Global Amphibian Declines: A Problem in Applied Ecology Author(s): Ross A. Alford and Stephen J.

Richards Reviewed work(s): Source: Annual Review of Ecology and Systematics, Vol. 30 (1999), pp. 133-165 Published by: Annual Reviews Stable URL: http://www.jstor.org/stable/221682 . Accessed: 04/05/2012 14:44
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Annual Reviews is collaborating with JSTOR to digitize, preserve and extend access to Annual Review of Ecology and Systematics.

http://www.jstor.org

Anniu.Rev. Ecol. Syst. 1999. 30:133-65 Copyright? 1999 by Annual Reviews. All rights reserved

A PROBLEM GLOBAL AMPHIBIAN DECLINES:

IN APPLIEDECOLOGY
RossA. Alfordand StephenJ.Richards
School of TropicalBiology and CooperativeResearchCentre for TropicalRainforest Ecology and Management,James Cook University,Townsville,Queensland 4811, Australia;e-mail: rossalford@jcu.edu.an

Key Words

conservation, frog, salamander, null hypothesis, metapopulation

Declines and losses of amphibian populations are a global problem with complex local causes. These may include ultraviolet radiation, predation, habitat modification, environmental acidity and toxicants, diseases, changes in climate or weather patterns, and interactions among these factors. Understanding the extent of the problem and its nature requires an understanding of how local factors affect the dynamics of local populations. Hypotheses about population behavior must be tested against appropriatenull hypotheses. We generated null hypotheses for the behavior of amphibian populations using a model, and we used them to test hypotheses about the behavior of 85 time series taken from the literature. Our results suggest that most amphibian populations should decrease more often than they increase, due to highly variable recruitment and less variable adult mortality. During the period covered by our data (1951-1997), more amphibian populations decreased than our model predicted. However, there was no indication that the proportion of populations decreasing changed over time. In addition, our review of the literature suggests that many if not most amphibians exist in metapopulations. Understanding the dynamics of amphibian populations will require an integration of studies on and within local populations and at the metapopulation level.

* Abstract

INTRODUCTION
The current wave of interest in amphibian population biology and in the possibility that there is a global pattern of decline and loss began in 1989 at the First World Congress of Herpetology (10). By 1993 more than 500 populations of frogs and salamanders on five continents were listed as declining or of conservation concern (189). There is now a consensus that alarming declines of amphibians have occuffed (30, 51,1725, 147, 192). Because most amphibians are exposed to terrestrial and aquatic habitats at different stages of their life cycles, and because they have highly permeable skins, they may be more sensitive to environmental toxins or to changes in patterns of temperature or rainfall than are other terrestrial vertebrate 0066-4162/99/1 120-0133$08.00 133

134

ALFORD . RICHARDS

TABLE 1 Techniquesused in 46 studies1to quantifypopulations of frogs salamanders.Many studiesused more thanone technique or studiedboth taxa, so the total numberof techniquesused does not equal the numberof studies cited. Technique Egg mass counts Counts of individuals Drift fence/pit trapcounts Calling male counts Dipnet samples for larvae Countsof individuals estimates Mark-recapture Aquatictraps Habitat2 B B B B B N N N Frogs 4 12 7 7 3 2 1 0 0 Salamanders 1 1 8 2 N/A 2 6 1 1

estimates3 B Mark-recapture

1Sourcesused in compiling table: 15, 19, 20, 21, 41, 49, 53, 54, 63, 64, 73, 83, 84, 86, 88, 90, 91, 92, 96, 102, 107, 111, 113, 117, 131, 136, 137, 142, 157, 158, 159, 160, 161, 165, 169, 170, 171, 174, 177, 178, 179, 180, 181, 194, 196, 202.
2B = Breeding, N = Non-breeding. 3Includes, in descendingorderof frequencyused, Jolly-Seber,Petersen,Manly-Parr,

Schnabel,and Zippintechniques.

groups (29, 190). The best-documented declines have occurred in Europe and North America, are usually associated with habitat modification (87, 116), and are often attributed to interactions among causal factors (114, 125, 147). The factors associated with population declines in relatively undisturbed habitats such as montane tropical rainforests have been more difficult to elucidate (131, 163). Although they have been the subject of many experimental and monitoring studies, the autecology of amphibians in nature is poorly understood (87). The majority of studies of ecology and population biology of amphibians (Table 1) have been conducted on aggregations at reproductive sites. Relatively little is known of their movements or activities away from breeding sites, or of rates of exchange between populations. Many authors have suggested that there is a need for long-term studies directed toward a combination of understanding ecological theory and increasing knowledge of the autecology of amphibians (17,39,79,80, 190). Simple longterm programs that monitor the fluctuations of single populations and associated environmental factors, and then apply standard population models, are unlikely to be useful for understanding the dynamics of amphibian populations as they have not worked for that purpose when applied to other terrestrial vertebrates (164). It appears likely that understanding the problem of amphibian declines will require much more information on the ecology of the metapopulations in which many species live (87).

GLOBALAMPHIBIANDECLINES

135

Our goals in this review are to summarizeand synthesize the literatureon potential causes of amphibiandeclines, and to use the literatureon amphibian population dynamics to develop a null hypothesis for the behavior of amphibian populations.We then use our null hypothesis and data from the literatureto determinewhetherthe incidence of declines has recently increased.Finally, we place the dynamicsof amphibianpopulationsand their declines in the context of metapopulation dynamics.

POTENTIAL CAUSES OF AMPHIBIANDECLINES Ultraviolet Radiation

Depletion of stratosphericozone and resultantseasonal increases in ultraviolet B (UV-B) radiationat the Earth's surface (119) have stimulatedinterest in the possible relationshipbetween resistanceof amphibianembryos to UV-B damage and population declines. Significant variationamong species in levels of phoDNA repair enzyme that repairs UV-B damage, is tolyase, a photoreactivating correlatedwith exposureof naturalegg deposition sites to sunlight (25,98). In a survey of 10 Oregonamphibianspecies, photolyase activity,and hence ability to repairUV damage, was lowest in declining species and highest in nondeclining
species (25). Field experiments demonstrated that embryos of Hyla regilla, a non-

declining species with high photolyase activity,had significantlyhigher hatching success thandid two declining species (Ranacascadae andBufo boreas) with low photolyase levels (25). A numberof other studies have demonstrated that ambient(6,24,28, 132) or enhanced(144) UV-B radiation reducessurvivalor hatchingsuccess of amphibian stresses betweenUV-B andotherenvironmental embryos. Synergisticinteractions such as pathogens(120) and low pH (133) may also significantlyincreaseembryonic mortality.Ranapipiens embryosthat are unaffectedwhen exposed to UV-B and low pH separatelyhave significantlyreducedsurvivalwhen exposed to these factors simultaneously(133). Otherstudieshaveproducedmoreequivocalresults. Rana aurorais a declining species with high levels of photolyase (98), and experimentalhatchingsuccess is unaffected by exposure to UV-B (26). The declining frog Litoria aurea from easternAustraliahas a lower photolyaseactivitythantwo sympatricnondeclining species, L. dentata and L. peroni, but thereis no significantdifferenceamong the three species in hatching success underUV-B exposure (187). In many aquatic habitatsUV-B radiation is largelyabsorbedin the firstfew centimetersof the water column (138), so increased UV-B may only affect species breeding in habitats with a narrowrange of chemical and physical parameters.Ecologically relevant levels of UV-B had no effect on embryos of several Canadianamphibians,and experimental protocolsused to test UV impactshave been questioned(85, 130). Most studiesthathave examinedthe relationship betweenUV-B andpopulation declines have focused theirattentionon species thatbreedin shallow,clear water,

136

ALFORD a RICHARDS

where exposure to UV-B is expected to be greatest (6,25,28,132). Exposure to intense UV-B in shallow high-altitudeponds may exclude amphibiansfrom these habitats(138). Even when UV-B causes higher embryonicmortalityin declining species, the ecological significanceof this at the populationlevel is difficult to assess. More needs to be understoodabout the basic naturalhistory of amphibianspecies that might be at risk. For example, inforrnation is needed on variationin oviposition site characteristics(depth, vegetation) within local populations. Experimentsat "natural ovipositionsites"using embryosof Ambystoma were conmacrodactylum ductedin shallowwateralthoughthis species lays eggs in a varietyof microhabitats (28). Loss of a large proportionof near-surfaceclutches to UV-B damage may have negligible impactson populationsif even a small numberof deeperclutches effects (1). survive,as the survivorsare releasedfrom density-dependent Even fewer data are availableto assess the indirecteffects of increasingUV-B on amphibianpopulations. Potential indirect effects include changes in water relationchemistryand food supplies, and shifts in competitiveand predator-prey ships with other UV-B affected species (143). Exposureto increasedUV-B may reduce survivalrates of adult amphibiansthroughdamageto eyes (77), increased frequencyof cancersor tumors(143), and immunosuppression (143).

Predation
Biotic interactionsamong amphibians,and between amphibiansand otherorganisms, can play a significantrole in determiningtheir distributionand population dynamics (1). Larvalamphibiansare extremely vulnerableto vertebrateand invertebratepredators(1), and the diversity of aquatic amphibianassemblages is frequentlyreducedin habitatscontainingpredatoryfish (1, 100). Larval amphibiansthat coexist with aquatic predatorshave evolved a range of antipredator mechanisms (4,48,118). However, widespreadintroductionsof predatoryfish have increasingly exposed native amphibiansto predatorswith which they havenot previouslyinteracted.Inappropriate responsesto novel predators may increasemortalityof native amphibians(82,121), leading to significant effects on populations. Colonization of normally fish-free water bodies by predatoryfish can result in rapidextinctionof amphibianassemblages (76). The allotopic distributionsof native frogs and introducedfishes in many high-elevation(>2500 m asl) Sierra Nevada lakes indicate that introducedpredatoryfishes have caused the extinction of local frog populationsthere (31). Sixty percent of lakes that frogs could forrnerlyoccupy now contain introducedfishes and no frogs. Fish introductions have had a particularly severe impacton Rana muscosa, which breedsin the deep has been lakes inhabitedby fishes (31). A similarpatternof allotopicdistributions recordedfor larvalnewts, Tarichatorosa, and an introducedfish (Gambusiaaffinis) and crayfish(Procambarusclarki) (both predatorsof newt eggs or larvae)in Californianmountainstreams(82).

GLOBAL AMPHIBIAN DECLINES

137

Introduced predatorsmay also have more subtle effects. Some Rana muscosa populationspersistingin fish-freeenvironments have become isolated from other populationsby surrounding aquatic habitatscontaining introducedfishes. This may eventuallylead to regional extinction by preventingmigrationamong local populations(35). North American bullfrogs (Rana catesbeiana) that have become established outsidetheirnatural rangehavebeen implicatedin declines of nativefrogs (76, 102, 127, but see 97). Adult bullfrogs consume native frogs and reach densities at which they are likely to have a severe impact on local amphibianpopulations (166). Experimental studieshave shown thatRana auroralarvaeexposed to adult or larvalbullfrogshave increasedlarvalperiods, smallermass, and, when exposed to both, lower survival(122). Humanshave devastatedfrog populationsin severalcountriesfor the frog-leg trade. Before 1995, abouttwo hundredmillion frogs were exportedannuallyfrom Asia. By 1990 India was still illegally exportingapproximatelyseventy million frogs each year,resultingin seriouspopulationdeclines (145).

HabitatModification
Habitatmodificationis the best documentedcause of amphibianpopulationdeclines. Habitatloss certainlyreduces amphibianabundanceand diversity in the areas directly affected (99, 101). Removal or modificationof vegetation during forestryoperationshas a rapidand severeimpacton some amphibianpopulations of matureforests in the southernAppalachians (8). Clearcutting has reducedsalamanderpopulationsby almost9%, or morethana quarter of a billion salamanders, below the numbersthatcould be sustainedin unlogged forests (149). Logging exposes terrestrial amphibiansto drasticallyalteredmicroclimaticregimes (9), soil compaction and desiccation, and reductionin habitatcomplexity (197). It exposes aquaticamphibians to streamenvironments with increasedsiltation(52) and reduced woody debris (43). Although populationsmay recover as regenerating forests mature,recoveryto predisturbance levels can take many years (9) and may not occur at all if mixed forests are replacedwith nionocultures(108). Drainingwetlandsdirectlyaffects frog populationsby removingbreedingsites (116), and by fragmentingpopulations (74,168), which increases the regional probabilityof extinction(e.g. 53). Modificationof terrestrial and aquatichabitats for urbandevelopmentcan reduce or eliminate amphibianpopulations. Populations of some amphibiansin urbanFloridadeclined after degradationof upland, dryseasonrefugesandmodificationof wetlandsused forbreeding(62). Protection of aquaticbreedingsites may be of little value if adjacentterrestrial habitatsused by amphibiansfor feeding and shelterare destroyed(167). More subtlealterations to habitatstructure can have severeimpactson amphibian populations. Bufo calamita populationsin Britain declined over a 40-year period due to shifts in land use practices that altered vegetation characteristics (13). Changingvegetation structureand an associated increase in shading were

138

ALFORD a RICHARDS

detrimental to B. calamita andprovidedconditionsunderwhich the common toad Bufo bufo became a successful competitor. Althoughhabitatalterationscan reduce amphibianpopulations,in some cases even severehabitatmodificationscan have little effect. The responseof a savanna woodlandfrog assemblageatWeipa,Queensland,Australiato stripminingappears in Figure 1 (200). The structureand floristics of the plant assemblage at 60 revegetated sites varywidely; none stronglyresemblethe originalnativewoodland.

Percentof sites

omatus U7nodynastes Bufo marinus caerulea litoria Litoria bicolor - * > ! novaehollandiae Cyclorana
Litoria rubellaIk

o.Y

Sphenophrtyne gracilipesUimncxynastes convexiusculus * $ litoriagracilenta rothiLitoria Cinia remotamimula- 0 Uperoleiea Litoria infrafrenata Litoria nasuta -

Figure 1 Profiles of percent of sites of five habitat types at Weipa, Queensland, Australia at which frogs of 14 species occurred. Species sorted in order of frequency of occurrence in native woodland. Dots and solid lines indicate native woodland habitat (13 sites). Diamonds and dashed lines indicate sites revegetated following strip mining. Density of diamonds reflects age of revegetation; from least to most dense this is: age <7 years, 7 years < age < 10.5 years, 10.5 years < age < 16 years, age > 16 years. There were 15 revegetation sites in each age group in the survey.

AMPHIBIAN GLOBAL DECLINES

139

The distancefrom revegetationsites to nativewoodlandvaries from a few meters to over 1 km. Despite this, within 7 years, the majorityof frog species occur at revegetationsites at frequenciesvery similarto the frequenciesat which they are found at sites in nativewoodland. This indicatesthatthe frogs in this assemblage flora, and structure are insensitive to radicalalterationsin the soil characteristics, of their terrestrial habitat,and they can recolonize rapidlyfollowing disturbances that have eliminated them over a relatively wide area. It is possible that other amphibianspecies assemblagesbehave similarly.

Acidity and Toxicants

The acidity of aquatichabitatshas majorimpacts on amphibiandistribution,reproduction,and egg and larval growth and mortality(78,79). Sensitivity to low pH varies among (79) and within (150) species and is influenced by complex aluminumconcenchemical interactionsamongpH andotherfactors,particularly tration(71,110,151). Mortalityoccurs in both the embryonic and larval stages of the yolk plug, arrested via severalmechanismsincludingincompleteabsorption development,and deformationof larvae(11, 79, 109). Sublethaleffects of acidification include delayed (109) or early (36) hatching,reducedlarvalbody size (36), disturbedswimmingbehavior(5), and slowergrowthratesresultingfromreduced responseto, and captureof, prey (155). Indirectsublethaleffects include changes to tadpolefood sourcesthroughimpactson algal communities(188), and shifting predator-prey relationshipsresultingfrom differentialmortalityof predatoryfish in acidifiedhabitats(104). and invertebrates The population-leveleffects of acidity are less well understood. It is possible that the effects of low pH, in combination with other abiotic factors, lead to decreased recruitmentinto adult populations (12). Acidic breeding sites often contain less diverse amphibianassemblages, at lower densities, than do less acidic sites (205). Long-termacidificationof ponds in Britainhas excluded Bufo calamita from many sites (16). Reduced pH and increasedmetal concentrations in an Appalachianstreameliminatedvirtuallyall salamander larvae, causing severe long-term declines in populationsof Desmognathusquadramaculatusand Euryceawilderae (124). Low soil pH also influencesthe distribution, abundance, and diversityof terrestrial amphibians(204, 205). Despite the well-documentedeffects of low pH on amphibians,there are few data to implicate acidificationin recent, unexplainedcatastrophic populationdeclines. Acid depositionwas proposedas a factorin the decline of tigersalamanders, Ambystomatigrinum,in the Rocky Mountains(96), but subsequentfield studies demonstrated thatmortalitydue to pond dryingwas equally likely to be the cause of this decline (201). Acid depositionis unlikely to be involved in populationdeclines of frogs and salamanders at high altitudesin the SierraNevada Mountains (33, 34,36) and Rocky Mountains(54,55, 188,201). Theremay be no rigorously documentedcases where acidificationof naturalhabitathas led to the extinction of an amphibianpopulation(71). However, studies of acid tolerance have been biased towardspecies thatare likely to have evolved toleranceto low pH (195).

140

ALFORD u RICHARDS

Similarly,althoughthereis an extensive literature on the toxic effects on larval amphibiansof metals and chemicals used in insecticides and herbicides(154), insufficientdataexist to determinetheirlong-termimpactson amphibian population dynamics (22). Environmental toxicants act directlyto kill animals, or indirectly by impairingreproduction, reducinggrowthrates,disruptingnormaldevelopment andreproduction (endocrinedisruption),or increasingsusceptibilityto disease by immunosuppression or inhibitionof immune system development(22,46).

Diseases
Little is known about the diseases of wild amphibians. Many disease agents are presentin healthyanimals,anddisease occurs when immunesystems arecompromiised(56,57). Declines in populationsof Bufo boreas boreas between 1974 and 1982 were associated with Aeromonashydrophilainfection, but Carey (47) suggested that environmentalfactor(s) caused sublethal stress in these populations, directly or indirectly suppressingtheir immune systems. A pathogenic fungus largely responsiblefor egg mortalityin one populationof Bufo boreas in Oregon may have been more virulentto embryosunderenvironmental stress (27). Epidemicscan cause mass mortalityof amphibians(123). In 1981 Aeromonas hydrophilakilled all larvalRana sylvatica in a RhodeIslandpond, andthreeyears later few adult frogs were breeding at this site (141). The same bacteriumwas implicated in a well-documenteddecline to local extinction of a population of Rana muscosa in California(32). A chytridomycetefungus found on moribund anuransin Australiaand Panamaduringmass mortalityis fatal to healthy frogs underexperimentalconditions (18). This fungus was proposedas the proximate cause of declines in these two regions (18), but this hypothesis is untested at present. Viruses have been isolated from dead and dying frogs during mass mortality events (59,60) and may be the primarycause of mortalityin animals where other infections such as bacteriahave been identified (57). Lauranceet al (129) arguedthat the patternof populationdeclines among Australianrainforestfrogs was indicativeof a "wave"of epidemic disease caused by an unidentifiedwaterborne virus. This interpretation was challenged (3, 106) on statistical grounds and because numerousother explanationsfor the observedpatternswere equally parsimonious.The involvementof a virus in these declines has now been largely discounted,but the possible involvementof a disease has not. The patternof population declines in CentralAmerica has also led to the suggestion that a wave of epidemic disease might be responsible(131).

Climate/Weather
of golden toads,Bufoperiglenes, the rainImmediatelypriorto the disappearance forests of Monteverde, Costa Rica, had the lowest twelve-month rainfall in 20 years. Toadswere forcedto shifttheirhabitatuse, andthe dryconditionsmay have interactedwith an unidentifiedfactor such as disease or a pulse of contaminants

DECLINES GLOBAL AMPHIBIAN

141

in cloud water to eliminate toad populations(152). Unusual weatherconditions were dismissed as a cause of declines of Australianrainforestfrogs (128). This resultrelies on seasonalrainfalltotals calculatedfor fixed groupsof months,while in northern Australiathe date of onset of the wet season is highly variable,occur(140). The analysis is thereforelikely to have ring between Octoberand February the variancesof rainfall,and a reexammissed the extremes and underestimated ination of the data seems warranted. Severe, short-termclimatic events such as Hugo violent stormscan alterthe dynamicsof amphibianpopulations. Hurricane in In the Puerto the short Rico 1989. extensive to forests of damage caused term, frog Eleutherodactylus richmondi decreasedby 83%, populationsof the terrestrial but increasedavailabilityof groundcover due to disturbanceled to a six-fold increase in the densities of E. coqui, followed by a long period of gradual population

decrease (202,203). Alterationsin local weather conditions caused by global climate change will in a numberof ways. The onset of spawningin influencethe ecology of amphibians Rana temporaria in Finlandbetween 1846 and 1986 shifted earlierby 2-13 days, and dates of snow cover loss (182). following shifts in air and watertemperature At some sites in Britaintherehas been a statisticallysignificanttrendtowardearlier first sighting and spawningof Bufo calamita, Rana esculenta, and R. temporaria between 1978 and 1994, correlatedwith changingpatternsof springtemperatures (14). Amphibiansin Canadaare affected by decreases in summerprecipitationand increasedtemperatures andwinterrainfall(105, 143). In the neotropics,increased extended dry seasons, and increasinginter-yearrainfallvariability temperatures, may affect litter species by reducing prey populations and altering amphibian on increasinglydry soil (67). Shiftingrainfallpatternswill affect the distributions species. Pondswill fill laterandpersist phenologiesof pond-breeding reproductive for shorterperiods,leading to increasedcompetitionandpredationas amphibians are concentrated at increasinglylimited aquaticsites (67). Frogs exposed to these stresses may also become more vulnerableto parasitesand disease (67). Interactions Among Environmental Factors

Most studies invoke multiple causes or interactionsamong factors. Increased UV-B exposure may alter species interactionsor vulnerabilityto pathogens or changes in pH. Predationmay eliminate local populationsand have larger-scale effects by alteringrates of migrationbetween populations. Outbreaksof disease may only occur when other stresses reduce immune function. Pesticides, pollutants, and environmentalacidity may interactto produceunforeseeneffects. All local effects may interactwith global climate change. Proving the existence of these complex effects in naturalpopulationswill requirewell-plannedprograms of observationand experimentation.To plan such studies, and to determinehow of the natureof the stresses affect populationbehavior,requiresan understanding It also requires and the of studied limitations study techniques. populationsbeing

142

ALFORDI RICHARDS populationsbehave the developmentof null hypothesesregardinghow amphibian in the absence of externalpressures.

DEFININGAND STUDYINGAMPHIBIANPOPULATIONS Monitoring and CensusingTechniques

We have summarizedthe techniquesused in 46 long-termpopulationstudies of in Table 1. All but one of the studies of frogs were carried frogs and salamanders out at breeding sites, while about half of the studies of salamandersincluded at habitat.The most least some datacollected on densitiesof animalsin nonbreeding commonlyused techniqueis directcounts, where animalsare locatedby intensive searching,localizationof calls, or by driftfences with pitfall traps. Many studies estimates. that reporteddirect counts also reportedthe results of mark-recapture estimates are very Frequently,however, the standarderrors of mark-recapture large, so that counts are regardedas being better estimates. The high standard estimatesof animals at breedingsites errorstypically obtainedin mark-recapture of breedingsites varieswidely probablyreflectthe fact thatthe degreeof attraction over time, as does the activity of individualanimalsnearthem.

Problemswith StudyingBreedingAggregationsas Populations

Most estimates of frog populationsare expressed in units such as total numbers of frogs attendinga pond, numbersof frogs per m2 of pond area, and maximum numbersof frogs at a pond on a single night duringthe breedingseason (70). The habitattend frog groupsfor whichdataareavailableon densitiesin nonreproductive to be species that do not aggregateat breedingsites (70, 115, 203). A few studies have examined aquatic-breeding habitats(156,206). species in nonreproductive The use of breedingaggregationsin populationstudiescan cause problemsin data interpretation. A simple illustrationof why censusing amphibiansat breedingsites can cause of populationdynamics appearsin Figure 2. The problems in the interpretation numberof frogs per unit of nonreproductive habitat(the entirerectangle)remains constantat 50. In Figure2a, neitherof the temporary pools in the habitatcontains water. If censuses are carriedout by visiting pool 1, a census at the time of Figure 2a would detect only 7-10 frogs in the habitat. A census carriedout afterpool 1 filled (Figure 2b) would detect 43 frogs, while one carriedout after both pools filled (Figure 2c) would detect only 23 frogs at pool 1. Similar variations in measureddensity could occur over time even if the pools remain filled, because behaviorof frogs. Changes in habitatavailability of changes in the reproductive and in the attractiveness of bodies of water are both likely to affect the numbers of amphibiansdetected(191), complicatingthe use of dataof this type as an indihabitat. cation of the size or density of the populationoccupying the surrounding and site attributes and temporally with weather vary seasonally Because behavior

GLOBAL AMPHIBIAN DECLINES

143

S.

bC

Ce

fordatainterpretation Figure2 An illustration of thedifficulties created by sampling An areaof habitat a fixednumber of animals in breeding contains animals aggregations. outatsite 1. In (a), animals arenot andtwopotential sites. Surveys arecarried breeding attracted to breeding sites, so a censusat site 1 wouldencounter only 9-10 individuals. In (b), animalsare attracted to site 1, and a census would find 43 animals. In (c), 22 animals. animalsareattracted to bothsites, anda censusat site 1 wouldencounter All of these situations may occurwithina few days in manyhabitats, makingcounts at breedingsites highlyvariable. andinternalrhythmsof the animals,the best hope for obtainingaccurateestimates using censuses at breedingsites is to performcensuses frequently.

Effectsof SamplingIntensity on Population Size Estimates

We used datacollected by one of us (SJ Richards,unpublisheddata)on the number of adultmales of the stream-breeding hylid Litoriagenimaculatato investigatethe effects of sampling intensity on the accuracy of count data taken at a breeding aggregation.A 60-m transectalong a rainforeststreamsite was visited an average of 19 times each year over seven years. All frogs on the transectwere counted, marked,andreleased. We used a resamplingtechniqueto explorehow less intense samplingmighthave affectedthe results. We set the probabilitythata visit would occur to 0.75, 0.5, and 0.25, and resampledthe data set 5000 times using each of these probabilities. It is apparent(Figure 3) that decreasing the intensity of samplingbelow about0.75 times the numberof visits actuallymade would greatly decrease the accuracyof the annualmeans obtained. At intensities of 0.5, 0.25,

144

ALFORD * RICHARDS

---

Actual nean
-

35-

30 S
co

\-

75%sampling 50% sampling 25% sampling

25 20-

"

1visitperyear

10 a = 0
0C_

1989 1990 1991 1992 1993 1994 1995 16 25 21 10 17 22 23 Year,

Number of visits
Figure 3 Results of resampling analysis on numbers of adult male Litoria genimaculata present on a 60-m transect on Birthday Creek, Mt. Spec, Queensland, Australia. Points indicate the mean number of individuals present taken over all visits in each year. The lines connecting the points are for illustration only. The dashed lines at increasing distances from the mean are the upper and lower 95% confidence limits for the location of the mean for each year at the given sampling intensity, as determined from 5000 resamplings of the data. Upper and lower solid lines are the 95% confidence limits for numbers obtained if the site was visited once annually; they are the minimum and maximum numbers encountered each year.

AMPHIBIAN DECLINES GLOBAL

145

or one visit per year,the observedpatternof change throughtime would probably have differedfrom the patternthat appearsin the full data set, and the coefficient of variationof the annualmean would have increasedsubstantially. of breedingsites can also occur Problemscreatedby the changingattractiveness habitat,because the attractiveness when animalsare censused in nonreproductive of patches changes. A populationof Rana arvalis in a 2000-m2 sampling area increasedmore thantenfoldover the period 1984-1988 (89); this mighthave been due to a decreasein the local availabilityof water,causing animalsto concentrate whereit was available(89). Densities of the neotropicallitterfrogs Bufo typhonius and Colostethusnubicolain nonreproductive habitatvariedmore thantenfoldover the course of a one-year study (183). This variationwas probably caused by differences in the activity level and catchabilityof frogs, ratherthan by changes in populationsize (183). Once a reasonableset of populationestimatesat a site is available,the next step in an analysis of the behaviorof the populationis to examine how it changes over time. It is unrealisticto expect perfect stabilityfrom any naturalpopulation;even if the populationis truly constant, sampling errorwill introducesome variation. Giventhatwe expectpopulationsto fluctuate,we shouldtryto determinehow they will behave over time in the absence of directionalpressurestowardexpansionor decline.

A Null Hypothesisfor the Behaviorof AmphibianPopulations

over Time
Therehas been substantial debatein recentyears(23, 146, 148, 169) regarding how "normal" populationsshouldchangethroughtime, but little theoretical amphibian workon this question. The problemhas been definedclearly: Fluctuatingpopulations of amphibianswill be eitherincreasingor decreasingat any time (146, 148). for bias if populationsare declaredto be "declinThereis a very largeopportunity ing" on the basis of shorttime series indicatingthatnumbersare decreasing. Amphibian Populations Fluctuate It is almost universallyagreedthat most local populationsof amphibiansare likely to fluctuateconsiderablyin size. This is highly variable(7, 15,20,44,65, 81, 185, 186, 194). occurs because recruitment Survivalratesof terrestrial stages often appearto be relativelyconstant,with some degree of fluctuation(7,44,73, 185, 186). The survivalrates of eggs, hatchlings, and larvae often vary over several orders of magnitude (1, 20,199). Pechmann & Wilbur(148) provided a useful review of approachesthat have been used to examinepopulationbehavior. thathas been used to Examining Changes of Numbers with Time One approach of numbers examinetrendswithinpopulationsof amphibians is simplecorrelations in the time series for six species of salamanders with time. Fifteen-to-twenty-year Appalachianmountainsof the United States showed no evidence for consistent

146

ALFORD . RICHARDS

trends in numbersover time (91). A lack of significantcorrelationsof numbers with years over a long-termstudy may indicatethat variationamong years represents fluctuationsin numbersratherthandeclines (146, 169). Significantnegative correlationsbetween populationsize and time have been interpreted as indicating thatpopulationsarein decline (115, 117). Of 16 studiesthatmonitoredamphibian populationsfor four or more years, five reportedthatpopulationswere declining, one that six populationshad become extinct, seven that populationswere fluctuating, and three thatthey were stable (30). Most of the evaluationsof population change were based on correlationsof populationsize with time, and some statementsthatpopulationswere not in decline werebasedon failureto finda significant correlation.Failureto finda significantcorrelationdoes not necessarilyimply that none exists (162); it can also resultfrom a lack of statisticalpower. In eight studies of amphibianpopulationsthatreportedno trendin populationsize over time, the power of a product-moment correlationwas insufficientto allow acceptance of the null hypothesis (162). Reed & Blaustein (162) suggested that studies should not conclude thatpopulationsarenot in decline unless thatconclusion can be supportedstatistically.Unfortunately, the poweranalysisthey used, relying on simple Pearsoncorrelations,is not statisticallyvalid, as the adjacentannualvalues in time series on a single populationarenot independent randomsamples. We develop an alternativeapproachbelow. A Simple Model of Population Behavior In orderto suggest how more appropriatetests might be carriedout, we exploredhow time series data on amphibian populationsshould be analyzed and interpreted,and what the "normal" patterns in such data might be. We constructeda simple verbalmodel of frog population dynamicsand comparedits predictionswith the behaviorof 57 time series of frog abundancesand 28 salamandertime series (all studies cited in Table 1; details are available on the World Wide Web in the SupplementalMaterials section of the main AnnualReviews site [www.AnnualReviews.org]). Finally we compared the behaviorof those 85 time series with thatof a simulatedpopulation.The simple verbalmodel is: Assumptions: 1. amphibianpopulationsoften persist for many generations, neitherdecreasingto extinctionnor exploding to infinity;2. survivalrates of terrestrial stages may vary,but thatvariationis typically over less than a single orderof magnitude;3. survivalrates of aquaticstages often vary over severalordersof magnitude.Deductions: (A) Assumption 1 implies that recruitment must on averagebe sufficientfor replacementto occur. (B) Assumptions 1 and 2 togetherimply thatmost variationin the size of terrestrial from the populationsmust be due to fluctuationsin recruitment aquaticstages. (C) Assumptions 1, 2, and 3 togetherimply that when aquaticsurvivalis high, populationsmust rapidlyincrease,but that aquatic survivalmust frequentlybe below replacementlevel, so thatthe increasesdo not tend to lead to sustainedexplosions. Deductions (A-C) takentogether

AMPHIBIAN GLOBAL DECLINES

147

suggest thatpopulationsshould decreasemore often thanthey increase, by slower because increasescan be very rapidand must be counterbalanced decreases. This model suggeststhatpopulationsof species withhighly variablerecruitment stage might be expectedto decreaseduringmore from the aquaticto the terrestrial point in the context of than 50% of time intervals. This is a potentiallyimportant the problemof amphibiandeclines: If populationsnaturallydecreasein numbers populationstudies may often more often thanthey increase,relativelyshort-term find thatthe populationor populationsstudiedappearto be in decline. Comparing the Predictions of the Model with Data To examine this problem in more detail, we extracteddatafrom a databaseof informationon time series of andavailable collatedfromthe literature populationsizes of frogs andsalamanders from the United States Geological Survey (USGS) (72). Most of the populations studiedwere not regardedas being in decline. We modifiedthe databasefrom its raw form in severalways. The databaseincludesa numberof time series collected using differenttechniqueson the same species at the same sites and times. We includedonly one time series of data on any species for any site and set of dates. We did not use data on larvae or juveniles in our analysis because they reflect input modifiedby highly variableaquaticmortalityrates and so are reproductive least likely to reflectthe dynamicsof adultpopulations.When datawere available for both male and female adults, we combinedthem. When count data and other valueswere available,we includedthe countdata estimatessuchas mark-recapture if they had been collected in a mannerthat was likely to be comparableamong years; otherwisewe used the estimatednumbers. When only dataon numbersof egg masses were available,we included them in our analyses, as they should be highly correlatedwith numbersof adult females. We included only time series takenover four or more years. We also included data from Cohen (50) that were not presentin the USGS database.A summaryof the original sources of the data we used appearsin Table 1, and details of the datafor each time series appearon the WorldWide Web in the SupplementalMaterialssection of the main Annual Reviews site (www.AnnualReviews.org). We classified each yearly populationestimate as either the initial numberin a series, an increase, a decrease,or no change. Three families of frogs (Bufonidae, Hylidae, and Ranidae) and two families of salamanders(Ambystomatidaeand Plethodontidae)contained sufficientnumbersof time series. We used nonparametricstatistics(Table2) to determinewhetherthe percentageof the time in which populationsdecreasedfrom one year to the next varied among families. In both orders, families differed in the percentageof the time that populationsdeclined between years. In the Caudata, populationsof species in the family Ambystomatidae, which are pool-breedingspecies with relativelylarge clutch sizes, decreased between years an average of 59.2% of the time. Species in the Plethodontidae, with smaller average clutch primarilyterrestrialegg layers and stream-breeders

148

ALFORD . RICHARDS TABLE 2 Means, standard deviations,and numbersof time series analyzed for the percentageof year-to-yearchanges in populationsize thatare decreases, for amphibiansof five families, and tests for significantdifferencesin the mean among families within orders. Percent of population changes that are decreases Family Mean Standard Deviation N

Salamanders.Wilcoxon 2-sample test, Z _ 2.43, P = 0.015 Ambystomatidae 59.2 18.1 12 Plethodontidae Bufonidae Hylidae Ranidae 42.0 16.1 14 19 15 23 Frogs. Kruskal-Wallis test, X2 = 8.82, 2 d.f., P = 0.012 55.5 19.8 47.3 60.6 6.7 13.0

sizes (see references in Table 1), declined between years only 42% of the time. In the Anura, species in the families Ranidae and Bufonidae decreased in more than 50% of intervals, while species in the family Hylidae decreased in slightly less than 50% of intervals. (Table 2). The ranids and bufonids included in the database generally produce larger clutches of eggs and often have highly variable offspring survival, while hylids produce smaller clutches and may have less variable rates of offspring survival (references in Table 1). The results of this analysis suggest strongly that the expected behavior of populations over time varies among families of frogs and salamanders, and that the adult populations of species that have more variable survival of premetamorphic offspring tend to decrease between years more often than they increase. This result has implications for the ability to draw conclusions about population trends from simple time series of numbers over years. A species in the family Plethodontidae, in which the mean population behavior is to decrease between 40% of years, has a probability of four successive decreases of only 0.031, and finding four successive decreases between years in a censused population might be cause for alarm. However, a ranid species would need to decrease in numbers six times in succession before the probability of that sequence of decreases was less than 0.05. Our data set included populations from many times and places, some of which might have been undergoing declines caused by external pressures. This would bias the results of our analysis. To erect a null hypothesis independent of the data from natural populations, we examined the behavior of a simulated population with known characteristics.

A Numerical Model of Population Behavior For comparisonwith our verbal

model and the results of our data analysis, we used a simulation model based on a

GLOBAL AMPHIBIAN DECLINES

149

long-termstudyof the populationdynamicsof Bufo marinus in northern Australia (2). Our model incorporatedsimple density-dependencein the adult stage and lognormalvariationin recruitment success. Meanrecruitment rateswere adjusted so that populationstended to fluctuateabout a mean size ratherthan explode or decline rapidly to extinction. The population arbitrarilystartedwith 100 adult female toads. Year-to-year survivalof adultfemales was set at the lower of 50% or 50 total. Each female produced9000 eggs, which survivedto reachmaturityas females in one year at variablerates equal to 0.00047ea+l, where a is a normally distributedrandomvariablewith mean 0, standarddeviation 1. The numberof adults was truncatedto an integer following survival and recruitmentin each year. We ran the simulation5000 times, for 1000 "years"each time. On average, the populationpersistedfor 423 years before extinction and contained296 adult females while extant. Withina run,recruitment differedon averageapproximately 300-fold between the lowest and highest years, while adult survivaldiffered 15fold as a consequence of density-dependence. While the population persisted, it decreasedin 56.32% of intervalsbetween years, a result that is in very close agreementwith the observation(Table 2) that populationsof bufonids declined between 55.5% of years. This suggests that our initial verbalmodel was correct: Whenpopulationfluctuations aredrivenby highly variablerecruitment, it is likely thatpopulationdynamicswill be characterized by occasionaloutbreaks with longer interveningperiodsof decrease,so thatthey are"indecline"morethan50%of the time. This resultis similarto the "storage" effect in open populations(45, 193). It is clearthata populationdecreasingin moreyears thanit increasesis not necessarily in decline. However,if therehas recentlybeen an increasein the general tendency of amphibianpopulationsto decline, there should be a correlationbetween the yearin which a studyendedandthe frequencyof decreasesin thatstudy. We tested this hypothesisusing the 85 time series in our database.We correlated the finalyearof each studywith the percentageof intervalsacrosswhichthe studied populationdeclined. We analyzedthe datafor each family of frogs and salamandersseparately, as combiningthemmighthaveconfoundedrealeffects of time with the effects of changes in the proportionsof studies that were carriedout on each taxon. We calculatedSpearman rankcorrelations because these straighten nonlinearrelationshipsand decreasethe effects of outliersand datathatarenot normally distributed. We found no evidence for any significantcorrelationof proportion decreaseswith time in any family (maximumrs = 0.327, minimum P = 0.234), an outcome supported by examinationof the data(Figure4), which do not suggest any strongtrendfor eitherfrogs or salamanders. The proportion of yearsin which a populationdeclines may be a weak indicator of trendsandis potentiallysubjectto difficultiesin decidinghow largea changein a realdecreaserather thannoise in the data. Correlations populationsize represents between populationsize and time might be a betterindicatorof populationstatus (30,148).

150

ALFORD . RICHARDS

w
-b

100-

o
0
en

80n
600~~~~~ 800 0 DO0 [I* ?0

40O~~~~~~~~K

?
o

L?
L bL

200

-tt

Li

0.

1980 1985 1990 1995 2000 Year (frogs)

1945 1955 1965 1975 1985 1995

Year(salamanders)
of theintervals fromone yearto thenextoverwhichpopulations Figure4 Proportion decreased studies(4 ormoreyears)of 85 localamphibian populations during long-term shownas squares, (28 of salamanders, 57 of frogs,shownas diamonds).The abscissa is the finalyearof each study.

CORRELATIONS OF POPULATION SIZE WITH TIME

If there is a general global phenomenonacting on all amphibiansand increasing their tendency to decrease in numbers,we would expect that recently completed studies would show more or strongernegative relationshipsbetween population size andtime thanwould olderones. Foreach of the 85 time seriesin ourdatabase, we regressedloglo(N+ 1) on yearof the series. We used loglo(N+ 1) becausethis

AMPHIBIAN DECLINES GLOBAL

151

stabilizesvariancesandbecausea populationincreasingor decreasingat a constant proportional rate will show a linearrelationshipof logl0(N + 1) with time. Years were adjustedwithin each time series so thatthe firstyear was year zero, to reduce possible influencesof roundingerrorson the regressions. We used the correlations and slopes of these regressionsin furtheranalysesto determinewhethertherewas any evidence for an overall trendtowardan increasein the incidence or intensity of negativerelationshipsbetween populationsize and time. In these analyses we examinedthe data for frogs and salamandersseparatelybecause trendsmight be of eitherthe slope presentin one butnotin the other.We firstlookedfor correlations or the correlationcoefficient with the year in which each time series ended, using Spearmanrankcorrelationsbecause of the unknownsamplingpropertiesof these two measures. None of the correlationsof the correlationcoefficient or the slope with final year were significant(salamanders:r, =-0.021, P = 0.917 and r, =
-0.024, P = 0.905; frogs: r, = 0.175, P = 0.174 and r, = 0.161, P = 0.232;

respectively). Althoughtherewas no trendin eitherorderfor changesovertime in therelationship of populationsize to year of study,it is still possible that a generaldeclining trendwas presentthroughoutthe period. To examine this possibility, we plotted the correlationsof populationsize with year and the slopes of the regressionsof populationsize on year against final year of the study (Figure 5a and 5b). Initial examinationof these figures could be a source of alarmbecause there are many more negative than positive correlationsand because most of the apparentlysignificantcorrelationsand slopes are negative. Furtheranalysis shows that both of these effects are probablyartifactsof the populationdynamicsof amphibiansand the fact that standardassessments of significance should not be applied to time seriesdata. In orderto morerigorouslyassess the significanceof these correlations to the resultsof ourpopulationsimulation. On each of the and slopes, we returned 5000 iterationsof the time series, we calculatedcorrelationsand slopes for regressions of log10(N + 1) on year for time series containing4 through9, 11, 12, 14, at year 20 of the simulatedtime series 15, 16, 23, and 28 years, startingarbitrarily to allow the effects of initial conditionsto disappear.This resultedin at least 4904 coefficients for each combinationof parameterand series length (less than 5000 because a few populationswent extinct before the final year of each simulated series). We sorted the vector of coefficients for each combinationof parameter and series length into ascendingorder,and we took the coefficients at 0.025 from the bottom and top of the series as the upperand lower 95% confidencelimits for the correlation and slope of regressionsof an amphibian populationon year. These confidence limits appearin Figure 5, c and d, with plots of the correlationsand slopes obtainedfrom the time series in our database.Using the criterionthatto be must fall outside these empirical95% confidence limits, significanta parameter only a single correlationand four slopes are significant,well within the number thatwould be expected due to Type I errorwhen 85 comparisonsare made. Althoughwe mustconcludethatthereis no evidencein the 85 time serieswe analyzed to suggest thatcorrelationsor slopes of regressionsof logl0(N + 1) against

152

ALFORD * RICHARDS

1.0
?

04 b
?- @? Q0.2-

0.50

??0

??

8?

ot?

??

t?

l?80e

X*,0

-1.0
I

*
1980
' i ' I ' I '

*_

1985 1990 1995 2000

I1-'-I

1980 1985 1990 1995 2000

_ '

_ 1 'I

_ I

_ '

Year (frogs)
19 19 1945 1955 1965 1985 1995 Year (salamanders)

Year (frogs)
1945 1955 19 1975 1985 1995 Year (salamanders)

E-.

Q0Q8

0.30

-084 13~~~~~~~~~~~~~~~. 0-0

rC>

<>

C>

02

-0.

10

20 I. ,

Years intime senes

,30

10 I

20

I,

Years intime series

Figure 5 (a) Product-moment correlation of 1og10(N + 1) with year of study for 85 time
series of four or more years of data on local populations of amphibians, plotted against final year of the study. Data for salamanders shown as squares, frogs as diamonds. Correlations that would be significant at ae = 0.05 using standard parametric criteria are indicated by filled shapes. (b) Slopes of regressions of 1og10(N + 1) on year of study, other details the same as in (a). (c) The same correlation coefficients as in (a), plotted against number of years in the time series from which the correlation was derived. Lines indicate upper and lower 95%tconfidence limits derived from correlations calculated on 5000 simulated time series of each length. Only the single correlation outside these confidence limits should be considered significant (indicated by the filled shape). (d) The same slopes as in (b), plotted against number of years in the time series from which they were derived. Lines indicate 95%tconfidence limits, derived as in (c). Only four of the 85 slopes should be regarded as significant at the 0.05 level (filled shapes).

AMPHIBIAN GLOBAL DECLINES

153

time have changedin recent years, we might still be concernedover the apparent excess of negative relationshipsof populationsize with time (Figure 5, a and b). of populationsize with time (67%)arenegative. Fifty-sevenof the 85 relationships The proportionof negativecorrelationsfrom our simulationincreasedwith length of the time series from 54.8% negative with a series length of 4 years to 57.5% with a series length of 28 years. We used these expected proportions,weighted by the numberof each of the real time series thatwere of each length, to calculate the expectednumbersof correlations thatshouldhave been greaterand less than0 (37.6 and 47.4, respectively),and comparedthese with the observednumbers(28 and57) using a chi-squared goodness-of-fittest (chi-squared= 4.42, P _ 0.037). This significantresult indicatesthatthe amphibianpopulationswe examinedhad a greaternumberof negative correlationswith time than would be expected, as comparedwith our simulatedpopulations. This could reflect a general tendency towarddecline, but it could also reflect the fact that our simulationmodel, while it probablyprovides a more realistic null hypothesis than the simple assumption that effects in both directionsshould be equal, does not perfectly reflectthe populationbehaviorof all amphibians.Using a greatervarietyof models to generate for each family, genus, or even species would null hypotheses more appropriate obviously be preferable.It would also be useful to examine the sensitivity of our conclusions to variationsin model parameters and form. The use of appropriate null hypotheses will allow more rigorousexamination of the behaviorof individualpopulations. However,many amphibiansappearto live in local populationsthat interactstronglywith otherpopulations,so that unthe implicationsof local populationdynamicsfor species persistence derstanding requiresa knowledge of theirmetapopulation biology.

AMPHIBIANMETAPOPULATION BIOLOGY
A metapopulationconsists of a group of local populations inhabiting more or differs from a collection less discrete patches of habitat(94). A metapopulation of independentpopulations in that there is substantialmigrationbetween local populations,so that no local populationis likely to remainextinct for any length of time. Migrationrates may be high enough to affect rates of local population increaseanddecrease(95, 139). A metapopulation differsfroma single subdivided populationby having sufficientlylow ratesof migrationbetweenlocal populations that they exhibit some degree of independencein their dynamics, including the possibility of declining to extinction(93).

MetapopulationStudies on Amphibians
Although it has been suggested that amphibiansare generally highly philopatric (30, 61, 172), many species departfrom this pattern. One of the problems that studiesof amphibian plagues mark-recapture populationsat single sites is the high

154

ALFORD * RICHARDS

rates at which animalsdisappearfrom local populations(2, 31,40, 111, 134, 156). Substantialrates of dispersal among local populations have been documented in many species (2,21,37,38,42,58,66,81,112,161,173,198)). Additionalevicomes fromexplicit metapopdence thatmanyamphibians live in metapopulations viridescens ulationstudies(175). Breedingpopulationsof thenewtNotophthalmus act as cells in a regional metapopulation(83). The Europeanpool frog Rana lessonae lives in spatiallycomplexmetapopulations; the qualityof potentialbreeding sites and their degree of isolation from other sites determines their probability of occupancy and the probabilityof local extinction (176). Increases in the isolation of habitatpatches due to naturalsuccession or habitat destruction decrease the persistence of local populations (176). Surveys of Rana clamitans occupancy at 160 ponds in three distinct regions (103) demonstratedthe existence of regional metapopulations.Colonization rates varied from 0 to 0.25 ponds* (pond occupied. year)-1, while local extinctionrates were between 0 and 0.5 ponds* (pond occupied. year)-. Small populationswere more proneto local extinctionthanwere large ones, andtherewas no overalltrendin occupancyrates when all threemetapopulations were considered(103). Tenotherspecies examined at 97 ponds in the same regions (101) also exhibited metapopulation dynamics, with rates of turnoverfrom 0.07 to 0.30 species. (pond. year)-. Pool size and isolation both affected species richness in 77 pools in the southernNetherlands (126) and 332 habitatsin Bavaria(68).

Modelsof AmphibianMetapopulationsandTheirImplications
Because amphibiansoften live in metapopulations,declines and extinctions of local populationsmay be common events. Detailed studies of local populations may give useful insights into the autecology of species, but they are of limited use in evaluatingthe status of regional metapopulations.One approachto examis to examine simple probabilisticmodels ining the behaviorof metapopulations for the frequencywith which local populationsmight change in status (184). A was used to probabilisticnull model for populationdeclines and disappearances examine whether the declines and disappearancesof frogs that occurredin the 1980s at Monteverde,Costa Rica, might be due to chance (153). Pounds et al (153) used long-term studies to estimate probabilities of disappearance.They then compared the numbers of species disappearancesat their study sites to the numbers of disappearancespredicted by the cumulative binomial distributhan would have been tion, and they found that far more species had disappeared predicted. The probabilisticapproach(153) seems to be a useful way to quantifythe idea at a rate that,when certainregions or taxa areconsidered,species aredisappearing "too great to be coincidental." Several studies have produceddata that could be examined using this technique. It seems likely that the disappearances of seven species from all sites at elevations above 400 m throughoutthe AustralianWet Tropics (135,163) would be shown to be extremely improbable, as would the

AMPHIBIAN GLOBAL DECLINES

155

disappearances of threespecies andlargedeclines in site occupancyof four others in the Yosemiteareaof California(69), the disappearance of Rana cascadae from the southernend of its range (75), and the disappearances of many species at Las Tablas,Costa Rica (131). A more complex analyticalmodel was used to predict the rates of extinction of local populationsof the common toad Bufo bufo and the crested newt Triturus cristatusin Europe,andto examinehow thoseratesshouldrespondto the size of the local habitatpatchandits distancefroma sourceof migrants(92). The persistence of populationsof both species should increase with the carryingcapacity of the local habitatand shoulddecreasewith increasingdistanceto a sourceof migrants. Over a wide range of carryingcapacities,the criticaldistances are approximately 10 times as great(-5 km) for toadsas they arefor newts (-500 m) (92). This study must be takeninto account suggested thatboth patch size and spatialdistribution when managing amphibianmetapopulations. Informationon patch occupancy from a geographicallyreferenceddatabase(168) indicates that small, relatively in the metapopulation isolated wetlandsare important dynamicsof amphibiansin SouthCarolina,USA. Loss of these habitatsmightlead to disproportionately large ratesof extinctionin regionalmetapopulations thatdependupon them as stepping stones in colonizationand as refuges from local extinctions(168). Delineating and monitoringthe status of metapopulationsrequiresextensive sampling, but because metapopulationdynamics are concerned mostly with the presenceor absenceof species in local populations,samplingof local populations the dynamicsof amphibian does not need to be intense (87). Fully understanding metapopulations will requiremuchmoreinformationon movementsanddispersal among local populationsthanis presentlyavailable.

CONCLUSIONS
It is clear that local populations of many amphibianspecies have declined in recentyears, andthereare severalwell-documentedcases of declines at and above factors can the level of regionalmetapopulations.Although many environmental of amphibians,few studies adverselyaffect the growth,survival,andreproduction thatthese effects altertheirpopulationdynamics. have convincinglydemonstrated Studies linking factors that negatively affect amphibiansin the laboratoryor in artificialfield trials with effects on populationdynamics in more naturalsettings are urgentlyneeded. Local populationsof amphibianstend to fluctuate,and our results show it is likely that many normallydecrease more often than they increase. It is therefore importantto develop realistic null hypotheses for their behavior. If we had not based our null hypothesison a simulationof frog populationdynamics,we would have reached very differentconclusions in our analysis of populationbehavior. Additionaldata on an ecologically diverse range of species will allow the development of more sophisticatedand specific null hypotheses for a greaterrange of

156

ALFORD * RICHARDS

populations. This is necessary to make rigorous tests of the responses of local factorspossible. populationsto environmental Many amphibianspecies occur as metapopulations,so the dynamics of local populationsmay be poor indicatorsof their status. Declines and extinctions of are likely to result from interactionsbetween changes in the dymetapopulations namics of local populations and habitat modificationor loss (93,94, 168). For many species, understandingthe factors affecting the status and dynamics of shouldthereforebe the ultimategoal of studiesaimingto prevent metapopulations requiresdifferentdatacollection or reversedeclines. Monitoringmetapopulations techniquesthanmonitoringisolated populations,so the first step in designing any monitoringprogramshould be to determinewhetherthe species of interestforms a metapopulation.Studies integratingresearchwithin local populationswith inlevel are most likely to discover the causes of vestigations at the metapopulation of amphibian diversity. declines andprovidea basis for the conservation amphibian ACKNOWLEDGMENTS The originaldataincludedin this reviewwere collected with the supportof funding from the AustralianResearch Council and the CooperativeResearch Centre for TropicalRainforestEcology and Management.We thankJ Winterand the Comfor permissionto includedatafromReference monwealthAluminiumCorporation 200, and S Droege for permissionto use dataextractedfrom the USGS amphibian count database(72). The manuscriptwas improvedwith the help of MJ Caley, CN Johnson,and L Schwarzkopf. Visit the Annual Reviewshome page at http://www.AnnualReviews.org
LITERATURE CITED 1. Alford RA. 1999. Ecology: resource use, competition, and predation. In Tadpoles: TheBiology of AnuranLarvae,ed. RW McDiarmid, R Altig. Chicago: Univ. Chicago Press. In press 2. Alford RA, Cohen MP, Crossland MR, HearndenMN, SchwarzkopfL. 1995. Populationbiology of Bufo marinusin northern Australia.In Wetland Research in the WetDry Tropicsof Australia,ed. CM Finlayson, pp. 173-181. Canberra:Commonwealthof Australia 3. Alford RA, RichardsSJ. 1997. Lack of evidence for epidemic disease as an agent in the catastrophicdecline of Australianrain forest frogs. Conserv.Biol. 11:1026-29 4. Altig R, Channing A. 1993. Hypothesis: functionalsignificanceof colourandpattern of anurantadpoles.Herpetol.J. 3:73-75 5. Andren C, Henrikson L, Olsson M, Nilson G. 1988. Effects of pH and aluminium on embryonic and early larval stages of Swedish brownfrogs Rana arvalis, R. temporaria and R. dalmatina. Holarctic Ecol. 11:127-35 6. Anzalone CR, Kats LB, GordonMS. 1998. Effects of solar UV-B radiationon embryonic developmentin Hyla cadaverina,Hyla regilla, and Tarichatorosa. Conserv.Biol. 12:646-53 7. Arntzen JW, Teunis SFM. 1993. A sixyearstudyon the populationdynamicsof the

GLOBALAMPHIBIANDECLINES crestednewt (Triturus cristatus)following the colonization of a newly createdpond. Herpetol.J. 3:99-1 10 Ash AN. 1988. Disappearance of salamanders from clearcutplots. J. Elisha Mitchell Sci. Soc. 104:116-22 Ash AN. 1997. Disappearanceand return of Plethodontid salamandersto clearcut plots in the southern Blue Ridge Mountains. Conserv.Biol. 11:983-89 BarinagaM. 1990.Wherehaveall the froggies gone? Science 247:1033-34 Beattie RC, Tyler-JonesR. 1992. The effects of low pH and aluminumon breedJ. ing success in the frog Rana temporaria. Herpetol. 26:353-60 Beattie RC, Tyler-Jones R, Baxter MJ. 1992. The effects of pH, aluminiumconcentration and temperatureon the embryonic development of the European common frog, Rana temporaria.J. Zool., London228:557-70 changeas Beebee TJ. 1977. Environmental a cause of NatterjackToad Bufo calamita declines in Britain. Biol. Cons. 11:87102 Beebee TJC. 1995. Amphibian breeding and climate. Nature374:219-20 Beebee TJC, Denton JS, Buckley J. 1996. Factors affecting population densities of adult natterjacktoads Bufo calamita in Britain.J. Appl. Ecol. 33:263-68 Beebee TJC, Flower RJ, Stevenson AC, PatrickST, Appleby PG, et al. 1990. DeToadBufo calamita cline of the Natterjack in Britain: paleoecological, documentary, and experimental evidence for breeding site acidification.Biol. Conserv.53:1-20 Beiswenger RE. 1986. An endangered species, the Wyoming toad Bufo hemiophrys baxteri-the importanceof an early warningsystem. Biol. Cons. 37:59-71 Berger L, SpeareR, Daszak P, Green DE, Cunningham AA, et al. 1998. Chytridiomycosis causes amphibianmortalityassociated with population declines in the rainforests of AustraliaandCentralAmer-

157

19.

8.

20.

9.

21.

10. 11.

22.

12.

23.

13.

24.

14. 15.

25.

16.

26.

17.

27.

18.

28.

ica. Proc. Natl. Acad. Sci., USA 95:903136 BertramS, Berrill M. 1997. Fluctuations in a northern populationof gray treefrogs, Hyla versicolor. See Ref. 87, pp. 57-63 Berven KA. 1990. Factorsaffectingpopuin larvalandadultstages lationfluctuations of the wood frog (Ranasylvatica).Ecology 71:1599-1608 Berven KA, GrudzienTA. 1990. Dispersal in the wood frog (Rana sylvatica): implicationsfor genetic populationstructure. Evolution44:2047-56 Bishop CA. 1992. The effects of pesticides on amphibians andthe implicationsfor determining causes of declines in amphibian populations.In Declines in Canadian AmphibianPopulations: Designing a NationalMonitoringStrategy,ed. CA Bishop, KE Pettit, pp. 67-70. Occas. Pap. No. 76, Can. WildlifeServ. Blaustein AR. 1994. Chicken Little or Nero's fiddle? A perspectiveon declining amphibianpopulations.Herpetologica 50: 85-97 Blaustein AR, EdmundB, Kiesecker JM, Beatty JJ, Hokit DG. 1995. Ambient ultravioletradiationcauses mortalityin salamandereggs. Ecol. Appl. 5:740-43 Blaustein AR, Hoffman PD, Hokit DG, Kiesecker JM, Walls SD, Hays JB. 1994. UV repairand resistanceto solar UV-B in amphibianeggs: a link to populationdeclines? Proc. Natl. Acad. Sci., USA 91: 1791-95 BlausteinAR, HoffmanPD, KieseckerJM, Hays JB. 1996. DNA repair activity and resistanceto solar UV-B radiationin eggs of the red-legged frog. Conserv.Biol. 10: 1398-1402 BlausteinAR, HokitDG, O'HaraRK, Holt RA. 1994. Pathogenic fungus contributes to amphibianlosses in the Pacific Northwest. Biol. Conserv.67:251-54 Blaustein AR, KieseckerJM, ChiversDP, Anthony RG. 1997. Ambient UV-B radiation causes deformities in amphibian

158

ALFORD * RICHARDS embryos. Proc. Natl. Acad. Sci., USA 94: 13735-37 Blaustein AR, Wake DB. 1990. Declining amphibianpopulations: a global phenomenon? TrendsEcol. Evol. 5:203-4 BlausteinAR, WakeDB, Sousa WP. 1994. Amphibiandeclines: judgingstability,persistence, and susceptibilityof populations to local and global extinctions. Conserv. Biol. 8:60-71 BradfordDF. 1989. Allotopic distributions fishesin high of nativefrogs andintroduced SierraNevada lakes of California: implications of the negative effect of fish introductions. Copeia 1989:775-78 DF. 1991. MassmortalityandexBradford tinction in a high-elevationpopulation of Rana muscosa. J. Herpetol. 25:174-77 Bradford DF, Gordon MS, Johnson DF, AndrewsRD, JenningsWB. 1994. Acidic deposition as an unlikely cause for amphibian population declines in the Sierra Biol. Conserv.69:155Nevada,California. 61 Bradford DF, Swanson C, Gordon MS. 1994. Effects of low pH and aluminumon amphibiansat high elevation in the Sierra Nevada,California.Can.J. Zool. 72:127279 Bradford DF, TabatabaiF, Graber DM. 1993. Isolation of remaining populations of the nativefrog, Rana muscosa,by introducedfishes in SequoiaandKing's Canyon National Parks, California.Conserv.Biol. 7:882-88 Bradford DF, Swanson C, Gordon MS. 1992. Effects of low pH and aluminum on two declining species of amphibiansin the SierraNevada, California.J. Herpetol. 26:369-77 Breckenridge WJ, Tester JR. 1961. Growth,local movementsand hibernation of the Manitoba toad, Bufo hemiophrys. Ecology 42:637-46 Breden F. 1987. The effect of postmetamorphicdispersal on the population genetic structureof Fowler's toad, Bufo woodhouseifowleri. Copeia 1987:386-95 39. BrooksRJ. 1992. Monitoringwildlife populations in long-termstudies. See Ref. 22, pp. 94-97 40. Brown WC, Alcala AC. 1970. Population ecology of the frog Rana erythraea in southern Negros, Philippines. Copeia 1970:611-22 41. Bruce RC. 1995. The use of temporary removal sampling in a study of populaDesmogtion dynamicsof the salamander nathusmonticola.Aust.J. Ecol. 20:403-12 42. Buckley D, Arano B, HerreroP, Llorente G. 1996. Population structureof Moroccan waterfrogs: genetic cohesion despitea distribution. J. Zool. Syst.Evol. fragmented Res. 34:173-79 43. Bury RB, Corn PS. 1988. Responses of aquaticand streamsideamphibiansto timber harvest:A review.In StreamsideManagement: Riparian Wildlifeand Forestry Interactions,ed. KJ Raedeke,pp. 165-8 1. Seattle: Univ. Was. Press 44. Caldwell JP. 1987. Demography and life history of two species of chorus frogs in SouthCarolina.Copeia (Anura:Hylidae) 1987:114-27 45. Caley MJ, CarrMH, Hixon MA, Hughes TP, Jones GP, Menge BA. 1996. Recruitment and the local dynamics of open marine populations. Annu. Rev. Ecol. Syst. 27:477-500 46. Carey C, Bryant CJ. 1995. Possible interrelationships among environmentaltoxidevelopment,anddecline cants,amphibian of amphibian populations.Environ.Health Perspec. 103(Suppl.4):13-17 47. Carey CL. 1993. Hypothesis concerning the causes of the disappearanceof borealtoadsfromthe mountainsof Colorado. Conserv.Biol. 7:355-62 48. Chovanec A. 1992. The influence of tadpole swimmingbehaviouron predationby dragonflynymphs.Amphibia-Reptilia13: 341-49 49. Clay D. 1997. The effects of temperature and acidity on spawning of the spotted

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

GLOBALAMPHIBIANDECLINES salamander, Ambystoma maculatum, in FundyNationalPark.See Ref. 87, pp. 22632 Cohen MP. 1995. Ecology of two populations of Bufo marinus in north-eastern Australia. PhD thesis, James Cook Univ., Townsville,Australia.203 pp. Corn PS. 1994. What we know and don't knowaboutamphibian declinesin the west. In SustainableEcological Systems: Implementingan Ecological Approachto Land Management,ed. LF DeBano, WW Covington,pp. 59-67. FortCollins, CO:USDA For. Serv., Rocky Mountain For. Range Exp. Station CornPS, BuryRB. 1989. Logging in western Oregon: responses of headwaterhabitats and stream amphibians. For. Ecol. Manage. 29:39-57 Corn PS, Fogleman JC. 1984. Extinction of montane populations of the Northern LeopardFrog (Ranapipiens) in Colorado. J. Herpetol. 18:147-52 Corn PS, StolzenburgW, Bury RB. 1989. Acid PrecipitationStudiesin Coloradoand Wyoming: Interim Report of Surveys of Montane Amphibians and Water Chemistry. Biol. Rep. 80, US Fish & Wildlife Serv.pp. 1-56 Corn PS, VertucciFA. 1992. Descriptive risk assessmentof the effects of acidic deposition on Rocky Mountainamphibians. J. Herpetol. 26:361-69 CrawshawGJ. 1992. The role of disease in amphibiandecline. See Ref. 22, pp. 60-62 CrawshawGJ. 1997. Disease in Canadian amphibianpopulations. See Ref. 87, pp. 258-70 CrumpML. 1986. Homingandsite fidelity in a neotropicalfrog, Atelopusvarius (Bufonidae). Copeia 1986:438-44 CunninghamAA, Langton TES, Bennett PM, LewinJF,DrurySEN, et al. 1993. Unusual mortalityassociated with poxviruslike particles in frogs (Rana temporaria). Vet.Rec. 133:141-42 CunninghamAA, Langton TES, Bennett

159

50.

61.

51.

62.

63.

52.

64.

53.

65.

54.

66.

67.

55.

68.

56. 57.

69.

58.

70.

59.

71.

60.

72.

PM, Lewin JF, Drury SEN, et al. 1996. Pathologicaland microbiologicalfindings from incidents of unusualmortalityof the common frog (Rana temporaria).Philos. Trans.R. Soc. Lond. Ser B. 351:1539-57 Daugherty CH, Sheldon AL. 1982. Agespecific movementpatternsof the frog Ascaphus truei.Herpetologica 38:468-74 Delis PR, Mushinsky HR, McCoy ED. 1996. Decline of some west-central Florida anuranpopulationsin response to habitat degradation. Biodiv. Conserv.5:1579-95 Dodd CK Jr. 1991. Drift fence-associated sampling bias of amphibiansat a Florida sandhillstemporary pond. J. Herpetol. 25: 296-301 Dodd CK Jr. 1992. Biological diversityof a temporarypond herpetofaunain north Floridasandhills.Biodiv. Conserv.1:125142 Dodd CK Jr. 1994. The effects of drought on populationstructure, activity,andorientation of toads (Bufo quercicus and B terrestris) at a temporarypond. Ecol. Ethol. Evol. 6:331-49 Dole WJ. 1965. Summer movements of adultleopardfrogs,RanapipiensSchreber, in northern Michigan.Ecology 46:236-55 Donnelly MA, Crump ML. 1998. Potential effects of climate change on two neotropical amphibian assemblages. Climate Change39:541-61 Dorn WMP,BrandlR. 1991. Local distribution of amphibians: the importanceof habitat fragmentation.Global Ecol. Biogeogr Lett. 1:36-41 Drost CA, FellersGM. 1996. Collapse of a regionalfrog faunain the Yosemiteareaof the CaliforniaSierraNevada, USA. Conserv. Biol. 10:414-25 Duellman WE, Trueb L. 1986. Biology of Amphibians.New York: McGraw-Hill. 670 pp. Dunson WA, Wyman RL, Corbett ES. 1992. A symposiumon amphibian declines and acidification.J. Herpetol. 26:349-52 Eagle P. 1999. AmphibianCountDatabase.

160

ALFORD * RICHARDS United States Geological Survey, http:II www.mp2-pwrc.usgs.gov/ampCV/ampdb. cfm Elmberg J. 1990. Long-term survival, length of breedingseason, and operational sex ratioin a borealpopulationof common L. Can.J. Zool. 68: frogs, Rana temporaria 121-27 ElmbergJ. 1993. Threatsto boreal frogs. Ambio 22:254-255 Fellers GM, Drost CA. 1993. Disappearance of the cascades frog Rana cascadae at the southernend of its range,California, USA. Biol. Conserv.65:177-81 FisherRN, ShafferHB. 1996. The decline in California's of amphibians GreatCentral Valley. Conserv.Biol. 10:1387-97 Fite KV, Blaustein A, Bengston L, Hewitt HE. 1998. Evidenceof retinallight damage in Rana cascadae: a declining amphibian species. Copeia 1998:906-14 FredaJ, Dunson WA. 1986. Effects of low pH and other chemical variables on the local distributionof amphibians. Copeia 1986:454-66 FredaJ, Sadinski WJ, Dunson WA. 1991. popuLong termmonitoringof amphibian lations with respectto the effects of acidic deposition. Water, Air,Soil Pollut. 55:44562 FreedmanB, ShackellNL. 1992. Amphibians in the context of a national environmental monitoringprogram.See Ref. 22, pp. 101-104 Friedl TWP, Klump GM. 1997. Some aspects of population biology in the European treefrog, Hyla arborea. Herpetologica 53:321-30 Gamradt SC, Kats LB. 1996. Effect of introduced crayfish and mosquitofish on CaliforniaNewts. Conserv.Biol. 10:115562 Gill DE. 1978. The metapopulationecology of the red-spotted newt, Notophthalmus viridescens (Rafinesque). Ecol. Monogr.48:145-66 Gittins SP. 1983. Populationdynamics of the common toad (Bufo bufo) at a lake in mid-Wales.J. Anim.Ecol. 52:981-88 Grant KP, Licht LE. 1995. Effects of ultravioletradiationon life-history stages of anurans from Ontario, Canada. Can. J. Zool. 73:2292-306 Green DM. 1992. Fowler's toads (Bufo woodhousei fowleri) at Long Point, Ontario: changing abundanceand implications for conservation.See Ref. 22, pp. 3743 GreenDM. 1997. Perspectiveson amphibian populationdeclines: definingthe problem and searching for answers. In Amphibians in Decline. CanadianStudiesof a Global Problem, ed. DM Green, pp. 291308. Herpetological Conserv.,Vol 1. Green DM. 1997. Temporal variation in abundanceand age structurein Fowler's toads,Bufofowleri, at Long Point,Ontario. See Ref 87, pp. 45-56 Gyovai F. 1989. Demographicanalysis of the moor frog (Rana arvalis olterstorffi Feje'rva'ry 1919) population in Fraxino pannonicae-Alnetumof the Tisza basin. Tiscia (Szeged)24:107-21 HairstonNG. 1983. Growth,survival,and reproductionof Plethodonjordani: tradeoffs between selective pressures. Copeia 1983:1024-35 HairstonNG Sr, Wiley RH. 1993. No decline in salamander(Amphibia:Caudata) populations: a twenty year study in the southern Appalachians. Brimleyana 18: 59-64 HalleyJM,OldhamRS, ArntzenJW.1996. Predicting the persistence of amphibian populations with the help of a spatial model. J. Appl. Ecol. 33:455-70 HanskiI. 1997. Metapopulation dynamics: from concepts and observationsto predictive models. In MetapopulationBiology: Ecology, Genetics, and Evolution, ed. I Hanski, M. Gilpin, pp. 69-91. San Diego, CA: Academic Press. 512 pp. HanskiI. 1998. Metapopulation dynamics. Nature396:41-49

85.

73.

86.

74. 75.

87.

76.

77.

88.

78.

89.

79.

90.

80.

91.

81.

92.

82.

93.

83.

94.

84.

GLOBALAMPHIBIANDECLINES 95. Hanski I, Pakkala T, KuussaariM, Lei G. 1995. Metapopulationpersistence of an endangeredbutterflyin a fragmented landscape.Oikos 72:21-28 96. HarteJ, HoffmanE. 1989. Possible effects of acidic depositionon a Rocky Mountain population of the tiger salamanderAmbystomatigrinum.Conserv.Biol. 3:14958 97. Hayes MP, Jennings MR. 1986. Decline of frog species in westernNorthAmerica: are bullfrogs (Rana catesbeiana) responsible? J. Herpetol. 20:490-509 98. Hays JB, Blaustein AR, Kiesecker JM, Hoffman PD, Pandelova I, Coyle D, RichardsonT. 1996. Developmental responses of amphibiansto solar and artificial UVB sources: a comparativestudy. Photochem.Photobiol. 64:449-55 99. HecnarSJ. 1997. Amphibianpond communities in southwestern Ontario. See Ref. 87, pp. 1-15 100. Hecnar SJ, M'Closkey RT. 1996. The effects of predatory fish on amphibian species richness and distribution.Biol. Conserv.79:123-31 101. Hecnar SJ, M'Closkey RT. 1996. Regionaldynamicsandthe statusof amphibians. Ecology 77:2091-97 102. HecnarSJ,M'CloskeyRT.1997. Changes in the composition of a ranid frog communityfollowing bullfrogextinction.Am. Midl. Nat. 137:145-50 103. HecnarSJ, M'Closkey RT. 1997. Spatial scale and determination of species status of the green frog. Conserv.Biol. 11:67082 104. Henrikson B-I. 1990. Predationon amphibian eggs and tadpoles by common predators in acidified lakes. Holarctic Ecol. 13:201-6 105. HermanTB, Scott FW. 1992. Assessing the vulnerability of amphibians to climatic warming.See Ref. 22, pp. 46-49 106. Hero J-M, Gillespie GR. 1997. Epidemic disease and amphibiandeclines in Australia. Conserv.Biol. 11:1023-25

161

107. HeyerWR. 1979. Annualvariationin larval amphibianpopulationswithin a temAcad. Sci. 69:65-74 peratepond.J. Wash. 108. Homolka M, Kokes J. 1994. Effect of air pollution and forestrypracticeon the of Salamandrasalarangeandabundance mandra.Folia Zoologica 43:49-56 109. Home MT, Dunson WA. 1994. Exclusion of the Jefferson salamander,Ambystomajeffersonianum,from some potential breeding ponds in Pennsylvania: andmetals on effects of pH, temperature, embryonic development.Arch. Environ. Contamination Toxicol.27:323-30 110. Home MT, Dunson WA. 1995. Effects of low pH, metals,andwaterhardnesson larArch.Environ.Contamival amphibians. nation Toxicol.29:500-5 111. HustingEL. 1965. Survivaland breeding structurein a population of Ambystoma maculatum.Copeia 1965:352-59 112. IshchenkoVG. 1989. Populationbiology of amphibians.Sov. Sci. Rev. F Physiol. Gen. Biol. 3:119-55 and 113. JaegerRG. 1980. Density-dependent causes of extinction density-independent of a salamander population.Evolution34: 617-21 114. Jennings MR, Hayes MP. 1994. Decline of native ranid frogs in the desert southwest. In Southwestern Herpetol. Soc., Spec. Publ. No. 5, ed. PR Brown, JW Wright,pp. 183-211 115. JoglarRL, BurrowesPA. 1996. Declining amphibianpopulationsin PuertoRico. In Contributionsto WestIndian Herpetology: A Tributeto Albert Schwartz, ed. R Powell, RW Henderson, pp. 371-80. SSARContrib.to Herpetol., Volume 12. 116. JohnsonB. 1992. Habitatloss and declining amphibianpopulations.See Ref. 22, pp. 71-75 117. KagariseShermanC, MortonML. 1993. Populationdeclines of Yosemite toads in the easternSierraNevadaof California.J. Herpetol. 27:186-98 118. Kats LB, Petranka JW, Sih A. 1988.

162

ALFORD * RICHARDS defenses andthe persistence Antipredator of amphibianlarvaewith fishes. Ecology 69:1865-70 KerrJB, McElroyCT. 1993. Evidence for large upwardtrendsof ultraviolet-Bradiation linked to ozone depletion. Science 262:1032-34 KieseckerJM, Blaustein AR. 1995. Synergism between UV-B radiation and a pathogen magnifies amphibian embryo mortalityin nature.Proc. Natl. Acad. Sci., USA92:11049-52 KieseckerJM, Blaustein AR. 1997. Population differences in responses of redlegged frogs (Ranaaurora)to introduced bullfrogs.Ecology 78:1752-60 Kiesecker JM, Blaustein AR. 1998. Effects of introducedbullfrogs and smallmouth bass on microhabitatuse, growth and survival of native red-legged frogs (Rana aurora). Conserv. Biol. 12:77687 Koonz W. 1992. Amphibians in Manitoba. See Ref. 22, pp.19-20 JW,Smith KuckenDJ, Davis JS, Petranka KC. 1994. Anakeestastreamacidification and metal contamination: effects on a J. Environ.Qual. salamander community. 23:1311-17 KuzminS. 1994. The problemof declining amphibianpopulations in the ComStates and admonwealthof Independent Alytes 12:123-34 jacent territories. LaanR, VerboomB. 1990. Effects of pool size and isolation on amphibiancommunities. Biol. Conserv.54:251-62 LannooMJ,LangK, WaltzT, PhillipsGS. 1994. An alteredamphibianassemblage: Dickinson County, Iowa, 70 years after FrankBlanchard'ssurvey.Am.Midl. Nat. 131:311-19 Laurance WF. 1996. Catastrophic declines of Australianrainforestfrogs: Is unusualweatherresponsible? Biol. Conserv. 77:203-12 LauranceWF, McDonald KR, Speare R. 1996. Epidemic disease and the catastrophic decline of Australianrain forest frogs. Conserv.Biol. 10:406-13 Licht LE, GrantKP. 1997. The effects of ultraviolet radiationon the biology of amphibians.Am. Zool. 37:137-45 Lips KR. 1998. Decline of a tropical montaneamphibianfauna.Conserv.Biol. 12:106-17 LizanaM, PedrazaEM. 1998. The effects of UV-B radiation on toad mortality in mountainousareasof centralSpain. Conserv. Biol. 12:703-7 Long LE, Saylor LS, Soule ME. 1995. A pH/UV-B synergismin amphibians.Conserv. Biol. 9:1301-3 Martof B. 1956. Factors influencing the size and composition of populations of Rana clamitans.Am. Midl. Nat. 56:22445 McDonald KR, Alford RA. 1999. A review of declining frogs in northern Queensland. In Declines and Disappearances of AustralianFrogs, National 1997, ed. A ThreatenedFrog Workshop Campbell. Canberra:EnvironmentAustralia.In press Meyer AH, Schmidt BR, Grossenbacher K. 1998. Analysis of three amphibian populations with quarter-centurylong time series. Proc. R. Soc. Lond., Ser. B :523-28 MierzwaKS. 1998. Statusof northeastern Illinois amphibians.In Status and Conservationof MidwesternAmphibians,ed. MJLannoo,pp. 115-24. Iowa City: Univ. Iowa Press Nagle MN, Hofer R. 1997. Effects of ultraviolet radiationon early larval stages of the Alpine newt, Triturus alpestris,undernatural andlaboratory conditions.Oecologia 110:514-19 Nee S, May RM, Hassell MP. 1997. Two-species metapopulationmodels. In Metapopulation Biology: Ecology, Genetics, and Evolution, ed. I Hanski, M Gilpin,pp. 123-47. San Diego, CA: Academic Press. 512 pp.

130.

119.

131.

120.

132.

121.

133.

134.

122.

135.

123. 124.

136.

125.

137.

126.

127.

138.

128.

139.

129.

GLOBALAMPHIBIANDECLINES 140. Nicholls N. 1984. A system for predicting the onset of the northAustralianwetseason. J. Climatol.4:425-35 141. Nyman S. 1986. Mass mortality in larval Rana sylvatica attributable to the bacterium,Aeromonashydrophilum. J. Herpetol. 20:196-201 142. Olson DH. 1989. Predation on breeding western toads (Bufo boreas). Copeia 1989:391-97 143. OvaskaK. 1997. The vulnerability of amphibiansin Canada to globalwarmingand increasedsolarradiation.See Ref. 87, pp. 206-25 144. Ovaska K, Davis TM, Flamarique IN. 1997. Hatching success and larval survival of the frogs Hyla regilla and Rana aurora underambientand artificiallyenhancedsolarultravioletradiation.Can.J. Zool. 75:1081-88 145. Oza GM. 1990. Ecological effects of the frog's legs trade.Environmentalist 10:3941 146. PechmannJHK,Scott DE, SemlitschRE, Caldwell JP, Vitt LJ, Gibbons JW. 1991. Declining amphibian populations: the problem of separating human impacts from natural fluctuations. Science 253: 892-95 147. Pechmann JHK, Wake DB. 1997. Declines and disappearancesof amphibian populations. In Principles of Conservation Biology, ed. GK Meffe, CR Carroll, pp. 135-37. Sunderland, MA: Sinauer. 2nd ed. 148. Pechmann JHK, Wilbur HM. 1994. Puttingdeclining amphibianpopulations in perspective: naturalfluctuationsand humanimpacts.Herpetologica50:65-84 149. PetrankaJW, Eldridge ME, Haley KE. 1993. Effects of timber harvesting on southernAppalachiansalamanders. Conserv. Biol. 7:363-70 150. Pierce BA, Wooten DK. 1992. Genetic variation in tolerance of amphibiansto low pH. J. Herpetol. 26:422-29 151. Portnoy JW. 1990. Breeding biology of

163

152.

153.

154.

155.

156.

157.

158.

159.

160.

161.

the spotted salamander Ambystomamaculatum(Shaw) in acidic temporary ponds at CapeCod, USA. Biol. Conserv.53:6175 Pounds JA, CrumpML. 1994. Amphibian declines and climate disturbance:the case of the golden toad and the harlequin frog. Conserv.Biol. 8:72-85 Pounds JA, Fogden MPL, Savage JM, Gorman GC. 1997. Tests of null models for amphibiandeclines on a tropical mountain.Conserv.Biol. 11:1307-22 Power T, ClarkKL, HarfenistA, Peakall DB. 1989. A review and evaluationof the amphibiantoxicological literature.Tech. Rep. 61, Can. WildlifeServ. PreestMR. 1993. Mechanismsof growth ratereductionin acid-exposedlarvalsalamanders,Ambystomamaculatum.Physiol. Zool. 66:686-707 Ramirez J, Vogt RC, Villareal-Benitez J-L. 1998. Populationbiology of a neotropicalfrog (Ranavaillanti).J. Herpetol. 32:338-44 Ramotnik CA. 1997. ConservationAssessment of the Sacramento Mountain Salamander. Gen. Tech.Rep. RM-GTR293. Fort Collins, CO: USDAFS, Rocky MountainFor.Range Exp. Stat. RaymondLR. 1991. Seasonal activity of Siren intermedia in northwesternLouisiana (Amphibia: Sirenidae).Southwestern Nat. 36:144-47 Raymond LR, Hardy LM. 1990. Demographyof a populationof Ambystoma talpoideum(Caudata:Ambystomatidae) in northwestern Louisiana.Herpetologica 46:371-82 Raymond LR, Hardy LM. 1991. Effects of a clearcuton a populationof the mole salamander,Ambystomatalpoideum, in an adjacentunalteredforest. J. Herpetol. 25:509-12 Reading CJ, Loman J, Madsen T. 1991. Breeding pond fidelity in the common toad, Bufo bufo. J. Zool. Lond. 225:20111

164

ALFORD * RICHARDS (Bufo of a natterjack toad metapopulation calamita). Oecologia 90:489-99 Sinsch U. 1996. Population dynamics of natterjacktoads (Bufo calamita) in the Rhineland: a nine-years study. Verh. DeutschenZool. Ges. 89:1-127 SjogrenP. 1991. Extinctionand isolation gradientin metapopulations:the case of the pool frog (Rana lessonae). Biol. J. Linn. Soc. 42:135-47 Sjogren Gulve P. 1994. Distributionand extinction patterns within a northern metapopulationof the pool frog, Rana lessonae. Ecology 75:1357-67 SredlMJ,Collins EP,HowlandJM. 1997. Mark-recapture of Arizonaleopardfrogs. InRanidFrog Conservation andManagement. Nongame and Endangered Wildl. Prog. Tech. Rept 121. Ariz Game Fish Dept, ed. MJ Sredl, pp. 1-20. StebbinsRC. 1954. Naturalhistoryof the salamandersof the Plethodontid genus Ensatina. Univ. Calif. Publ. in Zool. 54: 47-124 StrombergG. 1995. The yearly cycle of the jumping frog (Rana dalmatina) in Sweden. A 12 year study. Scientia Herpetologica 1995:185-86 Stumpel AHP. 1987. Distribution and presentnumbersof the tree frog Hyla arborea in Zealand Flanders, The Netherlands (Amphibia,Hylidae).Bijdragentot de Dierkunde57: 151-63 Taub FB. 1961. The distribution of the red-backedsalamander, Plethodon c. cinereus,withinthe soil. Ecology42:68198 TerhivuoJ. 1988. Phenologyof spawning of the common frog (Rana temporaria) in Finlandfrom 1846 to 1986. Ann. Zool. Fennici 25:165-75 Toft CA, RandAS, ClarkM. 1982. Population dynamicsand seasonalrecruitment in Bufo typhoniusand Colostethusnubicola (Anura).In The Ecology of a Tropical Forest: Seasonal Rhythmsand LongTerm Changes,ed. EG Leigh Jr,AS Rand,

162. Reed JM, Blaustein AR. 1995. Assessment of "nondeclining"amphibianpopulations using power analysis. Conserv. Biol. 9:1299-1300 163. RichardsSJ, McDonald KR, Alford RA. 1993. Declines in populations of Australia'sendemic tropicalrainforestfrogs. Pacific Conserv.Biol. 1:66-77 164. Sarkar S. 1996. Ecological theory and anuran declines. BioScience 46:199207 165. Schlupp I, Podloucky R. 1994. Changes in breeding site fidelity: a combined study of conservationand behaviour in the common toad Bufo bufo. Biol. Conserv. 69:285-91 166. Schwalbe CR, Rosen PC. 1988. Preliminaryreporton effect of bullfrogson wetland herpetofaunasin southeasternArizona. In Management of Amphibians, Reptiles and Small Mammals in North America,pp. 166-73. US Dep. Agric. 167. Semlitsch RD. 1998. Biological delinbufferzones for pondeationof terrestrial breedingsalamanders.Conserv.Biol. 12: 1113-19 168. SemlitschRD, Bodie JR. 1998. Are small, isolated wetlands expendable? Conserv. Biol. 12:1129-33 169. SemlitschRD, Scott DE, PechmannJHK, and dynamGibbonsJW. 1996. Structure ics of an amphibiancommunity.In LongCommunities, TermStudies of Vertebrate ed. ML Cody,JASmallwood,pp. 217-50. San Francisco:Academic Press 170. Shirose LJ, Brooks RJ. 1997. Fluctuations in abundanceand age structurein three species of frogs (Anura: Ranidae) in AlgonquinPark,Canada,from 1985 to 1993. See Ref. 87, pp. 16-26 171. Shoop CR. 1974. Yearlyvariationin larval survival of Ambystomamaculatum. Ecology 55:440-44 172. Sinsch U. 1991. Mini-review: the orientationbehaviour of amphibians. Herpetol. J. 1:541-44 173. Sinsch U. 1992. Structureand dynamics

174.

175.

176.

177.

178.

179.

180.

181.

182.

183.

GLOBALAMPHIBIANDECLINES DM Windsor,pp. 397-403. Washington, DC: SmithsonianInst. Press Travis J. 1994. Calibratingour expectations in studyingamphibianpopulations. Herpetologica 50:104-8 Turner FB. 1960. Populationstructure and dynamics of the western spotted frog, Rana p. pretiosa Baird & Girard,in Yellowstone Park,Wyoming.Ecol. Monogr. 30:251-78 Turner FB. 1962. The demography of frogs and toads. Q. Rev.Biol. 37:303-14 van de MortelT, ButtemerW, HoffmanP, Hays J, BlausteinA. 1998. A comparison of photolyase activity in three Australian tree frogs. Oecologia 115:366-69 VertucciFA, Corn PS. 1996. Evaluation of episodic acidification and amphibian declines in the Rocky Mountains. Ecol. Appl. 6:449-57 VialJL,SaylorL. 1993. TheStatusofAmphibian Populations: a Compilationand Analysis. IUCN/SSCDeclining Amphibian Populations Taskforce.Work.Doc. No. 1 Vitt LJ, Caldwell JP, WilburHM, Smith DC. 1990. Amphibiansas harbingersof decay.BioScience 40:418-18 Voris HK, Inger RF. 1995. Frog abundance along streamsin Bornean forests. Conserv.Biol. 9:679-83 WaldmanB, TocherM. 1998. Behavioral ecology, genetic diversity,and declining amphibian populations. In Behavioural Ecology and ConservationBiology, ed. T Caro,pp. 394-443. New York:Oxford Univ. Press WarnerRR, Chesson PL. 1985. Coexistence mediatedby recruitmentfluctuations: a field guide to the storage effect. Am. Nat. 125:769-87 A. 1991. BreedWaringer-Loschenkohl ing ecology of Rana dalmatina in lower Austria: a 7-years study. Alytes 9:12134 WassersugR. 1992. On assessing envi-

165

184.

196.

185.

197.

186. 187.

198.

199. 200.

188.

189.

201.

190.

202.

191.

203.

192.

204.

193.

205.

194.

206.

195.

ronmentalfactors affecting survivorship of premetamorphic amphibians.See Ref. 22, pp. 53-59 Weitzel NH, Panik HR. 1993. Longterm fluctuationsof an isolated population of the Pacificchorusfrog (Pseudacris regilla) in northwesternNevada. Great Basin Nat. 53:379-84 WelshHWJr.1990. Relictualamphibians and old-growthforests. Conserv.Biol. 4: 309-319 WernerJK. 1991. A radiotelemetryimplant techniquefor use with Bufo americanus. Herpetol.Rev. 22:94-95 Wilbur HM. 1980. Complex life cycles. Annu.Rev.Ecol. Syst. 11:67-93 WinterJE, Alford RA. 1999. Terrestrial VertebrateFauna in Regenerated Mine Areas at Weipa.Rep. to Commonwealth AluminiumCo. 150 pp. Wissinger SA, Whiteman HH. 1992. Fluctuationin a Rocky Mountainpopulation of salamanders:anthropogenic acidificationor natural variation?J. Herpetol. 26:377-91 LL. 1991. The impactof HurWoolbright ricane Hugo on forest frogs in Puerto Rico. Biotropica23:462-67 WoolbrightLL. 1996. Disturbanceinfluin the ences long-termpopulationpatterns PuertoRican frog, Eleutherodactylus coqui (Anura:Leptodactylidae). Biotropica 28:493-501 Wyman RL, Hawksley-Lescault DS. be1987. Soil acidityaffects distribution, havior and physiology of the salamander Plethodon cinereus. Ecology 68:181927 WymanRL, JancolaJ. 1992. Degree and scale of terrestrialacidificationand amphibiancommunitystructure. J. Herpetol. 26:392-401 Zug GR, Zug PB. 1979. The marinetoad, Bufo marinus: a naturalhistory resume of nativepopulations.Smithson.Contrib. Zool. 284:1-58