EnvironmentalResearchLetters
LETTER
Topography,drainagecapability,andlegacyofdrought differentiatetropicalecosystemresponsetoandrecoveryfrom majorhurricanes
MeiYu andQiongGao
DepartmentofEnvironmentalSciences,UniversityofPuertoRico,RioPiedras,SanJuan,PR,00936,UnitedStatesofAmerica
E-mail: meiyu@ites.upr.edu
Keywords: majorhurricane, tropicalforest, coastalmangrove, ecosystemresistance, ecosystemresilience, intensifiedclimatevariability, legacyofclimateevent
Supplementarymaterialforthisarticleisavailable online
Abstract
High-carbonsequestrationsoftropicalmontaneforestsandcoastalmangroveshavebeengreatly disturbedbyintensifiedextremeclimateeventssuchasalternatinghurricanesanddroughts. However,fewstudiesofthehurricaneimpacthavetakenintoconsiderationthelegacyofpast climateeventsandanalyzedtheheterogeneityofhurricaneimpactsbetweenmontaneforestsand coastalmangroves.Here,westudiedtheimpactofHurricanesIrmaandMariain2017oncoastal mangrovesanduplandforestsinPuertoRicoafteraseveredroughtduring2015–2016.We investigatedtheisland-wideimmediateimpactongreennessusingfusedvegetationindexfrom Sentinel-2andLandsat-8,andtheimpactonandtherecoveryof62homogeneousvegetation patchesbyderivinganimpactindexandone-yearrecoveryratio(RR).Alinearmixed-effectmodel wasappliedtoexplorerolesofhurricanewind,rainfall,topography,andbiologicalcomponentsin theimpactandtherecovery.Island-wide,theimmediateimpactishighlyspatial-heterogeneous. Althoughmostoftheislandwasbrowned,agreen-upstripinthedrysouthshowedbenefitsfrom thehurricanerainwhichrelievedthepriordroughtstress.Coastalmangrovesexperiencedthe greatestimpactandslowestrecoverywithrelativerecoveryof0.44comparedtorecoverygreater than0.70foruplandforests,andevergreenforestssustainedsignificantlymoredamagethan deciduousforests.Therecoveryofevergreenforestswasonaverage11daysearlierandfasterthan thatofmangroves.Mangroverecoverywasmostlylimitedbyinundation-relatedfactorssuchas elevation,slope,anddrainagecapacity.Whilehigherelevationrelatestoslowerrecoveryforupland forests,itfavorsmangroverecovery.Particularly,mangroverecoveryisfacilitatedbyriverpresence, explaining65%variationinRR.Thedifferentiatedresponse,recovery,andunderlyingmechanisms highlightedacomplicatedarrayofexternalforces,geophysical/biologicalmodulators,andlegacyof pastclimateeventsindeterminingandunderstandinghurricanes’impactontropicalecosystems.
1.Introduction
TropicalforestshavethelargestplantCpool(Pan etal 2013)andthehighestnetprimaryproductivity(Chapin etal 2012),andthusplayimportantroles insequestratingCgloballyandmodulatingchanges inclimate.Intropicalcoast,theforestedwetlands dominatedbymangrovessequesterespeciallylarge amountsofcarbonwithexceedinglyhighwateruse efficiencyandlowdecompositionrate,despitetheir limitedspatialextent(KirwanandMegonigal 2013, Hutchison etal 2014).However,thesustainability ofcarbonsequestrationcapabilityoftropicalcoastal wetlandischallengedbyboostedcoastaldevelopment,acceleratedrisingsealevel(Church etal 2013, Krauss etal 2014,Lovelock etal 2015),andintensified climatevariability(Cai etal 2014, 2015,Yi etal 2015) suchasfrequentseveredroughtsandhurricanes. Tropicalcyclones,especiallymajorhurricanes, arethemostimportantpulsedisturbanceaffecting coastalecosystemsintropicalandtemperateregions (Lugo 2000, 2008,Fisk etal 2013),andhaveboth
short-andlong-termimpactsonecosystemstructures,compositions,andfunctionssuchasCfluxes (Beard etal 2005,Vargas 2012,Fisk etal 2013,Hutley etal 2013).Forexample,HurricanesHugoin1989 andGeorgesin1998hitthetropicalmontaneforests inPuertoRicoandinducedbothimminentand delayedmassivetreemortality(Lugo 2000).Thehurricanesin2005and2008werethoughttocausethe significantlossofsaltmarshintheGulfofMexico in2004–2009(DahlandStedman 2013).Highwinds andexcessiverainfallareoftenthemajorforcesassociatedwithhurricanes.Forinstance,asingleevent oftropicalcyclonescanbring5%–40%annualrainfalltotheCaribbeanandUSsoutheasterncoast (ScatenaandLarsen 1991,Michener etal 1997).In addition,stormsurgesandheavyrunoffinundate thecoastallowland,andtheinundationcanlast forweeksandcausesignificantchangesincoastal vegetations(Michener etal 1997,Lugo 2008).Therefore,acomplexarrayofexternalforces,e.g.wind, rainfall,discharge,stormsurge,flooding,landslide, andhurricane-induceddrought(Miller etal 2019) imposesdiverseeffectsonecosystemsfromuplandto lowland.Recoveryoftheheavilyaffectedecosystems canlastfordecadesandmighttakevariousdirections (Lugo 2008).
Comparedtouplandforests,mangroveslivein low,saline,andanaerobicenvironments,andsurvive thisenvironmentbyexcluding80%–99%ofsaltsin theirroots,andbystoringsaltsinorsecretingsalts throughtheirleaves.Therefore,mangrovesexpend extraenergyonsaltexclusionandwateruptake (ReefandLovelock 2015,Lovelock etal 2016).Under thesaltyandanaerobicenvironment,therecovery ofdefoliatedmangroveswillbemuchharderthan aneighboringuplandforest,whichcanre-sprout quicklyafterhurricanes.Ontheotherhand,tocope withanoxia,allmangroveshaveaerialrootsofvarioustypestotakeinoxygenformetabolismand nitrificationofammoniuminsoil(Clough 2013). Mangrovescantolerateananaerobicenvironmentto somedegree.Forexample,partialsedimentburial mightinducemorpho-anatomicaladaptations,such asdevelopingnewroots,forenhancedsurvivorship (Okello etal 2020),andrisingsealevelmightleadto increasedmangrovecontributiontosoilorganiccarbon(Chen etal 2020).However,hurricane-induced prolongedinundationswithseveredamagesuchas massdefoliationmayimmersetheiraerialrootsand inducemortalitybeforenewrootscomeout.
Ingeneral,tropicalforestecosystemssuchaswet andrainforestswerenotconsideredasmoisturelimited;however,intensifiedclimatevariabilityhas broughtoutmorefrequentandseveredroughtevents whichwerereportedtomodulateCdynamicsand inducetreemortality(Gatti etal 2014,Doughty etal 2015).Forexample,droughtsuppressedphotosynthesisintheAmazonandalteredcarbonallocationtoreducethemaintenanceanddefenseagainst
herbivoresanddiseaseswhichmightbelinkedto increasedmortalityafterdrought(Doughty etal 2015).AsfoundinthetropicalforestsofElYunquein theCaribbeanduring1989–1998(Beard etal 2005), droughtscausedmoreseveredamagetofineroots thanhurricanesdid,andtherecoveryoffineroots afterdroughtswasmuchslowerthanthatafterhurricanes.Forcoastalwetlands,severedroughtexacerbatedthenegativeeffectsofsaltwaterintrusiononthe Pterocarpus freshwaterswampandinducedmassive mortalityofseedlingsandsaplings(Yu etal 2019). Althoughimpactsofhurricanesontropical forestsandmangrovesarewelldiscussedintheliterature(Michener etal 1997,Beard etal 2005,Lugo 2008),fewstudieshaveconsideredtropicalupland forestsandcoastallowlandmangrovesasaspatial continuum,andlinkedtheimpactsofthesequence ofElNiñodrought,majorhurricanes,andposthurricanedrought.Asatropicalmountainousisland intheCaribbean,PuertoRicoispronetostorms andhurricanes(L´ opez-Marrero etal 2019),andthe intensifiedclimatevariabilitypromotestheprobabilityofconsecutiveextremeclimateevents(Cai etal 2014, 2015,Runkle etal 2018).PuertoRico underwentaseveredroughtin2015–2016(Mote etal 2017)andtheannualrainfallin2015wasabout 45%lowerthanthelong-termaverageinthewet forest(Guti´ errez-Fonseca etal 2020).Thisdrought largelyreducedthecommunity-widereproductionin theGu´ anicadryforest(Lasky etal 2016),increased thelitterfallofthetropicalmontaneforestdueto physiologicalstress(Guti´ errez-Fonseca etal 2020), increasedtheCemissionthroughsoilrespiration (O’Connell etal 2018),andinducedlowrecruitment ofthe Pterocarpus freshwaterswampthroughelevatedsalinity(Yu etal 2019).Followingthesevere drought,PuertoRicoexperiencedmajorhurricanes IrmaandMariain2017,whichdefoliatedtheecosystems,snappedoruprootedtreesfromtop-mountain rainforests,wetforests,low-mountainmoistforests, dryforest,andcoastalmangroves.Thegroundsurvey inthemontaneforestreportedthatHurricanesIrma andMariabrokethreetimesthestemsanddoubled themortalitycomparedtopreviousstorms(Uriarte etal 2019).Themajorhurricanesafterseveredrought leaveanimportantopportunitytoinvestigatetheecosystemresistanceandresiliencefromuplandforests tolowlandmangrovesunderconsecutiveextreme climateevents.
Inthisstudy,weaddressthespatialdistributionsoftheimpactof2017hurricanesandthe recoveryafterinPuertoRicointhecontextof majorhurricanesafterseveredrought.Weassume thepriorseveredroughtmightreduceplantactivitiesincludingslowingdownindividualtreegrowth (Doughty etal 2015,Schwartz etal 2019),lower thesurvivorshipofseedlingsandsaplings(Yu etal 2019),andinducefine-rootmortality(Beard etal 2005),andthusmightamplifyecosystemresponses
tosubsequentmajorhurricanes.Wehypothesize thattheimpactandrecoveryarefunctionsofhurricanewindandrainfall,topography,andbiological componentssuchasvegetationtypeandfunctional traits(Lugo 2008).Specifically,wehypothesizethat coastalmangrovesarelessresistantandresilient, intermsof insitu survivorshipduringashort timeperiod,thanuplandvegetationduetotheir salineandanaerobicenvironmentthatexertsstresses onthemetabolismandgrowthofdefoliatedmangroves.Amonguplandvegetation,wehypothesize thathigher-elevationforestsexperiencemoresevere damageandslowerrecoverythanlower-elevation onesduetostrongerwindandthefactthathigherelevationplantsinwetandmoistenvironmentsare moresensitivetoandmorelikelytosufferfromthe pre-hurricanedrought.Forcoastalmangroves,drainagecapacityreducesprobabilityandperiodofinundation,andwehypothesizethatmangrovesinhigher places,onsteeperslopes,orhavingariverflowing throughtodrainexperiencelessdamageandfaster recoverythanthosewithpoordrainage.
2.Methods
2.1.Studyarea
PuertoRicoisamountainoustropicalarchipelagoin thenortheasternCaribbean(18◦15’Nand66◦30’W, figure 1),withelevationsfromslightlylowerthansea leveltomorethan1000minthecentralmountains (figure 2(a)).Thenortheasterlytradewindbrings moisturefromtheAtlanticOceananddepositsmore than4000mmofannualrainfallinwindwardmountainsabout1000mabovesealevelbutlessthan 1000mminthesouthwesternleewardcoast.Forest distributionvariesfromcloudforestandwetforest inwindwardmountains,tomoistforestinlowland, todryforestinleeward,andtomangroveand Pterocarpus swampincoastalwetlands(MillerandLugo 2009).Afteraseveredroughtin2015–2016,HurricaneIrmapassednearthenortheastoftheisland onSeptember6,2017withsustainingwindspeed of298kmh 1,andHurricaneMariatraversedthe islandonSeptember20withsustainingwindspeedof 249kmh 1 atitslandfall.Thelargelyreducedtranspirationduetomassdefoliationandtheincreases inBowenratiochangedtheregionalhydrology,and incurredpost-hurricanedrought(Miller etal 2019).
2.2.Datapreparation
WeretrievedthehurricanedatafromtheNOAA NationalHurricaneCenter(Pasch etal 2019),includinghurricanepaths,rainfall,andgustwindduringMariarecordedatstationsinPuertoRico. WecalculatedthedistancestopathsofMaria andIrma,andinterpolatedtherainfallandgust windbycokriging(figure 2,appendixS1(available onlineat stacks.iop.org/ERL/15/104046/mmedia)). Weacquiredsatellite-basedvegetationindexfrom
Figure1. LocationofPuertoRicointheCaribbean(top) andlandcoverin2010(bottom).HurricaneIrma(blue line)passedtothenortheastonSeptember6,2017while HurricaneMaria(purpleline)madealandfalland traversedtheislandfromsoutheasttonorthweston September20,2017.
Figure2. Elevationinm(a),normalannualrainfallinmm (1980–2010)(b),andwindgustsinms 1 during HurricaneMaria(c).
Landsat-8(Wulder etal 2019)andSentinel-2 (Chastain etal 2019,Pahlevan etal 2019)viaGoogle EarthEngine(Gorelick etal 2017).Normalizeddifferencevegetationindex(NDVI)andenhanced vegetationindex(EVI)werederivedafterclouddetectionandshadowremoval(Housman etal 2018). Todisentangleimpactonvariousvegetation, weobtainedlargehomogeneouspatcheswithout land-coverchangessince2000,guidedbyLandCover
Figure3. TypicaldynamicsofSentinel-2EVItimeseriesforpasture(PAST),deciduousforest(DECI),evergreenforest(EVER), andmangroveandassociates(MANG).Theredverticalbardepictstheperiodbetweenthetwohurricanes(September6to20, 2017).Thehorizontalgreenlinestotheleftandrightoftheredbarillustratethe1-yearmeanbeforeandafterthehurricanes, respectively.TheblacklineindicatestheminimumEVIafterthehurricanes,fromSeptember6,2017–March31,2018.
Mapsof2000,2010andaerialimagesof2010(Gould etal 2008,Wang etal 2017,Yu etal 2017).These include13mangroveandassociate(Pterocarpus),19 evergreenforest,11deciduousforest,and19pasturepatchesusedasareferencevegetationtypein comparisontowoodyvegetation.Sentinel-2EVIand NDVIwith5-dayrevisitareusedtocapturethehighfrequencychangesofvegetationgreennessfromJanuary2016–March2019.
2.3.Dataanalysis
2.3.1.Changesinvegetationindexonemonthbefore andafterthehurricanesovertheisland
Toassesstheimmediateimpacts,wecomparedNDVI beforethehurricanes(Aug.1–Sep.5,2017)andafter (Sep.21–Oct.31,2017)viamedian-compositeof fusedNDVIsfromLandsat-8andSentinel-2(Zhang etal 2018).TheperiodofAug–Octiswithinthewet season,whendeciduousforestshavetheirleaves.The ratioofNDVIaftertoNDVIbeforereflectstherelativechangesingreennessandtheimmediateimpacts. Wefurthercomputedtherelativechangesregarding variouslifezones,directiontoMaria’spath(rightof thepathversusleft,consideringthecounterclockwise spinandstrongerwindstotherightofhurricaneeye), anddistancefromthepath.
2.3.2.Analysisofimpactofhurricanesonvegetation patchesandtheirrecovery
Differenttypesofvegetationshowdifferentresistance andresilience,andweusedthehomogeneousand stablevegetationpatchestoassesstheimpactfromthe mountainstothecoast.Afterfillingmissingdataand filteringrandomnoiseofvegetationindicesforeach patch(ZeileisandGrothendieck 2005),weobtained
thetypicalpatternsofSentinel-2EVItimeseriesas showninfigure 3.
Foreachtimeseries,wecomputedthreestatistics:meanfromSeptember6,2016–September5, 2017asMEANBEFORE,meanfromSeptember 21,2017–September20,2018asMEANAFTER, andminimumfromSeptember6,2017–March31, 2018asLOWAFTERwithconsiderationofthe delayedmortality(figureS2).TheLOWAFTER dividedbyMEANBEFOREiscomputedasrelativelow(RL)toquantifytheimpactsofthehurricanesandpost-hurricanedrought/inundation (equation(1)),whereastheMEANAFTERdivided byMEANBEFOREisregardedasrecoveryratio (RR,equation(2)).TheRLreflectstheproportionof greennessremainingaftertheperturbations,whichis conceptuallysimilartotheresistanceindexproposed todepictdroughtimpactsusingtreerings(Lloret etal 2011).Weusedone-yearmeanbeforeorafterthe perturbationsasreferencestoreduceseasonaleffects. TheRRreflectstheproportionofaveragegreenness thatvegetationisabletoreachafteroneyear,andit isconceptuallysimilartotheresilienceindexinthe literature(Lloret etal 2011).AsRRdependsalsoon severityoftheimpact(RelativeLow),wecomputed relativerecoveryratio(RRR,equation(3))asthe differencebetweenMEANAFTERandLOWAFTER overthatbetweenMEANBEFOREandLOWAFTER. Thisindexdepictsthepercentageoflostgreennessthatvegetationregainedduringtheyearafter perturbations(figureS2). RelativeLow : RL = LowAfter MeanBefore (1)
Figure4. NDVIderivedfromLandsat-8andSentinel-2 overthemainislandofPuertoRico(a)beforeHurricanes MariaandIrma(Aug.1–Sep.5,2017)and(b)afterthe Hurricanes(Sep.21–Oct.31,2017).(c)Ratioofthechange inNDVI(ratioofNDVIaftertoNDVIbefore,lessthan1 indicatingdecreaseinNDVIwhilegreaterthan1indicating increaseinNDVIafterthehurricanes).Thepurpleline showsthepathofHurricaneMariaonSeptember20,2017.
Figure5. Boxplotsoftherelativelowafterthehurricanes (RL),recoveryratio(RR),andrelativerecoveryratio(RRR) ofMANG–mangroves,EVER–evergreenforests, DECI–deciduousforests,andPAST–pastures.
RecoveryRatio : RR = MeanAfter MeanBefore (2)
RelativeRecoveryRatio : RRR = MeanAfter LowAfter MeanBefore LowAfter (3)
RLorRRreflectspartlytheecosystemresistance orresilience,anddependsonthelocalstrengthof thehurricanes(Gardiner etal 2008, 2019)reflected aswindandrainfall,geophysicalvariablessuchas topography(Achim etal 2003,Ribeiro etal 2016), andvegetationtypes(Peltola etal 2000,Peterson etal 2019,ShibuyaandIshibashi 2019).Toquantifytheir effects,welog-transformedandregressedRLandRR onmeanDEM(DigitalElevationModel)andslope, meangust,meanrainfallduringMaria,andmean distancestothepathsofMariaandIrma,respectively.Linearmixed-effectmodelwasappliedforevergreen,deciduousforest,andpasturepatches,with vegetationtypeasthegroupingvariable.ConsideringthatmangrovepatcheshavemuchlowerDEM andslopesthanothers,andthatdrainageisespeciallyimportantforcoastalswamps,weaddedabinaryvariableRIVERtoindicatewhetherthereisariver flowingthroughtodraintheswamp,andperformed
amultipleregressionformangroves.StatisticalanalysesweredonewithR(RCoreTeam 2017)andspatialanalyseswithArcGIS(10.6,ESRI,Redland,CA). DetailsofthemethodologyareinappendixS1.
3.Results
3.1.Island-widechangesinvegetationindexone monthbeforeandonemonthafterthehurricanes Althoughthevegetationindexlargelydecreased island-wideonemonthafterthehurricanes,the changesarehighlyspatiallyheterogeneous(figure 4). Themontaneforestsclosetothehurricanepath encounteredthemostrelativedecrease.Changeratios are0.63 ± 0.33,0.66 ± 0.37,0.71 ± 0.23,0.71 ± 0.53, and0.81 ± 0.26inlowermountainwetforest(highest zonewith880ma.s.l.),subtropicalrainforest,subtropicalwetforest,lowermountainrainforest,and moistforest,respectively(figuresS1(a)and(b)).In contrasttogreennessreductionintheaforementionedzones,wefoundastripofleewardvegetation alongsouthcoastwithindryforestzone(figures 2(b) andS1(c)),withincreasedgreenness(figure 4(c)) asindicatedby1.50 ± 0.78(figureS1(a)).The patternofchangeratioshoweddecreasedimpact (increasedchangeratio)withdistancefromthehurricanepath(figuresS1(d)and(e)).However,abrupt
Table1. Mean ± SEandpairwisecomparisonofrelativelow(RL),recoveryratio(RR),andrelativerecoveryratio(RRR)among vegetationtypes.MANG–mangroves,EVER–evergreenforests,DECI–deciduousforests,andPAST–pastures.
Relativelow(RL)
changeinelevationintheeast(figure 2(a))modulatedthispattern(figuresS1(d)and(f)):Thehigh ElYunqueNationalForest(figures 2(a)andS1(c)) located50–60kmeastofthepathsufferedmore damagethanthoseareasnearthepath(figuresS1(d) and(f)).Directionalsoplaysanimportantrole becauseofthecounterclockwisespinandnorthwestwardmovementofMaria:strongerwindandlower changeratiowerefoundintheeast(right)ofthepath thaninthewest(left),especiallyforareasgreaterthan 30kmawayfromthepath(figureS1(d)).
3.2.Differentiatedhurricaneimpactonvarious vegetationtypes
For62homogeneousvegetationpatches,mangroves havethelowestandevergreenforestshavethesecond lowestaverageRL,whereaspastureshavethehighest averageRL(figures 5(a)and(b)).Averagedover NDVIandEVI,mangroves,evergreen,deciduous forests,andpastureshavemeanRLof0.56,0.65,0.76, and0.79,respectively.Thepairwiset-test(table 1) showedmangroveshavesignificantlylowerRLthan others,andevergreenforestshavesignificantlylower RLthandeciduousforestsandpastures.
ThetimingofLOWAFTERasthenumber ofdaysafterMariaindicatesperiodofdying-out ofleavesandbearstheinformationaboutduration ofhurricane-inducedstresses(figure 6).Duetothe largevariationwithinvegetationgroups,wedidnot findsignificantdifferenceinthemeantimingamong groups.TheaverageddelayofLOWAFTERis53.3 (11/12/2017),53.9(11/13/2017),69.8(11/29/2017), and110.7(1/9/2018)daysformangroves,evergreen forests,deciduousforests,andpastures,respectively. Themuchlongerdelaysforthelattertwomayrelateto post-hurricanedroughtandonsetofdryseasonstartinginlateDecember.EVIismoresensitivetochanges ingreennessforhighleafbiomass,andismore suitablefortropicsthanNDVI(Huete etal 2002). ThedelaysdetectedbyEVIshowedstrongercontrastamongvegetationgroupsthanthosebyNDVI (figure 6).Thelongestdelay,140.8d(2/8/2018),for pasturesshowsthattheimpactonthelowerprofile herbaceousismostlycausedbythedroughts.The shortestdelayof47.9d(11/7/2017)forevergreen forestsand11daysmore(11/18/2017)formangroves
Figure6. Meanplusstandarderrorofthenumberofdays afterSeptember20,2017forLOWAFTERtooccur. MANG–mangroves,EVER–evergreenforests, DECI–deciduousforests,andPAST–pastures. implythedamagetotheseareaswasmostlycausedby thewindsand/orinundationduetostormsurge/runoff.
Theapplicationoflinearmixed-effectmodelof RL(table 2)fornon-mangrovesshowedthatRLsignificantlydependsnegativelyonelevationandgust strength,whilepositivelyonslope.Higherelevation andgreatergustspeedmayindicatestrongersustainedwindandgreaterrainduringthehurricane, whichtendtobringmoredamage.Ontheother hand,steeperslopemightindicateeasierdrainageand attenuatewindspeedwhenblowingfromtheoppositedirection.Theinterceptsforthethreevegetation groupsfollowthepatternofRL(figures 5(a)and(b)).
TheregressionofRLofmangroves(table 3) showeddifferentpatternsoftheeffectsofgeophysicalvariables(table 2).Althoughnotsignificant,the coefficientsofDEMbecomepositiveashigherelevationmightindicatebetterdrainageandlesslikely inundation.Theabsolutevaluesofthecoefficientsof DEMaremuchlargerthanthoseintable 2 because elevationofmangrovesismuchsmallerthanthatof othervegetationtypes.Slopeissignificantonlyfor EVI,andtheeffectissimilartothatofDEMwith higherslopeforbetterdrainage.Gustissignificant onlyforEVIanditseffectisnegative.Inadditionto
Table2. Coefficientsoflinearmixed-effectmodelappliedto log-transformedrelativelow(resistance,lowestdividedbymean beforehurricanes)andlog-transformedrecoveryratio(resilience, meanafterhurricanesdividedbymeanbeforehurricanes)as functionsofgeophysicalvariables,withvegetationtypesasgroup variable.Thefirstthreerowsarerandominterceptsaddedonto thefixedintercepts.Independentvariables:DEM_A–meanDEM, SLOPE_A–meanslope,GUST_A–meanmaximumgustduring HurricaneMaria,DMARIA_A–meandistancetothepathof HurricaneMaria.Independentvariablesweredividedbytheir meanvaluesacross62patches.‘∗’indicatessignificanceat0.05 level.
RelativeLow RecoveryRatio
Table3. Coefficientsofmultipleregressionoflog-transformed relativelowandrateofrecoveryofmangrovepatchesfrom hurricanesongeophysicalvariables.DEM_A—meanDEM, SLOPE_A—meanslope,GUST_A—meanmaximumgustduring HurricaneMaria,DEM_S—standarddeviationofDEMwithin patches,RAIN_A—rainfallduringHurricaneMaria.RIVERisa logicalvariable,indicatingwhetherthereisariverrunning throughthemangrovepatch.‘∗’indicatessignificanceat0.05 level,and‘†’significanceat0.10level. RelativeLowRecoveryRatio
thevariablesfornon-mangrovetypes(table 2),rainfallhassignificantnegativeeffectonRL,ashurricane rainfalltendstoinundatethemangrovesystem.
3.3.Heterogeneousposthurricanerecovery Theone-yearrecoveryfromSep.21,2017toSep. 20,2018(figures 5(c)and(d),table 1)indicatesdistinctivenessbetweenmangrovesandothervegetation groups.MeanRRofmangrovesis0.75,whereasmean RRofothergroupsisnearorabove0.90.Pairwise t-testindicatesthatmeanRRofmangrovesissignificantlylowerthanthoseofothergroupswithEVI andNDVI.WithEVI,meanRRofevergreenforestsis significantlylowerthanthatofdeciduousforestsand pastures.
WhenRRofnon-mangrovevegetationpatches wereregressedonthecovariates(table 2),wefound thattheRRsignificantlydecreasedwithelevationand increasedwithslopeanddistancetothepathofMaria forbothvegetationindices.Vegetationinhighelevationexperiencedgreaterdamageduringthehurricanes,andcoarsersoilandpoorernutrientstatus mayalsolimittheirrecovery.Steeperslopesand greaterdistancefromhurricanepathtendtogivea betterbaseforrecoverybecauseofthelighterdamage withsteeperslopesandgreaterdistance.Theinterceptsforthethreevegetationgroupsalsofollowthe patternofRR(figures 5(c)and(d)).
RegressionofRRofmangrovepatches(table 3) showedthathighelevationsignificantlyfavored recoveryofmangrovesbecauseofthebetterdrainage.ThestandarddeviationofDEMalsoshowspositiveimpact,sothatcomplextopographyhelpsthe recovery.Thegustspeedtendstonegativelyaffect therecovery.Amongthe13mangrovepatches,we foundthat5patcheshaveriversflowingthrough.Our regressionindicatesRIVERsignificantlyandpositivelyimpactsrecoveryofmangroveforests.ANOVA
showedthatRIVERexplainedonaverage65%ofthe sumofsquares.
TheRRR(figures 5(e)and(f),table 1),averaged overEVIandNDVI,is0.44,0.70,0.73,and0.90 formangroves,evergreenforests,deciduousforests, andpastures,respectively.Thus,RRRfurtherdistinguishesmangroverecoveryfromothervegetation groups.Thepairwiset-test(table 1)showedthat mangroveindeedhassignificantlylowerRRRthan others.
4.Discussion
Tropicalforestsandmangrovesarevulnerableto changesinpressdriverssuchaswarming,drying, orrisingsealevel,andpulsedriverssuchashurricanes,whichintroduceslargeuncertaintiestotheir Csequestrationcapacityandfeedbacktoglobalclimate.Intensifiedclimatevariability(Cai etal 2014, 2015)islikelytobringmorefrequentextremeclimateevents,suchasseveredroughtandsubsequent majorhurricanes,toPuertoRico.FromMay2015 toNovember2016,theU.S.DroughtMonitorrecordedthelongestdroughtinPuertoRicosince2000 (www.drought.gov/drought/states/puerto-rico).This pre-hurricanedroughtwouldsuppressplantgrowth, suchasreducedindividualtreegrowthintheEl Yunquemontaneforestinhighaltitude,andleadto mortalityoffineroots(Beard etal 2005,Schwartz etal 2019),andthusmayhaveamplifiedtheresponse ofvegetationtothehurricanesin2017.Ontheother hand,theexcessivemoisturecarriedbythehurricanes immediatelyrelievedthedroughtstressonthevegetationindryforestandpastureinthesouthwestofthe island,whichencounteredlessdirectimpactsfrom hurricanewinds(figure 2(c)).
Ouranalysessupportourhypothesesonthegreat differencesinresistanceandresiliencebetweenlowlandmangrovesanduplandforeststomajorhurricanes(figure 5).Coastalwetlandsencounteredthe worstdamageandtheslowestrecovery,andevergreen forestsexperiencedthesecondworstdamage.The recoveryofevergreenforeststartedearlierandwas fasterthanthatofmangroves(figures 5(e),(f)and 6). Themechanismofimpactandrecoveryinvolveshurricanestrength(windandrainfall),topography,and biologicalcomponentssuchasecosystemproperties (tables 2 and 3).
Coastalmangrovesaremostvulnerable,andless resistantandresilienttohurricaneimpacts(figure 5). TheRRR,whichconsidersthevariationindamage, clearlyshowedmuchslowerrecoveryofmangroves (0.44)thanothervegetation(>0.7).Inadditionto negativeeffectofhurricanewinds,theimpactand recoveryofmangrovesareaffectedbyinundationrelatedfactorssuchasrainfall,elevation,slope,and drainagecapacity.Afterhurricanes,thesalineand anerobicenvironmentexertedhighecophysiological stressondefoliatedmangroves(Medina 1999,Reef andLovelock 2015).Astheymustdevotemuch moreenergytomaintainingthebalanceofwater, salt,oxygen,andnutrientsthanotherterrestrial plants,mangroveecosystemsaremuchmoresensitivetodisturbancesandlessresilient.AfterHurricane Andrewin1992,upto80%–95%ofmangrovesin theEvergladesweredamaged(Smith etal 1994),and mangroveforestsinsoutheastFloridaencountered greaterchangesinstructureandhighermortalitythan adjacentuplandforests(Baldwin etal 1995).After HurricaneMaria,mostofredmangroves(Rhizophoramangle)inPuntaTunaReserveclosetothe landfalldied(personalcommunication).Inundation duetostormsurge,downpourrainfall,andexcessive runoffsubmergedtheaerialrootsandblockedroot respiration.Prolongedinundationtendstoincrease themortalityofmangroves,especiallyblackmangroves(Avicenniagerminans L.),whichgrowinhigher salinityandneedmoremetabolicenergytomaintaintherootfunction.Evaporationofinundating waterincreasedsalinityofbasinmangroves,hurtingtheirrecovery.FieldobservationsinPuertoRico afterthehurricanesexhibitedmassivemortalityof blackmangrovesonthenortherncoastduetothe extendedinundationanddisfunctionofpneumatophores.Withonly1.1minmeanelevationofthe mangrovepatches,inundationafterhurricanesisthe majorthreattosurvivalandrecovery.Efficientdrainageisthekeyformangrovestosurvivetheinundation.OuranalysesofRLandRRshowedthathurricanegustspeedanddownpourstendtoweaken theresistanceandrecovery,butsteeperslopeand higherelevationhelpreducethedamageorfavorthe recovery(table 3).Particularly,ourmodelpredicted thatriverinetypetogetherwithelevationpromote recoveryofmangroves(table 3),allsupportingthe
hypothesisoftheimportanceofdrainagecapacityto mangroverecovery.
Evergreenforestsaremoresusceptibletohurricanedamagethandeciduousforestsandpastures (figure 5).Evergreenforestsgrowinamplerainfall,andthusarelocatedmostlyinthenorthand higherelevations(meanof404.4mforthepatches) closerindistancetoMaria’spath(meanof25.1km, figure 2(c)).Deciduousforestslieinlowerelevations (150.8m),furtherfromthepath(38.2km),and totheleftofthepath,andthusexperiencedless windspeed(figure 2(b)).Evergreentreeshavethicker leaveswithstrongertiestobranches,andareless likelytoshedleaves;thus,theyexperiencemoredrag forceanddamageduringhurricanesthandeciduoustrees.Allthesefactorsfortifiedthecomparison betweenevergreenanddryforestsintermsofresistancetohurricanewinds.InGu´ anica,dryforests inthesouthofPuertoRico,thetreemortalityafter HurricaneGeorgeswaslowerthanwhatwasexpectedaccordingtothemortalityofwetforestsinthe LuquilloExperimentalForestsafterHurricaneHugo duepartiallytothelowerstatureandstructureof dryforesttrees(VanBloem etal 2005, 2006).Also, duetomoreseveredamage,theevergreenforests recovermuchslowerthanthedeciduousforests(table 2).Fromthelandscapeperspective,higherelevations encounteredmoreerosionduringhurricanesandthe nutrientlossassociatedwithexcessiverunoffmight exacerbatethenutrient-poorstatusinhigherelevations.IntheLuquilloExperimentalForests,themontaneelfinforestsatthetopshowedmuchslowerrecoverythanthewetforestsatlowerelevationsafterHurricaneHugo(Walker etal 1996,Lugo 2008).Similarly,canopyrecoveryofmontaneforestswasbehind thatoflowlandforestsinJamaica(Wunderle etal 1992).
Forbothdeciduousforestsandpastures,drought isamoreimportantthreat.Hurricanerainbenefitsthedeciduousforest(figure 3).However,ground observationsindicatetheboostedgreennessmight alsobeattributedtoprosperousvinesandunderstory seedlingsorherbaceousplants,whichtakeadvantages ofthelightandnutrientresourcesoftheopenedcanopy.Ouranalysisindicatedthattheminimumvegetationindicesofdeciduousforestsandpasturesappear onaverage70and111d,respectively,afterthehurricanesincomparisonwiththemeandelayof54dfor evergreenforests.AfterHurricaneHugo,widesproutingandnewleavesoccurredwithin7weeksinthe LuquilloExperimentalForests(Walker 1991),which isclosetoourfindingof54d.Anannualmajordry periodfordeciduousforestsappearsfromDecembertoApril(Rivera 2009),andleavesemergebefore therainyseasonstarts(MurphyandLugo 1986).The minimumgreennessofdeciduousforestsandpasturesoccurredduringthepost-hurricanedrought andtheonsetofthedryseasonduetothewater scarcity.
Weinterpretedourresultsunderthecontextof consecutivepre-hurricanedrought(Schwartz etal 2019,Yu etal 2019),hurricane,andpost-hurricane drought.Thepriordroughtstressmightexacerbatehurricaneimpactsintermsofseveredamage inlowlandswampsandhigh-altitudeforests,orbe relievedbyhurricanerainsfordryforestandpastures.Thelegaciesofpastdisturbanceswerealso highlightedinthestudiesattheLuquilloExperimentalForestsafterHurricaneHugo(Beard etal 2005)andHurricaneGeorges(Ostertag etal 2005). Inadditiontothelegacies,ourconclusionsemphasizethatacomplicatedarrayofexternalforces,includingbutnotlimitingtowind,rainfall,stormsurge, andinundation,andthemodulatorsoftopography andbiologicalcomponentsfeaturedbyvegetation type,structure,andfunctionaltraits(Lugo 2008) altogetherformthebasistodetermineandunderstandtheimpactsofhurricanesontropicalecosystemsandtheirrecovery.Useofhigher-resolution images,suchasLandsatorSentinel2images,wassuggestedtoreducelargesubpixelatmosphericeffects suchasthosefromcoarse-resolutionMODIS,which togetherwithpossiblesensordegradation(Zhang etal 2017)mightleadtothedebatesondrought effectsdetectionintheAmazon(AsnerandAlencar 2010,Atkinson etal 2011,Schwartz etal 2019).We alsotookadvantageofthehighrevisit-frequencyof Sentinel2satellitestoderivetimeseriesofgreennessforanalyzingthehurricanes’impact.However, ourobserveddamageandrecoveryaddressonly changesingreennesswithoutdiscriminationofspeciescompositionorstructuralchangesuchasvine prosperity(Chinea 1999).Ecosystemstructurestake muchlongertorecoverthanecosystemfunctions suchasphotosynthesisanddecomposition,andthus structuralchangesarebetterindicatorsofhurricane disturbance(Beard etal 2005).Resistanceatspecies levelisalsorelatedtowoodtexture,withhardwoodspeciesexperiencinglessdamagethansoftwood(Ostertag etal 2005).Futurestudiesonforest resistanceandresiliencetohurricanedisturbance, basedonhyperspectralorLiDARimagescoupled withgroundsurveys,willbeabletodetectcanopystructuralchangesanddistinguishatthespecies level.
Acknowledgments
ThisresearchissupportedbyNASAEPSCoR NNX15AK43AandNOAAPuertoRicoSeaGrant. WethankChaoWangforhishelpwithinitial VIdatapreparationinGEE.TheEVIandNDVI dataareextractedfrompubliclyavailabledatasetsatUSGS(https://lpdaac.usgs.gov/)andGoogle EarthEngine(https://developers.google.com/earthengine/datasets).
Dataavailabilitystatement
Thedatathatsupportthefindingsofthis studyareopenlyavailableatthefollowing URL/DOI: https://developers.google.com/earthengine/datasets/catalog/COPERNICUS_S2.
ORCIDiD
MeiYu https://orcid.org/0000-0002-0283-9649
References
RCoreTeam2017R:Alanguageandenvironmentforstatistical computing.RFoundationforStatisticalComputing AchimA,NicollB,MochanSandGardinerB2003Windstability oftreesonslopes Int.Conf.‘WindEffectsonTrees’ (Karlsruhe:UniversityofKarlsruhe)
AsnerGPandAlencarA2010DroughtimpactsontheAmazon forest:theremotesensingperspective NewPhytol. 187 569–78
AtkinsonPM,DashJandJeganathanC2011Amazonvegetation greennessasmeasuredbysatellitesensorsoverthelast decade Geophys.Res.Lett. 38 L19105
BaldwinA,PlattW,GathenK,LessmannJandRauchT1995 Hurricanedamageandregenerationinfringemangrove forestsofSoutheastFlorida,USA J.Coast.Res. SI21 169–83
BeardKH,VogtKA,VogtDJ,ScatenaFN,CovichAP, SigurdardottirR,SiccamaTGandCrowlTA2005 Structuralandfunctionalresponsesofasubtropicalforestto 10yearsofhurricanesanddroughts Ecol.Monogr. 75 345–61
CaiWJ etal 2015IncreasedfrequencyofextremeLaNinaevents undergreenhousewarming Nat.Clim.Change 5 132–7
CaiWJ etal 2014IncreasingfrequencyofextremeElNinoevents duetogreenhousewarming Nat.Clim.Change 4 111–6
ChapinFS,MatsonPAandVitousekPM2012 Principlesof TerrestrialEcosystemEcology Secondedn(Berlin:Springer) ChastainR,HousmanI,GoldsteinJ,FincoMandTennesonK 2019EmpiricalcrosssensorcomparisonofSentinel-2Aand 2BMSI,Landsat-8OLI,andLandsat-7ETM+ topof atmospherespectralcharacteristicsovertheconterminous UnitedStates RemoteSens.Environ. 221 274–85
ChenJ,HuangY,ChenGandYeY2020Effectsofsimulatedsea levelriseonstocksandsourcesofsoilorganiccarbonin Kandeliaobovatamangroveforests ForestEcol.Manage. 460 117898
ChineaJD1999Changesintheherbaceousandvinecommunities attheBisleyExperimentalWatersheds,PuertoRico, followingHurricaneHugo Can.J.ForestRes. 29 1433–7
ChurchJA etal 2013Sealevelchange ClimateChange2013:The PhysicalScienceBasis.ContributionofWorkingGroupItothe FifthAssessmentReportoftheIntergovernmentalPanelon ClimateChange,edTFStocker,DQin,G-KPlattner, MTignor,SKAllen,JBoschung,ANauels,YXia,VBex andPMMidgley(Cambridge:CambridgeUniversityPress) pp1137–216
CloughB2013Continuingthejourneyamongstmangroves InternationalSocietyforMangroveEcosystems(ISME) Okinawa,Japan;)(Yokohama:InternationalTropical TimberOrganization(ITTO)
DahlTEandStedmanSM2013Statusandtrendsofwetlandsin thecoastalwatershedsoftheconterminousUnitedStates 2004to2009U.S.DepartmentoftheInterior,Fishand WildlifeServiceandNationalOceanicandAtmospheric Administration,NationalMarineFisheriesService DoughtyCE etal 2015Droughtimpactonforestcarbon dynamicsandfluxesinAmazonia Nature 519 78–82
FiskJP,HurttGC,ChambersJQ,ZengH,DolanKAand Negr´ on-Ju´ arezRI2013Theimpactsoftropicalcycloneson
thenetcarbonbalanceofeasternUSforests(1851–2000) Environ.Res.Lett. 8 045017
GardinerB,AchimA,NicollBandRuelJ-C2019Understanding theinteractionsbetweenwindandtrees:anintroductionto theIUFRO8thWindandTreesConference(2017) Forestry Int.J.ForestRes. vol 92 pp 375–80
GardinerB,ByrneK,HaleS,KamimuraK,MitchellSJ,PeltolaH andRuelJ-C2008Areviewofmechanisticmodellingof winddamagerisktoforests ForestryInt.J.ForestRes. 81 447–63
GattiLV etal 2014DroughtsensitivityofAmazoniancarbon balancerevealedbyatmosphericmeasurements Nature 506 76–80
GorelickN,HancherM,DixonM,IlyushchenkoS,ThauDand MooreR2017GoogleEarthEngine:planetary-scale geospatialanalysisforeveryone RemoteSens.Environ. 202 18–27
GouldWA,AlarconC,FevoldB,JimenezM,MartinuzziS, PottsG,QuinonesM,SolorzanoMandVentosaE2008The PuertoRicogapanalysisproject(SanJuan,PR) InternationalInstituteofTropicalForestry,USDAForest Service
Guti´errez-FonsecaPE,RamírezA,PringleCM,TorresPJ, McdowellWH,CovichA,CrowlTandP´erez-ReyesO2020 Whentherainforestdries:droughteffectsonamontane tropicalstreamecosysteminPuertoRico FreshwaterSci. 39 197–212
HousmanIW,ChastainRAandFincoMV2018Anevaluation offoresthealthinsectanddiseasesurveydataand satellite-basedremotesensingforestchangedetection methods:casestudiesintheUnitedStates RemoteSens. 10 1184
HueteA,DidanK,MiuraT,RodriguezEP,GaoXand FerreiraLG2002Overviewoftheradiometricand biophysicalperformanceoftheMODISvegetationindices RemoteSens.Environ. 83 195–213
HutchisonJ,ManicaA,SwetnamR,BalmfordAandSpaldingM 2014Predictingglobalpatternsinmangroveforestbiomass Conserv.Lett. 7 233–40
HutleyLB,EvansBJ,BeringerJ,CookGD,MaierSWand RazonE2013Impactsofanextremecycloneeventon landscape-scalesavannafire,productivityandgreenhouse gasemissions Environ.Res.Lett. 8 045023
KirwanMLandMegonigalJP2013Tidalwetlandstabilityin thefaceofhumanimpactsandsea-levelrise Nature 504 53–60
KraussKW,MckeeKL,LovelockCE,CahoonDR,SaintilanN, ReefRandChenL2014Howmangroveforestsadjustto risingsealevel NewPhytol. 202 19–34
LaskyJR,UriarteMandMuscarellaR2016Synchrony, compensatorydynamics,andthefunctionaltraitbasisof phenologicaldiversityinatropicaldryforesttree community:effectsofrainfallseasonality Environ.Res.Lett. 11 115003
LloretF,KeelingEGandSalaA2011Componentsoftree resilience:effectsofsuccessivelow-growthepisodesinold ponderosapineforests Oikos 120 1909–20
LovelockCE,KraussKW,OslandMJ,ReefR,BallMC etal 2016
Thephysiologyofmangrovetreeswithchangingclimate TropicalTreePhysiology-AdaptationsandResponsesina ChangingEnvironment,edGGoldsteinandLSSantiago (Berlin:Springer)pp149–79
LovelockCE etal 2015ThevulnerabilityofIndo-Pacific mangroveforeststosea-levelrise Nature 526 559–63
LugoAE2000EffectsandoutcomesofCaribbeanhurricanesina climatechangescenario Sci.TotalEnviron. 262 243–51
LugoAE2008Visibleandinvisibleeffectsofhurricanesonforest ecosystems:aninternationalreview AustralEcol. 33 368–98
L´ opez-MarreroT,Heartsill-ScalleyT,Rivera-L´ opezCF, Escalera-GarcíaIAandEchevarría-RamosM2019
Broadeningourunderstandingofhurricanesandforestson theCaribbeanislandofPuertoRico:whereandwhatshould westudynow? Forests 10 710
MedinaE etal 1999Mangrovephysiology:thechallengeofsalt, heat,andlightstressunderrecurrentflooding Ecosistemas DeManglarEnAmericaTropical,edAYañez-Arancibiaand ALLara-DomínguezUICN/ORMA,CostaRica, NOAA/NMFSSilverSpringMDUSA.380(ElHaya: InstitutodeEcologíaA.C)pp109–26
MichenerWK,BloodER,BildsteinKL,BrinsonMMand GardnerLR1997Climatechange,hurricanesandtropical storms,andrisingsealevelincoastalwetlands Ecol.Appl. 7 770–801
MillerGLandLugoAE2009Guidetotheecologicalsystemsof PuertoRico(SanJuan,PR)InternationalInstituteof TropicalForestry,USDAForestService
MillerPW,KumarA,MoteTL,MoraesFDSandMishraDR 2019PersistenthydrologicalconsequencesofHurricane MariainPuertoRico Geophys.Res.Lett. 46 1413–22
MoteTL,RamseyerCAandMillerPW2017TheSaharanair layerasanearlyrainfallseasonsuppressantintheeastern Caribbean:the2015PuertoRicodrought J.Geophys.Res. Atmos. 122 10966–82
MurphyPGandLugoAE1986Ecologyoftropicaldryforest Annu.Rev.Ecol.Syst. 17 67–88
OkelloJA,KairoJG,Dahdouh-GuebasF,BeeckmanHand KoedamN2020Mangrovetreessurvivepartialsediment burialbydevelopingnewrootsandadaptingtheirroot, branchandstemanatomy Trees 34 37–49
OstertagR,SilverWLandLugoAE2005Factorsaffecting mortalityandresistancetodamagefollowinghurricanesina rehabilitatedsubtropicalmoistforest Biotropica 37 16–24
O’ConnellCS,RuanLandSilverWL2018Droughtdrivesrapid shiftsintropicalrainforestsoilbiogeochemistryand greenhousegasemissions Nat.Commun. 9 1348
PahlevanN,ChittimalliSK,BalasubramanianSVandVellucciV 2019Sentinel-2/Landsat-8productconsistencyand implicationsformonitoringaquaticsystems RemoteSens. Environ. 220 19–29
PanY,BirdseyRA,PhillipsOLandJacksonRB2013The structure,distribution,andbiomassoftheworld’sforests Annu.Rev.Ecol.Evol.Syst. 44 593–622
PaschRJ,PennyABandBergR2019NATIONALHURRICANE CENTERTROPICALCYCLONEREPORT-HURRICANE MARIA(AL152017)September16-30,2017(National HurricaneCenter)
PeltolaH,KellomakiS,HassinenAandGrananderM2000 Mechanicalstabilityofscotspine,Norwayspruceandbirch: ananalysisoftree-pullingexperimentsinFinland Forest Ecol.Manage. 135 143–53
PetersonCJ,RibeiroGHPDM,Negr´ on-Ju´ arezR,MarraDM, ChambersJQ,HiguchiN,LimaAandCannonJB2019 Criticalwindspeedssuggestwindcouldbeanimportant disturbanceagentinAmazonianforests ForestryInt.J.Forest Res. 92 444–59
ReefRandLovelockCE2015Regulationofwaterbalancein mangroves Ann.Bot. 115 385–95
RibeiroGHPM etal 2016Mechanicalvulnerabilityand resistancetosnappinganduprootingforcentralAmazon treespecies ForestEcol.Manage. 380 1–10
RiveraOAM2009 VascularFloraoftheGuanicaDryForest, PuertoRico (Mayaguez:UniversityofPuertoRico)
RunkleJ,KunkelK,StevensL,ChampionS,EasterlingD,Terando A,SunL,StewartBandLandersG2018NOAANOAAState ClimateSummaries http://pubs.er.usgs.gov/publication/70197909
ScatenaFNandLarsenMC1991PhysicalaspectsofHurricane HugoinPuertoRico Biotropica 23 317–23
SchwartzNB,BudsockAMandUriarteM2019Fragmentation, foreststructure,andtopographymodulateimpacts ofdroughtinatropicalforestlandscape Ecology 100 e02677
ShibuyaMandIshibashiS2019Stand-levelwindthrowpatterns andlong-termdynamicsofsurvivingtreesinnatural secondarystandsafterastand-replacingwindthrowevent ForestryInt.J.ForestRes. 92 473–80
SmithTJIII,RobbleeMB,WanlessHRandDoyleTW1994 Mangroves,hurricanes,andlightningstrikes:assessment ofHurricaneAndrewsuggestsaninteractionacross twodifferingscalesofdisturbance Bioscience 44 256–62
UriarteM,ThompsonJandZimmermanJK2019Hurricane Maríatripledstembreaksanddoubledtreemortality relativetoothermajorstorms Nat.Commun. 10 1362
VanBloemSJ,LugoAEandMurphyPG2006Structural responseofCaribbeandryforeststohurricanewinds:acase studyfromGuanicaforest,PuertoRico J.Biogeogr. 33 517–23
VanBloemSJ,MurphyPG,LugoAE,OstertagR,CostaMR, BernardIR,ColonSMandMoraMC2005Theinfluence ofhurricanewindsonCaribbeandryforeststructureand nutrientpools Biotropica 37 571–83
VargasR2012Howahurricanedisturbanceinfluencesextreme CO2fluxesandvarianceinatropicalforest Environ.Res. Lett. 7 035704
WalkerLR1991TreedamageandrecoveryfromHurricaneHugo inLuquilloExperimentalForest,PuertoRico Biotropica 23 379–85
WalkerLR,ZimmermanJK,LodgeDJandGuzmangrajalesS 1996Analtitudinalcomparisonofgrowthandspecies compositioninhurricane-damagedforestsinPuertoRico J.Ecol. 84 877–89
WangC,YuMandGaoQ2017Continuedreforestationand urbanexpansioninthenewcenturyofatropicalislandin theCaribbean RemoteSens. 9 731
WulderMA etal 2019CurrentstatusofLandsatprogram, science,andapplications RemoteSens.Environ. 225 127–47
WunderleJMJr,LodgeDJandWaideRB1992Short-term effectsofHurricaneGilbertonterrestrialbirdpopulations onJamaica Auk 109 148–66
YiC,PendallEandCiaisP2015Focusonextremeeventsandthe carboncycle Environ.Res.Lett. 10 070201
YuM,GaoQ,GaoCandWangC2017Extentofnightwarming andspatiallyheterogeneouscloudinessdifferentiate temporaltrendofgreennessinmountainoustropicsinthe newcentury Sci.Rep. 7 4125610.1038/srep41256
YuM,Rivera-OcasioE,Heartsill-ScalleyT,Davila-CasanovaD, Rios-L´ opezNandGaoQ2019Landscape-level consequencesofrisingsea-leveloncoastalwetlands: saltwaterintrusiondrivesdisplacementandmortalityinthe twenty-firstcentury Wetlands 39 1343-55
ZeileisAandGrothendieckG2005Zoo:S3infrastructure forregularandirregulartimeseries J.Stat.Softw. 14 1–27
ZhangHK,RoyDP,YanL,LiZ,HuangH,VermoteE,SkakunS andRogerJ-C2018CharacterizationofSentinel-2Aand Landsat-8topofatmosphere,surface,andnadirBRDF adjustedreflectanceandNDVIdifferences RemoteSens. Environ. 215 482–94
ZhangY,SongC,BandLE,SunGandLiJ.(2017).Reanalysisof globalterrestrialvegetationtrendsfromMODISproducts: Browningorgreening?.RemoteSensingofEnvironment. 191 145–55
