Experimental Zoogeography of Islands - Plant Ecology at Syracuse

Experimental Zoogeography of Islands: Defaunation and Monitoring TechniquesAuthor(s): Edward O. Wilson and Daniel S. SimberloffSource: Ecology, Vol. 50, No. 2 (Mar., 1969), pp. 267-278Published by: Ecological Society of AmericaStable URL: http://www.jstor.org/stable/1934855 .Accessed: 18/01/2011 13:24Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unlessyou have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and youmay use content in the JSTOR archive only for your personal, non-commercial use.Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at .http://www.jstor.org/action/showPublisher?publisherCode=esa. .Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printedpage of such transmission.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact support@jstor.org.Ecological Society of America is collaborating with JSTOR to digitize, preserve and extend access to Ecology.http://www.jstor.org

Early Spring 1969 EXPERIMENTAL DEFAUNATION OF ISLANDS 267some survive in the desert even during years ofdrought, without any artificial irrigation with only80 mm of rainfall, and in the subtropical zone with200-300 mm of rain.These results and those obtained earlier (Gindel1968b, 1969) indicate that further study is neededto clarify to what extent augmented stomataldensity and the presence of stomata on the upperepidermis in woody xerophytes grown in thedriest climate and poorest soil serve to increasethe ability of the tree to absorb atmospheric moisture,when moisture tension in the soil and withinthe plant reach their highest values.ACKNOWLEDGMENTSThis work was accomplished with the financial assistanceof the J.N.F. I express my thanks to Mr. JosephWeitz, pioneer of the 50 years of Israeli afforestation, andto Prof. A. H. Halevy for his criticism of the manuscript.LITERATURECITEDFahn, A. 1964. Some anatomical adaptations of desertplants. Phytomorphology 14: 93-102.Gindel, I. 1952. Some anatomical features of the indigenouswoody vegetation in Israel. Res. Coun.Israel 2: 1-16.. 1957. Acclimatization of exotic woody plantsin Israel: The theory of phytoplasticity. MateriaeVegetabiles 11: 81-101.. 1961. The water balance in xerophytes. Intern.Union For. Res. Org. 13 Congress 21: 2-4.. 1964a. Seasonal flucuations in soil moisture un-der the canopy of xerophytes and in open areas. CommonwealthForest. Rev. 43: 219-234.. 1964b. Transpiration of Aleppo pine as a functionof environment. Ecology 45: 868-873.. 1966. Attraction of atmospheric moisture bywoody xerophytes in arid climates. CommonwealthForest. Rev. 45: 297-321.. 1967a. The correlation between cambium activityand transpiration rate in Aleppo pine. 14th I U F R 0Congress 4: 188-207.. 1967b. The relationship between transpirationand six ecological factors. Oecol. Plant. 8: 227-239.-. 1967c. Afforestation of the desert. "Ariel,"Jerusalem 19: 13-24.. 1968a. Some ecophysiological properties of threetree xerophytes grown in the desert. Oecol. Plant. 3:49-67.1968b. Dynamic modifications in alfalfa leavesgrowing in subtropical conditions. Physiol. Plant. 21:1287-1295.. 1969. Stomata constellation in the leaves of cotton,alfalfa, maize and wheat plants as a function ofsoil moisture and environment. Physiol. Plant. (inpress).Kohl, F. G. 1886. Transpiration der Pflanzen und ihreAuswirkung auf die Ausbildung pflanzlicher Gewebe.H. Bruhn, Braunsweig. 34 p.Sorauer, H. 1873. Einfluss der Wasserzufuhr auf dieAusbildung der Gerstenpflanzen. Bot. Z. 31: 145-159.Stocker, 0. 1960. Physiological and morphologicalchanges in plants due to water deficiency. Arid ZoneResearch, UNESCO, 15: 63-104.Zelitch, I. 1961. Biochemical control of stomatalopening in leaves. Proc. Nat. Acad. Sci. Wash. 47:1423-83.EXPERIMENTAL ZOOGEOGRAPHY OF ISLANDS: DEFAUNATIONAND MONITORING TECHNIQUESEDWARD 0. WILSON AND DANIEL S. SIMBERLOFF1The Biological Laboratories, Harvard University, Cambridge, Massachusetts 02138(Accepted for publication December 16, 1968)Abstract. In order to facilitat experiments on colonization, a technique was developedthat permits the removal of the faunas of very small islands. The islands are covered by atent and fumigated with methyl bromide at concentrations that are lethal to arthropods butnot to the plants.Seven islands in Florida Bay, of varying distance and direction from immigrant sources,were censused exhaustively. The small size (diameter 11-1i m) and ecological simplicityof these islands, which consist solely of red mangrove trees (Rhizophora mangle) with nosupratidal ground, allowed the location and identification of all resident species. The terrestrialfauna of these islands is composed almost exclusively of arboreal arthropods, with 20-50species usually present. at any given moment. Surveys of these taxa throughouthe FloridaKeys, with emphasis on the inhabitants of mangrove forests, were made during 1967 in orderto estimate the size and composition of the "species pool."The seven experimental islands were defaunated in late 1966 and early 1967, and the colonistswere monitored for 17-20 man-hours every 18 days. Precautions were taken to avoid artificialintroductions during the monitoring periods.INTRODUCTIONArthur and Wilson (1963, 1967), who postulatedThe first attempt to formulate a quantitative an equilibrium number of species on an islandtheory which would unite an ever-increasing mass determined by the intersection of immigration andof insular biogeographic data was made by Mac- extinction curves drawn as a function of the num-1Present address: Department of Biological Science, ber of species already present. In addition toFlorida State University, Tallahassee, Florida 32306 speculation on the forms of these curves and ef-

268 EDWARD 0. WILSON AND DANIEL S. SIMBERLOFF Ecology, Vol. 50, No. 2fects of varying both island area and distance fromthe source, they derived the equationwhere 9-mean -X -to1.oto-0 1.15 mean (1)the hypothetical equilibrium numberof species on an islandthe mean of 9i for a number of verysimilar islands ithe extinction rate (or turnover rate,since an equilibrium requires that extinction- immigration) in number ofspecies per time at equilibriumtime required for number of speciespresen to increase from 0 to 90% of D.Later the same authors elaborated upon the generalshapes of immigration and extinction curves,discussed colonization rates and curves, and predictedthe distribution of survival times of speciesthat succeed in colonizing an island (MacArthurand Wilson 1967).Data to test these hypotheses are scarce. Mostbiogeographic information, when combined withinformation furnished by the scant fossil record,can at best lead to descriptions of broad patternsof distribution and zoogeographic regions and tohypotheses about the essentially long-term processesresponsible for them (e.g., Darlington1957). Such theories, though occasionally sug-rrpqtil-p fnr% rnri*i e vr t-,L foi to orn- -- i the--1E I~... X E BXMb.existing taxonomic distributions and the underlyingcolonization mechanisms for most islands.They neglect the short-term, even daily events thatlargely determine the parameters of colonizationon many islands. Such events become decisivelymore importanthan evolution as distance of islandfrom source area decreases. In particular,the classical zoogeographic methods do not providea test for the existence of an equilibriumnumber of species or for the accuracy of equation( 1 ).The literature on biotic dispersal, both anec-dotal and systematic, provides information on therelative importance of active and passive transportand a wealth of data on the specific agents of passivetransport (e.g., Wolfenbarger 1946). Certaincases of overseas dispersal have even beentraced to specific meteorological events (French1964). For no island, however, is the time courseof colonization by even a large fraction of theinhabitants known. The large number of recordsof long distance dispersal implies that immigrationrates to islands are probably high, especially fororganisms capable of passive transport by wind.But the shape and determinants of the immigrationcurve cannot be deduced. At best a lowerlimit might be given.In the recent studies of Surtsey, a new volcanicisland near Iceland, there has been a deliberateattempto documenthe colonization process fromCAPCCFIG. 1. The southern tip of Florida and the Florida Keys. The rectangles enclose the experimental areasshown in detail in Figures 3-5.

Early Spring 1969 EXPERIMENTAL DEFAUNATION OF ISLANDS 269NW-FIG. 2. Upper: Island El, the second smallest island inthe experimental series. Lower: Island E9, the largest ii) that the animal diversity be sufficientoisland in the experimental series; note also the presenceof supratidal mud.allow statistical treatment, and the organisms bephysically large enough to insure recording inconthebirth of the island (Fridriksson 1967, Her- spicuous forms as well as conspicuous ones;mannsson 1967, Lindroth et al. 1967). But theiii) that the extremely small size of the islandsinfrequent natural occurrence of such events tendscompensates for their relative nearness to sourceto render quantitative biogeographic theory theareas and therefore produces a distance effect.product of abstract speculation and of enlarge- During an exploratory trip to the Florida Keysment and sophistication of untestable hypotheses. in 1966, we found that conveniently accessibleWhat is needed, clearly, is a method either of producingnew islands similar to natural ones or ofsterilizing preexisting islands.THE EXPERIMENTAL ISLANDSconsist entirely of one to several red mangrovetrees (Rhizophora mangle), with rarely a smallblack mangrove bush (Avicennia nitida). Occasionally,in areas of weak tide, the trees are surroundedby small areas of supratidal mud andsand. Such land is only intermittently supratidal,since it is completely flooded during prolongedwinds of 10 knots or more. Individual islandsmay differ in several minor respects such as thenumber of large trunks and amount of dead barkthey contain, but on the whole they are remarkablysimilar to each other in physical appearance.A fruitful defaunation experiment with theseislands requires the following conditions:i) that there be enough of the islands for replicationand variation in distance to the nearestsource area;small islands were numerous at most distances upto 200 m from the nearest large islands. Beyond200 m they were rare, although somewhat largerislands (diameter 50 m and greater) were plentiful.A very few small islands suitable for our purposeswere located 0.2-1.4 km from the nearestsource area.A set of such islands were chosen for experimentationand their faunas carefully surveyed.Since the breeding fauna of islands this size con-Along the Overseas Highway of southern Florida(U. S. Route 1) there exist a vast array ofsmall, approximately circular islands situated inshallow bay water (Fig. 1). Together, they constitutea potential natural "laboratory" for experimentalbiogeography. The islands with diameter sists almost entirely of species of insects and spi-10-20 m (Fig. 2) are 5-10 m tall and usually ders, we made a reference collection of land ar-TABLE 1. Parameters of experimental islandsDistance from Initial no.Island Diameter nearest source arthropodname (i) (i) species LocationEl 11 533 25 off Squirrel KeyE3 12 172 31 in Rattlesnake LumpsSeries 1 ST2 11 154 29 off Saddlebunch KeyE2 12 2 43 off Snipe KeysE65 15a 73a 30a in Johnston Key MangrovesE7 25 15 29 near Manatee CreekE8 18 1, 188b 29 between Calusa Keys andBob Allen KeysSeries 2 E9 18 379 41 between Calusa Keys andBob Allen KeysElope lha 37a 208 near Bottle KeyaControl islandb"Stepping Stone" (SI) 572 m away.

270 EDWARD 0. WILSON AND DANIEL S. SIMBERLOFF Ecology, Vol. 50, No. 2thropods of the Florida Keys with the assistance that bred on our islands during the course of theof Robert Silberglied. This collection enabled us experiment (1966-68) were one or two pairs eachto identify the arthropods on the experimental of the Green Heron, White-crowned Pigeon, andislands, and it provided a measure of the size andcomposition of the species pool of the presumedsource area, the entire Florida Keys.Animal diversity on the small mangrove islandsproved to be adequate for statistical purposes.About 75 insect species (of an estimated 500 thatinhabit mangrove swamps and an estimated totalof 4,000 that inhabit all the Keys) commonly liveon these small islands. There are also 15 speciesof spiders (of a total Keys complement of perhaps125 species) and a few scorpions, pseudoscorpions,centipedes, millipedes, and arboreal isopods.At any given moment 20-40 species of insects and2-10 species of spiders exist on each of the islands.Many birds visit the islands either to roost orto forage, but very few nest there. The only birdsGray Kingbird. Snakes (mostly water snakes ofthe genus Natrix) occasionally swim to smallmangrove islands, and raccoons visit islands locatedon shallow mud banks. These vertebrateswere not included in the censuses.On the basis of their distance and directionfrom the nearest source area, their size, and theiraccessibility, the islands listed in Table 1 werechosen for defaunation. Two islands (E6 andE10), one each in the vicinity of the two groupsof experimental islands, were selected to serveas control islands; and they were censused at thesame time as the experimental islands. In orderto ascertain whether any long-range change hadoccurred in the general fauna of small mangroveislands in Florida Bay during the course of theBarracuda )Ys~~M afford Key J e17< = OR Sqirref Kaydaldn&KeyNappjj JackKeDy/U S ~~~~~Drequez H iy

Early Spring 1969 EXPERIMENTAL DEFAUNATION OF ISLANDS 271Foxtrot X5ysCa/usa Xgsr6''.ESBob Allen K~ysFLORIDAB~.AY0 5000 Metersoriginal faunas of the distance effect as definedby MacArthur and Wilson (1967). Qualitativedifferences in faunas of increasingly distant islandswere even more striking than quantitative ones.For many groups-especially ants and spidersthe-distanceffectis so regular that one can guessnot only the species number but also the approxi-.mate species composition. For example, centipedesand pseudoscorpions were found almostexclusively on islands of less than 200 m distancefrom the nearest large island.Islands of this minute size can be remarkablylong lived. All Rhizophora mangle trees, andparticularly those on low, small islands, havenumerous aerial roots: both brace roots growingout from the main stem and prop roots growingFIG. 4. Map showing the location of the experimentalislands of series 2.down from branches. The latter can emerge fromas high as 4 m above the water (LaRue andMuzik 1954) and in time can become as thickMA NATEEas the main stem. Rhizophora trees reproduceviviparously, their seeds growing on the tree untilthey become large, long, pointed seedlings. WhenE7 Bqy Pointthey drop, frequently from the upper canopy, someevidently plant themselves where they fall (Davis1940). Mangrove swamps are often quite thick,and on small islands there may be several trees.Stems, thickened prop roots, and brace roots in-~~ Cross Key,0 1/2 2 Miles0 1000 2000 MetersFIG. 5. Map showing the location of the original testisland E7.experiment, a second census of the control islandswas made at the close of the experiment.The two series of experimental islands form twowidely separated groups. Series 1 lies within theGreat White Heron National Wildlife Refuge,in a region north of Sugarloaf Key (Fig. 3). Series2 is in the Everglades National Park near theCalusa Keys north of Islamorada (Fig. 4). E7,which was chosen primarily to test a method ofdefaunation, is located in Barnes Sound south ofthe Glades Canal (Fig. 5).The islands were also selected to show as muchvariation as possible in the degree of isolationfrom the immigrant sources. The coefficient ofcorrelation for island series 1 between "distancefrom nearest source" and "number of arthropodspecies" is -0.83, indicating the operation in theIKilometer; ~: S KHIVBf13ake~s K~yElIFIG. 6. Map showing El and surrounding region as itappeared in 1851-57.IIW~st;I AKilometer/,Sq uirreIl'~E1IFIG. 7. The same region shown in Figure 6 as it appearedin 1964.

272 EDWARD 0. WILSON AND DANIEL S. SIMBERLOFF Ecology, Vol. 50. No. 2termingle and reproduce themselves so that it isimpossible to distinguish how many trees arepresent, which are which, which is the initial one(even if it is still present), and which are the mainstems.Maps of the Florida Keys, especially of theSugarloaf Key-Key West area, have been remarkablydetailed and accurate with respect to smallmangrove islands for at least a century. Thesetell us that E7 and all the islands in series 1 haveexisted at least 25 years. The other experimentalislands were not mapped in sufficient detail todetermine longevity. The size and shape of individualislands may change, even drastically, butin general the islands survive the climatic catastrophesto which they are repeatedly subjected.For example, Figure 6 shows El (Blake's Key)and surrounding areas as it appeared in 1851-57,while Figure 7 shows the same island region onthe Coast and Geodetic Survey map of 1964. Ittherefore seems reasonable to assume that thefaunas found initially on the experimental islandswere the product of many years' colonization. Ourexperiments have subsequently shown that theminimum ages of the islands greatly exceed thetime required to reach species equilibrium from astart of zero species (Simberloff and Wilson1969).DEFAUNATIONOur first attemp to defaunate islands made useof an insecticide spray of short-lived residual effect.On July 9-10, 1966, two experimentalislands (El and E2) were sprayed until drippingwith 60 g parathion, 240 g diazinon and 180 gPylac sticking agent per 1001 fresh water. Whenthe islands were examined 1 day later, all thesurface and bark fauna and most borers weredead. However, the following live animals werefound in thin hollow twigs. El: one adult wasp,Scleroderma macrogaster (Bethylidae); workersand larvae of Crematogaster ashmeadi (Formicidae).E2: two larvae of ?Styloleptus biustus.(Cerambycidae); two colonies of Paracryptocerusvarians (Formicidae); and one colony of Tapinomalittoral (Formicidae). A more throughexamination 9 days after spraying produced thefollowing live and apparently healthy insects. El:two larvae of ?Styloleptus biustus (Cerambycidae).E2: one larvae of ?Styloleptus biustus(Cerambycidae); one lepidopteran larva; and twocolonies of Xenomyrmex floridanus (Formicidae).The large populations of the ant colonies andadvanced development of the beetle larvae precludedthese insects having immigrateduring the9 days following spraying.The results on El and E2 showed that a spraycannot be expected to penetrate all the hollowtwigs of a Rhizophora island. Some twig dwellerswill probably survive, particularly if they inhabitnarrow twigs (Tapinoma and Xenomyrmex) orelse pack their tunnels tightly with powdery excreta(Styloleptus). Had we attempted to killthese survivors by using a spray with a long-livedor more powerful residue, we would have postponedthe beginning of the colonization curve,since immigrants may be killed by the residue.Because we had to be certain that all colonists,or at least all but those belonging to a small numberof known taxa, would be destroyed immediately,we abandoned the spray technique.We turned next to the more difficult alternativetechnique of fumigation. The method has theobvious advantages that the residual effect of agas is negligible, and even the inhabitants of hollowtwigs are contacted. This is the standardmethod used by professional exterminators ontermites and other wood-boring insect pests.The fumigation of living plants is a relativelyuncharted area. The biochemical effects of fumigantson plants are poorly known, damage usuallybeing described in terms of the physical appearanceof the plant shortly after fumigation(Page and Lubatti 1963). To our knowledge, theonly prior fumigation of red mangrove trees wasby B. P. Stewart (pers. comm.), who used methylbromide on small plants. With 32 kg/1000 m3for 2 hr at 320C and 40 kg/1000 m3 for 2 hr at270C he found that all insects apparently werekilled, although some did not die until 48 hr afterfumigation. Mortality of cockroach eggs wasunrecorded in Stewart's study. Damage to theplant was limited primarily to young leaves andtwigs, a result that is in accord with the generalfinding of Page and Lubatti (1963) that the growingstages of both plants and insects are the mostsusceptible to fumigation.In our preliminary trials we were limited bythe fact that fumigants highly soluble in water,such as hydrogen cyanide, could not be used.These substances would be expected either toleave the fumigation tent through the water orto concentrate to an unknown extent in the waterunder the island. Either effect would greatlycomplicate the maintenance and measurement ofgas concentration. Consequently, the followingfour relatively insoluble fumigants were tried:Pestmaster Soil Fumigant-1 (methyl bromide98% wt, chloropicrin 2% wt); Acritet 34-66(acrylonitrile 34% vol, carbon tetrachloride 66%vol); Dow Ethylene Oxide (ethylene oxide100%) ; and Vikane (sulfuryl fluoride 95% wt,inert ingredients 5% wt). Field tests were performedon small Rhizophora trees in Matheson

Early Spring 1969 EXPERIMENTAL DEFAUNATION OF ISLANDS 273Hammock Park, Coral Gables, by National Ex- canopy. Instead, whole sections of the tree beterminatorsof 'Miami, Florida, under the direc- came completely brown while others remainedtion of Steve Tendrich in consultation with the almost unscathed. The upper canopy was alwaysauthors. Additional tests on the mortality of man- more severely browned, but there seemed to begrove colonists were performed at National Ex- no other consistent damage pattern.terminators' laboratory in Miami.Abscission of the dead leaves began within aWith the temperature outside the fumigation week and, often aided by wind, was usually comtentremaining at 240-290C, we determined which plete within 8 weeks. At this time damaged treesduration-concentration intensities caused no more looked superficially dead except for remainingthan bearable damage to the tree yet were lethal green sections. Closer examination showed thatto all the mangrove arthropods. Acritet, Vikane, the damaged trees could be conveniently dividedand ethylene oxide caused extensive, apparently into three classes, as follows:irreversible damage to red mangrove and were For those trees where browning of leaves andrejected. However, methyl bromide at 22 kg/ shoots appeared complete by 72 hr the damage1000 M3 for 2 hr caused browning of only about was irreversible. Within a few months bark began10% of the leaves 1 week after treatment, with to peel and twigs with green wood could not beno visible later effects. This duration and concen- found. A year later there was no evidence of life.tration killed all the mangrove inhabitants (in- Island E8 is in this class. In the second category,cluding such resistant forms as roach eggs and which includes only E7, a large majority of thelepidopteran pupae) with one possible exception. leaves and shoots were browned and died, whileWhen data from all tests at this duration-concen- a small section remained green. In the thirdtration are lumped together, only 95% of ceram- category, the trees were less severely damagedbycids and 80% of weevils were dead when dug (60% or fewer browned leaves). Green sectionsout of their burrows 1-10 days after fumigation. remained unchanged, while leaf abscission in mostThe remainder all died within 2 days of removal. of the damaged sections was accompanied by theIt is tempting to ascribe the death of these few sprouting of new shoots and leaves from almostlarvae to a delayed effect of the fumigant. How- all branches. Within 2 months the trees appearedever, a control experiment consisting of the re- normal except for occasional leafless patches inmoval of five unfumigated cerambycid larvae the extreme upper canopy. Island El, which 2(Styloleptus biustus) from their respective bur- days after fumigation appeared to have about 50%rows caused death within 5 days without any brown leaves, recovered after 1 month to the exadditionaltreatment. Monro and Delisle (1943) tent that it did not appear to have incurred morereport that many insects are active soon after damage than similar trees do during heavy storms,exposure to methyl bromide but succumb in time and 2 months after defaunation seemed quite nor-(some lepidopterous larvae surviving for 2 mal except for a small bare patch in the uppermonths), and that this delayed effect is especially canopy. At no time did its bark crack or peel.common in fumigations at the low concentrations In most damaged trees of the firs two classesusedin our own field tests. This peculiarity is where either the whole tree or all but a small secconfirmedby Chisholm (1952) for Japanese beetle tion was killed-there was copious oozing of saplarvae and other insects. Of at least equal im- for up to a week after fumigation. Such oozingportance, the results from our experimental islands was never observed in untreated trees, nor in those(Simberloff and Wilson 1969) strongly suggest fumigated trees which showed good leaf replacethatthe cerambycids were eliminated by the fumi- ment.gation treatment. We therefore conclude that even The action of methyl bromide on living plantsthe deep-boring beetle larvae were very probably has been studied rather extensively by severalkilled by the treatment.authors. Mainwaring (1%1) examined the phys-The overt physical damage to the trees con- iological responses of several plants to methylsisted of a browning of leaves and shoots and bromide, and his numerous observations implyoccasional heavy oozing of sap. At 61 kg/1000 that its major effect is auxin inhibition. Them3 the leaf browning was manifest upon tent re- damage we observed seemed too drastic and cermoval,but at 35 kg/1000 m3 and less there was tainly too immediate to be attributed solely tolittle or no immediate browning. The effect becameevident at about 24 hr and worsened in termsof fraction of leaves browned through 72 hr, bywhich time up to 90%o of the leaves were brownand after which no further browning occurred.Browning did not occur randomly throughoutheauxin inhibition, although the presence of such abiochemical action cannot be ruled out.Whatever the effect of methyl bromide onRhizophora mangle, it apparently had the expectedQ1o of 2-3. Our initial tests in Matheson Hammockwere conducted on small plants well shaded

Early Spring 1969 EXPERIMENTAL DEFAUNATION OF ISLANDS 275disperse the gas. Chloropicrin (tear gas), addedto the odorless methyl bromide as a safety measure,was allowed to drip into burlap sheets at thebottom of the tubs. Concentration of the methylbromide was measured frequently by a Gow-Macthermal conductivity meter with a self-containedpump attached to a testing station 2.5 m abovethe water and well away from the two shootingstations.The methyl bromide concentration was graduallybuilt up from 0 to 22 kg/1000 m3 over 30 min,then kept at 22-25 kg/1000 m3 for 2 hr. Outsideair temperature during this time remained steadyat about 270C. At the end of the fumigation thegas was released through seam openings for 45min, and the tent was then removed.Immediately following the dissipation of thegas, E7 was explored for a total of 6 man-hours.The results generally confirmed the evidence fromthe Matheson Hammock tests. Dead individualsrepresenting 11 of the 23 species observed beforefumigation were found, frequently in large numbersand in all life stages. The details are givenin Table 2. The only live animals encounteredwere larvae of two deep-boring beetles. Theseformed only a small fraction of the original population,as shown by the following recovery data:of 59 larvae of Styloleptus bustus (Cerambycidae),2 were alive and 57 dead; of 10 larvae ofPseudoacalles sp. (Curculionidae), 2 were aliveand 8 dead. One of the two live cerambycids wasextremely sluggish and died an hour after removal.The other cerambycid and the two weevils seemedhealthy, though one of the weevils became torpidand died 9 hr after removal.The tree was severely damaged as describedpreviously. Within 2 months 85%o of the foliagewas dead, although the single living section hasremained green until at least May 1968. Thusthe procedure worked essentially as expected, andthis aspect of the fumigation was left unchanged FIG. 9. Upper: the fumigation tent being opened overin subsequent defaunations, except that the work E2 from a tower support. Middle: E2 partially covered.Lozccr: E9 covered by the fumigation tent on a towerwas henceforth done at night in all but one in- support.stance (El). The cubical scaffolding frameworkhad disadvantages, chief among them being thelarge crew necessary to erect it (at least five men)and the number of pieces and weight of the equipment.This necessitated numerous boat trips betweenE7 and our staging area near ManateeCreek, since the shallow water along the route precludedthe use of large craft. This problem wasgreatly aggravated on the other islands, which arelocated considerably further from a convenientstaging area. We therefore set out to devise atower to be located in the center of the island andrising high enough so that most of the tent weightrested on it. This device proved to be more vul-*..-...~~~~~~~~~~~~~~~~~~~........:~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~..... ...........:........nerable to high winds, but it has the great ad antageof requiring less equipment and a smallercrew to assemble it. National Exterminatorscontracted a steeplejack to perfect the towermethod. After a month of experimenting on islandsnear Key Largo, the following procedurewas adopted and used for the remaining islands.A collapsible 10-m triangular steel tower waserected in the center of the island, either proppedon a large root or forced deep into the ground.The steeplejack then climbed to the top and threwout three guy wires which were fastened down

276 EDWARD 0. WILSON AND DANIEL S. SIMBERLOFF Ecology, Vol. 50, No. 2TABLE 3. Extent of damage to foliage and subsequent recovery on experimental islandsIsland Fumigation Time Extent ofdatedamageDamage location Recovery of foliage(burnt leaves)El.............. 3/13/67 day 50% general complete in 2 months,except for extremeupper canopyE2.............. 3/7/67 night 10% upper canopy complete in 2 monthsE3.............. 3/17/67 night 5% upper canopy complete in 2 monthsE7.............. 10/10/66 day 85% all except one nonesectionE8 .............. 4/17/67 night 100% complete noneE9.............. 4/7/67 night 20% upper canopy and little in upper canopyone othercomplete elsewhere insection2 monthsST2............. 4/20/67 night 5% scattered complete in 2 monthswith mudscrews or stakes, depending on the substrate.Using a block and tackle, one man on thetower and one on the island lifted the rolled tentuntil it formed a triangle over the island. Thetent was then carefully unrolled down the guidewires until the island was covered (Fig. 9). Itwas impossible to unfurl the tent from the towerin winds greater than 8 knots. Sandbags andstakes were used to keep the tent edges submerged.Equipment used in the tower method was iden-tical to that described already for the scaffoldingmethod, and was similarly situated. The procedurewas identical except that gas was adminis-Precautions were taken at all monitoring2 totered from 0.5-kg cans and concentration and prevent contamination of the experimental islands.duration were increased to 28-33 kg/1000 m3 for Boats were never brought in contact with the is-2.5 hours on all islands but El. After fumigation lands and only seldom tied to them; usually theythe methyl bromide was released through tent were anchored at least 12 m away. Before wadingseams as with scaffolding, but in the tower system to an island, experimenters examined clothing,tent removal was more difficult. At low wind equipment, and persons for animals, which werespeeds the tent could be laboriously refurled and destroyed. To lessen the chance of phoresy, ingraduallylowered to the boat. We fortuitously vestigatorsimply immersed themselves briefly ordiscovered that high winds, instead of causing sprayed themselves with "Off" insect repellent.extensive damage as we expected, could actually All equipment used on the island was sprayedaid tent removal. The windward side of the tent weekly with a short-lived insecticide.was lifted by poles to about 3 m above water, The absence of supratidal land, except aroundwhereupon sufficient wind entered the tent to lift E9, greatly simplifies the environment of the exitup over the tree and to drop it, partly extended, perimental islands. The arboreal substrate mayinto the water on the leeward side of the island. be arbitrarily divided into hollow twigs, livingE1 was fumigated in the daytime. For all re- branches and twigs, dead bark and tree holes,maining islands the tent was unrolled no earlier leaves and flowers, green shoots, and fruits. Bethan60 min before sundown, the fumigation was cause of the extremely small size of the experimenconductedbetween 8:00 and 12:00 pm, the gas tal islands it was possible to study most of thesewas removed by 1 :00 am, and the tent was either microhabitats all but exhaustively for the pre-deoffthe island or refurled within 90 min of dawn. faunation censuses. A typical example is the sur-Effects on the mangrove trees of the fumigated vey of E3, where we broke and examined 3,500islands are noted in Table 3.of an estimated 5,000 small hollow twigs, collectedAll islands were examined immediately aftertent removal. No living animals at all were foundon El, E2, E8, and E9. A single live curculionidlarva was found on E3. On ST2, 1 living andat least 100 dead polyxenid millipedes (Lopho-proctinus ?bartschi) were found. Although wehave no further direct evidence and know of nowork on methyl bromide fumigation of millipedes,we suspect that all millipedes not destroyed immediatelywere killed by the delayed effect of methylbromide discussed earlier.In sum, we feel that this fumigation techniquekilled all mangrove arthropods with the singleremotely possible exception of deep-boring beetlelarvae.MONITORINGTECHNIQUE? The term "monitoring" is used to designate one ormore of the censuses of species, conducted for a period of17-20 hr during the course of the experiment. In thepresent series of articles, monitoring, census, and censusingare used interchangeably to refer to the listing ofspecies.

Early Spring 1969 EXPERIMENTAL DEFAUNATION OF ISLANDS 27790% of the fruits, examined virtually all of the mately 2 days every 18 days. At each censusleaves, and looked under at least 80% of the pieces period approximately 10% of all hollow twigsof dead bark. Rhizophora has numerous limbs were broken, and with this treatmen the numandour islands were relatively low, so that we ber of hollow twigs seemed to remain constantwere able, by climbing carefully, to examine the throughout the experiment. Only when gentlecanopy right to the top.lifting failed to allow vision was dead bark re-The fumigation procedure itself gave added moved, and then only a part was peeled off. Smallevidence of the completeness of our original sur- amounts of carbon tetrachloride and ammonia, andveys. The fumigant included 2% chloropicrin, tapping with a steel probe were used to drive aniwhichdrives out even deep-boring and hollow- mals out from underneath bark and inside hollowtwig-dwelling arthropods (cf. Lauck 1965). Theanimals either fell into the water or remainedhanging from branches by spines or silk, tenuouslyand conspicuously. Those which fell generallyremained afloat and, when a tent seam was opened tive and from altering the proportions of theafter fumigation, flowed out in masses where they respective microhabitats on an island.were collected on a screen.Most colonists could be identified in the fieldThe chloropicrin technique was especially effec- because of the small subset of the Keys faunative on ES, E9, and ST2, and the results indicated which invades small mangrove islands and thegreater densities of both insects and spiders than completeness of our reference collection. Thosewe had suspected. Only on E8, however, were that could not be recognized with certainty wereanimals found that had not been noted in the treated in one of two ways. If the animal in quesoriginalsurveys. These consisted of four beetles:up) 25-.0'UZ3 2-0'Ua.Boa(/)Li.Staphylinidae: gen. sp.Carabidae: Bembidion sp. nr. contractumCarabidae: Tachys occulatorOedemeridae: Oxacis sp.w -2/ E9, IV-5-6-68m -0-E7, IV-4,8-68M 10-E3, IV-2,4-68z05-J5 l0 1 5 20 25 30'NUMBER OF MAN-HOURS SEARCHINGFIG. 10. Cumulative species counts during individual censuseson three of the experimental islands.twigs without damage to tree or colonist. Althoughthe pre-defaunation searching methods hadbeen drastic, care was taken following defaunationto prevent the censusing from being destruc-tion was obviously part of a sizable population,as with many psocopterans and thrips, a smallcollection was made that did not exceed 2%o ofthe estimated minimum size of the population.When an unknown arthropod was not part of alarge population, we photographed it.Of these the first three were apparently living in It was quickly discovered that the rate of disapatch of temporarily supratidal mud, which was covery of new species during each censusing pesubmergedagain 2 weeks later during a strong riod, although at first very high, declined approxiwind.The Oxacis were encountered only on the mately linearly to near 0 at about 14 man-hours offumigation tent, and even their status as "tran- searching. A very few species are expected tosients" is doubtful.appear from immigration within the monitoringDefaunated islands were censused for approxi- period. Figure 10 depicts the cumulative numberof species observed vs. time for three representa-30tive monitorings. Different sizes of islands,weather conditions, and especially varying populationsizes caused minor differences in the curvesfrom island to island and from time to time; butdespite great changes in the numbers of speciesthrough the course of the experiment, the curvesremained remarkably similar in shape. The censusingperiod was therefore chosen to last between17 and 20 man-hours.Because of the simplicity and very small sizeof the experimental islands, and because severalnocturnal censuses located no new species, we feltthat all colonists having crepuscular or nocturnalactivity were being recorded during the diurnalmonitoring. The anyphaenid spider Aysha sp.and ant Paracryptocerus varians are both strictlynocturnal, yet both were recorded during the dayin their retreats. Similarly the crepuscular braconidwasp Macrocentrusp. and largely nocturnalants Camponotus floridanus and Tapinoma littoralewere repeatedly observed.

278 DANIEL S. SIMBERLOFF AND EDWARD 0. WILSON Ecology, Vol. 50, No. 2ACKNOWLEDGMENTSWe thank the United States Department of the Interioror permission to fumigate the islands of this experiment.n particular, Jack Watson of the Great White Heronnational Wildlife Refuge, Roger W. Allin and William3. Robertson of the Everglades National Park, andZobert M. Linn of the National Park Service examinedhe plans of the experiment and concluded, with the auhors,that the total mangrove faunas would be unaffected.;teve Tendrich, Vice-President of National Exterminaors,Inc. (Miami) coped imaginatively and effectivelyvith the unique problems presented by the project. W. H.3ossert and P. B. Kannowski assisted in the originalensuses. Maps and tables for all parts of this seriesvere done by Mrs. Helen Lyman. The work was suportedby NSF grant GB-5867.17: 373-387.MacArthur, R. H., and E. 0. Wilson. 1967. The theoryof island biogeography. Princeton University Press.203 p.LITERATURE CITEDMainwaring, A. P. 1961. Some effects of methyl bro-Chisholm, R. D. 1952. Nature and uses of fumigants, mide on aphids and whitefly and their host plants.p. 358. In Insects, The Yearbook of Agriculture 1952. Doctoral thesis, London Univ., London, England.Darlington, P. J. 1957. Zoogeography: the geographi-Monro, H. A. U., and R. Delisle. 1943. Further applicaldistribution of animals. Wiley, New York. 675 p.cations of methyl bromide as a fumigant. Sci. Agr.Davis, J. H. 1940. The ecology and geologic role of23: 546-556.mangroves in Florida. Papers from Tortugas Lab.,Carnegie Inst. Wash. 32: 307-412.Page, A. B. P., and 0. F. Lubatti. 1963. FumigationFrench, R. A. 1964 (1965). Long range dispersal of of insects. Ann. Rev. Entomol. 8: 239-264.insects in relation to synoptic meteorology. Proc. Simberloff, D. S., and E. 0. Wilson. 1969. Experi-Int. Congr. Entomol. 12th (London). 6: 418-419. mental zoogeography of islands. The colonization ofFridriksson, S. 1967. Life and its development on thevolcanic island, Surtsey. Proc. Surtsey ResearchConf., 1967: 7-19.Hermannsson, S. 1967. Introduction. Surtsey ResearchProgress Report, III: 1.LaRue, D. C., and T. T. Muzik. 1954. Growth, regenerationand precocious rooting in Rhizophora mangle.Pap. Mich. Acad. Sci. Arts Lett. (I) 39: 9-29.Lauck, D. R. 1965. Chloropicrin for fast action witha Berlese funnel. Turtox News 43: 115.Lindroth, C. H., H. Andersen, H. Bodvarsson, and S. H.Richter. 1967. Report on the Surtsey investigationin 1966. Terrestrial invertebrates. Surtsey ResearchProgress Report, III: 59-67.MacArthur, R. H., and E. 0. Wilson. 1963. An equilibriumtheory of insular zoogeography. Evolutionempty islands. Ecology 50: 278-296.Wolfenbarger, D. 0. 1946. Dispersion of small organisms.Amer. Midland Naturalist 35: 1-152.EXPERIMENTAL ZOOGEOGRAPHY OF ISLANDS: THECOLONIZATION OF EMPTY ISLANDSDANIEL S. SIMBERLOFF1 AND EDWARD' 0. WILSONThe Biological Laboratories, Harvard University, Cambridge, Massachusetts 02138(Accepted for publication December 16, 1968)Abstract. We report here the first evidence of faunistic equilibrium obtained throughcontrolled, replicated experiments, together with an analysis of the immigration and extinctionprocesses of animal species based on direct observations.The colonization of six small mangrove islands in Florida Bay by terrestrial arthropodswas monitored at frequent intervals for 1 year after removal of the original fauna by methylbromide fumigation. Both the observed data and climatic considerations imply that seasonalityhad little effect upon the basic shape of the colonization curves of species present vs. time.By 250 days after defaunation, the faunas of all the islands except the most distant one ("El")had regained species numbers and composition similar to those of untreated islands even thoughpopulation densities were still abnormally low. Although early colonists included both weakand strong fliers, the former, particularly psocopterans, were usually the first to produce largepopulations. Among these same early invaders were the taxa displaying both the highestextinction rates and the greatest variability in species composition on the different islands.Ants, the ecological dominants of mangrove islands, were among the last to colonize, but theydid so with the highest degree of predictability.The colonization curves plus static observations on untreated islands indicate strongly thata dynamic equilibrium number of species exists for any island. We believe the curves areproduced by colonization involving little if any interaction, then a gradual decline as interactionbecomes important, and finally, a lasting dynamic equilibrium. Equations are given forthe early immigration, extinction, and colonization curves.Dispersal to these islands is predominantly through aerial transport, both active and passive.Extinction of the earliest colonists is probably caused chiefly by such physical factorsas drowning or lack of suitable breeding sites and less commonly by competition and predation.Present address: Department of Biological Science, Florida State University, Tallahassee, Florida 32306