Biol.Rev.(2017),pp.000–000. 1 doi:10.1111/brv.12343 An Amazonian rainforest and its fragments as a laboratory of global change William F. Laurance1,2∗, Jose´ L. C. Camargo2, Philip M. Fearnside3, Thomas E. Lovejoy2,4, G. Bruce Williamson2,5, Rita C. G. Mesquita2,3, Christoph F. J. Meyer2,6,7, Paulo E. D. Bobrowiec8 and Susan G. W. Laurance1,2 1CentreforTropicalEnvironmentalandSustainabilityScience(TESS)andCollegeofScienceandEngineering,JamesCookUniversity,Cairns 4878,Australia 2BiologicalDynamicsofForestFragmentsProject,NationalInstituteforAmazonianResearch(INPA)andSmithsonianTropicalResearch Institute,Manaus69067-375,Brazil 3DepartmentofEnvironmentalDynamics,NationalInstituteforAmazonianResearch(INPA),Manaus69067-375,Brazil 4DepartmentofEnvironmentalScienceandPolicy,GeorgeMasonUniversity,Fairfax,VA22030,U.S.A. 5DepartmentofBiologicalSciences,LouisianaStateUniversity,BatonRouge,LA70803,U.S.A. 6CentreforEcology,EvolutionandEnvironmentalChanges,UniversityofLisbon,1749-016,Lisbon,Portugal 7SchoolofEnvironmentandLifeSciences,UniversityofSalford,SalfordM54WT,U.K. 8BiodiversityCoordination,NationalInstituteforAmazonianResearch(INPA),Manaus69067-375,Brazil ABSTRACT Wesynthesizefindingsfromoneoftheworld’slargestandlongest-runningexperimentalinvestigations,theBiological DynamicsofForestFragmentsProject(BDFFP).Spanninganareaof∼1000km2incentralAmazonia,theBDFFPwas initiallydesignedtoevaluatetheeffectsoffragmentareaonrainforestbiodiversityandecologicalprocesses.However, overits38-yearhistorytodatetheprojecthasfartranscendeditsoriginalmission,andnowfocusesmorebroadlyon landscapedynamics,forestregeneration,regional-andglobal-changephenomena,andtheirpotentialinteractionsand implications for Amazonian forest conservation. The project has yielded a wealth of insights into the ecological and environmental changes in fragmented forests. For instance, many rainforest species are naturally rare and hence are either missing entirely from many fragments or so sparsely represented as to have little chance of long-term survival. Additionally, edge effects are a prominent driver of fragment dynamics, strongly affecting forest microclimate, tree mortality,carbonstorageandadiversityoffauna. Evenwithinourcontrolledstudyarea,thelandscapehasbeenhighlydynamic:forexample,thematrixofvegetation surrounding fragments has changed markedly over time, succeeding from large cattle pastures or forest clearcuts to secondary regrowth forest. This, in turn, has influenced the dynamics of plant and animal communities and their trajectories of change over time. In general, fauna and flora have responded differently to fragmentation: the most locally extinction-prone animal species are those that have both large area requirements and low tolerance of the modified habitats surrounding fragments, whereas the most vulnerable plants are those that respond poorly to edge effects or chronic forest disturbances, and that rely on vulnerable animals for seed dispersal or pollination. Relative to intact forests, most fragments are hyperdynamic, with unstable or fluctuating populations of species in response to a variety of external vicissitudes. Rare weather events such as droughts, windstorms and floods have had strong impacts on fragments and left lasting legacies of change. Both forest fragments and the intact forests in our study area appear to be influenced by larger-scale environmental drivers operating at regional or global scales. These drivers are apparently increasing forest productivity and have led to concerted, widespread increases in forest dynamics and plant growth, shifts in tree-community composition, and increases in liana (woody vine) abundance. * Addressforcorrespondence(Tel:+61740381518;E-mail:[email protected]) BiologicalReviews(2017)000–000©2017CambridgePhilosophicalSociety 2 WilliamF.Lauranceandothers Suchlarge-scaledriversarelikelytointeractsynergisticallywithhabitatfragmentation,exacerbatingitseffectsforsome species and ecological phenomena. Hence, the impacts of fragmentation on Amazonian biodiversity and ecosystem processesappeartobeaconsequencenotonlyoflocalsitefeaturesbutalsoofbroaderchangesoccurringatlandscape, regionalandevenglobalscales. Key words: Amazonia, biodiversity, carbon storage, climate change, drought, ecosystem services, edge effects, environmentalsynergisms,habitatfragmentation,naturereserves. CONTENTS I. Introduction .............................................................................................. 3 II. Larger-scaledrivers ...................................................................................... 3 (1) Landscape-scalephenomena ......................................................................... 3 (2) Regional-scalephenomena ........................................................................... 4 (3) Global-scalephenomena ............................................................................. 4 III. Studyareaandkeydatasets .............................................................................. 5 (1) Studyarea ............................................................................................ 5 (2) Uniquedatasets ...................................................................................... 6 IV. Changesinintactforests .................................................................................. 6 (1) Unexpectedtrends ................................................................................... 6 (2) Potentialenvironmentaldrivers ...................................................................... 6 V. Consequencesoffragmentsize ........................................................................... 7 (1) Sampleeffects ........................................................................................ 8 (2) Areaeffects ........................................................................................... 8 VI. Edgeeffects ............................................................................................... 8 (1) Foresthydrology ...................................................................................... 8 (2) Strikingdiversityofedgeeffects ...................................................................... 9 (3) Impactsofmultipleedges ............................................................................. 9 (4) Effectsofedgeageandadjoiningvegetation .......................................................... 9 VII. Forestisolationandthematrix ........................................................................... 9 (1) Matrixstructureandcomposition .................................................................... 10 (2) Factorsinfluencingthematrix ........................................................................ 10 (3) Narrowforestclearings ............................................................................... 11 VIII. Dynamicsofforestfragments ............................................................................. 11 (1) Raredisturbances .................................................................................... 12 (2) Hyperdynamism ...................................................................................... 12 (3) Divergingtrajectoriesoffragments ................................................................... 13 (4) Ecologicaldistortions ................................................................................. 13 (5) Forestcarbondynamics .............................................................................. 14 IX. Speciesresponsestofragmentation ....................................................................... 14 (1) Non-randomextinctions .............................................................................. 14 (2) Non-neutralextinctions .............................................................................. 14 (3) Keycorrelatesofanimalvulnerability ................................................................ 15 (4) Keycorrelatesofplantvulnerability .................................................................. 16 X. Horizonsfornewresearch ............................................................................... 16 XI. Generallessons ........................................................................................... 17 (1) Valuesoflong-termresearch ......................................................................... 17 (2) Trainingisvital ....................................................................................... 17 XII. Lessonsforconservation .................................................................................. 17 (1) TheBDFFPisabest-casescenario ................................................................... 17 (2) Reservesshouldbelargeandnumerous .............................................................. 18 (3) Nofragmentisunimportant .......................................................................... 18 (4) Woundedlandscapescanrecover .................................................................... 18 XIII. Fragmentationandlarger-scaledrivers ................................................................... 18 (1) Interactingdrivers .................................................................................... 18 (2) TheAmazonandclimatechange .................................................................... 19 XIV. Conclusions .............................................................................................. 20 XV. Acknowledgements ....................................................................................... 20 XVI. References ................................................................................................ 20 BiologicalReviews(2017)000–000©2017CambridgePhilosophicalSociety Amazonianfragmentsandglobalchange 3 I. INTRODUCTION The Biological Dynamics of Forest Fragments Project (BDFFP) is the world’s largest and longest-running experimental study of habitat fragmentation (Lovejoy etal., 1986; Bierregaard etal., 1992; Laurance etal., 2002, 2011). LocatedincentralAmazonia(Fig.1),theBDFFPhasevolved since its inception in 1979 into an epicentre for long-term research. Beyond this, its research mission has gradually broadenedtoincludenotonlyforestfragmentationbutalso studies of forest regeneration, landscape dynamics, climatic Fig. 1. Map of the Biological Dynamics of Forest Fragments variation, regional- and global-change phenomena and a ProjectincentralAmazonia. varietyofinterdisciplinaryresearchtopics. The BDFFP is strategically located at the heart of the Amazon, the world’s largest tropical forest. The Amazon being altered by large-scale agriculture (Fearnside, 2001; itselfliesattheintersectionofkeyquestionsinglobalchange, Gibbs etal., 2010), industrial logging (Asner etal., 2005), both for research and for action. It is believed to be one of proliferating roads (Laurance etal., 2001a; Fearnside, 2002, the major regions that will be most impacted by projected 2007; Killeen, 2007), increasing biofuel production (Butler climaticchange(Salazar,Nobre&Oyama,2007;Dai,2012; & Laurance, 2009), hydroelectric dams (Fearnside, 2016b) IPCC, 2013; Nobre etal., 2016). If effectively conserved andoil,gasandminingdevelopments(Fineretal.,2008). and managed, the Amazon has the potential to contribute LargeexpansesoftheAmazonhavealreadybeencleared, markedlytoeffortstolimitclimatechangeduringthenarrow resultinginconsiderablefragmentation.Bytheearly1990s, window of time we have remaining to avert ‘dangerous’ the area of forest that was fragmented (<100km2) or global warming (Fearnside, 2000, 2016a; Houghton, Byers vulnerable to edge effects (<1km from edge) was over & Nassikas, 2015). Because of its enormous carbon-storage 150% greater than the area that had been deforested capacity,itisalsooneoftheplacesonEarthwheresharply (Skole & Tucker, 1993). From 1999 to 2002, deforestation reducing greenhouse-gas emissions could be achieved by and industrial selective logging in Brazilian Amazonia, limitingforestlossanddegradation,therebydeliveringgreat respectively,created∼32000and∼38000kmofnewforest globalbenefitsforhumankind(Stickleretal.,2009). edgeannually(Broadbentetal.,2008).Prevailinglandusesin Today, the BDFFP is one of the most enduring, Amazonia,suchascattleranchingandsmall-scalefarming, influential and highly cited environmental investigations typicallyproducelandscapesdominatedbysmall(<400ha) in the world (Gardner etal., 2009; Peres etal., 2010; and irregularly shaped forest fragments (Fig. 2) (Cochrane Pitman etal., 2011). Its wide-ranging research has involved & Laurance, 2002; Broadbent etal., 2008). Such fragments hundreds of Brazilian and international investigators and areespeciallyvulnerabletoawidearrayofedgeeffectsand thousandsofstudentsandothertrainees.Herewesynthesize otherexternalvicissitudes(Bierregaardetal.,1992;Laurance the contributions of this singular project to the study of etal.,2002,2011). habitat fragmentation, including its broader consequences Changesinforestcovercanhaveimportanteffectsonlocal for Amazonian ecosystems and biota. We emphasize that climateandvegetation.Habitatfragmentationcanpromote many of the local impacts of fragmentation in the Amazon forest desiccation via phenomena such as the ‘vegetation arebeingmodifiedorexacerbatedbyenvironmentalchanges breeze’ (Fig. 3). This occurs because fragmentation leads occurring at wider landscape, regional and even global to the juxtaposition of cleared and forested lands, which scales.Weassertthattheeffectsoffragmentationcannotbe differ greatly in their physical characteristics. Air above fully understood without considering the influence of these forests is cooled by evaporation and especially plant larger-scalephenomena. evapotranspiration,butsuchcoolingisgreatlyreducedabove clearings(Avissar&Schmidt,1998).Asaresult,theairabove clearingsheatsupandrises,reducinglocalairpressureand drawing moist air from the surrounding forests into the II. LARGER-SCALEDRIVERS clearing. As the rising air cools, its moisture condenses into convectivecloudsthatcanproducerainfallovertheclearing (1) Landscape-scalephenomena (Avissar&Liu,1996).Theairisthenrecycledascool,dryair Thecorrelatedprocessesofforestlossandfragmentationare backovertheforest.Inthisway,clearingsofafewhundred among the greatest threats to tropical biodiversity (Lovejoy hectaresormorecandrawmoistureawayfromnearbyforests etal., 1986; Ewers & Didham, 2006; Laurance & Peres, (W.F.Laurance,2004b;Cochrane&Laurance,2008;Nobre 2006; Gibson etal., 2011). Amazonia harbours more than etal., 2016). In eastern Amazonia, satellite observations halfoftheworld’ssurvivingtropicalforest,andiscurrently of canopy water content suggest such desiccating effects BiologicalReviews(2017)000–000©2017CambridgePhilosophicalSociety 4 WilliamF.Lauranceandothers greaterwindspeeds(fromreducedsurfacedrag),whichcould affectmoistureconvergenceandcirculation(Walker,Sud& Atlas,1995;Sud,Yang&Walker,1996).Itisevenpossible that moderate forest loss and fragmentation could increase net regional precipitation in the near term, as a result of increasingconvectionalstormsdrivenbyvegetationbreezes, although the main effect would be to remove moisture fromforestsandredistributeitoveradjoiningclearings.The greatestconcernisthatifdeforestationreachessomecritical threshold,Amazonianrainfallmightdeclineabruptlyasthe regional hydrological system collapses (Avissar etal., 2002; Nobreetal.,2016). Massive smoke plumes produced by forest and pasture fires cause two additional effects of forest loss. Smoke hypersaturates the atmosphere with cloud condensation nuclei (microscopic particles in aerosol form) that bind with airborne water molecules and thereby inhibit the formation of raindrops (Rosenfeld, 1999). In addition, by absorbing solar radiation, smoke plumes warm the atmosphere,inhibitingcloudformation.Asaresultofthese two effects, large fires can create rain shadows that extend Fig. 2. Habitat fragmentation in eastern Amazonia caused for hundreds or even thousands of kilometers downwind by a (A) forest-colonization project (Tailaˆndia) and (B) cattle (Freitas,SilvaDias&SilvaDias,2000).Thiscanbeaserious ranching(Paragominas).Forestsareblackandclearedareasare threat to forests because tropical fires are lit during the grey. Each image shows an area of about 600km2 (adapted criticaldry-seasonmonths,whenplantsarealreadymoisture fromCochrane&Laurance,2002). stressedandmostvulnerabletofire. can penetrate 1.0–2.7km into fragmented forests (Briant, (3) Global-scalephenomena Gond & Laurance, 2010). This moisture-robbing function of clearings, in concert with frequent burning in adjoining Howwillglobal-changedriversaffecttheAmazon?Although pastures, could help to explain why fragmented forests are model predictions for future climates in Amazonia vary sovulnerabletodestructive,edge-relatedfires(Cochrane& considerably, it is generally expected that parts of the Laurance,2002,2008;Barlowetal.,2006). basin will become hotter and drier under projected global warming(IPCC,2013;Nobreetal.,2016).Whatthisportends for the Amazon is a matter of some controversy. Earlier (2) Regional-scalephenomena studies assuming CO concentrations about twice those in 2 Extensiveforestclearingreducestherateofevapotranspira- the pre-industrial atmosphere, notably by the UK Hadley tionbecausepasturegrassesandcroplandshavefarlessleaf Centre, projected disastrous forest die-offs (Cox etal., 2000, areaandshallowerrootsthandorainforests(Jippetal.,1998). 2004). However, this conclusion has been countered by Atregionalscales,decliningevapotranspirationcouldreduce new models from the same research group, suggesting the rainfallandcloudcoverandincreasealbedoandsoil-surface Amazonforestwillremainalmostentirelyintactatuptofour temperatures. Moisture recycling via evapotranspiration is timespre-industrialCO levels(Coxetal.,2013;Goodetal., 2 exceptionally important in the hydrological regime of the 2013; Huntingford etal., 2013). The main difference is that Amazon (Salati & Vose, 1984; Eltahir & Bras, 1994), espe- thenewermodelsincludeCO -fertilizationeffects(Kimball 2 ciallyduringthedryseason(Malhietal.,2008),becausethe etal.,1993),whichareassumedtoincreaseplantgrowthand forestisbothvastandfarfromthenearestocean. water-useefficiency.Thisisbecausethehigheratmospheric However, the regional consequences of large-scale CO concentrationshouldallowplantstoconservewaterby 2 deforestationarefarfromfullyunderstood.Somemodelling decreasing the duration of stomatal-opening periods while studies suggest that Amazonian deforestation could reduce stilltakinginadequateCO forphotosynthesis. 2 basin-wide precipitation by roughly 20–30%, but these Other global-change phenomena, such as extreme estimates rely on a simplistic assumption of complete, climatic events, could also potentially have important uniformforestclearing(e.g.Nobre,Sellers&Shukla,1991; impacts.Forinstance,droughtsintheAmazonarenormally Dickinson & Kennedy, 1992; Lean & Rowntree, 1993). associated with El Nin˜o events and are strongest in the Model results based on actual (circa 1988) deforestation southern, eastern and north-central Amazon – areas of the patterns in Brazilian Amazonia have been less dramatic, basin that already experience pronounced dry seasons. with deforested regions predicted to experience modest However, severe droughts in 2005 and 2010 arose from (6–8%) declines in rainfall, moderate (18–33%) reductions a completely different cause – exceptionally high Atlantic in evapotranspiration, higher soil-surface temperatures and sea-surface temperatures, which caused the rain-bearing BiologicalReviews(2017)000–000©2017CambridgePhilosophicalSociety Amazonianfragmentsandglobalchange 5 Fig.3. Thevegetation-breezephenomenon,whichcanpromoteforestdesiccationinthegeneralvicinityofpasturesandclearings (fromCochrane&Laurance,2008). inter-tropical convergence zone to shift northward (Lewis The study area includes three large cattle ranches etal.,2011).Theresultingdroughtsaffectednotjustthedrier, (∼5000haeach)containing11forestfragments(fiveof1ha, more seasonal parts of the basin but also its wettest areas fourof10haandtwoof100ha),andlargeexpansesofnearby in central and western Amazonia. Because plant species in continuousforestthatserveasexperimentalcontrols(Fig.1). thesewetareasareadaptedtoperenniallyhumidconditions, Intheearly1980s,thefragmentswereisolatedfromnearby thenewdroughtscausedmassiveplantmortality,killingtens intactforestbydistancesof80–650mthroughclearingand of millions of trees while releasing several billion tonnes of burning of the surrounding forest. A key advantage was atmospheric carbon emissions (Lewis etal., 2011; Marengo that pre-fragmentation censuses were conducted for many etal.,2012).Withmountingevidencethatclimaticextremes animalandplantgroups(e.g.trees,understoreybirds,small couldbecomemorefrequentandintenseinawarmingworld mammals,primates,frogs,manyinvertebratetaxa),thereby (Veraetal.,2006;Herringetal.,2015;Jime´nez-Mun˜ozetal., allowing long-term changes in these groups to be assessed 2016),couldtheAmazonbedrivenintoanewkindofclimatic far more confidently than in most other fragmentation dynamic – oneforwhichitsecosystemsandbiodiversityare studies. poorlyadapted? Because of poor soils and low productivity, the ranches surroundingtheBDFFPfragmentswerelargelyabandoned, especially after government fiscal incentives dried up from III. STUDYAREAANDKEYDATASETS 1988 onwards. Secondary forests – initially dominated by Vismia spp. in areas that were cleared and burned, (1) Studyarea and by Cecropia spp. in areas that were cleared without TheexperimentallandscapeoftheBDFFPspans∼1000km2 fire – proliferatedinmanyformerlyforestedareas(Mesquita in area and is located 80km north of Manaus, Brazil. etal.,2001).Someregeneratingareasinitiallydominatedby The topography is relatively flat (80–160m elevation) Cecropia later grew into structurally well-developed (>20m but dissected by numerous stream gullies. The heavily tall), species-rich secondary forests (Longworth etal., 2014). weathered, nutrient-poor soils of the study area are typical Vismia-dominated regrowth, however, which is relatively oflargeexpansesoftheAmazonBasin.Rainfallrangesfrom species poor, is maturing far more slowly (Norden etal., 1900 to 3500mm annually with a moderately strong dry 2011;Williamsonetal.,2014). seasonfromJunetoOctober.Theforestcanopyis30–37m Tohelpmaintainisolationoftheexperimentalfragments, tall, with emergent trees to 55m. Species richness of trees 100m-wide strips of regrowth were cleared and burned [≥10cm diameter at breast height (dbh)] often exceeds around each fragment on 4–5 occasions, most recently in 280speciesha−1, which is among the highest known tree 2013–2014.However,humandisturbancesthataffectmany diversityintheworld(DeOliveira&Mori,1999;Laurance fragmented landscapes in the Amazon, such as major fires, etal.,2010b).Comparablyhighlevelsofdiversityareseenin logging and hunting (Michalski & Peres, 2005), are largely manyotherplantandanimaltaxa. preventedattheBDFFP. BiologicalReviews(2017)000–000©2017CambridgePhilosophicalSociety 6 WilliamF.Lauranceandothers (2) Uniquedatasets controls (Lovejoy etal., 1986; Bierregaard etal., 1992). The firstisthatstandardizedcensusesofmanyplantandanimal The BDFFP sustains some of the longest-running and taxawereconductedineachexperimentalfragmentbeforeit highest-quality environmental data sets in the Amazon. wasisolatedfromthesurroundingforest.Thesecondisthat This includes a network of 69 1-ha forest-dynamics plots dozens of ‘control’ sites in nearby intact forests have been arrayed across intact and fragmented forests in the study monitoredforupto38years,toassessthetemporaldynamics area, which has been monitored since the early 1980s, and of these sites. The intact-forest sites were expected to vary apermanent25-haplotinintactforestestablishedin2005. randomly over time or respond to occasional vicissitudes Theseplotshavemadeimportantcontributionstoreducing suchasdroughts,butnottochangeovertimeinadirectional uncertaintiesinbiomassandcarbon-storageestimatesforthe manner. Amazon (e.g. Phillips etal., 1998; Nascimento & Laurance, Amajorsurprise,however,isthattheBDFFPcontrolshave 2002; Baker etal., 2004). For example, in comparison to changed in several concerted ways (Laurance etal., 2014a). the 3000 1-ha plots surveyed by the RADAMBRASIL Beforeinterpretinghowfragmentationhasalteredecological Project (Nogueira etal., 2008, 2015), the BDFFP plots communities in the BDFFP, it is first important to identify include data on nearly all other forest components such howtheintact-forestsiteshavechanged,asthesewidespread assmaller(1–30cmdiameter)trees,palms,lianas,strangler effects are presumably altering the forest fragments as well. figs, understorey vegetation and dead biomass (Nascimento Thelong-termmonitoringoftensofthousandsoftreesand & Laurance, 2002, 2004). These data allow assessment populations of many other plant and animal groups has of spatial variability in aboveground biomass with a high allowed researchers to identify synchronous changes in the degreeofconfidence.Forexample,theabovegroundbiomass undisturbedforestsattheintactsites,andtoattempttoinfer of trees varies considerably among the 69 1-ha plots in the BDFFP landscape (mean±S.D.=356±47Mgha−1; theirenvironmentalcauses. How have the intact forests changed? Over the past Laurance etal., 1999). This high variability demonstrates 2–3decades, we have found that (i) forest dynamics (tree aneedformanyplotsthatarespatiallystratified,ratherthan mortality and recruitment) have accelerated significantly onlyafewplotsof1haorsmallerscatteredirregularlyaround overtime(W.F.Lauranceetal.,2004b,2014a;S.G.Laurance the Amazon, for calibrating satellite imagery for biomass etal., 2009); (ii) tree-community composition has shifted, mapping,andforestimatinggreenhouse-gasemissionsfrom generallyinfavouroffaster-growingcanopytreesandagainst ongoingdeforestation(seeFearnside,2016a). shade-tolerantsubcanopytrees(W.F.Lauranceetal.,2004b, Floristic data from the BDFFP are exceptional for 2005);(iii)growthrateshaveincreasedforthelargemajority their high quality of species identification, allowing better (84%)oftreegenerainourstudyarea(Fig.4)(W.F.Laurance matching with plant functional and phylogenetic traits etal., 2004b); (iv) aboveground tree biomass has increased such as wood density and tree form (e.g. Fearnside, 1997; significantly over time (although tree-stem numbers have Nogueira, Nelson & Fearnside, 2005; Chave etal., 2006; not changed significantly; S.G. Laurance etal., 2009); and Nogueira etal., 2007; Souza etal., 2016). Given their broad (v) lianas have increased markedly in abundance (Fig. 5) spatial extent and temporal depth, these data have also (Lauranceetal.,2014a,b). contributedtoknowledgeofthediversityofAmazonianplant speciesandtheirrelationshipstosoiltextureandchemistry, (2) Potentialenvironmentaldrivers topography, forest dynamics and climatic variables at both landscape and regional scales (e.g. Bohlman etal., Why are the intact forests changing? The causes of such 2008; S.G. Laurance etal., 2009; Laurance, Andrade & changes are incompletely understood (Lewis, Malhi & Laurance, 2010a; Laurance etal., 2010b; ter Steege etal., Phillips, 2004a; Lewis etal., 2009b) and often controversial 2013). Biodiversity and ecosystem processes represent part (Clark, 2004; Fearnside, 2004). Nonetheless, the trends we of what is lost when the forest is destroyed or degraded. detected appear broadly consistent with those observed Understanding these processes is essential for assessing not elsewhere in many Amazonian (Phillips & Gentry, 1994; only the vulnerability of forests, but also their potential Phillipsetal.,1998,2002;Bakeretal.,2004;Lewisetal.,2004b; resilience in the face of global change and their rates of Schnitzer&Bongers,2011)andAfrican(Lewisetal.,2009b) recovery following various perturbations (Williamson etal., tropical forests. These trends are consistent with ecological 2014; Souza etal., 2016). Data sets for a number of faunal patterns expected from rising forest productivity, including groups, such as birds, amphibians, primates and major faster plant growth, increasing forest biomass, intensifying invertebratetaxa,areofcomparablequalityandduration. competitionleadingtogreaterplantmortalityandturnover, and increasing abundances of plant species that can attain highgrowthratesorareadvantagedindynamicforests(W.F. Lauranceetal.,2004b;Lewisetal.,2004b,2009b). IV. CHANGESININTACTFORESTS Themostfrequentlyinvokeddriverofrisingtropicalforest productivity is CO fertilization (e.g. Lewis etal., 2004a, (1) Unexpectedtrends 2 2009a). This is because many plants show faster growth Aspartofitsoriginalmissiontoassesslong-termchangesin under enriched CO (Oberbauer, Strain & Fletcher, 1985; 2 fragmentedforests,theBDFFPhastwotypesofexperimental Granados & Ko¨rner, 2002; Ko¨rner, 2004) and because BiologicalReviews(2017)000–000©2017CambridgePhilosophicalSociety Amazonianfragmentsandglobalchange 7 Fig.5. Increaseinabundanceoflianasinintact-forestplotsof Fig. 4. Rates of tree growth in intact forests of the theBiologicalDynamicsofForestFragmentsProject(BDFFP) BiologicalDynamicsofForestFragmentsProject(BDFFP)have (fromLauranceetal.,2014b).Thesolidlineshowsy=xwhereas acceleratedovertimeforthelargemajority(84%)oftreegenera thedottedlineisalinearregressionfittedtothedata. (fromW.F.Lauranceetal.,2004b).Datashownaremeanrates oftrunk-diametergrowthforgenerathatincreasedordecreased The notable increases in liana abundance in our intact significantlyinabundanceovertimeintheplots,aswellasthose forests (Laurance etal., 2014b) might arise because lianas thatshowednosignificanttrend.Interval1is1984–1991,and interval2is1992–1999. appear to exploit rising CO2 concentrations and drier conditions more effectively than do trees (Condon, Sasek &Strain,1992;Granados&Ko¨rner,2002;butseeMarvin atmospheric CO levels have risen rapidly, especially in 2 etal.,2015).Treeswithheavylianainfestationsareknownto recent decades. This view is supported by compelling exhibitelevatedmortalityandreducedgrowth(Ingwelletal., evidenceofalargecarbonsinkinthebiosphere(Ballantyne 2010).Notably,inourstudyarea,lianaabundanceisstrongly etal.,2013),asubstantialpartofwhichappearstobeonland and negatively correlated with live tree biomass (Fig. 6) (Sarmientoetal.,2010)andinthetropics(Lewisetal.,2009a; (Laurance etal., 2001b). Liana increases over time have Huntingfordetal.,2013). alsobeenobservedintropicalforestsinwesternAmazonia, Other explanations for the rising productivity, however, the Guianas, Central America and elsewhere (Schnitzer & are not implausible. For instance, droughts can influence Bongers, 2011), with rising atmospheric CO and possibly forestdynamicsandcompositionandappeartobeincreasing 2 increasing drought being the most frequent explanations in parts of the Amazon (Lewis etal., 2009a; Marengo etal., (see Laurance etal., 2014b and references therein). This 2011; Chou etal., 2013; Fu etal., 2013). The increase in potentially negative effect of CO enrichment on forest forest dynamics we observed in intact forests appears to be 2 biomass via increasing liana infestations is not included in driven primarily by rising tree mortality, with recruitment thelatestHadleyCentremodels(Coxetal.,2013;Goodetal., and growth often lagging behind periods of high mortality. 2013;Huntingfordetal.,2013),andcouldcanceloutsomeof Thesemortalitypulsesarepositivelyassociatedwithseveral thecarbon-storagebenefitssuggestedforahigh-CO future factors, including El Nin˜o droughts and increasing rainfall 2 (Ko¨rner,2004,2017). seasonality(S.G.Lauranceetal.,2009). Hence,forwhateverthereasonorreasons,itisapparent Additionally, multi-decadal shifts in solar radiation or that the intact forests in our study area are changing in cloudiness could potentially increase forest productivity, a variety of ways. Such changes are likely to interact although evidence for such shifts in the tropics is limited with, and potentially complicate or amplify, the impacts (Lewisetal.,2009a).Recoveryfrompastdisturbancehasalso offragmentationontropicalforestcommunities. beenhypothesizedtounderliechangesatsometropicalforest sites, but there is no evidence of widespread disturbance in our study area (W.F. Laurance etal., 2004b, 2005) aside from charcoal fragments that are at least four centuries old V. CONSEQUENCESOFFRAGMENTSIZE (Bassini & Becker, 1990; Fearnside & Leal Filho, 2001), possibly indicating major fires during past mega-El Nin˜o TheBDFFP’soriginalmissionfocusesonassessingtheeffects events(Meggers,1994). of fragment area on Amazonian forests and fauna, and on BiologicalReviews(2017)000–000©2017CambridgePhilosophicalSociety 8 WilliamF.Lauranceandothers even with constant sampling effort across all fragments. Such declines are evident in leaf bryophytes (Zartman, 2003), tree seedlings (Benítez-Malvido & Martinez-Ramos, 2003b), palms (Scariot, 1999), understorey insectivorous birds (Stratford & Stouffer, 1999; Ferraz etal., 2007), bats (Sampaio, 2000; Rocha etal., 2017), primates (Gilbert & Setz, 2001; Boyle & Smith, 2010b) and larger herbivorous mammals (Timo, 2003), among others. For such groups, smaller fragments (<100ha) are often unable to support viable populations. A few groups, such as ant-defended plants and their ant mutualists, show no significant decline in diversity with fragment area (Bruna etal., 2005). Fragment size also influences the rate of species losses, withsmallerfragmentslosingspeciesmorequickly(Lovejoy etal., 1986; Stouffer, Strong & Naka, 2008). Assuming that the surrounding matrix is hostile to bird movements and precludes colonization, Ferraz etal. (2003) estimated that a 1000-fold increase in fragment area would be needed to Fig.6. Negativeassociationbetweenlianaabundanceandthe slow the rate of local species extinctions by 10-fold. Even a aboveground biomass of live trees in Biological Dynamics of fragment of 10000 ha in area would be expected to lose a ForestFragmentsProject(BDFFP)forest-dynamicsplots(from substantialpartofitsbirdfaunawithinonecentury(Ferraz Lauranceetal.,2001b). etal., 2003). Similarly, long-term mark–recapture studies suggestthatverylargefragmentswillbeneededtomaintain keyecologicalandecosystemprocesses.Herewesummarize fully intact assemblages of certain faunal groups, such as major findings and conservation lessons that have been ant-following birds, which forage over large areas of forest gleanedtodate. (VanHoutanetal.,2007). (1) Sampleeffects Many species in Amazonian forests are rare or patchily VI. EDGEEFFECTS distributed. This phenomenon is especially pronounced in the large expanses of the basin that overlay heavily AnimportantinsightfromtheBDFFPistheextenttowhich weathered, nutrient-poor soils (e.g. Radtke, da Fonseca & edgeeffects – physicalandbioticchangesassociatedwiththe Williamson, 2008). In these areas resources such as fruits, abrupt,artificialmarginsofhabitatfragments–influencethe flowersandnectararetypicallyscarceandplantsareheavily dynamicsandcompositionofplantandanimalcommunities. defendedagainstherbivoreattack(Laurance,2001). Herewesummarizekeyfindingsfromthiswork. Herein lies a key implication for understanding forest fragmentation: given their rarity, many species may be (1) Foresthydrology absent from fragments not because their populations have vanished, but because they were simply not present at The hydrological regimes of fragmented landscapes differ the time of fragment creation – a phenomenon termed the markedly from those of intact forest (Kapos, 1989; Kapos ‘sampleeffect’(Wilcox&Murphy,1985).Suchsampleeffects etal., 1993). Pastures or crops surrounding fragments are the hypothesized explanation for the absence of many have much lower rates of evapotranspiration than do rare understorey bird species from fragments (Ferraz etal., forests, causing such areas to be hotter and drier than 2007).Inaddition,manybeetles(Didhametal.,1998a),bats forests (Camargo & Kapos, 1995). Field observations and (Sampaioetal.,2003;Farnedaetal.,2015;Meyeretal.,2015; heat-flux simulations suggest that desiccating conditions Rochaetal.,2017),ant-defendedplants(Bruna,Vasconcelos can penetrate up to 100–200m into fragments from &Heredia,2005)andtrees(Bohlmanetal.,2008;Laurance adjoining clearings (Malcolm, 1998; Didham & Lawton, etal.,2010b)attheBDFFPexhibithighlevelsofrarity,habitat 1999).Further,streamsinfragmentedlandscapesexperience specializationorpatchiness. greater temporal variation in flow rate than do those in forests, because clearings surrounding fragments have less evapotranspirationandrainfallinterceptionandabsorption (2) Areaeffects by vegetation (Trancoso, 2008). Rapid runoff promotes Understandingfragment-areaeffectshaslongbeenacentral localized flooding in the wet season and stream failure goal of the BDFFP (Lovejoy & Oren, 1981; Lovejoy in the dry season, with potentially important impacts etal., 1984, 1986; Pimm, 1998). The species richness of on aquatic invertebrates (Nessimian etal., 2008) and fish many organisms declines with decreasing fragment area, assemblages. BiologicalReviews(2017)000–000©2017CambridgePhilosophicalSociety Amazonianfragmentsandglobalchange 9 (2) Strikingdiversityofedgeeffects 1994), tree mortality (Laurance etal., 2006a), abundance and species richness of tree seedlings (Benítez-Malvido, Atleastoverthefirst3–4decadesafterisolation,edgeeffects 1998; Benítez-Malvido & Martinez-Ramos, 2003b), liana have been among the most important drivers of ecological abundance (Laurance etal., 2001b) and the density and change in the BDFFP fragments. The distance to which diversityofdisturbance-lovingpioneertrees(Lauranceetal., differentedgeeffectspenetrateintofragmentsvarieswidely, 2006a,b,2007).Theadditiveeffectsofnearbyedgesprobably ranging from 10 to 300m at the BDFFP (Laurance etal., help to explain why small (<10ha) or irregularly shaped 2002)andconsiderablyfurther(atleast2–3km)inareasof forest remnants are often so severely altered by forest theAmazonwhereedge-relatedfiresarecommon(Cochrane fragmentation(Zartman,2003;Lauranceetal.,2006a).Some &Laurance,2002,2008;Briantetal.,2010). fauna are likewise sensitive to multiple edges. For instance, Edge phenomena are remarkably diverse (Fig. 7). They the number of nearby forest edges was found to be an include increased desiccation stress, wind shear and wind important predictor of local bat abundance (Rocha etal., turbulence that sharply elevate rates of tree mortality and 2017). damage (Laurance etal., 1997, 1998a). These in turn cause wide-ranging alterations in the community composition of trees (Laurance etal., 2000, 2006a,b) and lianas (Laurance (4) Effectsofedgeageandadjoiningvegetation etal., 2001b). Such stresses may also reduce germination When a forest edge is newly created it is open to fluxes of (Bruna, 1999) and establishment (Uriarte etal., 2010) wind, heat and light, creating sharp edge-interior gradients of shade-tolerant plant species in fragments, leading to inforestmicroclimatethatstressorkillmanyrainforesttrees dramatic changes in the composition and abundance of (Lovejoyetal.,1986;Sizer&Tanner,1999).Astheedgeages, tree seedlings (Benítez-Malvido, 1998; Benítez-Malvido & however,proliferatingvinesandlateralbranchgrowthtend Martinez-Ramos,2003b). to‘seal’theedge,makingitlesspermeabletomicroclimatic Manyanimalgroups,suchasnumerousbees,wasps,flies changes (Camargo & Kapos, 1995; Didham & Lawton, (Fowler, Silva & Venticinque, 1993), beetles (Didham etal., 1999).Treedeathfrommicroclimaticstressislikelytodecline 1998a,b), ants (Carvalho & Vasconcelos, 1999), butterflies over the first few years after edge creation (D’Angelo etal., (Brown & Hutchings, 1997), understorey birds (Quintela, 2004) as the edge becomes less permeable, because many 1985; S.G. Laurance, 2004a) and gleaning predatory bats drought-sensitive individuals die immediately and because (Rocha,2016;Rochaetal.,2017),declineinabundancenear surviving trees may acclimate to drier, hotter conditions forest edges. Edge habitats of continuous forest and larger near the edge (Laurance etal., 2006a). Tree mortality from fragments (100ha) have fewer species of bats and higher wind turbulence, however, probably increases as the edge levels of dominance by a few common species (Rocha, ages and becomes more closed because, as suggested by 2016;Rochaetal.,2017).Negativeedgeeffectsareapparent wind-tunnelmodels,downwindturbulenceincreasesifedges even along narrow forest roads (20–30m width). Among arelesspermeable(W.F.Laurance,2004b). understoreybirds,for example,fiveof eightforagingguilds Regrowth forest adjoining fragment edges can also declined significantly in abundance within 70m of narrow lessenedge-effectintensity.Microclimaticchanges(Didham roads, evidently in response to increased light and forest & Lawton, 1999), tree mortality (Mesquita, Delamoˆnica disturbancenearroadedges(S.G.Laurance,2004a). & Laurance, 1999) and edge avoidance by understorey Somegroupsoforganismsremainstableorevenincrease birds (Develey & Stouffer, 2001; S.G. Laurance, Stouffer in abundance near edges. Leaf bryophytes (Zartman & & Laurance, 2004; S.G. Laurance, 2004a) and gleaning Nascimento, 2006), wandering spiders (Ctenus spp.; Rego, animal-eating bats (Sampaio, 2000; Meyer, Struebig & Venticinque & Brescovit, 2007; Mestre & Gasnier, 2008) Willig, 2016; Rocha, 2016; Rocha etal., 2017) are all and many frogs (Gascon, 1993) displayed no significant reducedwhenforestedgesarebufferedbyadjoiningregrowth response to edges. Organisms that favour forest ecotones forest,relativetoedgesborderedbycattlepastures.Mature ordisturbances,suchasmanyspeciesofgap-favouringand regrowth can be particularly benign for some fauna; for frugivorous birds (S.G. Laurance, 2004a), hummingbirds example,diverseassemblagesofaerial-feedinginsectivorous (Stouffer&Bierregaard,1995a),frugivorousbatsthatexploit bats showed similar activity patterns in primary forest earlysuccessionalplants(Sampaio,2000;Rochaetal.,2017), and in adjoining 30-year-old secondary forests (Navarro, light-loving butterflies (Leidner, Haddad & Lovejoy, 2010) 2014). and fast-growing lianas (Laurance etal., 2001b), increase in abundancenearedges,sometimesdramatically. VII. FORESTISOLATIONANDTHEMATRIX (3) Impactsofmultipleedges BDFFP research demonstrates that plots near two or more Unliketrueislandsencircledbywater,habitatfragmentsare edges suffer more severe edge effects than do those near surrounded by a matrix of modified vegetation that can be just one edge (Fig. 8). This conclusion is supported by highly variable in space and time. Here we highlight key studiesofedge-relatedchangesinforestmicroclimate(Kapos, factors that can influence the matrix and how, in turn, the 1989; Malcolm, 1998), vegetation structure (Malcolm, matrixinfluencesfragmentdynamicsandcomposition. BiologicalReviews(2017)000–000©2017CambridgePhilosophicalSociety 10 WilliamF.Lauranceandothers Fig.7. Thediversityofedge-effectphenomenastudiedattheBiologicalDynamicsofForestFragmentsProject(BDFFP)andthe distancetowhicheachwasfoundtopenetrateintofragmentinteriors(adaptedfromLauranceetal.,2002).PAR,photosynthetically activeradiation.Someenvironmentalphenomena(e.g.relativehumidity,soilmoisture)appearmorethanoncebecausetheywere measuredbydifferentinvestigatorsusingvaryingmethods,times,and/orstudylocations. (1) Matrixstructureandcomposition Gasnier, 2008), dung beetles (Quintero & Roslin, 2005), euglossine bees (Becker, Moure & Peralta, 1991) and The BDFFP landscape has experienced considerable monkeyssuchasredhowlersAlouattaseniculus,beardedsakis dynamism over time. In particular, secondary forests have Chiropotes satanas and brown capuchins Cebus apella (Boyle & gradually overgrown most pastures in the study area. This Smith,2010b),havealsorecolonizedsomeofthefragments. regrowthlessenstheeffectsoffragmentationforsomespecies, The surrounding matrix also has a strong effect on with the matrix becoming less hostile to faunal use and plant communities in fragments by reducing edge effects movements. Several species of insectivorous birds that had (see Section VI), influencing the movements of pollinators formerlydisappearedfromfragmentshaverecolonizedthem (Dick, 2001; Dick, Etchelecu & Austerlitz, 2003) and seed as surrounding secondary forests regenerated (Stouffer & dispersers (Jorge, 2008; Bobrowiec & Gribel, 2010; Boyle Bierregaard, 1995b; Stouffer etal., 2011). The rate of local &Smith,2010b)andstronglyinfluencingtheseedrainthat extinctionsofbirdshasalsodeclined(Stoufferetal.,2008). arrivesinfragments.Forinstance,pioneertreesregenerating The regenerating forest in the matrix now permits in fragments differed strikingly in composition between fragments as small as 100ha to support bird and bat fragments surrounded by Cecropia-dominated regrowth and assemblages similar to those in continuous forest (Wolfe those encircled by Vismia-dominated regrowth (Nascimento etal.,2015;Rochaetal.,2017).Forbats,matrixrecoveryhas etal., 2006). In this way plant and animal communities in resultedinmarkedcompositionalchangesinfragmentsand fragments may increasingly tend to mirror the composition shiftsintherankorderofthemostabundantspecies(Meyer of the surrounding matrix (Laurance etal., 2006a,b), a etal., 2016; Rocha, 2016). Gleaning animal-eating bats, phenomenon observed elsewhere in the tropics (Janzen, which formerly occurred at low abundances in fragments 1983;Diamond,Bishop&Balen,1987;Laurance,1991). (Sampaio,2000)andyoungregrowth(Bobrowiec&Gribel, 2010), have increased over the past 10–15years as the (2) Factorsinfluencingthematrix surrounding regrowth has expanded and matured (Meyer etal., 2016; Rocha, 2016; Rocha etal., 2017). A number Land-use history is a key driver of secondary succession in of other species, including certain forest spiders (Mestre & Amazonia, resulting in distinct trajectories of regeneration BiologicalReviews(2017)000–000©2017CambridgePhilosophicalSociety
Description: