The Auk 117(3):634-639, 2000

INFLUENCE OF ENVIRONMENTAL AND DENSITY-DEPENDENT FACTORS ON REPRODUCTION OF LITTLE EGRETS

ROBERTE. BENNETTS,1,3 MAURO FASOLA,2 HEINZ HAFNER, 1 AND YVES KAYSER1 •StationBiologique de la Tourdu Valat,Le Sambuc,F-13200 Aries, France; and 2Dipartimentodi BiologiaAnimale, Universit& Piazza Botta 9, 1-27100Pavia, Italy

ABSTRACT.--Weevaluated the influenceof environmentaland density-dependentfactors (intraspecificand interspecific) on clutchsize, brood size, and nesting success of Little Egrets (Egrettagarzetta) in the Camargueof southernFrance. We recordedthese reproductive pa- rametersin mostyears from 1970to 1998.We useda generalizedlinear modelingapproach (modelselection based on AIC) to examinethe environmentaleffects of springrainfall, win- ter temperature,and wind on theseparameters. We also examineddensity dependence of theseparameters based on the total numberof Little Egretsand the total number of tree- nestingherons nesting in thesemixed-species colonies. Clutch size was positively associated with rainfall and negativelyassociated with the number of Little Egret nestsin the Ca- margue.Brood size was negatively associated with thenumber of Little Egretnests, although rainfall was only significantas an interactioneffect with thesetwo effects.Nesting success was negativelyassociated with the numberof tree-nestingherons, the proportionof each colonyconsisting of Cattle Egrets(Bubulcus ibis), wind speed,and severalinteractions among thesevariables. Virtually all of the reproductiveparameters that we evaluatedwere nega- tively associatedwith the numberof Little Egretnests or the numberof tree-nestingherons. Anecdotalevidence suggests that CattleEgrets displace Little Egretsat somecentrally lo- catednest sites.Such sites are better protectedfrom strongwinds, which are a common causeof nesting failure. Received20 November1998, accepted23 November1999.

DENSITY-DEPENDENTeffects on avian repro- nest building, Cattle Egrets frequently steal duction are well known, althoughlittle evi- sticksfrom nearby nests(Valentine 1958), and dencefor sucheffects exists in wading birds sometimesthey eject nest contents (Blaker (Butler 1994). In contrast,density-independent 1969,Siegfried 1972). Given (1) an advantageof environmentalfactors are commonlyreported selectingcertain nest sites,(2) increasednum- to influencereproduction in this group.For ex- bers of Little Egretsand Cattle Egrets,and (3) ample,Hafner et al. (1994)found that rainfall aggressivebehavior of Cattle Egrets,we hy- in winter and springis correlatedwith clutch pothesizedthat intra- and interspecificdensity size and broodsize of Little Egrets(Egretta gar- dependence related to nest-site acquisition zetta).Extended periods of temperaturesbelow could occur Here, we evaluate the influence of freezing are correlatedwith wintering and environmental and density-dependentfactors subsequentbreeding populationsize of Little on clutchsize, brood size, and nestingsuccess Egrets (Hafner et al. 1994), althoughit is not of Little Egrets. known if subsequenteffects influence repro- ductive success. METHODS Sincethe mid-1980s,numbers of Little Egrets in the Camargueregion of southernFrance Studyarea.--Our study area was the Camargue have increasedsteadily (Hafner and Fasola (delta of the RhoneRiver in southernFrance), which 1997).Cattle Egret numbers(Bubulcus ibis) also comprises180,000 ha and includes several colony have increased, from one nest in 1967 to a re- sites and associatedfeeding areas used by Little centhigh of 3,532nests in 1996(H. Hafnerun- Egrets(Hafner 1977,Hafner et al. 1982).The flood- publ. data). In the Camargue,anecdotal evi- plain of the RhoneRiver has been confinedwithin dencesuggests that Cattle Egretsoccupy the leveeson eitherside of the deltasince the early1850s centerof colonies,where reproductivesuccess (Benoit 1933), and the habitat outsidethe delta com- of Little Egretsis higher(Hafner 1977).During plex generallyis unsuitablefor breedingor foraging (Hafner 1977). Thus, the spatial extent of the Ca- margue population of Little Egrets has remained 3 E-mail: [email protected] largely the same for severaldecades (Hafner 1977,

634 July2000] ReproductiveSuccess ofLittle Egrets 635

1982).This populationalso is relativelyisolated; the TABLE 1. Annual estimates of mean clutch size, nearestmajor breeding areasare in Italy and Spain mean brood size, and proportion of nestssuccess- more than 400 km from the Camargue. ful for Little Egretsin southernFrance. Number of Breedingpopulation size.--Little Egretsin the Ca- nestsis in parentheses. marguenest in trees,usually in mixed colonieswith Nesting Cattle Egrets, Black-crownedNight-Herons (Nycti- Year Clutch size Brood size success coraxnycticorax), and SquaccoHerons (Ardeolarallo~ ides).Each year from 1967 to 1998, all colonieswere 1970 4.30 (156) -- 0.96 (142) locatedthroughout the study area usinga small air- 1971 4.39 (57) -- 0.90 (58) 1972 4.34 (215) -- 0.91 (115) craft(Hafner et al. 1994).From May to mid-July,each 1973 ------colonywas censusedweekly by the sameobserver 1974 ------(Hafner197•). Each colony was partitioned into sec- 1975 -- 3.02 (176) -- tors of approximately0.1 ha. Completenest counts 1976 -- 3.30 (152) -- were conductedwithin each sector,and any new 1977 -- 3.19 (117) -- nestswere added to the previouscount during each 1978 -- 3.10 (80) -- subsequentvisit. Based on repeated counts and 1979 -- 3.05 (150) -- marking of nestsin a sampleof sectors,we believe 1980 -- 3.06 (178) -- that overestimationwas extremelyunlikely and that 1981 -- 2.90 (221) -- the extent of underestimation was not substantial 1982 4.16 (58) 2.89 (189) 0.95 (57) 1983 -- 3.09 (144) -- (e.g. <10% in the largestcolonies). 1984 3.80 (51) 2.60 (279) 0.91 (54) Environmental variables.--We used the cumulative 1985 -- 3.31 (185) -- absolutevalue of daily minimum temperaturesbe- 1986 4.71 (135) 3.35 (155) 0.94 (156) low 0øC(i.e. the sumof the negativevalues) as an in- 1987 4.43 (30) 3.27 (137) 0.88 (49) dicationof the severityof cold winter weather(Haf- 1988 4.20 (97) 3.12 (136) 0.92 (122) ner et al. 1992, 1994). In accordancewith Hafner et 1989 3.83 (100) 2.98 (160) 0.88 (117) al. (1994), we used total cumulative rainfall between 1990 4.17 (106) 2.97 (207) 0.85 (148) Februaryand April as an indicationof spring rain- 1991 3.86 (117) 2.79 (163) 0.63 (173) fall. A strongwind, known as the Mistral, occursin 1992 4.02 (120) 2.69 (199) 0.81 (155) 1993 3.85 (121) 3.02 (325) 0.89 (132) the Camarguethroughout the yearand canbe a sub- 1994 4.03 (121) 2.78 (297) 0.92 (122) stantialsource of nestingfailure (Valverde1955, Haf- 1995 4.13 (104) 2.63 (238) 0.78 (147) ner 1978). Thus, we used the cumulative sum of the 1996 3.92 (79) 2.66 (277) 0.78 (151) maximumdaily wind speedduring the main nesting 1997 3.45 (60) 2.42 (358) 0.79 (73) period (15 April to 15 June)as a measureof wind ef- 1998 3.99 (116) 2.47 (283) 0.86 (121) fects.Climatic data were provided by the Tour du Valat meteorologicalstation, which is locatedin the centerof the Camargue. linear models to assess effects of environmental and Reproductiveparameters.--We assessed clutch size, density-dependentexplanatory variables on clutch brood size, and nestingsuccess for a sampleof col- size and brood size.We alsoincluded time (year) and onies during most years (Table 1). Nests were colonyeffects as potential explanatoryvariables. We markedalong transectsand their contentschecked usedlogistic regression to explorethe factorsinflu- once a week. To reduce disturbanceearly in the encingnesting success. Model selectionwas based on breedingcycle, we recordedclutch size at 10% or Akaike's Information Criterion (AIC; Akaike 1973, fewer of the number of nestsin large coloniesand Burnhamand Anderson1998). We checkedthe valid- 20% or fewer in small colonies. We defined brood ity of a Poissonassumption using the model devi- sizeas the numberof chicksalive at 20 to 25 daysof ance divided by its degreesof freedom (SAS 1997, age.After this age,chicks can of escapeby walking Burnham and Anderson 1998). We also checkedfor and may not be presentat nests(H. Hafner unpubl. temporal autocorrelationby testing the association data). We considereda nest to be occupiedwith the between the residuals of fitted models at times t and laying of the first egg and successfulif at least one t-1. youngreached the ageat whichit wasable to escape by walking. Becausecomplete censuseswere con- RESULTS ducted weekly and nestswere marked early during incubation,we did not expect a bias from finding Our selected model indicated that mean nestsin latterstages. Thus, we did not usethe May- field (1961)estimator. Rather, nesting success was es- clutchsize was positively associated with rain- timatedas the proportionof occupiednests that was fall (X 2 = 11.67, df = 1, P < 0.001) and nega- successful. tively associatedwith the total numberof Little Dataanalysis.--Data were analyzedwithin a gen- Egret nests(X 2 = 8.85, df = 1, P = 0.003; Fig. eralized linear modeling framework.We used log- 1). The model did not exhibit a substantial 636 BENNETTSET AL. [Auk, Vol. 117

4.8 3.6

3.4 4.6 3.2 4.4 ß

4.2 ß 2.8 4.0 ß 2.6 ß ß 3.8 2.4

3.6 2.2 50 100 150 200 250 300 350 0 1000 2000 3000 4000 5000 6000 Total rainfall(January-April) Total Number of Little Egret Nests

4.8 FIc. 2. Annual brood size relative to totalnumber of Little Egretnests in the studyarea. 4.6

N •.• 4.4 our selectedmodel becauseof its significance • 4.2 in an interaction term with the number of Little Egret nests (X2 = 8.31, df = 1, P = 0.004). Our e) 4.0 final model indicated a small degreeof over-

3.8 dispersionwith respectto a Poissonassump- tion (deviance/df = 2.3); however,adjusting 3.6 for overdispersionusing a quasi-likelihood 0 1000 2000 3000 4000 5000 6000 framework (SAS 1997,Burnham and Anderson Total Number of Little Egret Nests 1998)did not changewhich modelwas select- FIG. 1. Annual clutch size relative to total rainfall ed, and the significanceof explanatoryeffects (upper)and total numberof Little Egretnests in the changedonly slightly.As with clutchsize, we study area (lower). found no evidenceof temporal autocorrelation of the residuals. Our final modelof nestingsuccess indicated amount of overdispersionwith respect to a an effect of the total number of tree-nesting Poissonassumption (deviance/df = 1.3), nor herons,the proportionof the colonyconsisting was there evidenceof temporalautocorrelation of Cattle Egrets,wind speed,and severalinter- of the residuals. actionsamong these variables (Table 2). The to- Brood size also was negatively associated tal numberof tree-nestingherons and the pro- with the total numberof Little Egretnests (X 2 portion of Cattle Egrets in colonieswere neg- = 18.36, df = 1, P < 0.001; Fig. 2). Cumulative ativelyassociated with nestingsuccess (Fig. 3). rainfall wasnot significantas a main effect(X 2 Wind speedwas not significantas a main ef- = 0.25, df = 1, P = 0.617) but was retained in fect, but it contributedstrongly to two inter-

TABLE2. Maximum-likelihoodanalysis of variancetable for final logisticregression model of nestingsuc- cessof Little Egretsin southernFrance.

Source X2a df P Year 8.48 I < 0.001 Totalpopulation size of tree-nestingherons (TPS) b 2.26 I 0.133 Proportionof colonyconsisting of Cattle Egrets(PCE) 18.29 1 <0.001 Wind b 0.87 1 0.350 TPS x PCE 33.15 1 <0.001 PCE x wind 29.56 1 <0.001 TPS x PCE x wind 41.86 1 <0.001

Basedon type 3 likelihood-ratiostatistics and thusnot dependenton orderwithin model(SAS 1987). Not significantat ct = 0.05 as main effect,but retainedbecause of significancein interactions. July2000] ReproductiveSuccess ofLittle Egrets 637

1 ing populationsize of Little Egretsin the Ca- margue(Hafner et al. 1994).However, our anal- ysis did not indicatea subsequentinfluence of winter temperature on reproductive parame- ters. An associationbetween rainfall preceding the breeding seasonand reproductivesuccess also has been reportedfor severalspecies of herons(e.g. Ogden et al. 1980, Bancroftet al. 1988, Bildstein et al. 1990, Maddock and Baxter ß

0.6 1991),including the Little Egret (Hafner et al. 2000 4000 6000 8000 1 oooo 1994).Rainfall was the primary environmental Number of Tree-Nesting Heron Nests influence we observed on clutch size and brood size.Years of higherrainfall correspond with a greater surface area of Little Egret foraging habitat (Hafner et al. 1994),and prolongedpe- riods without surface water in freshwater marshescan result in prey die offs (Frederick 0.9 andCollopy 1989). Although rainfall preceding the breedingseason may havepositive benefits 0.8 relatedto foraginghabitat, rainfall duringthe breeding seasonhas been known to result in 0.7 colonyabandonment in someNorth American wetlands (e.g. Frederick and Collopy 1989). 0.6 0 0.2 0.4 0.6 0.8 Such an effect was not apparent in the Ca- Proportionof Cattle Egrets margue. Wind had a significanteffect on nestingsuc- F1G.3. Annualproportion of neststhat weresuc- cess,but only as an interactionterm with den- cessfulrelative to total number of tree-nestingher- ons(upper) and proportionof eachcolony consisting sity- dependenteffects. High winds canresult of Cattle Egrets(lower). in collapsednest structures,particularly when nestsare locatedon the edge of a colonywhere they have a more direct exposurewithout the action effects. In contrast to clutch and brood sheltering effects of adjacent trees. Mistral size,the residualsof our fitted modelof nesting windsin the Camarguetypically are out of the success exhibited temporal autocorrelation north. Thus, nests on the north ends of colonies without the inclusionof an additionalyear ef- also are more likely to be susceptibleto wind fect, which was also supportedby AIC and therefore included in our final model. An damage.During someyears (e.g. 1991),high winds can cause a substantial decrease in over- anomalousevent (seen only once since 1967) all nestingsuccess. occurred during 1986 where Eurasian Jack- Density-dependenteffects.--Density-dependent daws (Corvusmonedula) nesting in proximity to reproductionhas been observedin many ani- one small colony destroyedour entire sample mal populations (Sinclair 1989, Woiwod and of nestsfor that colony.Because the jackdaws Hanski 1992, Holyoak and Baillie 1996). De- probablyused the disturbanceof our visit to spite evidencefor density-dependenteffects on gainaccess to nestswhile the parentswere off their nests,we excludedthis colonyfrom our reproductionin otherwaterbirds (Gaston et al. 1983, Birkhead and Furness 1985, Hunt et al. analysis.Removal of this colonyfrom analysis did not alterour generalresults but would have 1986),evidence generally has beenlacking for introducedan additionalcolony effect and col- sucheffects on ardeids.Recently, Butler (1994) ony x year interaction. noted the generallack of evidencefor density dependence in the reproduction of wading DISCUSSION birds. Although our data are correlative,and inferencesregarding causality are therefore Environmentaleffects.--Severely cold winters limited, they are in contrastto Butler'sassess- reportedlyaffect winter and subsequentbreed- ment in that they indicate density-dependent 638 BENNETTSETAL. [Auk,Vol. 117 effectson reproductionof Little Egrets in the ACKNOWLEDGMENTS Camargue.Virtually all of thereproductive pa- We appreciatethe helpful commentsof Ga•tan Le- rameterswe evaluatedwere negativelyassoci- febvre,Douglas Mock, Brigitte Poulin,and two anon- ated with the numberof nestingLittle Egrets ymous reviewers.This work was carriedout under and/or othertree-nesting herons. In eachcase, licenseof the FrenchMinistry of Environmentand the associationwas explainedbetter by num- the CRBPO(French National Ringing Center). We are bers of heronsthan by a linear trend overtime gratefulfor the supportprovided by StationBiolo- (i.e. clutchand brood size), or was in addition gique de la Tour du Valat. to a linear trend (i.e. nestingsuccess). Competitionfor nest sitesin the Camargue LITERATURE CITED may occurbecause suitable undisturbed stands of trees are quite limited (Hafner 1982). Given AKAIKE,H. 1973.Information theory and an exten- the substantial increase in the number of Cattle sionof the maximumlikelihood principle. Pages 267-281 in Proceedingsof the 2nd International Egretsin the Camarguesince 1967, the poten- Symposiumon InferenceTheory (B. N. Petran tial for interspecificcompetition for nest sites and E Csfiki,Eds.). Akad•miai Kiadi, Budapest, alsois likely to haveincreased, especially con- Hungary. sideringthe aggressivenature of CattleEgrets BANCROFT,G. t., J. C. OGDEN, AND B. W. PAttY. 1988. with respectto nest sites (Blaker1969, Sieg- Colony formation and turnover relative to rain- fried 1972).Our data suggestthat nestingsuc- fall in the CorkscrewSwamp area of Floridadur- cessof Little Egretsis densitydependent only ing 1992through 1985. Wilson Bulletin 100:50- when considered in the interaction with the 59. BENOit,F. 1933. La Camargue. Henri LaurensPress, proportionof Cattle Egretsin nestingcolonies. Paris. Prior to widespread distribution of Cattle BILDSTEIN, K. L., W. Post, J. JOHNSTON,AND P. C. Egretsin the Camargue,Hafner (1977) docu- FREDERICK.1990. Freshwater wetlands, rainfall, menteda relationshipbetween nesting success and the breedingecology of White Ibisesin of Little Egretsand the positionof the nestrel- coastal South Carolina. Wilson Bulletin 102:84- ative to the centeror edgeof the colony.Nests 98. on theperiphery of a colonywere more suscep- BIRKHEAD,T. g., AND g. W. FURNESS.1985. Regula- tible to damagefrom wind than thoselocated tion of seabirdpopulations. Pages 145-163 in Be- more centrally (Hafner 1977).Anecdotal obser- haviouralecology (R. M. Siblyand R. H. Smith, vationssuggest that Cattle Egretstend to oc- Eds.). Blackwell SciencePublishers, Oxford. BLAKER,D. 1969.Behaviour of the CattleEgret Ar- cupy the more centrallocations of coloniesin deola ibis. Ostrich 40:75-129. the Camargueand may even displaceLittle BURNHAM K. P., AND D. g. ANDERSON. 1998. Model Egretsfrom thesesites. This is currentlybeing selectionand inference:A practical-theoreticap- investigated,and if it occurs,it may providea proach.Springer-Verlag, New York. mechanismfor the negativeeffect on nesting BUTLER,R.W. 1994.Population regulation of wading success.The phenomenonalso would explain ciconiiform birds. Colonial Waterbirds 17:189- the interaction that we observed between the 199. number and proportion of Cattle Egrets and DHONDT, g. g., B. KEMPENAERS,AND E ADRIAENSEN. wind. 1992.Density-dependent clutch size causedby habitatheterogeneity. Journal Animal Ecology Dhondtet al. (1992)postulated that density- 61:643-648. dependentdecreases in reproductionoccurred FREDERICK,P. C., ANDM. W. COLLOPY.1989. Nesting becauseof more frequentoccupation of rela- successof five ciconiiformspecies in relation to tively poorhabitats at higherdensities (thereby water conditionsin the FloridaEverglades. Auk reducingoverall breeding success). The spatial 106:625-634. distribution of heron colonies within the Ca- GASTON, A. J., G. CHAPDELAINE, AND D. G. NOBLE. margue has remained essentiallythe same 1983.The growth of Thick-billedMurre chicks in colonies in Hudson Strait: Inter- and intra-col- throughout the study (Hafner 1977, 1982). ony variation. Canadian Journal Zoology 61: Thus, increasesin densityhave not resultedin 465-475. spatial expansionof coloniesinto previously HAFNER,H. 1977. Contribution a l'•tude de quatre unused habitats. However, our results do not esp•cesde herons (Egrettagarzetta L., Ardeola precludethe possibilitythat Little Egretsare ralloidesScop., Ardeola ibis L., Nycticoraxnycticor- usingpoorer nest sites within colonies. ax L.) pendant leur nidification en Camargue. July2000] ReproductiveSuccess of LittleEgrets 639

Ph.D.dissertation, University Paul Sabatier, Tou- HUNT, G. L., Z. A. EPPLEY, AND D.C. SCHNEIDER. louse, France. 1986. Reproductive performance of seabirds: HAFNER,H. 1978.Le succ•sde reproductionde qua- The importanceof populationand colonysize. tre esp•cesd'Ardeid•s Egrettag. garzettaL., Ar- Auk 103:306-317. deolar. ralloidesScop., Ardeola i. ibisL., Nycticorax MADDOCK,M., AND G. S. BAXTER.1991. Breeding n. nycticoraxL. en Camargue. Revue Ecologie successof egretsrelated to rainfall: A six-year (Terre Vie) 32:279-289. Australian study. Colonial Waterbirds 14:133- HAFNER,H. 1982.Creation of a breedingsite for tree- 139. nesting herons in the Camargue, southern MAYFIELD,H. F. 1961. Nesting successcalculated France.Pages 216-220 in Managing wetlands from exposure.Wilson Bulletin 73:255-261. and their birds (D.A. Scott,Ed.). International OGDEN, J. C., H. W. KALE, AND S. A. NESBITT. 1980. Waterfowl and Wetlands ResearchBureau, Slim- The influence of annual variation in rainfall and bridge,United Kingdom. water levelson nestingby Floridapopulations of HAFNER,H., V. Boy, AND G. GORY.1982. Feeding wading birds. Transactionsof the Linnaean So- methods,flock size and feedingsuccess in the ciety of New York 9:115-126. Little EgretEgretta garzetta and the Squacco Her- SASINSTITUTE. 1997. SAS/STAT software: Changes and enhancementsthrough release 6.12. SAS In- on Ardeola ralloidesin Camargue, southern France. Ardea 70:45-54. stitute,Inc., Cary, North Carolina. SIEGFRIED,W. R. 1972.Breeding success and repro- HAFNER,H., ANDM. FASOLA.1997. Long-term mon- ductiveoutput of the CattleEgret. Ostrich 43:43- itoring and conservationof heronsin Franceand 55. Italy. Colonial Waterbirds 20:298-305. SINCLAIR,A. R. E. 1989.Population regulation in an- HAFNER, H., Y. KAYSER, AND O. I•INEAU. 1994. Eco- imals. Pages197-241 in Ecologicalconcepts (J. logical determinants of annual fluctuationsin M. Cherrett, Ed.). Blackwell SciencePublishers, numbersof breedingLittle EgretsEgretta garzet- Oxford. ta in the Camargue,S. France.Revue Ecologie VALENTINE,J. M. 1958. The Cattle Egret at Chinco- (Terre Vie) 49:53-62. teague, Virginia. Raven 29:68-95. HAFNER,H., O. PINEAU,AND J.P. WALLACE.1992. The VALVERDEJ. A. 1955. Essai sur l'Aigrette garzette effectsof climateon the sizeof the CattleEgret (Egretta g. garzetta) en France. Alauda 23:147- (Bubulcusibis L.) populationin the Camargue. 171. Revue Ecologie(Terre Vie) 47:403-410. WOIWOD, I. P., AND I. HANSKI. 1992. Patterns of den- HOLYOAK, M., AND S. R. BAILLIE. 1996. Factors influ- sity dependencein mothsand aphids.Journal of encingthe detectionof densitydependence in Animal Ecology61:619-629. Britishbirds. Oecologia 108:54-63. AssociateEditor: J. S. Sedinger