Ecology and Evolution. 2017;1–12.  | 1www.ecolevol.org

Received:25January2017  |  Revised:10April2017  |  Accepted:26April2017DOI: 10.1002/ece3.3095

O R I G I N A L R E S E A R C H

Quantification and identification of lightning damage in tropical forests

Stephen P. Yanoviak1,2  | Evan M. Gora1 | Jeffrey M. Burchfield3 | Phillip M. Bitzer3 |  Matteo Detto2,4

ThisisanopenaccessarticleunderthetermsoftheCreativeCommonsAttributionLicense,whichpermitsuse,distributionandreproductioninanymedium,providedtheoriginalworkisproperlycited.©2017TheAuthors.Ecology and EvolutionpublishedbyJohnWiley&SonsLtd.

1DepartmentofBiology,UniversityofLouisville,Louisville,KY,USA2SmithsonianTropicalResearchInstitute,Balboa,Panama3DepartmentofAtmosphericScience,UniversityofAlabamainHuntsville,Huntsville,AL,USA4DepartmentofEcologyandEvolutionaryBiology,PrincetonUniversity,Princeton,NJ,USA

CorrespondenceStephenP.Yanoviak,DepartmentofBiology,UniversityofLouisville,Louisville,KY,USA.Email:steve.yanoviak@louisville.edu

Funding informationNationalScienceFoundation,Grant/AwardNumber:DEB-1354060,DEB-1354510,GRF-2015188266;NationalGeographicSociety

AbstractAccurateestimatesoftreemortalityareessentialforthedevelopmentofmechanisticforestdynamicsmodels,andforestimatingcarbonstorageandcycling.However,iden-tifyingagentsoftreemortalityisdifficultandimprecise.Althoughlightningkillsthou-sandsoftreeseachyearandisanimportantagentofmortality insomeforests,thefrequencyanddistributionof lightning-causedtreedeathremainunknownformostforests.Moreover,becauseallevidenceregardingtheeffectsoflightningontreesisnecessarilyanecdotalandposthoc,rigoroustestsofhypothesesregardingtheecologi-caleffectsof lightningareimpossible.Wedevelopedacombinedelectronicsensor/camera-basedsystemforthelocationandcharacterizationoflightningstrikestotheforestcanopyinnearrealtimeandtestedthesystemintheforestofBarroColoradoIsland,Panama.Camerasmountedontowersprovidedcontinuousvideorecordingsoftheforestcanopythatwereanalyzedtodeterminethelocationsoflightningstrikes.Weusedapreliminaryversionofthissystemtorecordandlocate18lightningstrikestotheforestovera3-yearperiod.Datafromfieldsurveysofknownlightningstrikelocations(obtainedfromthecamerasystem)enabledustodevelopaprotocolforreli-able,ground-basedidentificationofsuspectedlightningdamagetotropicaltrees.Inallcases,lightningdamagewasrelativelyinconspicuous;itwouldhavebeenoverlookedbyground-basedobservershavingnoknowledgeof theevent.We identified threetypesofevidencethatcanbeusedtoconsistentlyidentifylightningstrikedamageintropicalforests:(1)localizedanddirectionallybiasedbranchmortalityassociatedwithflashoveramongtreeandsaplingcrowns,(2)mortalityoflianasorsaplingsnearlianas,and(3)scorchedorwiltingepiphyticandhemiepiphyticplants.Thelongitudinaltrunkscarsthataretypicalof lightning-damagedtemperatetreeswereneverobserved inthisstudy.Giventheprevalenceofcommunicationstowersworldwide,thelightningdetectionsystemdescribedherecouldbeimplementedindiverseforesttypes.Datafrommultiplesystemswouldprovideanoutstandingopportunityforcomparativere-searchontheecologicaleffectsof lightning.Suchcomparativedataare increasinglyimportantgivenexpectedincreasesinlightningfrequencywithclimaticchange.

K E Y W O R D S

camera,disturbance,forestcanopy,landscape,mortality,Panama,tree

2  |     YANOVYAK et Yal.

1  | INTRODUCTION

Humans have a long history of fear and fascination with lightning(Andrews, Cooper, Darveniza, & Mackerras, 1992; Botley, 1951;Bouquegneau & Rakov, 2010; Franklin, 1769), and scientists havebeenexploringthephysicsandbiologicaleffectsoflightningformorethanacentury(e.g.,Anonymous,1898;Stone,1903;alsoseeRakov&Uman,2003).However,thespecificroleoflightningasanecologicaldisturbanceremainsoneofitsleast-studiedaspects.Thisisapartic-ularly important knowledge gap in the tropics,where lightning fre-quencyisrelativelyhigh,buttheecologicaleffectsoflightningstrikeshaveneverbeenquantifiedatlargespatialscalesinrealtime.

This gap in our understanding of the ecological importance oflightning mainly is attributed to two logistical obstacles. First, thespatial unpredictability and temporal unpredictability of lightningconstraindocumentationoftheimmediateecologicaleffectsoflight-ningstrikestoserendipitous,unreplicatedobservations(e.g.,Furtado,1935;Orville, 1968;Tutin,White, &Mackanga-Missandzou, 1996).Second, whereas ground-based and satellite-based lightning flashdetectionsystemsgenerateaccurateflashfrequencydataatregionalscales (Albrecht, Goodman, Buechler, Blakeslee, & Christian, 2016;Boccippio,Cummins,Christian,&Goodman,2001),theydonotpro-vide sufficient spatial accuracy to locate treedamagecausedby in-dividualflashes (Mäkelä,Mäkelä,Haapalainen,&Porjo,2016).Here,wedescribefield-basedmethodsthatcollectivelyenabletheaccuratequantificationof lightning strike frequency, lightning characteristics,andlightning-causedtreemortalityatthekm2scaleinnearrealtime(i.e.,withinhoursofastormevent).Wefocusontropicalforests,butthemethodscouldbeappliedtoanyterrestrialecosystem.

Discussionsof lightning-causedecologicaldisturbance invariablyfocusontrees.Indeed,lightningstrikesthousandsoftreesworldwideeachyear (Taylor,1974)and isan importantagentof treemortalityin some forests (Yanoviak etal., 2015). Ecologists commonly assesslightningdamagetotreesviaposthocsurveys(e.g.,oflightningscarsontrunks;Taylor,1964,1965).However,thisapproachis inherentlybiasedbecauselightningeffectsontreesareextremelyvariable—fromcatastrophic trunk shattering (Fernando, Mäkelä, & Cooray, 2010;Stone,1916;Taylor,1974)tonoobviousdamage(e.g.,Orville,1968).Moreover,thelethaleffectsoflightningoftenareprotracted(Furtado,1935)anddeathcommonlyoccursviaindirectmechanisms(e.g.,bee-tle and fungal infestations;DuCharme, 1972; Coulson etal., 1983).Thus, a considerable fraction of lightning-caused tree damage andmortality likely either goesunnoticedorultimately is not attributedtolightning.Finally,field-basedforestsurveysoftenareconductedatintervals>1year(commonly5years),whichgreatlylimitstheaccuracyoflightningdamageassessments,especiallygiventhatlightningscarscanbecomeunrecognizableovertimeduetolocalizedhealing,decom-position,orsecondaryinfections(SPYandEMG,pers.obs.).

Historically, post hoc lightning damage surveys havebeenmosteffectiveintemperatepineforests,inpartbecauselightningdamagetoconiferoustreescommonlyappearsasaconspicuouslongitudinalstripeonthetrunk(Gora&Yanoviak,2015;Outcalt,2008;Wadsworth,1943).Bycontrast,thebestavailablelandscape-scaledatafortropical

forestsarelimitedtosurveysofconspicuouslightninggaps(Anderson,1964;Brünig,1964;Magnusson,Lima,&deLima,1996).Theselargegroupmortalityeventspresumablyresultfromthemostintenselight-ning flashes (although thishasneverbeenverified) and representasmall fractionofthetotalnumberofstrikes inaforest.Collectively,theselimitationssuggestthatposthocsurveyssignificantlyunderesti-matelightning-causeddisturbanceinforests,especiallyinthetropics.

Ecologicalassessmentsoflightningdisturbancealsoarelimitedbyalackofinformationaboutthestrikeitself.Individuallightningflashescanhavepositiveornegativepolarityanddiffersubstantiallyininten-sity(measuredaspeakcurrent)andduration(Rakov&Uman,2003).Thereturnstrokes(thefamiliar,visibleportionofalightningdischarge)ofcloudtoground(CG)andgroundtocloud(GC)flashescauseinjuriestotrees.Manyflasheshavemultiplereturnstrokes,eachlastingjusttensofmicroseconds,butsome“continuingcurrent”(CC)strokesper-sistforhundredsofmillisecondsandlikelyinitiateforestfires(Bitzer,2017;Fuquay,Taylor,Hawe,&Schmid,1972).Itispossibletomeasurethepolarity,intensity,duration,andmultistrokenatureofalightningflash(Bitzeretal.,2013).However,toourknowledge,onlyonestudyhasattemptedtoassociatesuchcharacteristicswiththedirecteffectsoflightningontrees(Mäkelä,Karvinen,Porjo,Mäkelä,&Tuomi,2009);nosimilarstudiesexistfortropicalforests.

Theonlysolutiontotheproblemssummarizedaboveistolocatelightningstrikesandtomeasuretheircharacteristicsastheyhappen.Asnotedabove,satellite-basedopticaldetectionandland-basednet-works of electronic sensors can provide regional flash distributiondata innear real time (Cummins&Murphy,2009),but their limitedspatialaccuracymakeslocatinglightningdamagedifficultorimpossi-ble(Mäkeläetal.,2016).Themethodsdescribedhereovercomethisproblemandestablishabasisforaccuratequantificationoflightningstrikesat the landscapescale.This information, incombinationwithfield-basedassessmentoftreecondition,andknowledgeoftreetraits(e.g.,Gora&Yanoviak,2015)and lightningdischargecharacteristics,can provide comprehensive assessment of the ecological effects oflightninginforests.

Accuratelyquantifying lightning-causeddisturbance is importantbecausethefrequencyofCGlightningisexpectedtoincreasewithcli-maticchange(e.g.,Romps,Seeley,Vollaro,&Molinari,2014;Williams,2005).Specifically,modelspredictthatforeach1°Cincreaseinaver-agesurfacetemperature(oreachdoublingofatmosphericCO2 con-centration),CG lightningasa fractionof total lightningwill increasebyatleast10%(Price&Rind,1994a,1994b;Williams,1992,2005).Inthewettropics,thiswill likelyoccurviaincreasedstormintensity,prolongedinterstormintervals,increaseddryingbetweenrainevents(Price,2009), and increased lightning-initiated fires (whichcurrentlyare very rare in lowland rainforests). Models further suggest thatsmokefromagricultureandlightningfireswillincreasestormintensityandthuslightningfrequency,oftenatlocationsverydistantfromthesourcefire(Cochrane,2003;Goldammer&Price,1998;Price,2009;Price&Rind, 1994b).There is growing evidence that such changesarealreadyhappening(Norrisetal.,2016).Thus,accuratelymeasur-inglightning-causedtreemortalitywillberelevanttopredictingfutureforestdynamicsandstructureunderachangingclimate.

     |  3YANOVYAK et Yal.

Theprincipalobjectiveof this studywas toestablisha lightningmonitoringnetworkthatwouldenableustodeterminethelocationsof lightning strikes on a forest-wide scale (ca. 20km2), in near realtime, andwith high spatial accuracy (ca. 10m). Two secondary ob-jectiveswere to (1)demonstratehowecologically relevant lightningflashcharacteristics(intensity,polarity,duration,andnumberofreturnstrokes)canbequantifiedand(2)generateastandardizedlistofindi-catorsthatcouldbeusedtoreliablyidentifylightningdamageinthefieldposthoc.

2  | METHODS

FieldworkforthisprojectwasconductedincentralPanamaonBarroColorado Island (hereafter, BCI; 9.152°N, 79.846°W) and GigantePeninsula (hereafter,Gigante;9.128°N,79.856°W).BCI isa15-km2 island administered by the Smithsonian Tropical Research Institute(STRI) and is one of the best-studied patches of tropical forest onearth(Croat,1978;Leigh,Rand,&Windsor,1996).TheforestsonBCIandGigantearecategorizedasseasonallymoist,withawell-definedwetseasonspanningMaytoDecember.Muchof therainfall in theareacomesfromstormsassociatedwithtropicallow-pressurewaves.Manyofthestormsproducefrequent lightning;BCIreceivesca.40flashes km−2 year−1 and peak flash rates occur between mid-Julyandmid-August (from 1995 to 2012 satellite data; Christian etal.,2003). Roughly 25% of those flashes are CG lightning (Boccippioetal., 2001; Price&Rind, 1993) and thus are potentially damagingtotrees.Consequently,BCIcurrentlyreceivesca.150CGstrikesperyear.Sometropicallightninghotspots(e.g.,theCongo,theColombianChocó,andLakeMaracaibo;Albrechtetal.,2016)havelightningflashfrequencies ca. twice that of BCI. We focused on Panama (ratherthanalightninghotspot)becauseituniquelyoffershighlightningfre-quencyincombinationwithpoliticalstability,excellentinfrastructure,establishedlong-termforestresearchplots(Hubbell&Foster,1983),andeasyaccess.

2.1 | Video camera network

MeasuringCGlightningfrequency,distribution,anddamagetoindi-vidual trees requires continuousmonitoringof largeareasof forestcanopy at resolution of <20m,which is impossiblewith traditionallightningquantificationmethods(Mäkeläetal.,2016;Stall,Cummins,Krider, & Cramer, 2009). An alternative approach—placing a largenumber of electronic sensors in a forest—would provide adequatespatial resolution, but also would be logistically difficult and verycostly.Weestablishedavideo-basedlightningmonitoringsystemonBCI in2014 toovercome theseproblems.The systemconsistedofvideosurveillancecamerasmountedonaseriesofpreexistingguyedtowersoverlookingtheforestcanopy(Figures1and2,Table1;Kaysetal.,2011).Towerheightwasca.43mandextended>10mabovethesurroundingforestcanopyinallcases.

Eachvideocamerawasfittedwitha6mmf/1.2lensanda77mm3.0 neutral density filter and housedwithin a metal protective box

(Figure1andTable1).Thecamerasoperatedataninterlacedshutterspeedof1/30swithaperturesrangingfromf/4tof/8,dependingonlocallightconditionsandcameraorientation.Onceinstalled,eachcam-erarecordeddigitalvideoscontinuouslytoaDVRfittedwitha32GbSDmemorycard located inaweatherproofboxat thebaseofeachtower (Figure2andTable1).Thecameraandrecordingcomponentson each towerwere powered by a 12-V deep-cyclemarine batterychargedbytwoormoresolarpanelswiredinaparallelcircuittogen-erateupto90Wunderfullsunexposure(Figure2).Thepreliminarysystemestablishedin2014providedvisualcoverageofca.50%oftheislandbyonecameraand15%oftheislandbytwocameras(Figure3).Thesystemwasexpandedtofourcamerasin2015and2016topro-vide approximately the same amount of spatial coverage (Figure4),whileimprovingdirectionaldiversityandremovingthelogisticalcon-straintof frequent travelbyboat toGigante.Each camera recordedvideoforupto8daysbeforetheSDcardapproachedcapacity.The

F IGURE  1 Asurveillancecameramountedononeofthetowersusedinthevideonetworkforlightningmonitoring.A3.0neutraldensityfilterisattachedtothefrontofthecamera,andboththepowersupplyandvideofeedcablesarefittedwithin-linesurgeprotection.Thecompletesetupisprotectedwithinametalbox,whichisopenforthephotograph

F IGURE  2 Thesolarpanels(left)andelectronicequipment(right)usedtosupportindividualcamerasinthevideonetwork.Thepanelscollectivelyprovidedupto90Wofpowertorechargeadeep-cycle12-Vbattery(notshown)duringtheday,whilealsopoweringthecamera,GPSantenna,textoverlayunit,andDVR.Theselattercomponentswerethenpoweredbythebatterythroughthenight

4  |     YANOVYAK et Yal.

amountofdatarecordedperminutevariedbasedonindividualcamerasettings,sotheSDcardswereremovedfromtheDVRsandreplacedwithblankcardsevery3or4daystoavoiddataloss.

Weusedframe-by-frameanalysesofthevideosfromeachcameratoidentifylightningstrikes.Theseanalyseswereconductedautomati-callyviaanIDLprogramthatcomparedthebrightnessofeachframetothepreviousframeonapixel-by-pixelbasis(SupportingInformation).Specifically,ifmorethanonehundredpixelswerebrighterbyapartic-ularthreshold(typically,tenbytes),thentheframewasmarkedaspo-tentiallycontainingalightningstrike.Frameswithbrightnessexceedingthethresholdweresavedandsubsequentlyexaminedbytheauthorstoverify thepresenceof lightning flashes.TheSDcards collectivelycontainedhundredsofGbofdigitalvideoevenafterjust3days,soweexpeditedthevideoprocessingbytargetingtimeintervalssurrounding

knownstormevents.WeidentifiedthesetimeintervalsbasedonourownobservationsofelectricalstormsonBCI,andbypollingotherre-searchersresidingatthestation.WealsousedPanamaCanalAuthority(ACP)weatherradarimagestoverifythetimingofstormevents.Theradar images were captured automatically from the ACP website(http://www.pancanal.com/eng/radar/main.html)viaashortUnixshellscript. The program recorded the radar image every 5min, and thesubsequent imagestackwasaggregated,using IDL, intoanMPEG-4videoloopspanningtheprevious24hrat06:00everyday(VideoS1,Supporting Information).Upon identifying the time frameassociatedwith a specific storm event,we extracted the corresponding videosfromtheSDcardsforframe-by-frameprocessingasdescribedabove.

Wedetermined the approximate locationsof lightning strikes bycomparingimagesofthesameflashrecordedbytwoormorecamerasmountedondifferenttowers(Figures5and6).Theazimuthandfieldofviewofeachcameraprovidedareliableestimateofthestrikeloca-tion,whichwasthenmappedonGoogleEarth(Figure7)andverifiedby

Description Manufacturer Model number

Guyedtower ROHNProducts,LLC,Peoria,Illinois Rohn 25

Videocamera Watec,Inc.,Japan WAT-902H2Ultimate

Cameralens FujifilmOpticalDevices,Inc.,Wayne,NewJersey

DF6Ha-1B

Camerafilter SchneiderKreuznachOptics,Germany B+W65-1066177

Camerabox PelcoInc.,Clovis,California EH3512

Solarpanel SolarLandInc.,Grayslake,Illinois SLP030-12U

Chargecontroller SolarLandInc.,Grayslake,Illinois SLC-NR2420A

Textoverlayunit BlackBoxCameraCo.,London,UnitedKingdom GPSBOXSPRITE2

GPSantenna BlackBoxCameraCo.,London,UnitedKingdom ANT-555

DVR PinecomSurveillanceInc.,Covina,California PL0067

Distancemeter LeicaGeosystems,Norcross,Georgia DistoD5

TABLE  1 Sourcesofequipmentusedintheproject

F IGURE  3 Thefieldofviewofeachoftwocamerasusedinthefirstiterationofthecameranetwork(2014).Circles=approximatelocationsofthetwostrikesshowninFigure5

F IGURE  4 Thefieldofviewofeachoffourcamerasusedintheseconditerationofthecameranetwork(2015–2016).Underthisarrangement,ca.15%(225ha)ofBCIisviewedbyatleasttwocameras

     |  5YANOVYAK et Yal.

hikingtothesite.Thisapproachenabledustolocatestrikesiteswithinafewdaysoftheevent.Thesuccessfulimplementationofthecamera-basedsystemprovidedthelogisticalfoundationforthedevelopmentofthemoresophisticatedandefficientELSsystemdescribedbelow.

2.2 | Electronic network

In addition to the videomonitoring system described above,wewilldeploy twoadvancedelectronic lightningsensors (ELSs) in2017.Thesensorswereconstructed for thisproject in theAtmosphericSciencelaboratoryattheUniversityofAlabamainHuntsvilleandweretestedforbasicfunctionalityonBCIin2015and2016(Figure8).Thekeycompo-nentoftheELSisafieldchangemeterthatrecordsvariationinelectricalfieldsatextremelyfinetemporalscales(<1μs;Bitzeretal.,2013),pro-vidingintensity,duration,andpolarityforeachCGflash(Krider,1992;Figure8).Theradiationfromareturnstrokeproducesauniquesigna-turedetectedbytheELS,allowingthedifferentiationofCGstrokesfromotherlightningdischarges.Asexplainedbelow,theELSarraywillprovideveryaccurateCGflashlocationsformostofthestudysite.

EachELSrecordstheprecisemomentthatareturnstrokeforms(generally<10mabovethestrikepoint).ThetimedifferencebetweentwoELS recordings is thenplotted incombinationwith informationfroma single camera imageof the flash. Specifically, thedifferenceinarrivaltimesoftheradiationateachELSyieldshyperboliccurvesdescribingpossiblereturnstrokelocations.Theactuallocationwithinthat distribution is determined from the location of the lightningflash in thecamera imageand theazimuthof thecamera lensaxis.Collectively,this informationprovidesveryaccuratetreestrike loca-tionswithinminutes of the event and dramatically reduces depen-denceoncameras.

We tested theELSmethodologywith aMonteCarlo simulationofCGreturnstrokesreplicatedatmultiple100m×100mgridpointssuperimposedonamapofBCIandsurroundingareas(Figure9).Thearrivaltimeoftheradiationfromeachstrokeattwosensorswases-timatedwithnormallydistributederror,andtheresultinghyperboliccurvesprovidedadistributionofpossiblereturnstrokelocations.The

actualreturnstrokelocationwasdeterminedfromarecordedimageoftheattachmentpointofthestroketotheforestcanopy.Thehorizon-talpixelvalueofthestrikepointintheimageprovidedlocationdata(again,withnormallydistributederror)relativetothefieldofviewandazimuthofthecameralens,thusgeneratingtherequisitethirdpointof reference for triangulation of the strike location. Comparison oftheestimatedreturnstrokelocationswiththeiractuallocationsfromthesimulationshowedthat,onceinstalled,theELSsystemwillyield<12merrorinstrikelocationsover95%ofthestudysite(Figure9).

2.3 | Field assessment

Welocated,identified,andassessedtheconditionofstrucktreesandlianasineachstrikelocationfollowingestablishedguidelines(USDA1999).Werecordedupto25differentpiecesof informationabouteachstrikelocationandsketchedthespatialdistributionofdamage.Whennecessary,weusedthesingle-ropetechniquetoclimbnearbyundamagedtrees(Perry,1978)toverifyground-basedobservations.Treesnotonestablishedforestresearchplotsweremeasured(asdi-ameteratbreastheight;dbh)andidentifiedfromleafsamples.Eachstrikelocationwassurveyedassoonaspossibleaftertheevent,andagain3–4monthslatertodocumentmortalityandtheproductionofcoarse woody debris (CWD).We quantified CWD volume at eachstrikelocationbytallyingalldeadtrees>10cmdbh,crowndieback,anddead liana stems>2cmdbh.TheamountofCWDattributableto crown dieback was estimated based on general tree architec-tureandassociatedpatternsofallometryincrownarea(Bohlman&O’Brien,2006;Chaveetal.,2014;Malhi,Baldocchi,&Jarvis,1999;Montgomery & Chazdon, 2001; O’Brien, Hubbell, Spiro, Condit, &Foster, 1995). Although rough, these approximations provided anoverallestimateofCWDproduction foreachstrike location,whichwillprovideabasisforcomparisonwithongoingCWDinventoriesintheforestdynamicsplotsonBCI.

Togeneratea listofdiagnosticcues for theposthoc identifica-tionoflightningdamage,wequalitativelycomparedpatternsofveg-etationdamageinstrikesiteswithhaphazardlyselectedcontroltrees

F IGURE  5 StillvideoframesoftwolightningstrikestoBCItreeson14August2014,16:14localtime.ThetopimageswerecapturedbythecameraonGigante,andthebottomimagesarefromthecameraonBCI.Thetwoimagesontheleftareofthefirstflash(Example 1inthetext),andthoseontherightareofthesecondflash(Example 2).ThefieldofviewofeachcameraandtheapproximatestrikelocationsforthesetwoflashesareshowninFigure3.Allimageswerecroppedforclarity

6  |     YANOVYAK et Yal.

(>10cmdbh)locatedinpatchesofforestwithnoknownrecentlight-ningstrikes.Intotal,wesurveyed5,000controlcanopyandsubcan-opytreesdistributedacrossBCIandGigante.Thesesurveysfocusedonflashoverdamageastheprimarydiagnosticcue(describedbelow),becauseitwastheonlycueobservableateveryknownstrikesite.

3  | RESULTS

3.1 | Recorded lightning flashes

AsofOctober2016,thecamera-basedmonitoringsystemrecorded91lightningstrikestotheBCIforestandsurroundingmainland(e.g.,

Figures5and6).Themajorityof theseCGflashes (80%)werecap-turedbyonlyonecamerainthenetwork,andtheprecisestrikeloca-tioncouldnotbedeterminedinthesecases.Dueinparttoatypicallydrywet seasonsof2014and2015drivenbya strongENSOcycle(Sánchez-Murillo,Durán-Quesada,Birkel,Esquivel-Hernández,&Boll,2017),werecordedandlocatedonlyfourCGflashesproducedbyjustthreestormsthatpassedoverBCIwhenat leasttwocameraswerefullyfunctional(i.e.,werecordedatleastoneflashperstormwithinjust15%of theareaofBCI).Bycontrast,during the first4monthsof the 2016wet season, the cameras captured72 lightning strikestoBCItreesfrom25storms.Nineoftheseflasheswererecordedbymorethanonecamera,andallninestrikelocationswerelocatedandsurveyedwithindaysoftheevent.

Becausedatacollectionisongoing,hereweprovidethreeexam-plesofrecordedlightningstrikesthatrepresenttherangeofeffects

F IGURE  6 StillvideoframesofalightningflashcapturedbythreecamerasonBCIon21September2015,16:18localtime.ThecameraswerelocatedontowersassociatedwithBCItrailnamesasfollows:Zetek(topimage),AVA(centerimage),andDrayton(bottomimage).Allimageswerecroppedforclarity.AlsoseeFigure7

F IGURE  7 AnannotatedGoogleEarthimageofBCIillustratinghowtheapproximatelocationofthestrikerecordedinFigure6wasdeterminedbasedontheazimuthandfieldofviewofeachcamera.Thecoordinatesneartheintersectionofthethreelineswereusedtolocateandground-truththesite.ThelengthoftheAVAlineisca.650m

F IGURE  8 Timeseriesofanine-strokenegativelightningflashrecordedbythefieldchangemeteronBCI.RapidchangesinvoltagesindicateCGdischargesandtheirpolarity,andtheamplitudeofthechangeindicatesthepeakcurrent(i.e.,electricalintensity)ofeachdischarge

     |  7YANOVYAK et Yal.

observed todate.The first twoexample strikesoccurredwithin1sof each other on 14 August 2014 (Figure5; Video S1, SupportingInformation). The third examplewas triangulated from three towercamerason21September2015(Figures6and7).

Example 1—A mature Beilschmiedia pendula tree was killed by the flash shown in the two left panels of Figure 5 (see Figure 10). A large unidentified liana inhabiting the crown of this tree was severely damaged by the discharge, but was still alive 2.5 years later. This flash killed four other trees and conspicuously damaged 10 other trees at the site.

Example 2—A mature Sterculia apetala tree was dam-aged by the larger fork of the flash shown in the two right panels of Figure 5. Multiple large branches in the upper crown of the tree were shattered, and a large Arrabidaea sp. liana in the crown was severely damaged (Figure 11). However, there was no evidence of the strike on the trunk of the tree or on the ground in the immediate vicinity of the tree. Thus, this event would be completely unnoticed by a ground- based observer. Now, ca. 2.5 years later, the focal tree and the liana are alive and have extensive regrowth. No trees were killed by this event, and only one other tree was damaged.

Example 3—A large Ficus obtusifolia (dbh = 154 cm) and 18 other trees were damaged by this lightning flash

(Figures 6 and 12). One large canopy tree died within 2 weeks of the flash and another 6 trees (>10 cm) died over the following 10 months. In addition to the direct effects of lightning, many thousands of beetles attacked three of the lightning- damaged trees, likely contributing to their deaths. Although the damage caused by this event was extensive, the lightning flash did not con-spicuously fracture branches or cause observable trunk damage.

3.2 | Field assessment

Basedonsurveysof20strikesites,wegeneratedalistofstandard-ized diagnostic cues that can be used to identify lightning damageposthocintropicalforests(Table2).Wealsousedobservationsfrommultiplevisitstoeachsitetoapproximatethetimeperiodwheneachlightning diagnostic cue was typically observable (Table2). Finally,wemeasured the frequency of occurrence for each diagnostic cue(Table2)usingdataonlyfromstrikesthatwerecapturedontwoormorecameras.

Themostconsistentposthocdiagnosticcueofalightningstrikeevent was flashover damage (i.e., directionally biased vegetationmortality causedwhen electric current jumped fromone branch toanother across an air gap; Table2).Whereas tree trunks exhibitednoconspicuousdamagesuchas lightningscars(Taylor,1965),flash-overdamageoccurredinallstrikesites.Flashoverdamagewaseasilyrecognizedasclustersofdeadvegetationonoppositesidesofgapsbetweenneighboringplants(Figure13),distributedradiallybutasym-metricallyfromthecentralstrucktree,anddecreasing inmagnitudewith increasing distance. Dead leaves and branches occurred onlyamong individual plants thatwere in close proximity to each other(<1m),suchthatlargeportionsoftheunderstorywereunaffectedinareasbetweendamaged canopy trees.Thedistributionof flashover

F IGURE  9 Expectedspatialerror(m)forlightningstrikelocationsusingtheELSnetwork.DatafromfieldchangemetersonGigante(bluesquare)andtheBCIlaboratoryclearing(redsquare)pluscameraimagesfromacentraltoweronBCI(redX)providestrikelocationswith<12merrorfor95%ofthestudysite

F IGURE  10 Ground-basedviewofvegetationdamagefromthestrikedescribedasExample 1inthetext.Thisimagewastaken12monthspoststrike.Thefocaltreeisthelargestandingdeadtrunkascendingfromthelowerrightcorner.Notethedanglingdeadlianastems(lowerhalfoftheimage)andthedeadsaplingcrownsontheperimeteroftheimage.Thegreenfoliageinthecenteroftheimageisaportionofthelianathatsurvivedthestrike

8  |     YANOVYAK et Yal.

damage was more complicated when lianas were present, as theytended to conduct current and damage plants tens ofmeters fromthestrucktree(Figure14).Flashoverdamageoccurredonalltypesofvegetation,includingadulttrees,saplings,lianas,hemiepiphytes,epi-phytes,andpalms.Thecentralportionsoftreeshavingflashoverdam-ageoftensurvivedformonthsorlonger,butpalmsconsistentlydiedwithinweeksofthestrikewhenlocatedupto2mfromthenearestdamagedtreeorlianastem.

Inconjunctionwithflashoverdamage,damagedlianas,wiltinghe-miepiphytes,andbeetleinfestationswerediagnosticoflightningstrikesites (Table2). Lianasexhibitedcrowndieback inpatterns similar tothosedescribedfortreesabove.Hemiepiphytesgenerallywiltedanddiedwithinthefirst6weekspoststrike(Figure15).Colonizingbeetlesandtheirassociateddamage(accumulationsoftrunkentryholesandsawdust)werecommonca.2-4monthspoststrike.Vascularepiphytesgenerallywereunharmedbylightning,butepiphyticfernsoccasionally

exhibited blackening of their leaves during the first 3months post-strike,particularlywhentheywerepositionedbetweenthehosttreeandaneighboringstem(i.e.,inaflashoverpathway).Inallcases,theindiscriminate distributionof damage amongplant taxa and growthforms,incombinationwithclearflashoverpatterns,servedasreliableindicators of a lightning strike despite the absence of conspicuoustrunkdamageonanytreesatasite.

Comparison with dead and damaged vegetation in control sur-veysshowedthatthediagnosticcharacteristicsdescribedabovearereliableforground-basedlightningdamageidentification.Specifically,of the5,000haphazardly selectedcanopyandsubcanopy trees sur-veyed, damage resembling flashover among neighboring trees andother plantswas exceedingly rare;we never observed evidence re-semblingflashoverdamageamongmorethanthreeindividualplants.Wethereforesuggestthatposthocfieldidentificationoflightningintheabsenceofcameraimagesrequiresatleastfourindividualplantsthatexhibitflashoverdamage,preferablyincombinationwithanotherdiagnosticcuedescribedabove.

4  | DISCUSSION

Here,wedescribe a hybrid electronic sensor/camera-based systemforlocatinglightning-causeddisturbanceinnearrealtime,andwepro-videarecipeforthereliableposthocidentificationoflightningdam-agetotropicaltrees,lianas,andotherforestcomponents.Inconcertwitheachother,thesetwomethodsenabletheaccuratequantifica-tionoflightning-causedtreedamageandmortalityinforestsatlargespatialscales.Suchdatawillprovideanunprecedentedopportunitytoassesstheroleoflightningasanecologicaldisturbanceatmultiplelevelsofbiologicalorganization(Table3).Themonitoringsystemde-scribedherecouldbeestablishedinanyforestwithacanopythatcanbeviewedfromatleasttwoemergentpoints(e.g.,towers,buildings,

F IGURE  11 ASterculia apetalatreecrownwithanArrabidaea sp.liana14monthsbefore(topimage,June2014)and3daysafter(bottomimage,August2015)thelightningstrikedescribedasExample 2inthetext.Thetreecrownandthelianawerebadlydamaged,butwerestillliving2.5yearspoststrike.Noevidenceofthisstrikeorthesignificantcrowndamagewasvisiblefromtheground

F IGURE  12 Ground-basedimageofthefocaltree(Ficus obtusifolia)inthelightningstrikedescribedasExample 3inthetext.Thisphotographwastaken8monthsafterthestrike.Notethatmanyvascularepiphytesandepiphyticfernsarealiveinthetreecrown.By12monthspoststrike,allbrancheshadfallensuchthatonlythemaintrunkremainedstanding

     |  9YANOVYAK et Yal.

orexposedridges).Givenrecentgrowth inabundanceofcommuni-cationtowersinforestsworldwide,thisisfastbecomingarelativelyminorlogisticalhurdle.

Preliminarydatafromthissystemindicatethatlightningisanim-portant agent of tropical treemortality; indeed, lightning damagesandkillsmoretreesintheBCIforestthanweanticipatedatthestartoftheproject.Mostimportantly,thatmortalityisnotobviouslydueto lightningand isunlikely tobeattributed to lightningduringnor-malforestsurveyswithoutpriorknowledgeoftheevent.Asacasein point,Example 3 abovewas included in the 2015 survey of the50haforestdynamicsplotonBCI(Hubbell&Foster,1983),butonlythe focal treewasdeadat the timeof thesurveyandtheagentofmortalitywasnotidentified.Givenrapidchangesatthissiteoverthe12monthspoststrike,a2016plotsurveylikelywouldhaverecordedthe agent ofmortality for all trees killed at this site as blowdown.Thus,itisclearthatlightningdamageiseasilyoverlookedormisiden-tifiedinpractice.

TABLE  2 Diagnosticcharacteristicsoflightning-causedtreedamageincentralPanama.N=numberofknownstrikesitesrecordedontwoormorecamerasadjustedbasedonthepresenceofrequisiteorganismsforeachcharacteristic;Frequency=fractionofknownstrikesitesexhibitingagivencharacteristic;FirstObserved=rangeoftimepostflashthateachcharacteristicfirstbecomesclearlyevident;Persistence=approximatedurationpostflashthateachcharacteristicremainsobservable

Characteristic N Frequency (%) First observed Persistence

Flashoverdamageamongfoliage 11 100 1–4weeks Years

Lianasdamaged 11 91 1–4weeks Years

Hemiepiphytesscorchedandwilting 6 67 1–4weeks 3–6months

Beetleinfestation 11 55 2–4months Years

Epiphytesscorched 5 0 1–4weeks 2–4months

Lightningscarsontrunkorbranches 11 0 0days Years

F IGURE  13 Flashoverdamageappearsasscatteredpatchesofdeadandwiltingfoliage(arrows)innearbybranchesofneighboringtrees

F IGURE  14 Lightningdamageoftenappearsasdeadandwiltingfoliage(arrows)onunderstoryplantsgrowingincloseproximitytomaturelianasthathavegrownintoornearthecrownofthestrucktree

F IGURE  15 TypicallightningdamagetoahemiepiphyticAnthuriumsp.inPanamaThisphotographwastakenca.4weeksafterthestrike,whichkilledthefocalTabebuia guayacanandmultiplenearbysaplingsinMay2012

10  |     YANOVYAK et Yal.

Once the monitoring system described here has been estab-lished in a forest, the number of large-scale, integrative ecologicalquestions that canbeanswered immediately increases (Table3). Inparticular, measuring the ecosystem-level relevance of lightning-caused disturbance requires accurate information about its effectsonindividualtreesandtreespeciesonaforest-widescale,andthissystemisthefirsttoprovidesuchinformationwithahighdegreeofaccuracy.Specifically,thisnetworkovercomestwogeneralandverypertinentproblemsovershadowingallhistoricalresearchconcerningtheecologicaleffectsoflightning:(1)overwhelmingdependenceonanecdotalandposthocobservationsand(2)agenerallackofinfor-mation concerning lightning characteristics (intensity, polarity, andflashduration).Posthocassessmentsoflightningdamageinforestsarebiasedtowarddetectionofconspicuousevidence.AsdescribedforExample 3above,suchevidence iseasilyoverlooked,at least intropicalforests.

GiventheestimatedflashfrequencyforBCIdescribedabove(i.e.,150CGflashesperyear),thenumberofstrikesrecordedbyatleastonecamerainthenetworksince2014ismuchlowerthanexpected.Specifically,alliterationsofthenetworkprovidedca.50%coverageofBCIbyatleastonecamera,suggestingthat>200strikesshouldhavebeenrecordedsince2014.Thereasonsforthisdifferencearetwofold.First, due to logistical constraints (including humidity and lightningdamagetoequipment),thecameraswerenotoperatingcontinuouslyduring all months of each wet season. Cumulative recording timewasca.2monthsin2014,4monthsin2015,and5monthsin2016.Second,due toa strongENSOevent (Sánchez-Murilloetal., 2017),thefieldsitedidnotexperienceanyelectricallyactivestormsduring4of the6months that the systemwasactiveduring the2014and2015wetseasons.Finally,althoughthevarietyofstrikeswerecordedindicatesthatmostflashesarecapturedbythissystem,itisimpossibletotestforfalsenegativesusingonlyvideocameras.Weexpectthatfactors limitingtheconsistencyandeffectivenessofthissystemwillbeovercomebythedeploymentoftheELSsystemandadvancesinremotesensingandvideotechnologies.

Finally,growthintelecommunicationsinfrastructurehasledtoanabundanceof towers in forests. Such towersmay functionas light-ningattractors;theyareoftenonhilltops,extendwellabovetheforestcanopy,andareequippedwithlightningrods.RakovandUman(2003,p.51)estimated that a towerwithaneffectiveheightof60m (i.e.,exceedingtheheightsofsurroundingobjectsby60m)inanareawith10CGflasheskm−2 year−1willgetstruckonceeverytwoyears.Thus,iflightningplaysasignificantroleinforestdynamics,suchtowersarepotentially changing those dynamics. Coincidentally, canopy cranesusedaroundtheworldtostudyforestecology(e.g.,Parker,Smith,&Hogan,1992) likelyhavesimilareffects.Thestudysiteforthisproj-ecthasbothhighlightningfrequencyandarelativelyhighconcentra-tionofmetaltowersextending>10mabovethesurroundingcanopy.Ourobservationsaretoopreliminarytoevaluatetheeffectsofthesetowers on the frequency of lightning damage to nearby trees, butthemonitoringsystemdescribedinthisstudyisgeneratingdatathateventuallywillresolvethisissue.

5  | CONCLUSIONS

Here,wedemonstratearelativelysimpleandinexpensivesystemthatprovidesaccuratelightningstrikelocationandflashcharacteristicin-formationinrealtime(withinminutestohours)andoverlargespatialscales.Consequently,thissystemprovidesanunprecedentedoppor-tunity to track the effects of both conspicuous and inconspicuouslightningdamage,thusenablingstructuredandunbiasedtestsofhy-pothesesrelatedtotheecologicaleffectsoflightningforthefirsttime.Moreover,bypairingthissystemwithawell-studiedforest,suchasaforestdynamicsplot,itispossibletoconductlong-termanddetailedinvestigations using historic information aboutmonitored individualtrees.Becausetheecologicaleffectsoflightningarelargelyunstudied,weexpectthatincreaseduseofthisandsimilarsystemswillleadtotheinitiationofmanyunanticipatedavenuesforfutureresearch.

TABLE  3 Exampleresearchquestionsthatcouldbeansweredatthreedifferentlevelsoforganizationgivensufficientlightningstrikelocationdataatthelandscapescale

Individual & Population

1.Aresometreespeciesmoreattractivetolightningthanothers?

2.Dosometreeshavetraitsthatresistthedamagingeffectsoflightning(Goraetal.unpub.)?

3.Howdononlethallightningstrikesaffecttreefitnesscorrelates(e.g.,lifespan,totalreproductiveoutput,herbivoredamage)?

4.Istherearelationshipbetweentreeconditionatthetimeofastrikeandtheamountofdamagethatoccurs?

5.Doeslightning-causedmortalitymediatecompetitionamongtrees?

Community

1.Whatistheroleoflightningdisturbanceingap-phaseforestdynamics(Brokaw,1985)?

2.Dolightningstrikesmodifythestructureoflocalsoilmicrobialcommunities?

3.Whataretheeffectsoflightningstrikesonnontreeforestcomponents(e.g.,lianas,epiphytes,endophyticfungi)?

4.Dolianasprotecttreesagainstlightningdamage(Yanoviak,2013;Goraetal.unpub.)?

5.Whataretheeffectsoflightningstrikesontreeinhabitants,especiallyants?

6.Arelightning-damagedtreesspecificallyattractivetocertaintypesoffungiorherbivores?

Ecosystem

1.Howmuchcarbon(ascoarsewoodydebris)isproducedbylightningperhectare?

2.Howdoestheecologicalimportanceoflightningchangewithecosystemtype?

3.Dolightninggapsdifferfromothertypesofdisturbanceintermsoflocalenvironmentalconditions?

4.Doestheamountoflightningdamagevarypredictablywithlandscape-scalechangesinsoilmoistureorcomposition?

5.Howdotallmetaltowersandotherstructuresaffectthelandscape-scaledistributionoflightningdamage?

     |  11YANOVYAK et Yal.

ACKNOWLEDGMENTS

BenjaminAdams,RobertaEthington,NoahGripshover,RileyKneale,andAlyssaStarkassistedinthefield.WethankOrisAcevedo,MelissaCano,CesarGutierrez,BelkysJimenez,andthestaffoftheSmithsonianTropicalResearchInstituteforlogisticalsupport.Commentsfromtwoanonymous reviewers improved themanuscript. This research wassupportedbyNSFgrantsDEB-1354060 toSPYandDEB-1354510toPMB,andNSFgrantGRF-2015188266andaNationalGeographicSocietygranttoEMG.

CONFLICT OF INTEREST

Nonedeclared.

AUTHOR CONTRIBUTIONS

Allauthorsassistedwiththe installation, implementation,andmain-tenanceofthecameraand/orELSnetworks.SPYcollectedfielddata,wrotethemanuscript,andmanagedtheproject.EMGcollectedfielddata,processedvideos,andmanagedfieldoperations.JMBassistedinthedevelopmentofthecameraandELSnetworks,processedvideos,andcomposedrelevantcomputercode.PMBconceived,developed,andtestedtheELSnetwork;composedrelevantcomputercode;andmanagedtheproject.MDcollectedfielddata,processedvideos,andcomposedrelevantcomputercode.Allauthorscontributedcriticallytothedraftsandgavefinalapprovalforpublication.

REFERENCES

Albrecht,R.I.,Goodman,S.J.,Buechler,D.E.,Blakeslee,R.J.,&Christian,H.J.(2016).WherearethelightninghotspotsonEarth?Bulletin of the American Meteorological Society,97,2051–2068.

Anderson,J.A.R.(1964).ObservationsonclimaticdamageinpeatswampforestinSarawak.Commonwealth Forestry Review,43,145–158.

Andrews, C. J., Cooper, M. A., Darveniza, M., & Mackerras, D. (1992).Lightning injuries: Electrical, medical, and legal aspects (p. 208). BocaRaton,FL:CRCPress.

Anonymous, (1898). Which trees attract lightning? Monthly Weather Review,26,257–258.

Bitzer,P.M.(2017).Globaldistributionandpropertiesofcontinuingcur-rent in lightning. Journal of Geophysical Research: Atmospheres, 122,1033–1041.

Bitzer, P. M., Christian, H. J., Stewart, M., Burchfield, J., Podgorny, S.,Corredor,D.,…Franklin,V. (2013).CharacterizationandapplicationsofVLF/LFsourcelocationsfromlightningusingtheHuntsvilleAlabamaMarxMeterArray.Journal of Geophysical Research: Atmospheres,118,3120–3138.

Boccippio,D.J.,Cummins,K.L.,Christian,H.J.,&Goodman,S.J.(2001).Combined satellite-and surface-based estimation of the intracloud-cloud-to-ground lightning ratio over the continental United States.Monthly Weather Review,129,108–122.

Bohlman,S.,&O’Brien,S. (2006).Allometry,adultstatureandregenera-tionrequirementof65treespeciesonBarroColoradoIsland,Panama.Journal of Tropical Ecology,22,123–136.

Botley,C.M.(1951).Thelightningdischargeinmythandlegend.Weather,6,277–279.

Bouquegneau,C.,&Rakov,V.(2010).How dangerous is lightning?.Mineola,NY:DoverPublications.

Brokaw, N. V. L. (1985). Gap-phase regeneration in a tropical forest.Ecology,66,682–687.

Brünig,E.F.(1964).AstudyofdamageattributedtolightningintwoareasofShorea albidaforestinSarawak.Commonwealth Forestry Review,43,134–144.

Chave,J.,Réjou-Méchain,M.,Búrquez,A.,Chidumayo,E.,Colgan,M.S.,Delitti,W.B.C.,…Vieilledent,G.(2014).Improvedallometricmodelstoestimatetheabovegroundbiomassoftropicaltrees.Global Change Biology,20,3177–3190.

Christian,H.J.,Blakeslee,R.J.,Boccippio,D.J.,Boeck,W.L.,Buechler,D.E.,Driscoll,K.T.,…Stewart,M.F.(2003).Globalfrequencyanddistri-bution of lightning as observed from space by theOpticalTransientDetector.Journal of Geophysical Research,108,4005.

Cochrane, M. A. (2003). Fire science for rainforests. Nature, 421,913–919.

Coulson,R.N.,Hennier,P.B.,Flamm,R.O.,Rykiel,E.J.,Hu,L.C.,&Payne,T.L.(1983).TheroleoflightningintheepidemiologyoftheSouthernPineBeetle.Zeitschrift für angewandte Entomologie,96,182–193.

Croat,T.B. (1978).Flora of Barro Colorado Island. Stanford,CA: StanfordUniversityPress.

Cummins,K.L.,&Murphy,M.J.(2009).Anoverviewoflightninglocatingsystems:History,techniques,anddatauses,withanin-depthlookattheU.S.NLDN.IEEE Transactions on Electromagnetic Compatibility,51,499–518.

DuCharme,E.P. (1972).Lightninganddeclineofcitrustrees inFloridagroves. Proceedings of the Florida State Horticultural Society, 85,80–84.

Fernando,M.,Mäkelä, J.,&Cooray,V. (2010). Lightning and trees. InV.Cooray(Ed.),Lightning protection(pp.843–858).London:InstitutionofEngineeringandTechnology.

Franklin, B. (1769). Experiments and observations on electricity, made at Philadelphia in America.London:DavidHenry.

Fuquay,D.M.,Taylor,A.R.,Hawe,R.G.,&Schmid,C.W.Jr(1972).Lightningdischargesthatcausedforestfires.Journal of Geophysical Research,77,2156–2158.

Furtado,C.X. (1935).Lightning injuriestotrees.Journal of the Malaysian Branch of the Royal Asiatic Society,13,157–162.

Goldammer,J.G.,&Price,C.(1998).PotentialimpactsofclimatechangeonfireregimesinthetropicsbasedonMAGICCandaGISSGCM-derivedlightningmodel.Climatic Change,39,273–296.

Gora,E.M.,&Yanoviak,S.P.(2015).Electricalpropertiesoftemperatefor-esttrees:Areviewandquantitativecomparisonwithvines.Canadian Journal of Forest Research,45,236–245.

Hubbell,S.P.,&Foster,R.B. (1983).Diversityofcanopytrees inaneo-tropicalforestandimplicationsforconservation.InS.L.Sutton,T.C.Whitmore,&A.C.Chadwick(Eds.),Tropical rain forest: Ecology and man-agement(pp.25–41).Oxford:BlackwellScientific.

Kays, R., Tilak, S., Crofoot, M., Fountain, T., Obando, D., Ortega, A., …Wikelski,M.(2011).Trackinganimallocationandactivitywithanau-tomatedradiotelemetrysysteminatropicalrainforest.The Computer Journal,12,bxr072.

Krider, E. P. (1992).On the electromagnetic fields, Poyntingvector, andpeakpowerradiatedbylightningreturnstrokes.Journal of Geophysical Research,97(D14),15913–15916.

Leigh,E.G.Jr,Rand,A.S.,&Windsor,D.M.(1996).The ecology of a tropical forest.Washington,DC:SmithsonianInstitution.

Magnusson,W.E.,Lima,A.P.,&deLima,O.(1996).Grouplightningmor-tality of trees in aNeotropical forest. Journal of Tropical Ecology,12,899–903.

Mäkelä,J.,Karvinen,E.,Porjo,N.,Mäkelä,A.,&Tuomi,T.(2009).Attachmentofnaturallightningflashestotrees:Preliminarystatisticalcharacteris-tics.Journal of Lightning Research,1,9–21.

Mäkelä,A.,Mäkelä,J.,Haapalainen,J.,&Porjo,N.(2016).TheverificationoflightninglocationaccuracyinFinlanddeducedfromlightningstrikestotrees.Atmospheric Research,172–173,1–7.

12  |     YANOVYAK et Yal.

Malhi,Y.,Baldocchi,D.D.,&Jarvis,P.J.(1999).Thecarbonbalanceoftropical,temperate,andborealforests.Plant, Cell & Environment,22,715–740.

Montgomery,R.A.,&Chazdon,R.L. (2001).Foreststructure,canopyar-chitecture,andlighttransmittanceintropicalwetforests.Ecology,82,2707–2018.

Norris,J.R.,Allen,R.J.,Evan,A.T.,Zelinka,M.D.,O’Dell,C.W.,&Klein,S.A.(2016).Evidenceforclimatechangeinthesatellitecloudrecord.Nature,536,72–75.

O’Brien, S.T.,Hubbell, S. P., Spiro, P.,Condit, R.,&Foster, R.B. (1995).Diameter,height,crown,andagerelationshipineightNeotropicaltreespecies.Ecology,76,1926–1939.

Orville, R. E. (1968). Photographof a close lightning flash.Science,162,666–667.

Outcalt,K.W.(2008).Lightning,fireandlongleafpine:Usingnaturaldis-turbancetoguidemanagement.Forest Ecology and Management,255,3351–3359.

Parker,G.G.,Smith,A.P.,&Hogan,K.P.(1992).Accesstotheupperforestcanopywithalargetowercrane.BioScience,42,664–670.

Perry,D.R.(1978).Amethodofaccessintothecrownsofemergentandcanopytrees.Biotropica,10,155–157.

Price,C.(2009).Willadrierclimateresultinmorelightning?Atmospheric Research,91,479–484.

Price,C.,&Rind,D.(1993).Whatdeterminesthecloud-to-groundlightningfractioninthunderstorms?Geophysical Research Letters,20,463–466.

Price,C.,&Rind,D.(1994a).Theimpactofa2XCO2climateonlightning-causedfires.Journal of Climate,7,1484–1494.

Price,C.,&Rind,D.(1994b).Possibleimplicationsofglobalclimatechangeongloballightningdistributionsandfrequencies.Journal of Geophysical Research,99,10823–10831.

Rakov,V.A.,&Uman,M.A.(2003).Lightning: Physics and effects.Cambridge:CambridgeUniversityPress.

Romps,D.M.,Seeley,J.T.,Vollaro,D.,&Molinari,J.(2014).Projectedin-creaseinlightningstrikesintheUnitedStatesduetoglobalwarming.Science,346,851–854.

Sánchez-Murillo,R.,Durán-Quesada,A.M.,Birkel,C.,Esquivel-Hernández,G.,&Boll,J.(2017).Tropicalprecipitationanomaliesandd-excessevolutionduringElNiño2014-16.Hydrological Processes,31,956–967.

Stall,C.A.,Cummins,K.L.,Krider,E.P.,&Cramer,J.A.(2009).Detectingmultiplegroundcontactsincloud-to-groundlightningflashes.Journal of Atmospheric and Oceanic Technology,26,2392–2402.

Stone,G.E.(1903).Injuriestoshadetreesfromelectricity.Hatch Experiment Station of the Massachusetts Agricultural College,91,1–21.

Stone,G.E.(1916).Shadetrees,characteristics,adaptation,diseasesandcare. Bulletin of the Massachusetts Agricultural Experiment Station,170,123–264.

Taylor, A. R. (1964). Lightning damage to forest trees in Montana.Weatherwise,17,61–65.

Taylor,A.R.(1965).Diameteroflightningasindicatedbytreescars.Journal of Geophysical Research,70,5693–5695.

Taylor,A.R.(1974).Ecologicalaspectsoflightninginforests.Proceedings of the Tall Timbers Fire Ecology Conference,13,455–482.

Tutin,C.E.G.,White,L.J.T.,&Mackanga-Missandzou,A.(1996).LightningstrikeburnslargeforesttreeintheLopeReserve,Gabon.Global Ecology and Biogeography Letters,5,36–41.

USDAForestService (1999).Forest health monitoring 1999 field methods guide. Research Triangle Park, NC: USDA Forest Service, NationalForestHealthMonitoringProgram.

Wadsworth,F.H. (1943). Lightningdamage inponderosapine standsofnorthernArizona.Journal of Forestry,41,684–685.

Williams, E. R. (1992).The Schumann resonance:A global tropical ther-mometer.Science,256,1184–1187.

Williams, E. R. (2005). Lightning and climate: A review. Atmospheric Research,76,272–287.

Yanoviak, S. P. (2013). Shockvalue:Are lianas natural lightning rods? InM.Lowman,S.Devy,&T.Ganesh(Eds.),Treetops at risk: Challenges of global forest canopies(pp.147–154).NewYork,NY:Springer.

Yanoviak,S.P.,Gora,E.M.,Fredley,J.,Bitzer,P.M.,Muzika,R.-M.,&Carson,W.P.(2015).Directeffectsoflightningintemperateforests:Areviewandpreliminarysurveyinahemlock-hardwoodforestofthenorthernUnitedStates.Canadian Journal of Forest Research,45,1258–1268.

SUPPORTING INFORMATION

Additional Supporting Information may be found online in thesupportinginformationtabforthisarticle.

How to cite this article:YanoviakSP,GoraEM,BurchfieldJM,BitzerPM,DettoM.Quantificationandidentificationoflightningdamageintropicalforests.Ecol Evol. 2017;00:1–12. https://doi.org/10.1002/ece3.3095