, published 24 September 2012, doi: 10.1098/rstb.2012.0245367 2012 Phil. Trans. R. Soc. B

Daniel C. Reuman and Mark E. LedgerGuy Woodward, Lee E. Brown, Francois K. Edwards, Lawrence N. Hudson, Alexander M. Milner,

alters food web size structure in a field experimentClimate change impacts in multispecies systems: drought

Supplementary dataml http://rstb.royalsocietypublishing.org/content/suppl/2012/09/18/rstb.2012.0245.DC1.ht

"Data Supplement"

References

http://rstb.royalsocietypublishing.org/content/367/1605/2990.full.html#related-urls Article cited in:

http://rstb.royalsocietypublishing.org/content/367/1605/2990.full.html#ref-list-1 This article cites 34 articles, 10 of which can be accessed free

Subject collections (526 articles)ecology

Articles on similar topics can be found in the following collections

Email alerting service hereright-hand corner of the article or click

Receive free email alerts when new articles cite this article - sign up in the box at the top

http://rstb.royalsocietypublishing.org/subscriptions go to: Phil. Trans. R. Soc. BTo subscribe to

on April 22, 2014rstb.royalsocietypublishing.orgDownloaded from on April 22, 2014rstb.royalsocietypublishing.orgDownloaded from

Climate change impasystems: drought al

ew

M

E.

n Ml SB1ds,BX1ooofUn

1. INTMost esystemtem p[1,2], oveys, wlong-teeffectscausalcan ide[4]. Areachewhichto simu

Thetheory

ducedortantoftenangesurcesc cas-tifyinge andfers ao per-ght bequan-ia thepeciesmass-ed via-scalen befromfood

chains and coarser-grained attributes of the wholefood web [20,21].

Two fundamental measures are used here, withothers derived from them, to quantify trophic size

* Autholedger@

Electronic supplementary material is available at http://dx.doi.org/10.1098/rstb.2012.0245 or via http://rstb.royalsocietypublishing.org.

One contribution of 17 to a Theme Issue Climate change in size-

on April 22, 2014rstb.royalsocietypublishing.orgDownloaded from structured ecosystems.We imposed 2 years of intermittent drought on stream channels in a replicated field trial, to measurefood web responses to simulated climate change. Drought triggered widespread losses of species andlinks, with larger taxa and those that were rare for their size, many of which were predatory, beingespecially vulnerable. Many network properties, including sizescaling relationships within foodchains, changed in response to drought. Other properties, such as connectance, were unaffected.These findings highlight the need for detailed experimental data from different organizationallevels, from pairwise links to the entire food web. The loss of not only large species, but alsothose that were rare for their size, provides a newly refined way to gauge likely impacts that maybe applied more generally to other systems and/or impacts.

Keywords: allometric scaling; ecological networks; experimental mesocosms; stream ecosystems;tritrophic food chains; trivariate food webs

RODUCTIONmpirical studies of climate change inmultispeciess have focused on community structure or ecosys-rocesses in space-for-time or temporal surveysr laboratory experiments [3]. Unfortunately, sur-hich are correlative, are often confounded (e.g. byrm change in other stressors or biogeographicin space-for-time surveys) and unable to discernrelationships, whereas laboratory experimentsntify mechanisms but suffer from limited realismcompromise between realism and control may bed in larger-scale field experiments, several ofhave recently demonstrated ecological responseslated climate change [510].re is a growing body of empirical evidence andthat larger organisms suffer disproportionately

from climate change [4,7,1114]. Their reabundance or extinction has particularly impimplications for aquatic food webs, which arestrongly size-structured [1521]. For instance, chin the mass and abundance of consumers and resocan trigger dramatic changes, including trophicades and secondary extinctions [4,10,17]. Quanallometric relationships between the abundancbody mass of interacting species therefore ofpotentially powerful means of gauging responses tturbations and assessing how structural changemilinked to dynamical properties [20]. One way totify the size structure of trophic networks is vconstruction of trivariate food webs, in which spopulations are plotted on loglog bodyabundance (MN) axes as nodes that are connectfeeding links [15,17,18,2022]. Within this logMN space allometric scaling relationships caquantified at different organizational levels,feeding links between species populations, to

rs for correspondence (g.woodward@qmul.ac.uk; m.e.bham.ac.uk).Experimental data from intergenerational field manipulations of entire food webs are scarce, yetsuch approaches are essential for gauging impacts of environmental change in natural systems.structure in a fiGuy Woodward1,*, Lee E. Bro

Lawrence N. Hudson5, Alexander

and Mark1School of Biological and Chemical Sciences, Quee

2School of Geography, Earth and EnvironmentaBirmingham

3School of Geography, University of Lee4Centre for Ecology and Hydrology, Maclean

Wallingford O5Imperial College London, Silw

6Institute of Arctic Biology, University7Laboratory of Populations, Rockefeller2990cts in multispeciesters food web sizeld experimentn2,3, Francois K. Edwards2,4,

. Milner2,6, Daniel C. Reuman5,7

Ledger2,*

ary University of London, London E1 4NS, UKciences, University of Birmingham, Edgbaston,5 2TT, UKWoodhouse Lane, Leeds LS2 9JT, UKuilding, Benson Lane, Crowmarsh Gifford,0 8BB, UKd Park, Ascot KT2 6SH, UKAlaska, Fairbanks, AK 99775, USAiversity, New York, NY 10065, USA

Phil. Trans. R. Soc. B (2012) 367, 29902997

doi:10.1098/rstb.2012.0245This journal is q 2012 The Royal Society

ing algae, detritus and invertebrates) through a feeder

Drought alters food web size structure G. Woodward et al. 2991

on April 22, 2014rstb.royalsocietypublishing.orgDownloaded from structure [20,21,23]. First, the length of a trophic linkbetween a consumer (C) and its resource (R) is thenumber of orders of magnitude of difference in bodymass plus the number of orders of magnitude of differ-ence in population density between R and C. Second,the link angle relates to the rate of change in biomass,population productivity, and population consumptionfrom R to C. See 2 for precise definitions andinterpretation. Tritrophic chains, which have beenthe focus of considerable research on indirect effectsof consumerresource interactions in food webs (e.g.trophic cascades), contain an intermediate species (I)between R and C. Derived measures include thebetween-angle in a tritrophic chain, which describesthe change of angle in the upper link (I, C) relativeto the lower link (R, I) [20]. At the whole-networklevel, the webs allometric slope is the scaling coefficientof log(N) regressed on log(M) for all the trophicallyconnected species: a steepening of this slope mayimply a weakening of top-down effects, as consumersbecome smaller and/or rarer relative to their resources.Data on links and angles at different levels of organiz-ation within and across the food web can therefore beused to make inferences about the impacts of pertur-bations, especially where the size spectrum is notaffected evenly.

Climate change effects on freshwater ecosystemsinclude predictions that droughts will increase in fre-quency and intensity in the near future [22,24,25].Even partial or temporary drying can threaten thelocal survival of many species, especially those thatare large, rare, and/or high in the food web[1,2628]. It is also plausible, though not previouslydemonstrated, that species that are rare for their sizecould be especially vulnerable to local extinction,even if they are not rare absolutely. Rarity-for-sizeshould be evident from residuals from the food websgeneral log(N)-versus-log(M) regression, which typi-cally describes the average or expected abundance ofa species (given its size) within a web [29]. Taxacould be rare for their size because the ambientenvironmental conditions are already sub-optimal forthem: hence, they may be particularly vulnerable tothe imposition of additional pressures. Currently,almost nothing is known about food web responsesto drought, as the few studies conducted to datehave been either correlational surveys or have notmeasured species-level interactions explicitly in atrophic network context. There is great potentialfor drought to alter size structure at multiple levelsof food web organization: we carried out the firstlong-term (i.e. intergenerational) replicated fieldexperiment to assess this.

Eight artificial stream channels were exposed toeither intermittent drought (6 days of dewatering permonth) or left as permanently flowing controls, tomimic the patchy drying of natural river beds duringextreme low flows [26,27]. The experiment ran for2 years, allowing intergenerational responses andindirect food web effects to be manifested. At theend, eight replicate food webs were constructed(four per treatment). Earlier work in this systemrevealed that the control channels contained foodwebs of realistic complexity [30], and that biomassPhil. Trans. R. Soc. B (2012)pipe from an adjacent stream. Each channel (width0.33 m, length 12 m, depth 0.30 m) was controlled byupstream input valves and drained under gravity via anoutlet 10 cm above a downstream channel. Channelswere filled to 20 cm depth with stony substrate, provid-ing physical refugia for suitably adapted species duringdrought [26,27]. Physicochemistry and biotic assem-blages were similar among channels and to the sourcestream prior to drought [26,27,3234].

Stream water was diverted into all channels in theinitial two months. A drought treatment (intermittentflow; 6 day flow cessation per month) was subsequentlyapplied to one channel per block, with the second chan-nel in each block acting as a control. The droughttreatment mimicked the repeated, patchy dewateringthat occurs during severe supra-seasonal droughts[26]. Under drought, surface flows ceased and dryingof exposed substrata occurred in patches over the 6days, whereas wetted subsurface interstices and smallpools persisted [26,27]. Flows were continuous in thecontrol channels throughout the experiment.

At the end of the experiment, we collected theentire invertebrate assemblage in each mesocosm andconstructed food webs by direct observation of feedinglinks via analysis of dissected gut contents (of 4305individuals in total; five fields of view per individualat 200 magnification). Gut contents were identifiedto genus or species where possible. Food webs wereproduction of large, long-lived consumers was stronglyreduced by drought, with some local extinctions,whereas some smaller short-lived consumers flour-ished [27,28]. Given that large consumers canimpose strong top-down control on their prey inaquatic systems [9,10,28], we anticipated significantdrought impacts on food web size structure. Wetested the following hypotheses: (i) drought causeslosses of not only large taxa, but also those that wererare for their body mass; (ii) drought steepens the allo-metric slope of the web, and reduces the area occupiedby the community in loglog MN space (the con-straint space sensu [31]) due, respectively, to relativeincreases in the abundances of smaller taxa and loss ofrare-for-size species that deviated from the generalMNscaling relationship; (iii) drought causes the collapse ofmany tritrophic food chains into pairwise feeding links,owing to loss of larger predators, with maximum trophiclevel and mean food chain link count declining accord-ingly; (iv) loss of rare-for-size species homogenizeslink angles and between-angles within the web, asthese will have links that deviate most strongly fromthe angle of the general community log(N)-versus-log(M) regression slope; (v) indirect effects (i.e. thosebeyond the direct effects of species loss per se) will bemanifested and detectable via changes in the corecommunity common to both treatments.

2. METHODSThe experiment ran for 2 years (March 2000February2002) in four blocks of outdoor stream channel meso-cosms at the Freshwater Biological Association RiverLaboratory, UK (5084004800 N, 281100600 W) [31],which received water and suspended particles (includ-

2992 G. Woodward et al. Drought alters food web size structure

on April 22, 2014rstb.royalsocietypublishing.orgDownloaded from constructed independently for each replicate channel(after Brown et al. [30]). The body mass (mg) of eachprocessed animal was calculated from body length(mm), measured to the nearest 0.1 mm, using lengthmass regressions (see Ledger et al. [26] and referencestherein), macroinvertebrate abundances,N, were deter-mined from Surber samples (0.025 m22) collected fromeach replicate mesocosm at the end of the experiment.Animals from these samples (n 3049) were identifiedto the lowest practicable taxonomic unit (usually speciesor genus), counted andmeasured. Abundance and indi-vidual body mass were similarly derived for algalresources [30].

Logistic regressions were used to ascertain whetherabsolute rarity and rarity-for-size were important deter-minants of extinction risk of species from the food webs.Here, two predictors were computed per species inthe control replicate of each block: (i) log(M) itself,accounting for larger species having higher extinctionrisk; (ii) the residual from the log(N)-versus-log(M)regression of the replicate. Base-10 logarithms wereused throughout. The first predictor represents rarity,because size and rarity are very strongly correlated inour webs (and many others). The second predictor israrity-for-size: species which are rare for their size willhave substantially negative residuals. This gave twonumbers for each species in each replicate per block. Ifthe same species was present in the control replicatesof two blocks, separate numbers were derived. Theresponse variable was whether or not a species wentextinct (true, false) in the paired drought treatmentreplicates. Logistic regression was performed usingtwo models: one with predictor log(M) only, and onewith both log(M) and residuals as predictors, to seewhether rarity-for-size provided any additional explana-tory power for loss from the food web. Logisticregression is a standard technique whereby the trans-formed risk of an event occurring is written as a linearfunction of predictor variables. Because absolute rarityis often highly correlated with size and rarity-for-sizeis independent of size for webs with homoskedasticlog(N)-versus-log(M) regressions (i.e. most webs [29],including ours), the latter measure is a new possibledeterminant of extinction risk.

MN trivariate webs were produced for each replicatechannel by overlaying links between consumers andresources on the log(N)-versus-log(M) scatterplot oftaxa: several community metrics were defined from thisplot (after Cohen et al. [20]). The community-wide allo-metric slope is the slope of the log(N)-versus-log(M)ordinary linear regression through all taxa connected tothe web by a trophic link. When expressed as an anglebetween 2908 and 908, measured anticlockwise fromthe positive horizontal axis, this is the allometric angle;for instance, slope 21 corresponds to angle 2458. Thecommunity span is the range of log(M), from the smallestto the largest taxa, plus the range of log(N), from therarest to the most abundant taxa, over all connectedtaxa. We also derived the area of the minimum con-vex hull (a polygon) in MN space that bounded allthe connected species within each web (after Leaper &Raffaelli [31]).

The link length between a consumer (C) and itsresource (R) was defined as jlog(MC) 2 log(MR)j

Phil. Trans. R. Soc. B (2012)jlog(NC) 2 log(NR)j [20], i.e. the l1 distance orManhattan distance from mathematics. The firstterm summand is the absolute log body mass ratio,i.e. the number of orders of magnitude of differencein body mass between R and C. The second summandis the absolute log density ratio, i.e. the number oforders of magnitude of difference in population den-sity. When plotting a link as a vector from R to C, itslength is the distance from R to C (l1 distance).Its angle is the anticlockwise turn to the link from ahorizontal arrow parallel to the horizontal (log(M))axis, starting from R and pointing right (between21808 and 1808, where 21808 is allowed but 1808 isnot). If the link angle equals 2458, then its slopeequals 21 and, resource biomass BR MRNR equalsconsumer biomass BC MCNC [20].

We also calculated several higher-level measures ofnetwork size structure. A 2-chain is a tritrophic inter-action consisting of three taxa (R, intermediate taxonI and C) and two links [20]. On MN plots, theupper link lies below and to the right of the lowerlink if body mass increases and abundance declinesmoving up the chain, as in many chains. The 2-spanis the l1 distance from R to C (see the electronic sup-plementary material, figure S1). Within each chain,Llower and Alower describe the length and angle of thelower link (from R to I); Lupper and Aupper describethe length and angle of the upper link (from I to C).The between-angle of a 2-chain is the angle, between21808 and 1808 (including 21808 but not 1808),from the vector of the lower link (R, I) to that of theupper link (I, C). Positive angles are anticlockwiserotations (e.g. if the lower link is 2508 and the upperlink 2358, then the between-angle is 158). A positivebetween-angle means that biomass, population prod-uctivity and population consumption increase fasterin the upper link (from I to C) than in the lower link(from R to I): this may imply increased transfer effi-ciency moving up the chain, as might arise whencomparing a carnivorous upper link with a herbivorouslower link, for instance. It could also imply greaterpotential for top-down cascades, as the top consumeris larger and/or more abundant than would be expectedfrom simple extrapolation from the lower link. The dis-tribution of between-angles thus describes how log bodymass ratios and log population density ratios varyamong the tritrophic chains within a food web.

Maximal food chains (chains henceforth) from abasal to a top taxon were counted as any chain passingfrom resource to consumer at each link, but not includ-ing the same taxon twice (cannibalistic links wereexcluded and cycles were not traversed completely).The chain span is the l1 distance between a chains topand basal taxa. Food chain link count is the averagenumber of links contained within all the chains in theweb. Several more familiar whole-network parameterswere also calculated, namely: web size (S, the numberof nodes), number of feeding links (L) and directedconnectance (C L/S2).

Two additional scenarios were also considered:(i) extinct species in the drought treatment wereexcluded from controls; and (ii) the core communityalone was considered, i.e. only species that werecommon to each pair of drought and control webs

6(

(

10(N

Drought alters food web size structure G. Woodward et al. 2993

on April 22, 2014rstb.royalsocietypublishing.orgDownloaded from 15 10 5 0

0

2

4

log8

10

12

0

2

4

6

8

10

12(a)

(c)

log 1

0(N) (

m2 )

) (m

2 )were considered. This was to gauge potential emergentor indirect effects beyond those due directly to speciesloss or gain. Between-treatment differences in foodweb parameters were tested using paired t-tests, withblocks providing the pairings. Computations wereperformed in R [35].

3. RESULTSAs predicted (hypothesis 1), body mass influenced vul-nerability to drought, with a second-order effect ofrarity-for-size (figure 1). The coefficient (20.22) forlog(M) in our logistic regressions demonstrates thatlarger species were more likely to be lost. The positivecoefficient (0.794) for residuals shows that species thatwere rare for their size were additionally vulnerable(p, 0.0001; electronic supplementary material,table S1).

Several of the fine-grained measures revealedmarked changes within the food web, whereas othermeasures, including some commonly used ones (e.g.connectance), were unaffected (table 1). Droughtsignificantly reduced the numbers of species andlinks. In agreement with predictions (hypothesis 2),allometric slopes steepened slightly but significantlyfrom 20.50 to 20.52, reflecting reductions in large,rare taxa and increases in some of the smaller taxa:these patterns were also evident in the two additional

log10(M ) (mg)

Figure 1. Food web nodes, plotted using the body mass (log10(Mshows a comparison between a replicate control food web and itssent in both webs, diamonds denote species that were in the drspecies in the control but lost from the drought webs. Ordinar

and so were fitted to species denoted by circles and triangles on

Phil. Trans. R. Soc. B (2012)b)

d)

15 10 5 0scenarios that accounted for species loss or gain in thedrought treatment (see the electronic supplementarymaterial, table S2). Convex hull area decreased, butnot in either of the additional scenarios. Hypothesis 3was also supported: as species were lost and/or hadtheir links stripped away, the maximum trophic levelof the web decreased as the number of links fromthe base to the highest predator declined. The pro-portion of intermediate nodes declined, basal nodesincreased and top-level nodes remained the same (seethe electronic supplementary material, table S2): inter-mediate nodes were lost either via extinction or bypromotion to the termini of chains (figure 2). Conse-quently, the total number of tritrophic food chainsdeclined, with many collapsing into simple pairwiselinks, even though mean food chain link count did notdecline significantly (table 1). Hypothesis 4 was sup-ported: owing to the loss of rare-for-size species, linkangles and between-angles were more tightly con-strained in the drought treatments than in the controls(figures 3 and 4). Between-angles also became morenegative in the drought treatment owing to declines inboth body mass and abundance of top predators intritrophic chains.

Although there were some significant differencesbetween treatments in link lengths and angles, allometricslope, and species richness, many other parameters wereunaffected by drought in the two additional scenarios

log10(M ) (mg)

)) and abundance (log10(N)) of each taxon. Each panel (ad)paired drought treatment: circles denote taxa that were pre-ought treatment but not in the control and triangles denotey linear regression lines were used to assess extinction risk

ly. Panels correspond to blocks.

2 343637383940414243444546484951525354555657585960616569

2

2

829

3031

3436

3738

3941

4244

4546

4849

5152

5355

5657

5859

6061

6364

6566

6869

7173

67

2994 G. Woodward et al. Drought alters food web size structure

on April 22, 2014rstb.royalsocietypublishing.orgDownloaded from 1 1 2 3 4 5 6 7 8 91012131415161819202223242526272829303133

d1 d

12

34

56

78

910

1213

1415

1617

1819

2022

2324

2526

272

2 353738394142444549515457585960616365697470

3437

3841

4245

5051

5859

6061

6263

65

70

trop

hic

leve

l le

vel70

6773 c1 c

62

70for detecting indirect effects (see the electronic supple-mentary material, table S2). Thus, indirect effects weregenerally modest when compared with the direct effectsof species change, and the core community remainedrelatively intact: hypothesis 5was therefore not supportedfor most measures. Overall, drought tended to simplifyand homogenize network size structure, primarilyvia the direct effects of the loss of larger, rarer (andpredominantly, but not always, predatory) species.

4. DISCUSSIONThis is the first replicated study of the impact of acomponent of climate change on the architecture of

1 1 2 3 4 5 6 910121316171920212223242526272829303133 1 2 3 4 5 6 7 8 910121314151618192021222324252627282930313

trop

hic

Figure 2. Food webs from the manipulative field experiment, in wintermittent drought (d) or permanent flow (c). The webs are orpredators. Numbers correspond to species identifiers (see figumaterial for codes and taxonomic identities).

Table 1. Mean+ s.e. community structure measures for the conspecies within the food web. See 2 for details. Paired t-tests(d) from zero. Results for additional scenarios testing for indirematerial, table S2.

control webs

pairwise links, tritrophic interactions and food chainsmedian link angle 227.17+0.mean link length 18.23+0.median Alower 226.67+0.median Aupper 231.80+1.log10 number of tritrophic chains 2.05+0.median Abetween 28.99+3.mean two-span 19.2+0.mean chain span 19.4+0.mean food chain link count 1.49+0.trophic level of apex predator (chain length) 2.53+0.

community scaling and whole-network propertiesallometric slope 20.50+0.community span 28.1+0.constraint space area (MN convex hull area) 60.23+2.S, the number of connected food web nodes 60+1.log10 L, number of links 2.48+0.C, directed connectance 0.08+0.

Phil. Trans. R. Soc. B (2012)c3 c4

d3 d4

74

12

34

56

78

910

1213

1415

1617

1819

2021

2223

2425

2627

2829

3031

3436

3738

3941

4243

4445

4950

5152

5354

5759

6061

6365

6668

6972

71

706774

73

12

34

56

78

910

1112

1314

1516

1819

2021

2223

2425

2627

2829

3031

3233

3436

3738

3941

4243

4445

4647

5152

5354

5758

5960

6162

6364

6668

697148

70

6774

73

3435

3839

4142

4551

5455

5758

5960

6165

74

70

3438

3941

4244

4547

5051

5457

5859

6061

6263

6569

70food web size structure in a long-term field experi-ment. We found clear evidence that droughttriggered the widespread loss of species and links andthe homogenization of aspects of size structure.Because larger species and those that were rare fortheir size were lost, perturbed webs were boundedwithin a smaller constraint space and fitted moretightly to MN scaling relationships than did controls,causing changes in the finer-grained network proper-ties (e.g. among the webs pairwise links, tritrophicinteractions and food chains). Drought caused a win-nowing of the web (cf. [36]), as nodes (and links)were stripped out to leave a skeleton outline withinthe same community span and only slightly steeper

21

23

45

67

89

1011

1213

1415

1617

1819

2021

2223

2425

2627

2829

3132 1

23

45

67

89

1012

1314

1516

1819

2021

2223

2425

2627

2829

3031

3233

hich eight replicate stream channels were exposed to monthlydered vertically by trophic level, from basal resources to apexre 1 for symbols legend; see the electronic supplementary

trol (c1c4) and drought (d1d4) treatments for connected

were performed to test for significance of mean differencesct effects (see 2) are given in the electronic supplementary

drought webs d t p

24 227.59+0.24 0.42 3.52 0.03918 18.62+0.11 20.40 22.11 0.12540 225.76+1.02 20.91 21.04 0.37509 2142.30+9.67 110.5 10.46 0.00220 1.29+0.01 0.77 4.05 0.02793 2116.53+9.78 107.5 13.40 0.00119 18.05+0.11 1.11 6.21 0.00817 18.7+0.10 0.69 6.06 0.00917 1.09+0.01 0.39 2.32 0.10305 2.16+0.04 0.37 7.64 0.005

006 20.52+0.002 0.017 3.71 0.03417 27.7+0.08 0.33 1.94 0.14845 43.09+1.86 17.14 4.76 0.0183 46.5+1.3 13.5 11.34 0.00105 2.31+0.04 0.17 3.20 0.050008 0.09+0.008 20.01 20.87 0.448

Drought alters food web size structure G. Woodward et al. 2995

on April 22, 2014rstb.royalsocietypublishing.orgDownloaded from control

50

50

150

150

50

50

150

drought

A upp

erA u

pperoverall MN slope (reflecting the relative increasein some of the smaller taxa). The thinning effectexplains reductions in community biomass and sec-ondary production reported previously [26]. Otherweb properties (e.g. connectance) were apparentlyeither insensitive to drought or too coarse-grained toreveal a response. The core food web containingspecies common to both treatments did not respondstrongly to drought for most parameters (see the elec-tronic supplementary material, table S2), revealing alack of obvious emergent effects beyond the directimpacts of species loss. There was no compelling evi-dence for the widespread cascading effects orsecondary extinctions that might be expected due tostrong top-down control. More subtle indirect effectsof drought were, however, evident on the slopes ofthe core webs and their respective links.

Large size and absolute rarity are both associatedwith increased extinction risk and are often correlatedwith one another [14]. We also found evidence of asecond-order, but important rarity-for-size effect thatwas distinct from overall rarity effects. Ecological driftis unlikely to be driving this, as rare species werecommon in the source stream and recolonized disturbed

150 50 0 50 150

150

Alower

Figure 3. Upper angle (Aupper) versus lower angle (Alower) ofall 2-chains within food webs from the control and droughttreatments. Vertical and horizontal solid lines representmedian lower and upper angles for all 2-chains (see 2).

One representative web (c4, d4) per treatment is shownhere; all eight (c1c4, d1d4) are shown in the electronicsupplementary material, figure S3.

Phil. Trans. R. Soc. B (2012)control

5

10

20

30

0

5

10

20

30

drought

two s

pan

two s

panpatches repeatedly over the 2-year experiment [27], i.e.rather than being opportunistic visitors they were per-manent residents that were numerically rare in thesystem. Species below the general MN-scaling linewere especially vulnerable, being already rarer inthe controls than expected based on their size. Suchspecies may already be in sub-optimal conditions(hence their relative rarity) and the imposition ofadditional stress on the system might be sufficient topush them to extinction. Rarity-for-size might thereforehelp to identify especially vulnerable taxa, withoutnecessarily requiring detailed a priori knowledge oftrophic position or environmental tolerances. Thiscould be useful for assessing impacts of stressors in eco-logical networks in general, and warrants furtherexploration in systems exposed to perturbations orenvironmental stress whereM andN data exist [17,18].

The larger consumers that were lost were alsopredominantly aquatic throughout their life cycle,whereas the surviving large insect species possessed aterrestrial adult phase, enabling them to (re)colonizedenuded patches. Many of the smaller taxa survivedthe drought, most likely in patches of wetted refugia,and some even flourished (e.g. certain midge larvaeand small oligochaete worms) suggestive of releasefrom competition and/or predation from the larger

150 50 0 50 1500

Abetween

Figure 4. Network substructure in control and drought treat-ments: two-span as a function of between angle (Abetween)in tritrophic chains within each food web (see 2). One

representative web (c4, d4) per treatment is shown here;all eight (c1c4, d1d4) are depicted in the electronicsupplementary material, figure S4.

Research Council (NERC) postdoctoral fellowship toM.E.L. and NERC grant no. NER/B/S/2002/00215. L.N.H.

2996 G. Woodward et al. Drought alters food web size structure

on April 22, 2014rstb.royalsocietypublishing.orgDownloaded from was supported by Microsoft Research, and D.C.R. waspartially supported by NERC grant nos. NE/H020705/1,NE/I010963 and NE/I011889/1. The participation of G.W.at the SIZEMIC Workshop in Hamburg was supported bythe German Research Foundation (JA 1726/3-1) as well asthe Cluster of Excellence CliSAP (EXC177), University ofHamburg funded through the DFG. We thank everyone whohelped collect the field data, especially Rebecca Harris,Brian Godfrey, Bethan Ledger and John Murphy. We alsothank Ulrich Brose and two anonymous referees, whoseinsightful comments improved the manuscript considerably.

REFERENCES1 Ledger, M. E. & Hildrew, A. G. 2001 Recolonization bythe benthos of an acid stream following a drought. Arch.Hydrobiol. 152, 117.

2 Woodward, G. et al. 2010 Sentinel systems on the razorsedge: effects of warming on Arctic stream ecosystems.Glob. Change Biol. 16, 19791991. (doi:10.1111/j.1365-2486.2009.02052.x)

3 Petchey, O. L., McPhearson, P. T., Casey, T. M. &Morin, P. J. 1999 Environmental warming alters food-web structure and ecosystem function. Nature 402,6972. (doi:10.1038/47023)

4 Woodward, G. et al. 2010 Ecological networks in achanging climate. Adv. Ecol. Res. 42, 72138. (doi:10.1016/B978-0-12-381363-3.00002-2)

5 Liboriussen, L. et al. 2011 Effects of warming and nutri-ents on sediment community respiration in shallow lakes:

an outdoor mesocosm experiment. Freshw. Biol. 56,437447. (doi:10.1111/j.1365-2427.2010.02510.x)

6 Yvon-Durocher, G., Montoya, J.M., Trimmer, M. &Woodward, G. 2011 Warming alters the size spectrumand shifts the distribution of biomass in freshwater eco-

systems. Glob. Change Biol. 17, 16811694. (doi:10.1111/j.1365-2486.2010.02321.x)

7 Grieg, H. S. et al. 2012 Warming, eutrophication, andpredator loss amplify subsidies between aquatic and ter-

restrial ecosystems. Glob. Change Biol. 18, 504514.(doi:10.1111/j.1365-2486.2011.02540.x)taxa [26,27]. This might also indicate at least someindirect food web effects are modulated by drought,via apparent competition for enemy-free space, asrevealed by the modest but significant steepening inthe allometric slope of the web (even in the scenarioswhere we accounted for species loss or gain).

Identifying which food web parameters are most sen-sitive to perturbations is key to assessing the impacts ofenvironmental change in natural systems: focusing onthe more commonly used properties (e.g. connectance)would have missed subtle yet informative structuralchanges. The next move towards understanding climatechange impacts will necessitate modelling the dynam-ical consequences of structural change, if we areultimately to predict impacts on food web stability[4,37]. One important pattern emerging from recentresearch is that large, rare organisms high in the foodweb seem to suffer disproportionately from environ-mental warming [4,13] (but see [2]) and becausedrought had similar impacts, these two components ofclimate change could combine to produce potentiallylethal synergies in the near future [4].

The Freshwater Biological Association (FBA) and the Centrefor Ecology and Hydrology supported this research. Theproject was funded by a FBA/Natural EnvironmentalPhil. Trans. R. Soc. B (2012)8 Suttle, K. B., Thomsen, M. A. & Power, M. E. 2007Species interactions reverse grassland responses to chan-ging climate. Science 315, 640642. (doi:10.1126/science.1136401)

9 Jochum, M., Schneider, F. D., Crowe, T. P., Brose, U. &OGorman, E. J. 2012 Climate-induced changes inbottom-up and top-down processes independentlyalter a marine ecosystems. Phil. Trans. R. Soc. B 367,29622970. (doi:10.1098/rstb.2012.0237)

10 Shurin, J. B., Clasen, J. L., Greig, H. S., Kratina, P. &Thompson, P. L. 2012 Warming shifts top-down andbottom-up control of pond food web structure and

function. Phil. Trans. R. Soc. B 367, 30083017.(doi:10.1098/rstb.2012.0243)

11 Lurgi, M., Lopez, B. C. & Montoya, J. M. 2012 Novelcommunities from climate change. Phil. Trans. R. Soc.B 367, 29132922. (doi:10.1098/rstb.2012.0238)

12 Eklof, A., Kaneryd, L. & Munger, P. 2012 Climatechange in metacommunities: dispersal gives double-sided effects on persistence. Phil. Trans. R. Soc. B 367,29452954. (doi:10.1098/rstb.2012.0234)

13 Daufresne, M., Lengfellner, K. & Sommer, U. 2009

Global warming benefits the small in aquatic ecosystems.Proc. Natl Acad. Sci. USA 106, 12 78812 793. (doi:10.1073/pnas.0902080106)

14 Raffaelli, D. 2004 How extinction patterns affect ecosys-tems? Science 306, 11411142. (doi:10.1126/science.1106365)

15 Woodward, G. et al. 2005 Body-size in ecological net-works. Trends Ecol. Evol. 20, 402409. (doi:10.1016/j.tree.2005.04.005)

16 Brose, U. et al. 2006 Consumerresource body-sizerelationships in natural food webs. Ecology 87, 24112417. (doi:10.1890/0012-9658(2006)87[2411:CBRINF]2.0.CO;2)

17 OGorman, E. & Emmerson, M. 2010 Manipulating

interaction strengths and the consequences for trivariatepatterns in a marine food web. Adv. Ecol. Res. 42,301419. (doi:10.1016/B978-0-12-381363-3.00006-X)

18 Layer, K., Riede, J. O., Hildrew, A. G. & Woodward, G.2010 Food web structure and stability in 20 streams

across a wide pH gradient. Adv. Ecol. Res. 42,265301. (doi:10.1016/B978-0-12-381363-3.00005-8)

19 Gilljam, D. et al. 2011 Seeing double: size-based versustaxonomic views of food web structure. Adv. Ecol. Res.45, 67134. (doi:10.1016/B978-0-12-386475-8.00003-4)

20 Cohen, J. E., Schittler, D. N., Raffaelli, D. G. & Reuman,D. C. 2009 Food webs are more than the sum of theirtritrophic parts. Proc. Natl Acad. Sci. USA 106, 22 33522 340. (doi:10.1073/pnas.0910582106)

21 Reuman, D. C. & Cohen, J.E. 2004 Trophic links lengthand slope in the Tuesday Lake food web with speciesbody mass and numerical abundance. J. Anim. Ecol.73, 852866. (doi:10.1111/j.0021-8790.2004.00856.x)

22 IPCC 2007 Contribution of Working Group II to the

Fourth Assessment Report of the IntergovernmentalPanel on Climate Change. In Climate change 2007:impacts, adaptation and vulnerability (eds M. L. Parry,O. F. Canziani, J. P. Palutikof, P. J. van der Linden, andC. E. Hanson), p. 976. Cambridge, UK: Cambridge

University Press.23 Cohen, J. E., Jonsson, T. & Carpenter, S. R. 2003 Eco-

logical community description using food web, speciesabundance, and body-size. Proc. Natl Acad. Sci. USA100, 17811786. (doi:10.1073/pnas.232715699)

24 Overpeck, J. & Udall, B. 2011 Dry times ahead. Science328, 16421643. (doi:10.1126/science.1186591)

25 Vorosmarty, C. J. et al. 2010 Global threats to humanwater security and river biodiversity. Nature 467,555561. (doi:10.1038/nature09440)

26 Ledger, M. E., Edwards, F. K., Brown, L. E., Milner,A.M. &Woodward, G. 2011 Impact of simulated droughton ecosystem biomass production: an experimental test in

stream mesocosms. Glob. Change Biol. 17, 22882297.(doi:10.1111/j.1365-2486.2011.02420.x)

27 Ledger, M. E., Harris, R. M. L., Armitage, P. D. &Milner, A. M. In press. Community stability dependson disturbance frequency: evidence of drought impacts

in stream mesocosm experiments. Adv. Ecol. Res.28 Power, M. E., Parker, M. S. & Dietrich, W. E. 2008

Seasonal reassembly of a river food web: floods,droughts, and impacts of fish. Ecol. Monogr. 78,263282. (doi:10.1890/06-0902.1)

29 Reuman, D. C. et al. 2009 Allometry of body size andabundance in 166 food webs. Adv. Ecol. Res. 41, 144.(doi:10.1016/S0065-2504(09)00401-2)

30 Brown, L. E., Edwards, F., Milner, A. M., Woodward,

G. & Ledger, M. E. 2011 Food web complexity and allo-metric scaling relationships in stream mesocosms:implications for experimentation. J. Anim. Ecol. 80,884895. (doi:10.1111/j.1365-2656.2011.01814.x)

31 Leaper, R. & Raffaelli, D. 1999 Defining the abundance

body-size constraint space: data from a real food web.Ecol. Lett. 2, 191199. (doi:10.1046/j.1461-0248.1999.00069.x)

32 Ledger, M. E., Harris, R. M. L., Armitage, P. D. &Milner, A. M. 2009 Realism of model ecosystems: anevaluation of physicochemistry and macroinvertebrate

assemblages in artificial streams. Hydrobiologia 617,9199. (doi:10.1007/s10750-008-9530-x)

33 Harris, R.M. L.,Milner, A.M., Armitage, P.D. &Ledger,M. E. 2007 Replicability of physicochemistry andmacroinvertebrate assemblages in stream mesocosms:

implications for experimental research. Freshw. Biol. 52,24342443. (doi:10.1111/j.1365-2427.2007.01839.x)

34 Ledger,M. E., Harris, R.M. L., Armitage, P. D. &Milner,A. M. 2008 Disturbance frequency influences patch

dynamics in stream benthic algal communities. Oecologia155, 809819. (doi:10.1007/s00442-007-0950-5)

35 R Development Core Team. 2008 R: A language andenvironment for statistical computing. Vienna, Austria: RFoundation for Statistical Computing. See http://www.

R-project.org.36 de Ruiter, P.C., Wolters, V., Moore, J. C. & Winemiller,

K.O. 2005 Food web ecology: playing Jenga and beyond.Science 309, 6871. (doi:10.1126/science.1096112)

37 Brose, U., Dunne, J. A., Montoya, J. M., Petchey, O. L.,

Schneider, F. D. & Jacob, U. 2012 Climate change insize-structured ecosystems. Phil. Trans. R. Soc. B 367,29032912. (doi:10.1098/rstb.2012.0232)

Drought alters food web size structure G. Woodward et al. 2997

on April 22, 2014rstb.royalsocietypublishing.orgDownloaded from Phil. Trans. R. Soc. B (2012)

Climate change impacts in multispecies systems: drought alters food web size structure in a field experimentIntroductionMethodsResultsDiscussionThe Freshwater Biological Association (FBA) and the Centre for Ecology and Hydrology supported this research. The project was funded by a FBA/Natural Environmental Research Council (NERC) postdoctoral fellowship to M.E.L. and NERC grant no. NER/B/S/2002/00215. L.N.H. was supported by Microsoft Research, and D.C.R. was partially supported by NERC grant nos. NE/H020705/1, NE/I010963 and NE/I011889/1. The participation of G.W. at the SIZEMIC Workshop in Hamburg was supported by the German Research Foundation (JA 1726/3-1) as well as the Cluster of Excellence CliSAP (EXC177), University of Hamburg funded through the DFG. We thank everyone who helped collect the field data, especially Rebecca Harris, Brian Godfrey, Bethan Ledger and John Murphy. We also thank Ulrich Brose and two anonymous referees, whose insightful comments improved the manuscript considerably.References