Rev. Ecol.Evol. Syst. 43: 429448 , doi: 10.1146/annurev-ecolsys-110411-160251

THERNEAU , T. M., A TKINSON , B., AND R IPLEY , B. (author of Rport). 2012. rpart: Recursive partitioning. R package version3.1-54 [accessed 06 July 2012]. Available from http://CRAN.R-project.org/package 5 rpart

T O TH , L. G., AND N. P. Z A NKAI . 1985. Feeding of Cyclops vicinus(Uljanin) (Copepoda: Cyclopoida) in Lake Balaton on thebasis of gut content analyses. Hydrobiologia 122: 251260 ,doi: 10.1007/BF00018286

, , AND O. M. M ESSNER . 1987. Alga consumption of four dominant planktonic crustaceans in Lake Balaton(Hungary). Hydrobiologia 145: 323332 , doi : 10.1007/BF02530294

VERDY , A., AND P. A MARASEKARE . 2010. Alternative stable statesin communities with intraguild predation. J. Theor. Biol. 262:116128 , doi: 10.1016/j.jtbi.2009.09.011

WAGNER , C., AND R. A DRIAN . 2009. Exploring lake ecosystems:Hierarchy responses to long-term change? Glob. Change Biol.

15: 11041115 , doi: 10.1111/j.1365-2486.2008.01833.xW ILHELM , S., T. H INTZE , D. M. L IVINGSTONE , AND R. A DRIAN .2006. Long-term response of daily epilimnetic temperatureextrema to climate forcing. Can. J. Fish. Aquat. Sci. 63:24672477 , doi: 10.1139/f06-140

WOOD , S. N. 2012. mgcv: Mixed GAM computation vehicle withGCV/AIC/REML smoothness estimation. R package version1.7-20 [accessed 06 July 2012]. Available from http://CRAN.R-project.org/package 5 mgcv

YUE , S., P. P ILON , B. PHINNEY , AND G. C AVADIAS . 2002. The influenceof autocorrelation on the ability to detect trend in hydrologicalseries. Hydrol. Process. 16: 18071829 , doi: 10.1002/hyp.1095

ZUUR , A. F., E. N. I ENO , AND C. S. E LPHICK . 2010. A protocol fordata exploration to avoid common statistical problems.Methods Ecol. Evol. 1: 314.

Associate editor: Michael R. Landry

Received: 25 July 2012Accepted: 10 December 2012Amended: 17 December 2012

http://dx.doi.org/10.1890%2F06-1525.1http://dx.doi.org/10.1890%2F06-1525.1http://dx.doi.org/10.1086%2F286018http://dx.doi.org/10.1086%2F286018http://dx.doi.org/10.1086%2F286018http://dx.doi.org/10.1111%2Fj.1365-2427.2005.01417.xhttp://dx.doi.org/10.1111%2Fj.1365-2427.2005.01417.xhttp://dx.doi.org/10.1111%2Fj.1365-2427.2005.01417.xhttp://dx.doi.org/10.4319%2Flo.2008.53.1.0001http://dx.doi.org/10.4319%2Flo.2008.53.1.0001http://dx.doi.org/10.1111%2Fj.1365-2427.2005.01430.xhttp://dx.doi.org/10.1111%2Fj.1365-2427.2005.01430.xhttp://dx.doi.org/10.1111%2Fj.1365-2427.2005.01430.xhttp://dx.doi.org/10.1007%2FBF00016806http://dx.doi.org/10.1007%2FBF00016806http://dx.doi.org/10.1016%2FS0924-7963%2897%2900064-Xhttp://dx.doi.org/10.1016%2FS0924-7963%2897%2900064-Xhttp://dx.doi.org/10.1016%2FS0924-7963%2897%2900064-Xhttp://dx.doi.org/10.1007%2Fs10452-005-9008-0http://dx.doi.org/10.1007%2Fs10452-005-9008-0http://dx.doi.org/10.1890%2F0012-9658%281999%29080%5B0752%3APIPAPD%5D2.0.CO%3B2http://dx.doi.org/10.1890%2F0012-9658%281999%29080%5B0752%3APIPAPD%5D2.0.CO%3B2http://dx.doi.org/10.1890%2F0012-9658%281999%29080%5B0752%3APIPAPD%5D2.0.CO%3B2http://dx.doi.org/10.1086%2F321321http://dx.doi.org/10.1086%2F321321http://dx.doi.org/10.1146%2Fannurev.es.20.110189.001501http://dx.doi.org/10.1146%2Fannurev.es.20.110189.001501http://dx.doi.org/10.1146%2Fannurev.es.20.110189.001501http://dx.doi.org/10.1007%2FBF00016825http://dx.doi.org/10.1007%2FBF00016825http://dx.doi.org/10.1093%2Fplankt%2F7.2.279http://dx.doi.org/10.1093%2Fplankt%2F7.2.279http://dx.doi.org/10.1093%2Fplankt%2F7.2.279http://dx.doi.org/10.2307%2F5803http://dx.doi.org/10.2307%2F5803http://dx.doi.org/10.1093%2Fplankt%2F16.2.171http://dx.doi.org/10.1093%2Fplankt%2F16.2.171http://dx.doi.org/10.1038%2F35098000http://dx.doi.org/10.1038%2F35098000http://dx.doi.org/10.1038%2F35098000http://dx.doi.org/10.1111%2Fj.1365-2486.2008.01806.xhttp://dx.doi.org/10.1111%2Fj.1365-2486.2008.01806.xhttp://dx.doi.org/10.1111%2Fj.1365-2486.2008.01806.xhttp://dx.doi.org/10.1146%2Fannurev-ecolsys-110411-160251http://dx.doi.org/10.1146%2Fannurev-ecolsys-110411-160251http://dx.doi.org/10.1146%2Fannurev-ecolsys-110411-160251http://dx.doi.org/10.1007%2FBF00018286http://dx.doi.org/10.1007%2FBF00018286http://dx.doi.org/10.1007%2FBF00018286http://dx.doi.org/10.1007%2FBF02530294http://dx.doi.org/10.1007%2FBF02530294http://dx.doi.org/10.1007%2FBF02530294http://dx.doi.org/10.1016%2Fj.jtbi.2009.09.011http://dx.doi.org/10.1016%2Fj.jtbi.2009.09.011http://dx.doi.org/10.1111%2Fj.1365-2486.2008.01833.xhttp://dx.doi.org/10.1111%2Fj.1365-2486.2008.01833.xhttp://dx.doi.org/10.1139%2Ff06-140http://dx.doi.org/10.1139%2Ff06-140http://dx.doi.org/10.1002%2Fhyp.1095http://dx.doi.org/10.1002%2Fhyp.1095http://dx.doi.org/10.1002%2Fhyp.1095http://dx.doi.org/10.1139%2Ff06-140http://dx.doi.org/10.1111%2Fj.1365-2486.2008.01833.xhttp://dx.doi.org/10.1016%2Fj.jtbi.2009.09.011http://dx.doi.org/10.1007%2FBF02530294http://dx.doi.org/10.1007%2FBF02530294http://dx.doi.org/10.1007%2FBF00018286http://dx.doi.org/10.1007%2FBF00018286http://dx.doi.org/10.1146%2Fannurev-ecolsys-110411-160251http://dx.doi.org/10.1146%2Fannurev-ecolsys-110411-160251http://dx.doi.org/10.1111%2Fj.1365-2486.2008.01806.xhttp://dx.doi.org/10.1111%2Fj.1365-2486.2008.01806.xhttp://dx.doi.org/10.1038%2F35098000http://dx.doi.org/10.1038%2F35098000http://dx.doi.org/10.1093%2Fplankt%2F16.2.171http://dx.doi.org/10.2307%2F5803http://dx.doi.org/10.1093%2Fplankt%2F7.2.279http://dx.doi.org/10.1093%2Fplankt%2F7.2.279http://dx.doi.org/10.1007%2FBF00016825http://dx.doi.org/10.1146%2Fannurev.es.20.110189.001501http://dx.doi.org/10.1146%2Fannurev.es.20.110189.001501http://dx.doi.org/10.1086%2F321321http://dx.doi.org/10.1890%2F0012-9658%281999%29080%5B0752%3APIPAPD%5D2.0.CO%3B2http://dx.doi.org/10.1890%2F0012-9658%281999%29080%5B0752%3APIPAPD%5D2.0.CO%3B2http://dx.doi.org/10.1007%2Fs10452-005-9008-0http://dx.doi.org/10.1016%2FS0924-7963%2897%2900064-Xhttp://dx.doi.org/10.1016%2FS0924-7963%2897%2900064-Xhttp://dx.doi.org/10.1007%2FBF00016806http://dx.doi.org/10.1111%2Fj.1365-2427.2005.01430.xhttp://dx.doi.org/10.1111%2Fj.1365-2427.2005.01430.xhttp://dx.doi.org/10.4319%2Flo.2008.53.1.0001http://dx.doi.org/10.1111%2Fj.1365-2427.2005.01417.xhttp://dx.doi.org/10.1111%2Fj.1365-2427.2005.01417.xhttp://dx.doi.org/10.1086%2F286018http://dx.doi.org/10.1086%2F286018http://dx.doi.org/10.1890%2F06-1525.1

Documents

Jurnal Kece Plankton!