ebook img

Structure and composition of altered riparian forests in an agricultural Amazonian landscape PDF

14 Pages·2015·0.64 MB·English
by  
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Structure and composition of altered riparian forests in an agricultural Amazonian landscape

EcologicalApplications,25(6),2015,pp.1725–1738 (cid:2)2015bytheEcologicalSocietyofAmerica Structure and composition of altered riparian forests in an agricultural Amazonian landscape R. CHELSEANAGY,1,2,6 STEPHENPORDER,1CHRISTOPHERNEILL,1,2 PAULOBRANDO,3,4 RAIMUNDOMOTAQUINTINO,5 ANDSEBASTIAˆOAVIZDONASCIMENTO5 1DepartmentofEcologyandEvolutionaryBiology,BrownUniversity,Providence,RhodeIsland02912USA 2EcosystemsCenter,MarineBiologicalLaboratory,WoodsHole,Massachusetts02543USA 3InstitutodePesquisaAmbientaldaAmazoˆnia(IPAM),Brası´lia-DF71503505Brazil 4DepartmentofGlobalEcology,CarnegieInstitutionforScience,Stanford,California94305-4101USA 5InstitutodePesquisaAmbientaldaAmazoˆnia(IPAM),Canarana-MT78640000Brazil Abstract. Deforestationandfragmentationinfluencethemicroclimate,vegetationstructure, andcompositionofremainingpatchesoftropicalforest.InthesouthernAmazon,atthefrontier ofcroplandexpansion, forestsareconvertedandfragmentedinapatternthatleaves standing riparianforestswhosedimensionsaremandatedbytheBrazilianNationalForestCode.These alteredriparianforestssharemanycharacteristicsofwell-studieduplandforestfragments,but differbecausetheyremainconnectedtolargerareasofforestdownstream,andbecausetheymay experiencewettersoilconditionsbecausereductionofforestcoverinthesurroundingwatershed raises groundwater levels and increases stream runoff. We compared forest regeneration, structure,composition,anddiversityinfourareasofintactriparianforestandfourareaseachof narrow,medium,andwidealteredriparianforeststhathavebeensurroundedbyagriculturesince theearly1980s.Wefoundthatseedlingabundancewasreducedbyasmuchas64%andsapling abundancewasreducedbyasmuchas67%inalteredcomparedtointactriparianforests.The most pronounceddifferences between altered andintact forestoccurrednear forestedges and within the narrowest sections of altered riparian forests. Woody plant species composition differedanddiversitywasreducedinalteredforestscomparedtointactriparianforests.However, despitebeingfragmentedforseveraldecades,largewoodyplantbiomassandcarbonstorage,the numberofliveordeadlargewoodyplants,mortalityrates,andthesizedistributionofwoody plantsdidnotdiffersignificantlybetweenalteredandintactriparianforests.Thus,eveninthese relatively narrow forests with high edge:area ratios, we saw no evidence of the increases in mortalityanddeclinesinbiomassthathavebeenfoundinothertropicalforestfragmentstudies. However, because of the changes in both species community and reduced regeneration, it is unclear how long this relative lack of change will be sustained. Additionally, Brazil recently passedalawintheirNationalForestCodeallowingnarrowerriparianbuffersthanthosestudied hereinrestoredareas,whichcouldaffecttheirlong-termsustainability. Key words: agriculture; Amazon; Brazil; composition; diversity; forest structure; fragmentation; riparianecosystems. INTRODUCTION tez-Malvido1998,Lauranceetal.1998,2002,Gasconet al.2000,OosterhoornandKappelle2000,Guariguataet Deforestationoftropicalforestsforagricultureisoneof al.2002,Higuchietal.2008). themajorforcesshapingtheEarth’ssurface(Lambinetal. Comparedtointactforests,forestfragmentsinupland 2003,LambinandMeyfroidt2011).Inadditiontodirect lossofforestcover,landconversionfragmentsremaining areas often exhibit greater mortality of canopy and forestbycreatingisolatedforestpatchesandforestedges emergent trees (Kapos 1989, Williams-Linera 1990, throughout landscapes (Fearnside 2005, DeFries et al. Oosterhoorn and Kappelle 2000, Laurance et al. 2002), 2008,Nepstadetal.2008,Asneretal.2010).Uncertainty inhibition of seedling and sapling regeneration at forest regardingtheabilityofforestfragmentstopersistinthe edges (Benitez-Malvido 1998, Gascon et al. 2000), and landscape and provide ecosystem services has led to a increased abundance of lianas and disturbance-adapted, numberofstudiesonfragmentationandedgeeffectsinthe light-demanding,andabioticallydispersedtreesnearthe Neotropics, from lowland evergreen forests of Brazil to forest edges (Oliveira-Filho et al. 1997, Laurance et al. montane cloud forests of Costa Rica (Kapos 1989, 1998, 2002, 2006, Oosterhoorn and Kappelle 2000). Williams-Linera 1990, Oliveira-Filho et al. 1997, Beni- Reduced regeneration and increased mortality of native vegetationintropicalforestfragmentsmaybecausedbya combination of microclimate changes such as increased Manuscript received 10 September 2014; revised 7 January air temperature, reduced humidity, and greater wind 2015;accepted12January2015.CorrespondingEditor:Y.Pan. 6E-mail:[email protected] damageandlightexposure,orbychangestofireregimes, 1725 1726 R.CHELSEANAGYETAL. EcologicalApplications Vol.25,No.6 which may reduce fragment areas over time from the muchofBrazil (Lima andGascon 1999,Leesand Peres edges inward (Gascon et al. 2000, Laurance et al. 2002, 2008, Macedo et al. 2013). This is because the Brazilian D’Angeloetal.2004).Fragmentationmayresultinlosses Forest Code mandates that land clearing practices leave ofcarbon(C)storedinbiomassbecauseofhigherratesof standingatleast30mofforestonbothsidesofstreams tree mortalitycoupledwithchangesinplant community lessthan10mwideto‘‘conservehydrologicalfunctions, composition, especially when lianas and pioneer species prevent soil erosion, support frontier defence, guarantee with low wood density replace slow-growing old-growth publichealth,protectsitesofnaturalbeautyandprovide treeswithhighwooddensity(Lauranceetal.2002,2006, protection for rare native species of flora and fauna’’ 2007), leading to more flammable fuel (Brando et al. (Stickleretal.2013).TheBrazilianstateofMatoGrosso 2012). Tree species diversity is often reduced in upland has one of the world’s highest rates of deforestation forestfragments(TurnerandCorlett1996,Turner1996, (DeFries et al. 2008) and is now the Amazon’s biggest Terborghetal.2001,Lauranceetal.2002),yetsomeforest and fastest-growing agricultural frontier (Brando et al. fragments isolated more than 100 years ago in the 2013).MostofthelandinMatoGrossocurrentlybeing Brazilian Atlantic Forest still maintain high species used for agriculture was cleared first for pasture in the diversity(Higuchietal.2008).Thedegreeorintensityof 1980s and converted to intensive soybean (Glycine max) past disturbance (Oliveira-Filho et al. 1997) and the cultivationafter2000(Brandoetal.2013,Galfordetal. physicalarrangement,particularlytheedge:arearatio,of 2013,Macedoetal.2013).Thetotalareainmechanized forestfragments(Kaposetal.1993,Lauranceetal.1998, agricultureinMatoGrossoalmostdoubledinthedecade Chambersetal.2007)canhaveanimportantinfluenceon from 2001 to 2011, growing from 38850 to 69421 km2 theextentofalterationofforestproperties.Additionally, (VanWey et al. 2013). Understanding the ability of somefragments andforestedges haveshown decreasing altered riparian forests to provide ecosystem services in fragmentation and edge effects over time as vegetation thisagriculturallandscapewillbeincreasinglyimportant regrew and the forest edge filled in (Oliveira-Filho et al. asthesepatternsoflandconversioncontinue. 1997, Didham and Lawton 1999, Laurance et al. 2002), The altered riparian forests of the southern Amazon the presence of pioneer and light-demanding species may experience similar microclimatic shifts to those in decreasedovertime(Higuchietal.2008),andrecruitment fragments of upland tropical forests, but they differ in exceeded mortality, suggesting recovery in very old severalotherwaysthatmaybeimportantdeterminantsof fragments(Oliveira-Filhoetal.1997). forestdynamics.Theyarelong,thinstripssurroundedon Insometropicallandscapes,forestfragmentsalsoexist threesidesbyclearedlandbutconnectedtolargerareasof as patches of forest left along stream channels. Increas- intact forest further downstream. In addition, lower inglyimportantinareasofcroplandexpansion,theseare evapotranspiration in cropland results in more water in some cases mandated by land-use regulations. Com- reaching riparian zones and small stream channels in pared to forest fragments in upland regions, we know agriculturalwatersheds(Brownetal.2005,Gordonetal. much less about the dynamics of these altered riparian 2008, Hayhoe et al. 2011, Neill et al. 2013). This may forests. Because the latter are typically thin strips with influencegroundwaterandsoilmoistureconditionsboth high edge:area ratios, changes to both forest microcli- in the riparian forest and in adjacent agricultural fields. mateandtreespeciescompositionanddynamicsmaybe Therefore,trees inthesealteredriparianareasmay have pronounced, and they are often highly altered and greater accesstosoilwater,whichcouldcompensatefor degraded(Capon etal. 2013,Strombergetal. 2013). As greater water losses through transpiration caused by theconfigurationandsizeofafragmentcaninfluencethe hotteranddrierairconditions,andincreasefireresistance. degree of change following fragmentation, the width of Alternatively, this higher water table could cause addi- theriparianbufferleftbehindinalteredlandscapes may tional stress if the roots are inundated for long time similarly influence changes in forest structure and periods.Itisalsopossiblethatalteredriparianforestsin composition. However, there is some evidence that agriculturallandscapesareresistanttochangeandsimilar riparian ecosystems may be more resilient than other insomewaystostreamsideorgalleryforeststhatexistas systems because of their inherently heterogeneous envi- natural fragments in drier landscapes, although the ronmental conditions, which include alternating periods mechanismofisolationandtimesinceisolationcertainly of wet and dry conditions. Understanding the fate of differ. Studies in Belize and Venezuela indicated that tropical riparian forests in increasingly fragmented galleryforestscontainhightreespeciesdensitiesandrates landscapes is important because these forests are highly oftreegrowthandturnoversimilartothoseincontinuous valuedforprovidingcriticalecosystemservices,including forests (Meave et al. 1991, MacDougall and Kellman refugia for plant and animal species in aquatic, semi- 1992,Kellmanetal.1998). aquatic, and terrestrial areas, and protecting stream We studied riparian forests in the Brazilian state of resourcesbymaintainingstreamtemperatureandreduc- Mato Grosso in the southern Amazon to see if they ingerosionofsedimentsandnutrients(Meaveetal.1991, respond similarly to upland forest fragments or natu- Naimanetal.2005,Malhietal.2008). rally isolated gallery forests because they share charac- Altered riparian forests are now common in agricul- teristics with both. We compared riparian forests turallandscapesofthesouthernAmazonandthroughout surrounded by agricultural cropland with riparian September2015 ISOLATIONOFAMAZONIANRIPARIANFORESTS 1727 FIG.1. MapoftheFazendaTangurofarmboundaries(MatoGrosso,Brazil)withaltered(surroundedbysoybeans;Glycine max)andintact(surroundedbyforest)riparianforesttransectlocations.ThebackgroundimageisaLandsat5TMfalsecolor satelliteimage(bands5,4,and3)fromJune2011.Theinsetsshowthestudydesignofvegetationanalysisinalteredriparianforests: plots(5310 m; for trees) andsubplotswithin the plots(131 m; for herbaceouscover). Intactriparian forestswere sampled similarly,withtransectlengthsmirroringthosefoundinthemediumwidth(W2;185–210mwide)alteredriparianzones. forests within large areas of remaining intact tropical 52823008.8500 W). The mean annual temperature is 258C forest to examine how fragmentation of riparian forest and mean annual precipitation from 2005 to 2011 was influences forest vegetation dynamics and ecosystem 1770mm/yr(Rochaetal.2013).Theregionexperiencesa services including C storage and biodiversity. We distinctdryseasonfromMaytoAugust,whenrainfallis addressed the following questions: (Question 1) Do ,10mm/month(Rochaetal.2013).Today,50%ofthe altered riparian forests have altered microclimates, ranch remains in closed-canopy evergreen forests inter- including hotter and drier conditions, and are changes mediate in stature between the more humid rain forests tomicroclimategreatestnearforestedges?(Question2) tothenorthandCerrado(Braziliansavanna)vegetation Have altered riparian forests degraded over time to the south (Ivanauskas et al. 2004, Balch et al. 2008). through loss of large trees and biomass, reduced tree TherestoftheareaonFazendaTangurowasdeforested regeneration, and decreased species richness? (Question for pasture in the early 1980s. During forest clearing, 3) Are changes to altered riparian forest structure some riparian buffers surrounding streams were left in greater in narrower compared with wider forest rem- place.Impoundmentswerealsoconstructedonheadwa- nants? (Question 4) Has increased water availability ter streams throughout the region to provide water for causedbyahigherwatertablebufferedalteredriparian cattle (Macedo et al. 2013). To initiate soybean forests against the microclimatic changes typically cultivation beginning in 2003, remaining woody vegeta- associatedwithfragmentation? tion was piled into rows and burned and the soil was Thiswork,particularlyempiricalinformationderived tilled,afterwhichsoybeanswereplantedandmaintained from Question 3, can inform policy decisions regarding usingno-tillpractices(Riskinetal.2013). forest management practices and mandated riparian buffer widthin Brazil. Experimental design We studied riparian forests along four headwater METHODS streamsflowingthroughsoybeanfields(alteredriparian Site description forests), and four streams flowing through intact forest Fazenda Tanguro is an 800-km2 soybean farm in the (Fig. 1). We use the term ‘‘edge’’ to mean the forest Brazilian state of Mato Grosso (13804035.3900 S, fragmentedgeadjacenttothesoybeanfield,nottheless- 1728 R.CHELSEANAGYETAL. EcologicalApplications Vol.25,No.6 pronouncedforestedgethatexistsalongthestream.For We counted seedlings (,5 cm diameter at breast eachofthestreamssurroundedbysoybeans,wecreated height[dbh]and,30cmheightandlargeenoughtobe three transects perpendicular to the stream in the seenbythenakedeye),saplings(,5cmdbhand(cid:2)30cm remaining riparian vegetation at three forest widths: height),andlianasandestimatedpercentcoverofgrass, narrow(W1;75–90mwidefromedgetoedge),medium forbs,andfernsinfive1-m2subplotsarrayeddiagonally (W2;185–210m),andwide(W3;210–325m).Thethree across each of the larger vegetation plots (Fig. 1). We riparian forest widths were included to look for the measured all live and dead woody plants (including presence of a threshold (Question 3) below or above trees, lianas, and palms) (cid:2)5 cm dbh in each 5310 m which riparian forest properties change to inform plot (Question 2). To confirm dead vs. living trees, the management decisions on mandated riparian widths. bark was cut and examined for water flowing in the All three of these width classes exceed the width xylem in trees suspected to be dead. To calculate the mandated by the Brazilian Forest Code (30 m on both biomassofeachtree,weappliedanallometricequation sides of the stream). The altered riparian forests were from Chave et al. (2005) developed for tropical moist asymmetrical;thedistancefromtheedgeoftheriparian forestswitha pronounceddryseason vegetationtothestreamwasnotequalonthetwosides lnðAGBÞ¼aþb3lnðdbhÞþlnðqÞ of the stream. All transects in altered riparian forests wereatleast200mdownstreamfromimpoundmentsto incorporating aboveground biomass (in kg; AGB), dbh limittheir effectonriparianvegetation. Riparianforest (incm),andwoodspecificgravity(q;ing/cm3);aandb plotsshowednosignofpreviousloggingasindicatedby areconstantsforatropicalmoistforestequalto(cid:3)1.864 stumps or other signs beyond the current maintenance and 2.608, respectively. We chose the wood mean of farm-field edges. specific gravity for the site (Mato Grosso, Brazil) to be 0.61g/cm3,basedonCarvalhoetal.(2001).Tocalculate We mirrored the design of the altered riparian forest the aboveground biomass of each palm, we used an transects in the intact riparian forests. We established equationfrom Goodmanet al.(2013) fourtransectsthatcrossedsimilar-sizedstreams(2.1–5.7 and 2.7–4.7 m wide in intact and altered forests, lnðAGBÞ¼(cid:3)3:3488þ2:74833lnðdbhÞ: respectively) and had the same length as the transects in the fourW2altered riparian forests. To calculate the aboveground biomass of each liana, Along each of the 16 transects, we sampled the weused anequationfrom Gehringet al.(2004) vegetation in six 5 3 10 m plots (300 m2/transect) to lnðAGBÞ¼(cid:3)7:114þ2:2763lnðDÞ characterize stream, mid-riparian zone (mid), and edge vegetation(Fig.1).Severalsmallerplotswerechosenas where D is diameter at 30 cm height (in mm). To opposed to one larger plot because we expected calculatethelianadiameterat30cmheight,wefollowed variation in both vegetation (Kellman et al. 1998) and the equationfrom Gehringet al.(2004) drivers of change (e.g., altered microclimate) within D¼1:235dbhþ0:002dbh2: riparian forests, with vegetation differences most pro- nouncedatforestedges(Question1).Thisplotsizeand WecalculatedCcontentoftrees,palms,andlianasas configuration maximized the available width of altered 50%ofthebiomassofeachwoodyplant.Wecalculated riparianforestswhilekeepingplotsizeconsistentamong the basal area (cross-sectional area) of all large woody sites at different riparian widths. Plot length was plantswith theequation constricted by (1) meandering streams and riparian (cid:2)dbh(cid:3)2 buffer vegetation that followed these contours, (2) BA¼p maintaining a minimum distance of 200 m from each 2 of the impoundments, and (3) maintaining a minimum whereBAisthebasalarea.Were-censusedtreesoneyear distance of 0.5 km between transects to discretize sites after initial plot establishment and calculated annual among the three width categories. Each transect had a mortality as the number of newly dead trees per plot complete set of the plot locations (edge, mid, and dividedbythetotalnumberoftreesperplot.Weassayed stream) on each side of the stream (Fig. 1). Because canopy cover using leaf area index (LAI) measurements therewasnocomparableedgeinintactriparianforests, from a LAI-2000 plant canopy analyzer (LI-COR, theplotswerelocatedatdistancesfromthestreamsuch Lincoln,Nebraska,USA)every10malongeachtransect thattheyreplicatedthedistancestothestream,mid,and intheintactandW2alteredriparianforestsonly. edge in the W2 altered riparian forest transects. In We identified plants (cid:2)5 cm dbh (including trees, additiontothecategoricalparametersofedge,mid,and palms,andlianas)tospecies,whenpossible.Thefloraof stream, we measured the distance to the stream and the region is well known, and local botanists from the distance to the nearest edge to use as continuous Instituto de Pesquisa Ambiental da Amazoˆnia (IPAM) explanatory variables. The distance of each plot center identified the trees. If the species was unknown, trees to the nearest edge (altered riparian forests only) or to were identified to the genus level. If the genus was the stream wasmeasured inthe field. unknown, trees were included in an ‘‘unknown’’ September2015 ISOLATIONOFAMAZONIANRIPARIANFORESTS 1729 category that encompassed 41 of 862 or 4.75% of the Madrid, Spain). We used pressure data from the water total number of trees in all plots. Unknown trees were levelloggersandconvertedittothewaterdepthbelowthe excludedfromcompositionalanalyses,butwereinclud- soilsurface.Tomapaprofileofthewatertablefromwell ed in structural analyses (size distribution, biomass, towellalongeachtransect,wesubtractedthedepthofthe etc.).Toreducetheinfluenceofextremelyrarespecies,if watertablebelowthesoilsurfacefromthewell’srelative therewasonlyonetreeofaknownspeciesfoundinany elevation(relativetothemostuplandwellinthetransect). of the plots (i.e., singletons) we removed that species Statistical analysis fromthecompositionalanalyses(Totietal.2000).Asa metric of a species’ contribution to the community, we AllstatisticalcomparisonsweredoneinRversion3.0.1 calculated the importance value index (IVI) for each (RCoreTeam2013),andweassessedsignificanceatP, speciesateach siteusingthe following equation: 0.05.Allvariablesweretestedfornormality;ifthedatadid not approach normality, a suite oftransformations were IVI¼RBAþRDþRF employed(log,squareroot,etc.)andifnormalitywasstill where IVI is the importance value index, RBA is the notachieved,nonparametricstatisticaltestswereused. relativebasalarea((basalareaofeachspecies/totalbasal We compared differences in temperature, relative area of all species) 3 100), RD is the relative density humidity, VPD,and light at forestedges (0m along the ((number of individuals per species per plot/total transect)inW2alteredandintactriparianforestswitha number of individuals of all species per plot) 3 100), Wilcoxon signed rank test. We analyzed the water table andRFistherelativefrequency((numberofoccurrenc- depth in paired wells in intact and W2 altered riparian es of each species/number of occurrences of all species siteswithapairedttest. combined) 3 100)), following Dangol and Shivakoti WeusedaGtestofindependencewitharowbycolumn (2001)andBautista et al.(2014). (R3C)contingencytabletoanalyzewhethercounts of Wemonitoredairtemperature,relativehumidity, and seedlings and saplings differed among forest types and locations (Sokal and Rohlf 1995). We also analyzed light(wetseasononly)simultaneouslyalongaW2altered counts of seedlings, saplings, lianas, and live and dead riparian forests and its paired intact riparian forest trees among forest types (intact riparian forest and W1, transect using HOBO Tidbit v2 water temperature data W2,andW3alteredriparianforests)andlocations(edge, loggers,HOBOU23Prov2temperature/relativehumidity mid, stream) using generalized linear models with a dataloggers,andHOBOPendanttemperature/lightdata PoissondistributionandHSDtests(Rpackageagricolae; loggers (Onset, Bourne, Massachusetts, USA). We Mendiburu 2014) when we found significant differences collected data from three different transect pairs over amonggroups(KindtandCoe2005).WeusedaKruskal- thecourseofthreeweeksinthedryseason(18June–9July Wallisranksumtesttoanalyzepercentgrasscoveramong 2013, measured every 30 min for six continuous days in foresttypes(intact,W1,W2,W3).Weanalyzedmortality each altered–intact pair), and for one transect pair over among forest types and locations using Kruskal-Wallis thecourseofoneweekintherainy season(8–14March ranksumtests. 2013,measuredevery30min).Weplacedloggersroughly Inadditiontoanalyzingplotsindiscretecategoriesof every 20 m along each transect, depending on the total location (edge, mid, stream), we compared counts of transect length. We calculated vapor pressure deficit seedlings, saplings, lianas, and live and dead trees in (VPD;inPa)withthefollowingequation altered riparian forests to the distance from the stream (cid:2) RH(cid:3) and distance from the nearest edge using generalized VPD¼ 1(cid:3) 3SVP 100 linearmodelswithaPoissondistribution(KindtandCoe 2005).Wecomparedmeantreedbhtothedistancefrom where RH is the relative humidity (%) and SVP is the the stream and distance from the nearest edge using saturatedvaporpressure(Pa).SVPwascalculatedfrom multipleregression. SVP¼610:7310ð7:5T=237:3þTÞ Tocomparesizedistributionsofalltreesamongforest types(intactvs.W1,intactvs.W2,andintactvs.W3),we whereTisthetemperature(8C;MonteithandUnsworth usedFisher’sexacttests(Kindtand Coe2005).Wealso 1990). compared tree size distributions, with binned tree To measure water table depth along the transects, we diameters (e.g., 5–10 cm, 10–15 cm), of all four forest installedfourtosixwellsalongeachintactandW2altered types(intact,W1,W2,W3)simultaneouslyusingaGtest riparianforesttransect.WeplacedHOBOU20waterlevel ofindependence(SokalandRohlf1995). loggersinwellsofeachtransectpair(n¼threepairs)for WeanalyzedabovegroundlargewoodyplantCstocks roughly one-weekintervalsinthedry season(18 June–9 using ANOVA on log -transformed data. We used a t 10 July2013),whenweexpectedthedifferencesinwatertable testtocompareLAIalongtransectsinintactandaltered heightbetweenforesttypestobegreatest.Wedetermined (W2 only) riparian transects and linear regression to the change in relative elevation along the transect using compareLAIanddistancetothenearestedge. traditionalsurveyingmethodsusinganAT-G2automatic Foranalysisofspeciescompositionanddiversity,we level, tripod, and surveyor’s rod (Topcon Positioning, excludedthemidlocationsofeachtransectbecausemid 1730 R.CHELSEANAGYETAL. EcologicalApplications Vol.25,No.6 FIG.2. Vaporpressuredeficit(Pa)duringthedryseasonatforestedgesinFazendaTanguro(0malongtransect;atforestedge) ofpairedsites:(a)intact1andaltered1,18–24June2013;(b)intact2andaltered2,3–9July2013;(c)intact3andaltered3,25 June–1July2013. plots ranged from well-drained uplands to waterlogged Bray-Curtis distances are more appropriate for ecolog- near-stream zones, depending on the configuration of ical data than metrics using Euclidian distances (e.g., theriparianforestandthelocationofthestreamwithin principal components analysis; Faith et al. 1987, Kindt the relatively flat area on either side of the stream and Coe 2005). To make the PCoA graph, we used the channel.Incontrast,theedgeandstreamlocationswere ggplot2(Wilson 2009)package features. always upland or adjacent to the stream channel, To study large woody plant species diversity, we respectively. To examine the effect of forest type and calculatedthespeciesrichness,Shannondiversityindex, plot location on species composition, we grouped the Simpson index, and inverse Simpson index for each two sides of each stream while keeping each stream foresttype(intact, W1,W2,W3). separate. Thus for all analyses of species composition RESULTS anddiversity,weanalyzedatotalof32sites:16adjacent to a stream, and 16 at the forest edge (or the same Microclimateandhydrology distance from the stream in intact forest as one of the During the dry season, edges of altered riparian edgesin the altered riparian forest). forests were significantly hotter and had significantly To analyze species composition of all large woody lower relative humidity (RH) than edges of intact plants with dbh (cid:2)5 cm, we selected a similarity metric, riparian forests (mean values of 22.68C vs. 21.28C and Kulczynski distance, and the constrained ordination 92.9% RH vs. 83.8% RH for altered and intact, method, principal coordinates analysis (PCoA) with respectively). Mean VPD was significantly higher at Bray-Curtis distance method, in the BiodiversityR alteredriparianforestsedgesthanintactriparianforests (KindtandCoe2005)andRcmdr(Fox2005)packages edges in the dry season (606 Pa compared to 226 Pa), to plot differences in species abundances among sites with the greatest differences just after midday (Fig. 2). (Kindt and Coe 2005). Both the Kulczynski and the Duringthedryseason,thewatertablewassignificantly September2015 ISOLATIONOFAMAZONIANRIPARIANFORESTS 1731 cantly(meanvaluesof25.58Cand26.18Cforintactand alteredriparianforestedges,respectively).Wewerenot abletocollectRHdatafromforestedgesduringthewet season and therefore cannot calculate VPD during this time.Therewassignificantlymorelight(asmeasuredin lux)attheedgesofalteredriparianforeststhanatedges of intact riparian forests during the wet season (mean valuesof6490lxforalteredriparianforestsand3200lx for intact riparianforests). Seedlings andsaplings There were significantly fewer seedlings in narrow (W1), medium (W2), and wide (W3) altered riparian forests compared with intact riparian forests (Table 1). These differences in seedling numbers were most pronounced among edge plots (Table 2). There were significantly fewer saplings in W1 and W2 altered riparian forests compared with W3 altered and intact riparianforests(Table1).Saplingsweresignificantlyless abundant in W1 and W2 sites in both edge and stream plots (Table 2). The number of seedlings and saplings increased with increasing distance from the stream in altered riparian forests (Table 3). There was no difference in counts of seedlings and saplings across all forest type (intact, W1, W2, W3) and location (stream, mid,edge)simultaneouscombinations(datanotshown). FIG.3. Meanadjustedwatertabledepth(mbelowground) Foreststructure andfunctionalcomposition relativetothestartingtransectelevationinwellsforalteredand LAI at forest edges (0 m along the transect) did not intact paired sites.(a) Intact1 andaltered 1;(b) intact 2 and altered 2; (c) intact 3 and altered 3. The light gray dashed differbetweenintactandalteredriparianforests(4.56 verticallinerepresentsthestreamlocationalongeachtransect. 0.4 for intact and 4.4 6 0.5 for [W2] altered riparian Datashownaremeanvaluesovertheentire;1-weekintervalat forests; all means shown 6 SE), but was significantly eachsitepairinthedryseason(18June–9July2013). lower along entire transects in altered than intact riparian forests (Fig. 4). There was more variation in closertothesoilsurfaceinalteredthaninintactriparian LAI among streams in altered than in intact riparian forests,sometimes byseveral meters (Fig.3). forests (Fig. 4). LAI was positively and significantly Duringthewetseason,thetemperatureattheedgesof related to the distance from the nearest edge (data not intact vs. altered riparian forests did not vary signifi- shown). TABLE1. Descriptivestatisticsforlargewoodyplants(cid:2)5cmdiameteratbreastheight(dbh),smallplants,5cmdbh,andgrass coveratFazendaTanguro,an800-km2soybeanfarminMatoGrosso,Brazil. Alteredforest Vegetationcomponent Units Intactforest W2 W3 W1 Plots Alllargewoodyplants no./50m2 8.1a(0.7) 9.1a(0.7) 8.4a(0.6) 9.3a(0.7) Deadwoodyplants no./50m2 0.7a(0.2) 0.8a(0.2) 0.4a(0.2) 0.6a(0.2) Treesonly no./50m2 7.1c(0.5) 8.8ab(0.7) 8.1b(0.6) 9.0a(0.8) Palmsonly no./50m2 0.6a(0.4)(cid:2) 0.0a(0.0)(cid:2) 0.3a(0.2)(cid:2) 0.2a(0.1)(cid:2) Largelianas no./50m2 0.4a(0.1)(cid:2) 0.4a(0.1)(cid:2) 0.0a(0.0)(cid:2) 0.2a(0.1)(cid:2) Subplots Seedlings no./m2 21.2a(4.0) 12.6c(1.6) 13.1c(1.9) 15.7b(3.6) Saplings no./m2 5.7a(0.5) 2.4b(0.2) 2.5b(0.4) 5.3a(0.8) Smalllianas no./m2 0.8b(0.1) 1.6ab(0.4) 2.0a(0.4) 1.9a(0.4) Grass %cover 2.8a(1.0) 6.5a(3.0) 9.4a(2.3) 8.1a(2.5) Notes:Meansareshownwithstandarderrorinparentheses.Significantdifferencesinmeanvaluesareindicatedbydifferent superscriptedletters.Plotswere50m2,subplotswere1m2.Foresttypeswereintactriparianforest,andnarrow(W1;75–90mwide fromedgetoedge),medium(W2;185–210m),andwide(W3;210–325m)alteredriparianforest. (cid:2)Asignificantdifferencewasfoundamongmeans,butpairwiseposthoccomparisonsdidnotfindadifferenceamonggroups. 1732 R.CHELSEANAGYETAL. EcologicalApplications Vol.25,No.6 TABLE2. Numberofofseedlingsandsaplings. Groups Intact W1 W2 W3 Devianceexplained Seedlings Edge 31.1a(10.7) 11.3c(1.7) 16.5b(3.7) 26.5a(9.3) 21.3% Mid 22.0a(3.5) 15.1b(2.7) 15.0b(3.8) 13.5b(3.6) 10.8% Stream 10.5ab(2.1) 11.9a(3.7) 8.0bc(1.6) 7.2c(1.3) 10.1% Saplings Edge 6.9a(3.2) 2.3b(0.3) 2.7b(0.8) 5.3a(1.1) 50% Mid 6.1a(1.1) 2.8b(0.5) 3.2b(0.9) 5.8a(2.1) 21.3% Stream 6.2a(0.8) 2.0b(0.2) 1.7b(0.2) 4.8a(1.1) 41.3% Notes:Meansareshownwithstandarderrorinparentheses.Whenanalyzedamongallgroups (foresttypeandlocation)atonce,therewerenosignificantdifferencesamonggroups.Significant differenceswithinedge,mid,orstreamplotsonlyareindicatedbydifferentsuperscriptedletters. Of the 862 total large woody plants (all trees, palms, altered riparian forests (Table 1). There was also a andlianas(cid:2)5cmdbh)inallplots,63.9%wereinthe5– significant difference in the number of large lianas and 10cmdbhsizeclassandonly1.7%weregreaterthan40 palms among locations along the transect, although cmdbh.Thesizedistributionoflargewoodyplants(Fig. pairwise comparisons did not show differences among 5) did not vary significantly among altered and intact the edge, mid, and stream groups (mean large liana riparian forests. The number of large woody plants per counts of 0.1 6 0.1, 0.4 6 0.1, and 0.2 6 0.1 for edge, plot did not vary significantly with forest type (intact, mid, and stream, respectively; mean palm counts of 0.0 W1, W2, W3; Table 1) or with location (edge, mid, 6 0.0, 0.4 6 0.3, and 0.4 6 0.2 for edge, mid, and stream; data not shown). The number of dead large stream,respectively).LargewoodyplantbiomassandC woody plants did not differ between intact and altered content did not differ significantly among forest types; riparianforests(Table1),andamongalteredsitesonly, the mean aboveground C content was 13.2, 8.3, 15.3, dead large woody plant abundance was not related to and 15.8 kg C/m2 for intact, W1, W2, and W3, distance from nearest forest edge (Table 3). Annual respectively. mortality did not vary significantly among forest types Thereweresignificantlymoresmalllianasinthewider (3.9%61.7%,4.2%61.7%,6.5%63.4%,and2.7%6 altered riparian sites (W2 and W3) than in intact 1.1%forintact,W1,W2,andW3,respectively).Within ripariansites,butthenumberofsmalllianasinnarrow the large woody plants, there were significantly fewer altered riparian (W1) sites did not differ significantly treesinintactthanalteredriparianforests(Table1).The from any of the other sites (Table 1). Small liana number of trees did not vary with location (data not abundance was significantly and negatively related to shown). Poisson regression indicated significant differ- the distance from stream in altered riparian forests ences in the number of large lianas and palms among (Table 3). Grass cover was low in all plots and did not forest types, butpairwise comparisons did notallow us vary significantly among forest types (Table 1). Grass to determine which groups differed. Numerically, large cover did vary significantly with location: grass cover lianas were most abundant in intact and W1 and least was significantly lower in edge plots than in mid or abundantinW2alteredriparianforests,andpalmswere stream plots (0.5% 6 0.2%, 9.7% 6 1.3%, and 9.9% 6 most abundant in intact and least abundant in W1 2.4%for edge,mid, andstream, respectively). TABLE 3. Regressionmodelsbetweennumberoflargewoodyplants,deadwoodyplants,small lianas, seedlings, saplings, and mean woody plant size (for woody plants (cid:2)5 cm dbh) and distancefromstream(dst)andnearestedge(dne)foralteredriparianforestsonly. Explanatoryvariable Modeltypeanddependent Devianceexplained variable First P Second P bymodel Poisson No.largewoodyplants dne((cid:3)) 0.83 dst 0.89 0.1% No.deadwoodyplants dne 0.28 dst 0.77 2.0% Meanno.smalllianas dne((cid:3)) 0.13 dst((cid:3)) ,0.001 20.8% Meanno.seedlings dne(þ) 0.15 dst(þ) ,0.001 21.4% Meanno.saplings dne(þ) 0.22 dst(þ) 0.001 6.5% Multiple log(meandbh) log(dne) 0.92 log(dst) 0.43 1.0% Notes: Significant relationships (P , 0.05) are indicated in bold and the direction of the relationship(þ)or((cid:3))isindicatedfollowingtheexplanatoryvariable.Thedistancetonearestedge orstreamwasmeasuredinthefield. September2015 ISOLATIONOFAMAZONIANRIPARIANFORESTS 1733 FIG.4. Leafareaindex(LAI;m2leafarea/m2groundarea)inintactandalteredripariansites.(a)Meanandstandarderror acrossallintactandalteredsites(differencessignificantatP,0.05markedwithdifferentlowercaseletters),and(b)boxplotsof LAIatindividualsites.Theboldlinewithineachboxisthemedianvalueforthatsite.Theboxendpointsarethelower(25%)and upper(75%)quartiles,thewhiskersextendtothemostextremedatapointwhichisnomorethan1.5timestheinterquartilerange fromthebox,andthecirclesoutsidethewhiskersareoutliers.Measurementsweretakenevery10malongthetransects. Tree species compositionanddiversity forests shared one of the six most important species with all altered riparian forests, two species with the Among all plots, there were 110 known species of W2andW3forests,andnospecieswiththeW1forests large woody plants (trees, palms, and lianas (cid:2)5 cm (Appendix: Table A1). dbh). The species with the highest IVIs among all altered riparian forests were (ranked from the most Forallwoodyplantswithdbh(cid:2)5cm,theKulczynski important) Nectandra cuspidata, Maprounea guinanen- distance metric indicated significant differences among sis, Tapirira guianensis, Qualea witrockii, Protium forest types (P ¼ 0.001), locations (P ¼ 0.004), and spruceanum, and Inga heterophylla (Appendix: Table individualstreams(P¼0.001).Thefirsttwodimensions A1).ThespecieswiththehighestIVIsinintactriparian ofthePCoAanalysisamongthe32sitesexplained44% forests were Miconia pyrifolia, Protium guianense, ofthevarianceandseparatedtheintactriparianforests Miconia sp., P. spruceanum, Xylopia amazonica, and fromallalteredriparianforestswithnooverlap(Fig.6). Euterpe sp. (Appendix: Table A1). Intact riparian Species richness was highest in W2 altered riparian FIG.5. Thenumberoflivetrees(frequency)inbinnedsizeclasses(diameteratbreastheight,dbh)inintactandnarrow(W1; 75–90mwidefromedgetoedge),medium(W2;185–210m),andwide(W3;210–325m)alteredriparianforestsplots. 1734 R.CHELSEANAGYETAL. EcologicalApplications Vol.25,No.6 forests,butthreediversityindices,theShannondiversity forestcanopiesinforestfragmentsandnearforestedges index, the Simpson diversity index, and the inverse (Williams-Linera 1990, Kapos et al. 1993, Laurance et Simpsonindex,foundhigherdiversityinintactthanall al. 1998, Oosterhoorn and Kappelle 2000, Chambers et altered riparian sites(Table 4). al.2007) wasonlyweakly supportedatoursites. Increased small liana abundance in wider (W2 and DISCUSSION W3)alteredriparianforests(Table1)isconsistentwith Hotter and drier understories in altered compared to some previous studies in upland forest fragments intact riparian forests, which produced conditions (Laurance et al. 2002, 2007), but the pattern did not conducivetogreaterplantwaterlossfromleavesduring holdforthenarrow(W1)alteredriparianforestsorfor thedryseason(Fig.2),wereconsistentwiththefindings larger((cid:2)5cmdbh)lianas.Williams-Linera(1990)found of other studies that compared fragmented and intact increasedabundanceoflarge((cid:2)5cmdbh),butnotsmall upland forests (Kapos 1989, Williams-Linera 1990, (,5 cm dbh) lianas near forest edges. In fragmented Murcia 1995, Gascon et al. 2000, Laurance et al. forests (Laurance et al. 2002) and non-fragmented 2002). Altered microclimatic conditions at forest edges forests (Phillips et al. 2002), and among a variety of can penetrate up to 100 m toward the forest interior forest systems (Oliveira et al. 2013), lianas have been (Laurance et al. 2002). We found these differences correlated with higher tree mortality and changes in permeated the entirety of the forest fragment (185–210 forest structure and species composition. We found no m wide) in the dry season. Changes to microclimate of evidencethatgreaterabundanceofsmalllianasaffected thismagnitudeareassociatedwithplantstressandhave forest structure by increasing tree mortality as found been linked to higher tree mortality in other forest elsewhere (Lauranceetal. 1998, 2002). fragments (Kapos 1989, Benitez-Malvido 1998, Laur- OriginalforestclearingatFazendaTangurooccurred ance et al. 2002) in plants lacking drought-tolerant in the early 1980s and land was used for cattle pasture characteristics. for ;20 years until conversion to soybean cultivation Reduced abundances of seedlings and saplings in from2003to2008(Riskinetal.2013).Itispossiblethat altered riparian fragments (Tables 1 and 2) were also changestoforeststructuretookplaceclosertothetime consistent with findings in upland forest fragments of initial land clearing, but that over time these effects (Benitez-Malvido 1998, Gascon et al. 2000, Santo-Silva have lessened as altered riparian forest edges fill in, as et al. 2012). The mechanisms that caused reduced has occurred in some forest fragments (Didham and seedlingandsaplingabundancearenotknown.Greater Lawton1999,Lauranceetal.2002;butseeLauranceet reduction in seedling (edge locations only) and sapling al. 1998). This would cause microclimate differences to (alllocations)abundancesinnarrow(W1)thanmedium lessen at forest edges. We found that LAI at riparian (W2) and in medium than wide (W3) altered riparian forest edges was not significantly different between zones may indicate a tolerance threshold (Question 3) altered and intact sites, but that microclimatic differ- beyondwhichtheestablishmentofseedlingsorsurvival ences existed at forest edges, lending only partial of saplings is reduced. Inhibited regeneration mayhave support to this idea. We did not have the long-term consequences for the future of altered riparian zones, recruitment data and relative rates of recruitment vs. but it remains to be seen if reductions in seedling and mortalityneededto furtherevaluatethis possibility. saplingabundancewillpropagatetolargersizeclassesof Alternatively, there are several reasons why larger trees. Because tree longevity in tropical forests is high, woodyplantsinalteredriparianforestsmaybebuffered reduced seedling and sapling abundance suggests the against the changes to forest structure that often result possibility of an extinction debt, or inevitable future fromfragmentation.A higherwater tablerelativetothe extinctionscausedbypastevents(Magnagoetal.2013). groundsurfaceinalteredriparianforestscomparedwith Thesizedistributionandnumberofallstandinglarge intact riparianforests (Fig. 3) mayalleviate some of the woodyplants,deadlargewoodyplants,annualmortal- stressfromthehotter,drierconditionsandgreaterwater ity, and biomass and C storage in altered and intact lossattheleaflevelexperiencedby plantsatforest–field riparian forests were similar. The lack of significant edges(Question4).Insomelocations,maturetreesnear differences in the structure of the large woody plant streamsuseprimarilydeepwater(DawsonandEhleringer community 30 years after forest clearing contrasts with 1991),andinsomecases,accesstosoilwatercansustain increases in tree mortality and reduced tree biomass plantproductivityunderconditionsofincreasedtemper- associated with fragmentation and edge creation in aturesandlowerhumidity(LobellandGourdji2012). upland areas (Kapos 1989, Williams-Linera 1990, It is also possible that microclimate changes on the Lauranceetal.1998,2002).Itwasmoreconsistentwith orderofthosewemeasuredhavelittleornoinfluenceon resultsfrom naturallyisolated galleryforests(Meaveet tree condition. Kapos (1989) found that in the rainy al.1991,MacDougallandKellman1992,Kellmanetal. season,ahotteranddriermicroclimatenearforestedges 1998). Mean LAI was lower in altered riparian forests did not cause plant water deficits in understory shrubs thanintactriparianforests,butthisresultwasdrivenby compared with interior forest shrubs, suggesting that onelargeopenpatchalongonestream(alteredstream2; theseplantsreducedtheirstomatalconductanceandthus Fig. 4). Thus, the pattern of more and larger gaps in controlled water loss despite microclimatic differences.

Description:
R. CHELSEA NAGY,1,2,6 STEPHEN PORDER,1 CHRISTOPHER NEILL,1,2 PAULO structure, composition, and diversity in four areas of intact riparian forest and four areas each of forest by creating isolated forest patches and forest edges . less than 10 m wide to ''conserve hydrological functions,.
See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.