Choristoneura Fumiferana Clem.) Louis-Etienne Robert Universite De Montreal

Natural Resource Ecology and Management Natural Resource Ecology and Management Publications

2017 Landscape host abundance and configuration regulate periodic outbreak behavior in spruce budworm (Choristoneura fumiferana Clem.) Louis-Etienne Robert Universite de Montreal

Brian R. Sturtevant U.S. Department of Agriculture

Barry J. Cooke Ontario Ministry of Natural Resources and Forestry

Patrick M. A. James Universite de Montreal

MFoallorie-Jw othisees a Fndor taindditional works at: http://lib.dr.iastate.edu/nrem_pubs UnivPerasitrty ofof TtheoronEtonvironmental Indicators and Impact Assessment Commons, Forest Management Commons, Natural Resources and Conservation Commons, and the Natural Resources See next page for additional authors Management and Policy Commons The ompc lete bibliographic information for this item can be found at http://lib.dr.iastate.edu/ nrem_pubs/248. For information on how to cite this item, please visit http://lib.dr.iastate.edu/ howtocite.html.

This Article is brought to you for free and open access by the Natural Resource Ecology and Management at Iowa State University Digital Repository. It has been accepted for inclusion in Natural Resource Ecology and Management Publications by an authorized administrator of Iowa State University Digital Repository. For more information, please contact [email protected]. Landscape host abundance and configuration regulate periodic outbreak behavior in spruce budworm (Choristoneura fumiferana Clem.)

Abstract Landscape-level forest management has long been hypothesized to affect forest insect outbreak dynamics, but empirical evidence remains elusive. We hypothesized that the combination of increased hardwood relative to host tree species, prevalence of younger forests, and fragmentation of those forests due to forest harvesting legacies would reduce outbreak intensity, increase outbreak frequency, and decrease spatial synchrony in spruce budworm (Choristoneura fumiferana Clem.) outbreaks. We investigated these hypotheses using tree ring samples collected across 51 sites pooled into 16 subareas distributed across a large ecoregion spanning the international border between Ontario (Canada), and Minnesota (USA). This ecoregion contains contrasting land management zones with clear differences in forest landscape structure (i.e., forest composition and spatial configuration) while minimizing the confounding influence of climate. Cluster analyses of the 76-years time-series generally grouped by subareas found within the same land management zone. Spatial nonparametric covariance analysis indicated that the highest and lowest degree of spatial synchrony of spruce budworm outbreaks were found within unmanaged wilderness and lands managed at fine spatial scales in Minnesota, respectively. Using multivariate analysis, we also found that forest composition, configuration, and climate together accounted for a total of 40% of the variance in outbreak chronologies, with a high level of shared variance between composition and configuration (13%) and between composition and climate (9%). At the scale of our study, climate on its own did not explain any of the spatial variation in outbreaks. Outbreaks were of higher frequency, lower intensity, and less spatially synchronized in more fragmented, younger forests with a lower proportion of host species, with opposing outbreak characteristics observed in regions characterised by older forests with more concentrated host species. Our study is the first quantitative evaluation of the long-standing "silvicultural hypothesis" of spruce budworm management specifically conducted at a spatio-temporal scale for which it was intended.

Keywords spruce budworm; harvest disturbance; landscape ecology, forest management legacies, dendrochronology, outbreak synchrony

Disciplines Environmental Indicators and Impact Assessment | Forest Management | Natural Resources and Conservation | Natural Resources Management and Policy

Comments This is a manuscript of an article published as Robert, L.-E., Sturtevant, B. R., Cooke, B. J., James, P. M. A., Fortin, M.-J., Townsend, P. A., Wolter, P. T. and Kneeshaw, D. (2017), Landscape host abundance and configuration regulate periodic outbreak behavior in spruce budworm (Choristoneura fumiferana Clem.). Ecography. doi: 10.1111/ecog.03553.

Rights Works produced by employees of the U.S. Government as part of their official duties are not copyrighted within the U.S. The onc tent of this document is not copyrighted.

This article is available at Iowa State University Digital Repository: http://lib.dr.iastate.edu/nrem_pubs/248 Authors Louis-Etienne Robert, Brian R. Sturtevant, Barry J. Cooke, Patrick M. A. James, Marie-Josee Fortin, Philip A. Townsend, Peter T. Wolter, and Daniel Kneeshaw

This article is available at Iowa State University Digital Repository: http://lib.dr.iastate.edu/nrem_pubs/248 7 II.Iowa Ames, USA. 50011 6 5 Street. Toron 4 Ontario 3 Service,K. Rhinelander, WI 5450 5985,Highway 2 Centre 1 FortinJosée Louis spruce fumiferana budworm (Choristoneura configuration and abundance host Landscape as [10.1111/ doi: c Please Record. of Version the and version this between differences to lead may which process, proofreading and pagination typesetting, copyediting, the through been not has but review peer full undergone and publication for accepted been has article This Decision date: H3C 3J7. Centre succursale 6128, C.P. Montréal, de Université author Corresponding Centre ete ’td d l frt CF. nvrié u ubc à Québec du Université (CEF). forêt la de d’étude Centre Labs,Russell Linden 1630 25 Toronto. of University Biology, Evolutionary and Ecology of Department Forestry and Resources Natural of Ministry Ontario succursale 6128, C.P. Montréal, de Université biologiques, sciences de Département Accepted Management. and Ecology Resource Natural of Dept. University, State Iowa Article Forest USDA Station, Research Northern Studies, Ecosystem Applied for Institute - Etienne Robert Etienne - - ville. ville. Ville. , Canada.

Email: 4 , Philip A.Townsend to, Ontario, Canada.to, Ontario,3B2 M5S Mo Mo 13 ntréal, Québec, Canada.H3C3J7 ecog.03553 ntréal, Québec,H3C 3P8 Canada. louis P6A 6V8 P6A - ‘This article is protected by copyright. All rights reserved.’ Nov 1 - , Brian R. Sturtevant R. Brian , [email protected] - 2017 :

Louis

Drive, Madison, USA. WIDrive, Madison, 53706 USA. ]

- ten Robert Etienne 5 , Peter T.Wolter 2

, Barry J. Cooke BarryJ. ,

Clem.

regulate

6 Département de sciences biologiques, biologiques, sciences de Département

, Daniel Kneeshaw ) . ,

- ville. ville. 0 otr Dr Foster 70

periodic outbreak behavior in behavior outbreak periodic 3 ,

Patrick M. A. James A. M. Patrick Montréal, C.P. 8888, Succ. Succ. 8888, C.P. Montréal, otél Qée, Canada. Québec, Montréal, 7 . ,

al Se Marie Ste. Sault ite this article this ite 339 Sc 339 Willcocks 1 , Marie ience - ,

accounted for a total of 4 of total a for accounted analysis fine at managed lands and spa S time minim while structure landscape forest in differences contains ecoregion This (USA). Minnesota ecoregion large using treering samplescollected across 51sites budworm in synchrony spatial decrease and frequency, outbreak increase intensity, outbreak forests those of fragmentation and forests, hardw increased of combination elusive. remains evidence empirical but dynamics, outbreak Landscape Abstract legacies, dendrochronology, outbreaksynchrony. Keywords conducteda spatio at long the concentrated more with forests species synchronized spatially outbreak in variation spatial (9%). climate and composition between and (13%) configuration and composition between variance shared

Accepted covariance nonparametric patial Article tial synchrony tial - series , with opposing outbreak characteristics observed characteristics outbreak opposing with , w as found also we , -

tnig sliutrl yohss o src bdom aaeet specifically management budworm spruce of hypothesis” “silvicultural standing (Choriston - : spruce budworm; harvest disturbance; landscape ecology, forest management forest ecology, landscape disturbance; harvest budworm; spruce : ee frs mngmn hs og en yohszd o fet oet insect forest affect to hypothesized been long has management forest level generally grouped generally zn h ofudn nlec f climate. of influence confounding the izing of pnig h itrainl odr ewe Otro Cnd) and (Canada), Ontario between border international the spanning At the scale of our study, climate study, our of scale the At ‘This article is protected by copyright. All rights reserved.’ - eura fumiferana Clem.) fumiferana eura spruce budworm outbreaks budworm spruce temporal scaletemporal

in more fragmented, younger forests with a lower proportion of host of proportion lower a with forests younger fragmented, more in 0 % of the variance in outbreak chronologies, with a high level of level high a with chronologies, outbreak in variance the of % that forest composition, configuration, and climate together together climate and configuration, composition, forest that s . spatial

Outbreaks were of higher frequency higher of were Outbreaks

by host species. host o rltv t hs te seis peaec o younger of prevalence species, tree host to relative ood

subareas analysis analysis scales for wh for

(i.e., forest co forest (i.e.,

ich it was intended. ich it n Minnesota in indicated

found within the same the within found due to to due contrasting land management zones zones management land contrasting Our study is the first quantitative evaluation of of evaluationquantitative first the is study Our outbreaks.

pooled into16subareas

were found were forest harvest forest that on its own its on mposition and mposition

respectively. , We investigated these hypotheses these investigated We

the highest and lowest degree of degree lowest and highest the in C luster analyses of the 76 the of analyses luster regions characterised by older by characterised regions within unmanaged wilderness wilderness unmanaged within

e yohszd ht the that hypothesized We ing did not explain explain not did ,

lower intensity lower land management zone management land

legacies would reduce would legacies spatial

distributed distributed sn m Using

configuration) any ultivariate with clear with , and less less and , across a

spruce spruce - of the the of year s .

(Berryman 2002, Turchin 2003) Turchin 2002, (Berryman top the Although 2009) al. et (Haynes ( caterpillar tent forest ( budworm pine jack the including systems, 2008) forest of concepts formalizeddynamics earliest the to central was periodic drives eruption partially population least at insects and forest between feedback reciprocal that notion The disturbances such as fire succession structure scale global a at climate affect influence can which of can all cycling, carbon and atmospheric and hydrology, processes cycling, nutrient biogeochemical affect outbreaks Insect consequences. economic disturbanceforestmajorhave thata areinsects boreal forest Outbreaks of Introduction outbreaks broad the for known native glauca defoliator T there are important interactions among levels trophic affecting outbreak dynamics. landscap forest to linked be to appears irruptions population to respond to communities enemynatural of ability the Further, species ( budmoth its and deadwood available example, For bottom to attributable

Acceptedhe Article pue budworm, spruce . An important role of the forest on outbreaks has been demonstrated in multiple defoliator multiple in demonstrated been has outbreaks on forest the of role important An .

insect is a strong flier strong a is insect

Mec) Voss (Moench) induce

(i.e., (Blais 1983, Jardon et al.2003, Royama 1984, Williams and Liebhold 2000) (Holling 1973) (Holling in North America North in erpea griseana Zeiraphera Bsevle 1975) (Baskerville (Kurz et al. 2008, Dymond e Dymond 2008, al. et (Kurz d impact stan

epicenter formation and travelling waves within outbreaks within waves travelling and formation epicenter

- on fet of effects down d

- ‘This article is protected by copyright. All rights reserved.’ , and winter moth ( moth winter and , scale spatial synchrony spatial scale species - up host up f h Erpa src br bel ( beetle bark spruce European the of big seily vulnerable especially being ) e conditions e aaooa disstria Malacosoma , and continues to be a central theme in ecosystems ecology ecosystems in theme central a be to continues and , hrsoer fumiferana Choristoneura (James et al.2017) spatial spatial , with balsam fir ( fir balsam with , opsto and composition - capable of long distance dispersal distance long of capable tme production timber , plant effects plant , there is also evidence that some of this outbreak behaviour maybe outbreakbehaviour this of some evidencethat also is there , )

n the in

distribution aua eeis yial udri cci otra behaviour outbreak cyclic underlie typically enemies natural (Cappuccino et al. 1998, Roland and Taylor 1997) Taylor and Roland 1998, al. et (Cappuccino Operophtera brumata Operophtera European t al. 2010) al. t (Price et al. 1984, Nealis and Lomic 1994, White 20 White 1994, Lomic and Nealis 1984, al. et (Price . , comparatively low frequency, and long duration long and frequency,low comparatively ,

Choristone

) spatial Abies balsamea Abies (Lausch et al. 2013) al. et (Lausch Rln 2005) (Roland

ls where Alps, Cag t l 2012) al. et (Chang .

Clem.,

Mcen 90 Hnia e a. 2008) al. et Hennigar 1980, (MacLean Insect outbreaks also affect forest landscape forest affect also outbreaks Insect configuration) ura pinus ura is is p typographus Ips )

gpy oh ( moth gypsy , (L.) Mill (L.) h ms dmgn frs insect forest damaging most the (Wesołowski and Rowiński 2006) Rowiński and (Wesołowski ) gradients in density of of density in gradients (Greenbank (Volney and McCullough 1994) McCullough and (Volney .

Another example is the the is example Another ih osqecs o forest for consequences with ) and white spruce ( spruce white and )

ad h rs o future of risk the and , significantecologicaland (Johnson et al. 2004) al. et (Johnson et al. 1980) al. et

) is is ) yati dispar Lymantria -

insect insect determined .

(Raffa et al. et (Raffa , suggesting, ecosystem

ot tree host and .

Picea of its of larch This best 12)

by ) . . . ,

nrae audne f e prsti seis associat species parasitoid key of abundance increased 1996) al. et Su 2013, Gray 2008, al. et Campbell 1995, budworm on fir balsam damage broad that and strong,althoughwas 1983) concentrated that regional 20 the in budworm spruce the Inpar hypothesized 1999) al. et principle a equilibrium”modelsof cause the Th processes outbreaks 1953 (Moran synchro density delayed by produced by century a nearly for 2016) al. et debate Pureswaran empirical intense of subject a been has top of importance relative The outbreaks species) tree nonhost vs host of arrangement spatial and convincing include to hypothesis 2003) al. et Quayle 1998, al. et (Cappuccino

Accepted Articlebudworm spruce e Royama (1984) Royama . allel, t allel, Baskerville , y s rdcd y regionally by produced is ny -

scale commercial harvesting of spruce of harvesting commercial scale emphasized the lack of supporting evidence supporting of lack the emphasized .

(Cooke al.et 2007, Régnièreand Nealis 2007, Sturtevant et al.2015) defoliation defoliation

may emerge from a combination of combination a from emerge may f cross of bottom evidence for the for evidence he “silviculturalhypothesis”spruceofbudworm management Wa ws isn from missing was What .

, effect of the forest on the budworm.

budworm oaa 2005) Royama - cld ocnrtos f aue asm i wr mr vleal t budworm to vulnerable more were fir balsam mature of concentrations scaled preferred - , ( up

1975) - who argued that the core populationunderlyingcoreoscillationoutbreaksthe argued who isperiodic that cl eutv dynamics eruptive scale ‘This article is protected by copyright. All rights reserved.’ MillerRusnockand

consistently the effect of effect the driver enabling budworm escape from escape budworm enabling driver inspired much of the foundational thinking about forest about thinking foundational the of much inspired spruce .

outbreak The argue host abundance and increased outbreak extent and intensity and extent outbreak increased and abundance host

- penultimate dependent mortality due to natural enemies, and that spatial outbreak spatial that and enemies, natural to due mortality dependent ) idea th . budwormhypothesizedoutbreaksfoabundance host that of - d

down versus bottom versus down H century was accentuated by “man’s influence in the forest” the in influence “man’s by accentuated was century

that the reciprocal feedback between feedback reciprocal the that

wvr m owever, tend forest composition on natural enemy interactions. S enemyinteractions.natural on composition forest

dynamics of forest of - s orltd atr sc a wahr ie, h Mrn effect Moran the (i.e., weather as such factors correlated

these to be lower lower be to

( effect of forest landscape structure landscape forest of effect 1993) - . driven feedback to to feedback driven r rcn rsac sget that suggests research recent ore al models early M Ldi e a. 1978) al. et (Ludwig . , whoforestbudworm, of impact accepted the the on

both top both r recent ore This has has This (

Picea

when - up up

for any influence of of influence any for

spp.) - processes , and , onad bottom and down hardwoods are abundant are hardwoods led

empirical was on ed top , and ,

o bodnn o te silvicultural the of broadening a to (2) (2)

the budworm outbreaks was outbreaks budworm - robust down ih hs adod component hardwood this with

this effect is due is effect this in budworm outbreak dynamics outbreak budworm in its replacement by replacement its dynamics .

studies

ned t Indeed, control control the the miia eiec fr the for evidence empirical

posited that the impactthe posited that of Sutvn e a. 2015, al. et (Sturtevant budworm and budworm - up recipr up such concentration such (i.e., the age, amount, age, the (i.e., ha .

- f pue budw spruce of (Jones 1977, Hassell1977, (Jones

ope cycling complex insect feedback insect

e al “multiple early he ve

(Bergeron et al. et (Bergeron hw that shown ,

ocal feedback ocal in part, in till missingi till fir balsam (Blais 1965, (Blais dismissed the the liage wasliage – forest to an to :

orm s of s s

i.e.,

(1) of as – s

oet land forest th extend we paper, this In silvicultural hypothesis. synchronized less but frequent, more were that younger (wilderness) forests fragmented forest a with aggregated. spatially than rather diffuse was harvesting particular,found they that outbreak affect to tended century data ring one to related functionally are another that characteristics outbreak of response multivariate coherent 1995) Lysyk through synchrony, 2007) al. et Eveleigh 1984, (Royama on effects its through cover forest host reducing Considering AcceptedMethods hypothesis more generally 2011) Sturtevant and Anderson 1980, al. et (Greenbank dynamics long the 1996) al. et Su 2013, Gray 2005, Fleming and Candau 2008, al. et Campbell effects forest of from outbreaks above outlined and frequency, cycle host rates, (disturbance including hypothesize Article ,

- and

term forests managed rm nai ad Minnesota and Ontario from at scale at forest forest cp srcue on structure scape

the the modified

n cce amplitude cycle and that landscape ie, vr utpe outbreaks) multiple over (i.e., . the effects of effects the landscape full We opsto (ot bnac, adod otn, oet age) forest content, hardwood abundance, (host composition s attributes of

effect

local fewer trees fewer connectivity relevant tutr feedback structure containing a greater proportion of proportion greater a containing structure onthe characteristics asingle of outbreak (Miller and Rusnock 1993) e might

study

synchrony climate on uwr dsesl success dispersal budworm forest landscape structure related to landscape management practices, the the spatio exhibited infrequent, spatially synchronous outbreaks. In contrast, In outbreaks. synchronous spatially infrequent, exhibited long to tha to affect of Robert et al et Robert of evidence intensitywas in managed landscapes than i managedthanlandscapes in Bror 90 Coe t l 2007) al. et Cooke 1990, (Barbour )

. One might also expect a reduction in host to to host in reduction a expect also might One . , (Gray 2008, 2013) 2008, (Gray -

o dis to - range influence budworm outbreak behavior outbreak budworm influence eprl aito in variation temporal ) t of dispersing of t

o hw ht pue uwr otras uig h 20th the during outbreaks budworm spruce that show to the according to the the to according

, both empirical and theoretical and empirical both , igih h efcs f forest of effects the tinguish frequency and frequency dispers to

effect landscape ubek dynamics outbreak

reducedareaspatternwhereforestspatialof the in . –

(201

al success al l i acrac wt te rdcin o the of predictions the with accordance in all . s

spruce budworm spruce 2 f forest of

. Although other studies other Although . The )

to quantitatively to expectations structure. Nai 2016) (Nealis intensity

se hardwood species hardwood n cnitn wt te silvicultural the with consistent and

(Peltonen et al. 2002, Régnière and Régnière 2002, al. et (Peltonen uwr otra dynamics outbreak budworm results were qualitatively consistent qualitatively were results n unharvestedwilderness n adcp structure landscape O .

Robert et al. et Robert of budworm outbreak cycles outbreak budworm of

tras ihn le, less older, within utbreaks

of a multi a of (i.e., (i.e., .

n on and hs ol rpeet a represent would This , investigate , landscape one our goal is to examine to is goal our (Bergeron et al. 1995,

10’s to 100’s of 100’s to 10’s (i.e.,

exhibited outbreaks exhibited have examined theexamined have n configuration and would - ( variate response variate 2012) enemies natural cycle intensity, cycle reduce n outbreak on on structure the effect of effect the expect that expect

used

areas

spatial .

tree km; We . In . - ,

ewe hs mngd ein le n prxmtl 1 approximately an lies regions managed these Between harvestofratesthe though forests Canadian managed border. the of sides both harvestingmechanizedon of advent the with coinciding ( pines” “big Logging the of harvest selective on focused and century 19th the entire The strobus Michx. ( fir balsam Lamb 1973) (Heinselman edge potential address to buffer movement 50km a includes BLL the of extent transition 1 (Fig. (Canada) Ontario and (USA) Minnesota between border (BLL) Landscape Lakes Border The Study area Acceptedcut consistent least and zones, managed patch disturbance 2012) al. documented been has BLL the in zones management between disturbances of patterning spatial the in Divergence Forest l balsamkills fir 2001) henceforth t no where Minnesota AreaWildernessin Canoe BoundaryWaters the and Ontario in ParkProvincial Quetico Article - blocks ), ), , enhancing succession toward spruce and fir and spruce toward succession enhancing , )

) L. black , as well as as well as ,

atrs iegd hrl bten Minne between sharply diverged patterns ( andscape structure of the study area ), red), pine ( between the Great Lakes Great the between Appendix

BLL as “Wilderness” as were A me hret a ocre sne h ery 1970s early the since occurred has harvest imber Adro ad treat 2011) Sturtevant and (Anderson .

spruce ( spruce balsamea in thisin region

shares a common early common a shares oh ml i size in small both

- bya time size distributions size

with a high proportion of boreal of proportion high a with several species near the northern limit of their range, such as white pine ( pine white as such range, their limitof northern the near species several ‘This article is protected by copyright. All rights reserved.’ 1 Pinus resinosa A) Picea mariana Picea ), .

Prior analyses of these data these of analyses Prior birch ( birch are are have been historic been have - .

series of series Forest fire Forest (Frelich and Reich 1995) an order of magnitude larger in size than managed American forests,Americanmanaged than size in larger magnitude of order an Betula papyrifera Betula - in the BLL were the most consistent through time in Minnesotain timethrough consistent most the were BLL the in

St. Lawrence mixed Lawrence St. Ait spatially and temporally and spatially Landsat n more and

contains (Mill.) B.S.P), (Mill.) s .

logging ), and), red maple ( have decreased over the last century last the over decreased have - ally derived lre ( large a diffuse .

history. Region history. similar

oet opsto i mxd na boreal” “near mixed is composition Forest . P . Marsh.), oa n Otro prxmtl 7 approximately Ontario and sota . tre white spruce white land cover mapsfromcover land (1975

eriodic defoliation by spruce budworm often budworm spruce by defoliation eriodic

hn the than e ~ (James et al. 2011, Sturtevant et al. 2014)(JamesSturtevantal.al. 2011,et et - 20 000 km 000 20 wood and boreal and wood 00 km 000 0 species Acer rubrum Sturtevant et al. (2014) al. et Sturtevant trembling aspen ( aspen trembling within - wide aggregated ; (e.g., Hisla 1996) (Heinselman ( Sutvn e a. 2014) al. et (Sturtevant 2 Picea 2

the the ) wilderness area wilderness

logging

ecoregion L. jack pine ( pine jack Wilderness ). ). - P fet rltd o insect to related effects glauca

forest regions. forest

strobus (Ahlgren and Ahlgren and (Ahlgren large

Populus tremuloides Populus

started at the end of end the at started

Cut

- that straddles that

(Moench) Voss (Moench) 2000, Pinus banksiana Pinus showed that showed - sized ; (2) ; , -

block sizes in sizes block rfre to referred , P that yas ago, years 0 .

Minnesota (Wolter et(Wolter resinosa c ut The full The includes )

- t the at Pinus block :

the (1) ). ), , .

o rvd sfiin sml sz ( size sample sufficient provide to site persampled meter. 1 of height a at tree spruce each from taken were cores Two s white canopy large old, fifteen surviving probabilityof because sites 51 from sampled cores tree using histories outbreak budworm spruce reconstructed We Sampling deciduous forest relative to contain Minnesota) and (Ontario landscape fir) and zones management imagery. TM ( sensin remote via mapped also was area) basal species tree of exception the with rare, Ontario in patterns noprtd Bomil, e Yr, S) Wdh wr cross were Widths USA). York, New Bloomfield, Incorporated, uni Velmex al. et Robert methods the using data ring tree spruce from reconstructed were outbreaks Budworm Outbreak history using progressively finer grit 150,(80, and 220). local DBH) cm 30 > pine; red and pine jack (i.e., km 15 area study the across gradients climatic for “W”), = West (O region sub 16 created Accepted ArticleAppendix climatic variation. All cores were stored in plastic straws and were later mounted and sande and mounted later were and straws plastic in stored were cores All variation. climatic

n temxmm distance maximum the and , especially , it ntario=“O”, 1 s as is - slide measuring table with an accuracy of 0.001 mm connected to a computer (Velmexcomputer a to mmconnected 0.001 of accuracyan with measuringtable slide B design ( ) and also and ) and latitud and 2012). aee analysis Wavelet area

(Table 1)(Table susceptible

s that represent a stratified sampling of the BLL based on forest management forest on based BLL the of sampling stratified a represent that s () n (3) ; between Minnesota and Minnesota between differed ‘This article is protected by copyright. All rights reserved.’

S reconstruction W ite attack circa ilderness e . Groups of three neighbouring sites werethreeneighbouringofaggregated“sub then into. Groups sites atural -

ee crnlge wr bit sn t using built were chronologies level (North =“N”, South =“S”) South =“N”, (North the the a few a somewhat in terms in somewhat

s asm i t dflain y pue budworm spruce by defoliation to fir balsam as distributed

(Hennigar et al. 2008)al. et (Hennigar 1985 ( 1985 Wilderness ,

pruce (>30 cm DBH) on flat or mid or flat on DBH) cm (>30 pruce

stand applied to the the to applied “” Minnesota=“M” =“W”, recent

n was 250 km 250 was

P.T. 1 tes fr ubek eosrcin Fg ) I ttl we total In 1). (Fig reconstruction outbreak for trees) 15 > - elcn dsubne in disturbances replacing

large wind large throughout the BLL (Fig. 1). 1). (Fig. BLL the throughout

Wolter, unpublished Wolter, (James al.et 2011) ed . The minimum distance between sub between distance minimum The . Wilderness

per subarea per similar forest types forest similar

of iiu fffen non fifteen of minimum A . 2002

the basal area of spruce budworm hosts (spruce hosts budworm spruce of area basal the and fire disturbances fire and . . The goal of this stratification was stratification this of goal The .

We targeted mesic sites that contained five tofivecontained that sitesmesic targeted We dataset

(

Appendix ) were also cored to serve as a control for control a as serve to cored also were longitud , g by g .

manuscript Jms t l 2011) al. et (James - ree ring analyses of white spruce tree spruce white of analyses ring , the

but otr t al. et Wolter - 1 (at “” Cnrl “C”, = Central “E”, = (East e ig its esrd sn a using measured widths ring B Wilderness Between 5 and 10 trees 10 and 5 Between - had ). The managed parts of the of parts managed The ). slope topographic positions. topographic slope - ae uig h program the using dated White spruce was selected was spruce White .

) using archived Landsat archived using ) more early more Forest composition (i.e., composition Forest - host coniferous trees coniferous host

but , area

( 2008) were

indicated - centroids was centroids successio has described in described

comparably

circa to higher a account areas

2002

were that nal, d ”

they would they reductions growth causing of capable are that region the in known insects defoliating rec 1996) Swetnam and (Holmes scale, and year given 2006) mean the from deviations standard1.28 than greater reduction a showed year one least at and years 5 than more for observed o tree, individual an of level the At used developto chronologies for each sub 2012) al. et Robert 2006, al. et reconstruction outbreak budworm spruce for standard 1985) cubic a with above errors. measurement COFECHA Acceptedspecies tree deciduous and spruce), and fir of estimates continuous 2012) al. et (Wolter land use data sensor Landsat temporal Forest Forest Predictor variables synch were shown outbrea detected validate to used only were species host to solely applied when efficient more Article ntuto ad a dtc otras hl aodn cnonig osqecs f other of consequences confounding avoiding while outbreaks detect can and onstruction used rony (described below). . Outbreak. cut response frequency 50% a to set were parameters Spline . . Index chronologies were calculated using the program ARSTANprogram the usingcalculated were chronologiesIndex . ). ). Landscape Structure landscape structure was structure landscape T oe tm series time cover he percentage affected trees trees affected percentage he as the response variable response the as (Holmes et al. 1986) al. et (Holmes be identified as an outbreak. an as identified be mohn pie odtedte eisadrmv age remove and series the detrend to spline smoothing

hronologies were hronologies ee eosrce uig hs prmtr ad h porm OUTBREAK program the and parameters these using reconstructed were and basal areas basal

chronology based on the percentage of trees classified as affected by an outbreak at a at outbreak affectedasclassifiedan bypercentage trees theof on based

maps ( previously been has OUTBREAK .

2) (Table 2) (Table

. Between 13 and 24 trees 24 and 13 Between . spatially

oet composition forest which was also u also was which t five at for cluster analysis cluster for (m utbreaks were defined when a when defined were utbreaks ring width ring 2 : by outbreak by

ha then quantified ( - area er nevl fo 1975 from intervals year 1) Outbreak detection with the program OUTBREAK was OUTBREAK program the with detection Outbreak . Composition was quantified at quantified was Composition . - 1 ) forest configuration forest aggregated into their respective sub respective their into aggregated s y iul oprsn f h got pten ( pattern growth the of comparison visual by ks

for for (Table 1).

balsam fir, combined budworm host species (i.e., species host budworm combined fir, balsam (Boulanger and Arseneault 2004, Bouchard et al. et Bouchard 2004, Arseneault and (Boulanger sed to locate missing or false rings and to test fortest to and rings false missing or locate to sed ec ya tete h st rsbra scales subarea or site the either at year each s in two ways, each each ways, two in (Bouchard et al. 2006) al. et (Bouchard a qatfe using quantified was , redundancy analysis, redundancy , (Boulanger and Arseneault 2004 Arseneault and (Boulanger

(mean = (mean was quantified using quantified was used used 19.8 trees; sd = 3.2 = sd trees; 19.8 tree – over - 2000 in spruce budworm outbreakbudworm spruce in using related growth trends growth related - - ring f o 6 years 60 of off (Cook 1985, Holmes 1992) Holmes 1985, (Cook

. No other defoliators aredefoliatorsother No . multiple . Non .

only use maps of of maps use ie, pit i time) in points 6 (i.e., separate sets of multi of sets separate and analysis of spatialof analysis and growth reduct growth

- two points in time in points two area host chronologies host years s

discrete as described as

trees) trees) , Bouchard , wih is which , such that such spatially ion was ion (Cook were land not -

age budworm to outbreak dynamics. the limiting limitations clear with period, time each at age forest for surrogates available best the were variables latter these datasets, available the of constraints the mature p the measured species tree deciduous all of combination the below)(see 2000 non and time metric each of specifics were configuration and composition forest Both averaged across sites within radii sample of range a quantified we conditions, Because species ( COHESIONwere calculated using FRAGSTATS radi sample all across area of proportion and (AWMPS) size patch mean weighted disturb non interval year one during “Disturbed” classifications. land “open” metric onbasal areabased fragmentation the as well as patches host budworm of connectivity Forest Accepted Article200 and 1985 - . forest land cover type cover land forest Forest composition was represented was composition Forest d oe types cover ed

a Shmkr 1996) (Schumaker forest configuration models - host host circle defined by a given sample radius for each site, and site and site, each for radius sample given a by defined circle we 5 -

year intervalyear (Sturtevant et al. 2014) al. et (Sturtevant do in 1 in ) species 2 .

roportion of forest forest of roportion

(Wolter et al. unpublished manuscript and ) due to the availability of forest plot data necessary to develop the underlying tree underlying the develop to necessary data plot forest of availability the to due ) onciiy f i ad spruce and fir of Connectivity . o know not 975 andremained 2000 in so

, within asubarea

i calculated was and

o siae uuaie rgetto o forest of fragmentation cumulative estimate to (i.e., 0.5, 1, 5, 10, 20, 30 km) 30 20, 10, 5, 1, 0.5, (i.e., vrgd vr all over averaged were , but , but our . s

the proportion of either of proportion the

quantified F priori a (i.e., ignoring water and open wetlands) open and water ignoring (i.e., rs openings orest

forest contingent on the corresponding data available at different points in points different at available data corresponding the on contingent were thenwere the a the in in

landscape

, representing the disturbance rate disturbance the representing , 1975 and 1975

verage basal areaveragebasal h scale the (approximately 3sites per subareas) using classified as open, grass, brush, or regenerating in the nextregeneratinggrass,or brush, the classified in open, as land 5 using the using -

er intervals year (non adcp pattern landscape

included for each for as ouras estimate of structure s then the proportion of p of proportion the then -

i pths a cluae fr ah ape radius sample each for calculated was patches fir t hc bdom ubek rsod o forest to respond outbreaks budworm which at - (McGa hrceie using characterized forest) forest) 1985 and2002

forest area ( area forest

amount basal area of of area basal measured was variable predictor each where ,

of of

sample radius for radius sample pixels predictor rigal al. et 2012)

Wolter et al. were balsam fir, sprucefir,andwithcombined balsam fir,

between for both for and that were that

circa ersne using represented old forest size ) metrics degree of inference linking forest linking inference of degree variables at multiple scales multiple at variables

using COHESION the landscape combinations of budworm host host budworm of combinations 1975

disturbed and open categories open and disturbed of forest of 1975) or 1975) (Sturtevant et al. 2014) al. et (Sturtevant ( multiple . 2008) ixels

. . classified “Open” land “Open” -

level variables were then were variables level .

both W that

and coverage

cluae te area the calculated e that were classified as classified were that ) .

openings (i.e., openings 2000 1985 and 200 and 1985

capture “ metrics old forest old as “ . d in in isturbed AWMPS and AWMPS mature combined 2000. w , the the ” ee the here

area

spatial 2 Given ” forest forest using

with

and W all in e - common interval 1928 interval common trees individual five least at on information contained series time dynamics trees affected percentage analysis cluster linkage complete Using T Statistical al. 2012) of model a outbreaks a of influence the examined budworm spruce 2007) al. et spring) summer, (winter, seasonal W Climate pr,ad yial rdcssot ea atrs fo patterns decay smooth produces typically and apart, 2009) nonparametric W Spatial R;Daniellin conduc we chronologies T identification was R in package 2012) Legendre and using assessed was clustering identified 2003) Kasetty and (Keogh ime dtrie h dge o proiiy n te rmr feuny f silto i outbreak in oscillation of frequency primary the and periodicity of degree the determine o

Acceptede Articlee assessed th

characterized - . series NF esrs h cvrac aogt ar o sts ls tgte vru toe further those versus together close sites of pairs amongst covariance the measures SNCF Synchrony

averaged Prior to clustering, t clustering, to Prior budworm

clustering (Table 2). (Table . Here, the the Here, . Analysis kernel e covariance function covariance

potential outbreaks

used as the the ; between 1970 and 2000. This This 2000. and 1970 between

- spatial 2005 Minimum

implemented using the the using implemented matrix of Euclidean distances was used as the response variable and clusterand variableresponse the as used distanceswasEuclidean of matrix ted spectral analyses on the cluster the on analyses spectral ted effect a w , predictor.

(Swetnam and Lynch 1993, Candau and Fleming 2005, 2011) 2005, Fleming and Candau 1993, Lynch and (Swetnam . spruce budworm spruce using Clustering was based on the the on based was Clustering e ime series were trimmed so that each that so trimmed were series ime synchrony parameter of climat dniid sub identified seasonal

historical s

the the development

(SNCF)

hclust e

Koh n Ksty 2003) Kasetty and (Keogh distance

= of of on outbreaks within our study area

temperatures

3 (1901 , ) site

population sipeetd nte c pcae n R in package ncf the in implemented as .

ucin n R in function adonis area - and - level level generation survival probability survival generation - ae rdnac aayi (dbRDA; analysis redundancy based 2000) 2000) s in response to to response in

ht exhibit that function in the vegan the in function outbreak r pairs of sites further apa further sites of pairs r have been been have growth index index growth temperature Euclidean distance between time between distance Euclidean - level . h saitcl infcne f the of significance statistical The

time time , resulting in series spanning the spanning series in resulting , ed

previously climatic

year examined with a sub a with examined year applied to to applied

- for - similar series series series (Table 2 (Table each each

variation (Oksanen et al. 2017) al. et (Oksanen using (function (n=51)

sub itrcl outbreak historical shown , time series of of series time Appendix index is is index area rt. meanminimum

using NF were SNCFs (Régnière et (Régnière

to influen to spec.pgram (McKenney (Bjørnstad (Legendre . based on based We also We 1 spatial - C series ) on ) area raw ce

using the R adjusted and criterion selection selection forward using type predictor each within predictors of all subset analysis, to Prior 2). (Table standard dates were multiple predictors cases some in and calculated, were these whichpredictors at scales spatial multiple the due composition and configuration forest for included ( P annual variation forest how model to used was (RDA)ordination Constrained Multivariate analysis landscape covariance) zones. replicates (i.e., area study of pairs all for computed grgt rsos o mlil otras o ihr vrg srcue r tutr a ay specific any at structure or structure average either to outbreaks multiple of response aggregate strongly more respond may outbreaks multiple first the is analysis Our function in veganthe package R in variance partitioning using assessed was climate) and configuration, composition, forest (i.e., type predictor vif.cca model the from VIF removed were a 10 than with greater Predictors interpretation. and direction coefficient confound can and models biased col model. Once n tnil p otential Accepted Article ( composition forest 2) =48); linear linear predictor

% of trees affetrees of % W function in the vegan package in R in package vegan the in function W based on based forward.sel e s in per site e

eitr aibe ( variables redictor (M, O) predicted within ute rdcd h rtie peitr b rmvn toe ht ee dniid as identified were that those removing by predictors retained the reduced further (Peres subarea a subset global (5 .

. - Da e a. 07 Osnn t l 2017) al. et Oksanen 2007, al. et (Dray the variance inflation factor (VIF). Collinearity among predictors can result in result can predictors among Collinearity (VIF). factor inflation variance the the eo et Neto - ‘This article is protected by copyright. All rights reserved.’ This This

15 trees) 15

function in the cted by outbreak that s

less

were SNCF zd o eo en n ui variance. unit and mean zero to ized to sites objectiv ,

- o ay ie dsac class, distance given any for

eae forest relate disturbed l 2006) al. to to budwormoutbreak

to maximizeto dniid o ec tp, e combined we type, each for identified oa n Total within ) . chronologies n oet adcp srcue t h tm o ec otra ta the than outbreak each of time the at structure landscape forest We used site level datalevel site used We 4) ad ) lmt ( climate 3) and =48); e (Oksanen et al.2017) presents packfor , more , each 11 al 2, ee f he tps 1 oet configuration forest 1) types: three of were 2), Table =101; . Variance partitioning was undertaken using the the using undertaken was partitioning Variance . landscape management zone ( managementzone the the

(Borcard et al. 2011) al. et (Borcard . package in R. host two O number (Zuur et al. 2009) al. et (Zuur ur multivariate response matrix consisted of of consisted matrix response multivariate ur ,

- 1928 motn challenge important rich tutr t bdom ubek yais over dynamics outbreak budworm to structure

- W of pairwiseof 200 instead of subareasof instead . n ilderness

5. getr ubr f rdcos are predictors of number greater A =5).

hr wud be would there 5 , Frad eeto ws implem was selection Forward .

landscape for each sub each for M, W, M, e is ietfe a parsimonious a identified first We .

. VIFs were calculated using the using calculated were VIFs . comparisons (W) The relative importance of importance relative The O) . is, niiul outbreaks individual First, s. them than in in than , and for all sites within the within sites all for and , structure area rae synchrony greater despite no fnl ordination final a into

(rows;

within managementwithin the

and climate affectclimate and relativelyfewer aae forest managed n =16; Fig. 1).Fig.=16; 2 varpart

s the as ented each (i.e. the ,

consistent with consistent it because analysis the in used peak cycle the at trees host of 65% than more affecting Cycle region a yielding o The Outbreak Results earlier cycle multi aggregate, the either than that 1958 windows w drivers, outbreak infer to ability cycle time in point outbreaks synchronized poorly was Minnesota, in entirely and area, observedoutbreakswere lasting single were ( significant T Cluster its in timing 40% than more timing variable db Accepted16hesub Article RDA; ; the variance explained by the predictors applied to the latest cycle (i.e., IV) would be greater be would IV) (i.e., cycle latest the to applied predictors the by explained variance the - . 1983 (cycle II (cycle 1983 also verall h ms comprehensive most the I sub

) o etr nesad o otra tmn ad the and timing outbreak how understand better To time.

ing ing of n ls ( last and wt rltvl hg snhoy nro cniec itra) n lse 3 n lower and 3 cluster in interval) confidence (narrow synchrony high relatively with ,

f ogl eul length equal roughly of F area oriented area clusters =

, but , but s. chronology contained chronology Temporal 2

Second,

. 44

in in chronologiesAppendix(see of trees in any one year one any in trees of ( rvos studies previous peaks affected O - ; the years 39 of periodicity cycle of estimate wide

p peak utbreak I on the basis of raw outbreak chronologies outbreak raw of basis the on ln a along

central data ~ 1910, 191 1910, ~ y), 1984 y),

~ just over 4 within Patterns 199 on .

southwest The clusters correspond clusters The part of the study area, in northern in area, study the of part Time

forest forest 1

- eoe sensing remote wa - Cycle , (Fig. cluster - exhibited irregular, high irregular, exhibited yl aayi o dsgrgtd nlss applie analyses disaggregated or analysis cycle Bas 1954) (Blais 2005 ec sann oe ubek yl: 1928 cycle: outbreak one spanning each ,

s defined by very few very by defined s e r e four complete, major outbreak cycles outbreak major complete, four 6 0% of trees at the cycle peak -

, and 193 and , landscape Series 4 epeated

C (cycle I (cycle

to to ) (Fig. . ) IV The third cycle ( cycle third The . north Cluster 2 Cluster o tee or yls ee nes, n synchronous, and intense, were cycles four these of ) 2

4 .

for subarea raw 1 the RDA procedure outline procedure RDA the D V h scn cce ( cycle second The , spanning a long interval, but never rising to affect to rising never but interval, long a spanning , structure becomes more limited more becomesstructure east data ) , 22y) , .

Clusters 3 and 4 each experienced three periodic three experienced each 4 and 3 Clusters , xs (Fig. axis ed restricted to the southwestern part of the study the ofsouthwestern the part restricted to s associated is .

. According to our logic above, w above, logic our to According large

While While

- III; peak ~ 196 ~ peakIII; frequency, low frequency, sample trees sample ly to management zone, zone, management to ly viaiiy f pta dt afc our affect data spatial of availability - level chronologieslevel

3 the first outbreak first the fprettes affected trees percent of . B

Minn

). (Fig. II ih h ms recent most the with ae ndsrt ple with pulses discrete in came ) lse 1 Cluster esota ,

2 an outbreak at this time this at outbreak an ) 1 - . d amplitude behaviour that behaviour amplitude between ) ,

The first ( first The , where just two, long two, just where , above was - 97 cce I (cycle 1957

was as one goes one as ) more synchronous more

t ete o the of either to d (Cycle grouped

for three shorter three for 1850 co although they although peak pie of mprised e predicted e

I and (Fig. budworm ) into

~ was not was I back in back 28y), ,

187 2005, four 3 A 2 is a ) - ,

any given distance class distance given any t sites between co Spatial Outbreak 1940 Thi cycles. 1990s), and 1930s the non exhibited years 33 and 33, 40, at peak spectral single a exhibited the to applied analyses Spectral studythe area. b cannot that waves) travellingoutbreaks(e.g.,expansionof thedirectionality insome potential indicatessimplyphasing evidenced acrossin of confidenc (wider synchrony was repr alone, years I 10km and20km balsam at disturbance forest three the among distributed T Forest Landscape Structure and Climate greater than Wilderness he three management zones (Fig zones management three he n total, thismodelcaptured 4 total, n e ia RA model RDA final he -

Accepted Article late years three occurring consistently cycles 3 cluster with phase, - hrd mn predictors among shared . 19

0 o te oa vrac ws trbtbe o oet ofgrto ad oet composition forest and configuration forest to attributable was variance total the of 20% fir 90 the 90 the of % trees affected in variance

esenting roughly half of the variance explained. variance the of half roughly esenting at scale at

rsle i wa proiiy n h 12 the in periodicity weak in resulted s (Fig 40 km Spatial Synchrony , whether measured at the scale of the whole study area study whole the of scale the at measured whether , -

stationary behavior, stationary . Climate was 5 B) th B) s , wherecovariance actually in outbreak in

- of ‘This article is protected by copyright. All rights reserved.’ correlation tabulationcorrelation conducted scale e further evaluated without analysis of spatial dynamics at scales larger than larger scales at dynamics spatial of analysis without evaluated further e an in the other two areas (Fig areas two other the in an 5 and km ple t te ul hoooy (1928 chronology full the to applied was s

of of e interval) in cluster 4 (Fig. 4 cluster in interval) e ormnrcce ocriga eua nevl nbtente major the between in intervals regular at occurring cycles minor four predictor and 1km and

0 higher included as the 30km

wt hl o ta (13%) that of half with , % of the variance outbreakofin the % occurrence mean chronologies for the clusters the for chronologies mean 5 including

and 5km and cluster 2

(i.e., more synchronous) more (i.e., type in the year 2005, and 2005, year the in

s curves). black C, generally .

F two . Forest composition was composition Forest . budworm never rose higher than 40%. orest configuration was configuration orest increased major cycles major Influence on Outbreak Dynamics decreased with decreased - a posteriori 5 15 A, ,

growth index

with with distance thein respectively 4

y A, B). Clusters 3 and 4 were 4 and 3 Clusters B). A, C). ears

The other half of the explained the of half other The

hrd between shared expected, As the proportion of old forest in 2005 at 2005 in forest old of proportion the pattern This was This

in the in at the start and end of the series (in series the of end and start the at rqec range. frequency -

2005) 2005) ( r lag3 showed that clusters 1, 3, and 4 and 3, 1, clusters that showed

among the 16 subareasamong16 the increasing geographic distance geographic increasing

(Fig

less ( (Fig. inclu Fig.

included as included = 0.891). especially ta cutr cce, as cycles, 4 cluster than r contained . -

disturbed, more disturbed, spatial

4 6 ded as ded 5 ) ) Wilderness , red curves), or within or curves), red , . oet opsto or composition forest .

Cluster 2 Cluster

This minorThis shift in covariance within covariance

vr h interval the Over true for distances for true the proportion of proportion the the basal area ofarea basal the seven , in contrast, in , slightly

(Fig.

variables host variance

over 5

B). - out rich 76

outbreaks that of frequency the lower the landscape, that structure landscape results Our Discussion none of the variance attributed climateto alone. remainder the and (inclusive) by explained variance the of 40% with predictors, of classes three all IV pattern outbreak in variance configuration ( our each analysing In Climate a and configuration, Cycle Cycle frequency, amplitude, synchronyand shaping of area, study structured forest to due that from, indistinguishable was variance this climate, to due was histories outbreak reproduction optimal interior progressively whereas rates, could Superior Lake to closest climate cooler the that indicates spring cooler vary not was climate outbreak on climate of effect important an expect not did We 1928

Accepted Article ( models 1984 h g the somewhat - 1957 - lone lone accounted for none of the 2005 reater the the reater sp Fig. ; captured indicate ruce ruce budworm outbreak ; s

are are

no other no ; Fig.; with distance from Lake Superior Lake fromdistance with n summer and sufficiently more intense, more extensive, of longer duration, and greaterof synchrony. outbreak cycle outbreak 7 most of of most A that resulted from resulted that ‘This article is protected by copyright. All rights reserved.’ 7 a greater proportion of variance as we movedforwar we asvariance ofproportiongreater a aa ae o hs te seis and species tree host of area basal ) C

that Rgir e al et (Régnière

it appears that appears it ) predictors were retained werepredictors 14%

66 historical %

variable within our study area. However, area. study our within variable s h ohr af ( half other the was

f h vrac i otra pattern outbreak in variance the of shared between climate and forest and climate between shared of theof variance outbreakin pattern ( Fig.

explained by forest forest by explained independently

7 dynamics sites and sites south of our study area study our of south sites and sites ubek yais ee fetd y differences by affected were dynamics outbreak 2012) . C forest different land management legacies in the BLL. the in legacies management land different ) variation (Fig. .

outbreak ned te pue budworm spruce the Indeed, landscape landscape 9

%) ( .

Appendix

( During

C hrd between shared such that such s ycles II ycles

n te higher the and composition and configuration and composition structure.

6 structure is more important than climate in climate than important more is structure ). Cycle III Cycle

- C consistently IV subareas the the ). ;

Fig. lhuh oe f h vrain in variation the of some Although behavior Thus

rprin f le forest older of proportion was landscape produce higher population higher produce ( forest 1958

2 ol b epand by explained be could that are that a te cl o or uniquely our of scale the at , explained ) their , we , we - hrd ih ad therefore and with, shared

1983 because forest d in time in d composition population learned structure variables, withvariables, structure found that found mltd, eutn in resulting amplitude, closer to the lake have lake the to closer may be too warm for warm too be may ; Fig. ;

by landscape

we predicted we . that a combination of

During 7 . During . B growth , as expected, as , climate ) n clim and

46% 46% We found We n forest in structure Cycle IICycle

growth the on

Cycle the of forest index does that at e .

forest fragmentation of host forests in Minnesota, the including in forests host offragmentation forest What cluster 1. However th with consistent anomalously I (Fig. for evidence found We zones. other clusters contrast The harmonic oscillations in the spruce budworm system. 4 we (Fig. peaks outbreaks budworm in 4 this landscape structure that and way, systematic a in another We disturbance as Minnesota but where disturbances were more aggregated in space, in aggregated more were disturbances where but Minnesota as disturbance disturbed 2; budworm host process. synchronization the overcome to enough management forest in contrasts selected purposefully study Our dispersal legacies synchrony outbreak in contrast The remov dynamics hypothesis” outbreak “silvicultural the supporting decades two last the in emerged turn in This cover. host of t is possible possible is that ourt statistical test climaticof effects on outbreak patterning was weak, andthatthe ).

Accepted Article e re Fig uh tog eidct ad ycrn ae lo osset with consistent also are synchrony and periodicity strong Such predicted Wilderness 4 might al of of al . xtensive,spanning C: 1 subarea), and low and subarea), 1 C:

are 4 influence , , ) budworm spruce the of capacity

I cnrs, the contrast, In . exhibited the highest degree of synchrony of degree highest the exhibited hs ol nt xli te poie atr big bevd n Minnesota’s in observed being pattern opposite the explain not would this susceptible host forest from landscape the 4 in Minnesota. in explain th explain in A, low that outbreak cycle frequencies, amplitudes, and synchro and amplitudes, frequencies, cycle outbreak that exhibited the lowest degree of synchrony threemanagement the of zones

observed

B). They were well were They B).

n i Ontario in and amplitude spatial e , andthis precisely is whatwe observed.

niomn big o wr t promote to warm too being environment o e tras were utbreaks s e

o large scale large 75% W synchrony (Fig. synchrony divergent Here utbreak would be would est and central areas central and est and high frequency high and Wi - amplitude, high amplitude, of the , a greater diversity of cycling behavior was observed was behavior cycling ofdiversity greater a ,

lderness, was behavior

- patterns within a singlemanagementa zonewithin patterns high wholly synchronized at th at synchronized spatial co spatial subareas among consistent periodic

(i.e., -

amplitude, low amplitude,

where (Anderson and Sturtevant 2011, Greenbank et al. 1980) al. et Greenbank 2011, Sturtevant and (Anderson 5 ). between clusters 1 and 2 and 1 clustersbetween consistent with the growing number of results that havethat results numbergrowingof the consistentwith

aaeet zones management oe apiue n hge frequency higher and amplitude lower

This result is interesting given the the given interesting is result This - (Fig , and evenly spaced, with with spaced, evenly and , frequency cycling in cycling frequency rrelation of outbreak of rrelation multi

of of ih ot other most with host of Minnesota that Minnesota of .

irregular 3 -

. B variate response would be related to to related be would response variate (Fig

: 7 subareas in cluster7 subareas3,5subareas cluster in : in a ms audn and abundant most was e scale of both clusters (Fig. clusters both of scale e - frequency, 40 frequency, . 4 Outbreak c behavior within - cycling within Minnesota’s cluster 2 is 2 cluster Minnesota’s within cycling 6 ) . Ontario, which had which Ontario, . decreased

niae that indicates tree regular, eais t pta sae large scales spatial at legacies cluster 2 cluster dynamics contain was strikingwas - ig ae suis f spruce of studies based ring - Royama ye ny would be related to one to related be would ny abundance and abundance ar ar 33 high ed cycles in tiny cluster 1 cluster tiny in cycles wih ugss that suggests which , (Fig.

y is is

created by the cycle the by created - 's the as ewe cycle between ears forest forest mltd cycling. amplitude spatial variationin spatial

oet wr least were forests known even though boththough even ( was 1984) 4 least amount of amount least D: 3 subareas) 3 D: similar rates of rates similar

5 ) relative to the to relative A,

management lusters 3and intermediate

hog the through ( i.e.,

long

B), which B), contagion theory of of theory adjacent c - forest range luster . .

Contrast with the 2008) Canada southeastern (i.e.,documented been have damage budworm and content hardwood between relationships negative where locations geographic the 5km radius, were area basal deciduous and area basal correlationsfir balsam Pearsonbetween other, each with correlated negatively be could factors these that speculate communities, enemy natural on non of effect the not and success, dispersal on forest host of effect the is it that suggests fac compositional important the as emerged content, de than rather host, budworm preferred of connectivity and concentration the study, our In 2011) boundariespolitical necessarilyfollow not does that variablecontinuous a as structure study area host more contained synchrony of degree its in cycles that (Fig. to structures that scales spatial Quebec southern with consistent is which was that cycle stable relatively legacies management budworm land divergent spruce the in landscape structured patterns spatially a reported other to well literature, dendroecology compare report we patterns The

Accepted estimate reliably Article interesting described a s 2 ail ycrn vre te ot mn mngmn zns ih ifrn forest different with zones management among most the varied synchrony patial ) .

. , of defoliation intensity, whereonly major the cycles were counted “outbreaks”as 4 Further research will be required to identify which of the mechanisms are responsible for responsible are mechanisms the of which identify to required be will research Further ) is is ) .

( Our study supports i.e., Clusters 3and 4 respectively further consistent with a with consistent further , a ,

asynchronous high s relatively . did

. s

but are perhaps more perhaps are but low It tudies

Buagr t al. et (Boulanger (Fig is is possible that balsam the fir is limiting factor in this region, relative to Jardon et al. outbreak . ;

Fig. 5 fairly - hog which through frequency oscillation ih clear with

( ) 2003) . Notably, outbreak dynamics of subareas from Minnesota that Minnesota from subareas of dynamics outbreak Notably,

3 small Sutvn e a. 2014) al. et (Sturtevant ) e also observed persistent asynchrony at finer temporal and temporal finer at asynchrony persistent observed also e fr el synchronized well

–

, in in , equency

u osrain f w a of observation Our de ( - 2003) scale monstrating importance of the treating landscape forest ;

their contrasts Fig. 2012) directly

’s yln bhvo i modified. is behavior cycling (~67 years) multi large 6

i ter small their in , ) assertion

tors in our RDA analyses (Fig. analyses RDA our in tors tended to follow the dynamics of the broader the of dynamics the follow to tended s - - oet landscape forest

interpretable because we specifically chose specifically we because interpretable scale study from study scale century observed ;

(Cappuccino hogot h bl o te td area, study the of bulk the throughout that , but , but low ( low .

study in Maine in study n h oe hand one the On a study in in d range ide with - r c cl, multi scale, = luster 2 0.08 and 0.08 alternating et al. 1998, Campbell et al. et Campbell 1998, al. et across across area needs be to extensive tutr ascae with associated structure

f ubek intensities outbreak of .

(Fraver et al. 2007) al. et (Fraver r Quebec - = hl oe might one While etr suy in study century

0.34 at 1km and 1km at 0.34 major major e bevd a observed we 6 ). ). (James et al. et (James . . - T

host forest host and minor We found We landscape his result his ciduous

rpi lvl o a tri a of levels trophic cycle outbreak re consider should community research the that suggests this 2004); al. et (Johnson alps Swiss 2009) (Roland 1993,Taylor 1997) Roland and in operating t budworm, spruce the beyond Looking down due strongto transitions in forest landscape structure that suggest results focus factors of fluctuation factor one any that concluded and cycle, outbreak the in times different at bottom including study recent pheno ubiquitous a is budworm effect Moran the that inferred have and distance, large on reported have investigators Previous Accepted 1984). (Royama data survey aerial and ring tree as such proxies impact using detectable be not may that processes underlying the of understanding local best the provide studies outbr budworm with associated damage with tree the occurring at maybe dynamicspopulation that in variation damage defoliation indicate to sufficient growth radial with trees of percentage the of use the Indeed, proxy. perfect a not is it location, given a within While Study limitations accordance with silviculturalthe hypothesis. some be long may there manipulating behaviour, cycling in implicated be to seem levels trophic Article aerial ed ; tree jack pine budworm in central North America North central in budworm pine jack

on the effect of slow changes in forest composition and structure on cyclingbehavior on structure and composition forest in changes slow of effect the on (Williams and Liebhold 2000)

- defoliation surveys suggests that suggests surveys defoliation ring

other f ycrn i te pue budworm spruce the in synchrony of ae rvn tity y eircl edak curn i ete h upr r lower or upper the either in occurring feedback reciprocal by strictly driven are s

(Bouchard et al. et (Bouchard data is a good proxy to quantify the temporal dynamics of budworm outbreaks budworm of dynamics temporal the quantify to proxy good a is data - up factors such as host tree cone production cone tree host as such factors up - em ubek yais y aiuaig adcp srcue wih s in is which structure, landscape manipulating by dynamics outbreak term periodic in spite of of spite in ‘This article is protected by copyright. All rights reserved.’

eae to related - rpi (ot plant (host trophic eoitr ytm, namely: systems, defoliator the Moran the

odaalblt and availability food 2017) and win and . ; gypsynortheasternthe mothin; United States

and otherandforest systems insect While their study focused on the effect of high frequency high of effect the on focused study their While eaks here is evidence for evidence is here effect t er moth in Fennoscandia in moth er -

herbivore

tree ring data ring tree - (Bouchard et al. 2006) al. et (Bouchard scale patterns in outbreak synchrony as a function of function a as synchrony outbreak in patterns scale , the process of outbreak synchronization can break can synchronization outbreak of process the , (Nealis and Lomic 1994) Lomic and (Nealis - demonstrate aua eey itrcin Weee both Wherever interaction. enemy) natural budworm survival on synchrony, our study our synchrony, on survival budworm a nt atr sm of some capture not may . f

rs tn ctrilr n oel Canada boreal in caterpillar tent orest do

synchrony is unlikely to be caused by caused be to unlikely is synchrony

reciprocal insect reciprocal capture larger scale variation scale larger capture - – vlaig h ie ta population that idea the evaluating

can contribute to the Moran effect Moran the to contribute can - s level

that (Ims et al. 2004) al. et (Ims . Clearly, intensive population intensive Clearly, (Peltonen al.2002) et . Nonetheless,. ifrn cua factors causal different eo i bt spruce both in menon ; larch budmoth larch -

There are, however, are, There host plant host

potnt for opportunity h fine the . (Haynes etal.

a comparisona Collectively,

feedback

- . in tree in scaled . Our . in the in M ore –

(Frelich and Reich 1995) Reich and (Frelich disturbance budworm to sensitive clearly is fir balsamNonetheless, 1980s. and 1970s mechanized government provincial road traditionally been have Ontario northern surroundingmanaged centurylastthan forests the during 2014) al. Sturtevant2003, et Host and (White disturbance harvest clearcut diffuse T composition species Scanner Multispectral Landsat of absence the forest of effects outbreak examine explicitly to first the is study Our outbreaks. to to necessary scale spatiotemporal available methods practical no analyses, and disti data quality landscape 2014) al. et (Sturtevant asstructure landscapedivergent outbreaks time through variables structure landscape forest relevant generate to used data sensing remote of quality and quantity increased analys (b) and cycles, all of analysis aggregated the (a) both than response forest the of estimate robust W the fir managementforestactivities of legacydominant hematic Acceptede Article gih ewe tee w pasbe xlntos o te eut fo te outbreak the from results the for explanations plausible two these between nguish found

i (James al. et 2011) s - derived spatial derived of of dynamic

M structure moving that the analysis specific to cycle IV cycle to specific analysis the that

ale cce Cce I n III) and II (Cycle cycles earlier of population abundance and landscape structure landscape and abundance population of apper forestry operations prior to 1975 were markedly different from those observed in the in observed those from different markedly were 1975 to prior operations forestry ,

indeed, they not are mutually exclusive. detailed s

sensor owr truh time through forward across across multiple outbreak cycles. as it diverges among management zones through time through zones management among diverges it as ‘This article is protected by copyright. All rights reserved.’ Moe n Bur 1990) Bauer and (Moore (Rem sensors historic data historic , and it is likely that this that likely is it and , Ta is, That . . i 18 ad 1984 and 1982 in s data data

pel et al. 1997) al. et pel xed back extends d the legacies of legaciesthe d o hv te pcrl eouin eesr t dsigih tree distinguish to necessary resolution spectral the have not id .

evaluate landscape evaluate Alternatively, s uwr otra bhvo my ipy e tracking be simply may behavior outbreak budworm

on have been applied consistently in Minnesota since the 1930s the since Minnesota in consistently applied been have quantify forest conditions forest

(Fig (Fig.

– only .

. so there is no reason to suspect the spatial pattern of pattern spatial the suspect to reason no is there so

and H forest managementforestactivities

which was which 7 6

(i.e., the most recent; Figure recent; most the (i.e., because spruce budworm is simultaneouslykillingbudworm is spruce because - owever, ) a , uh fine such iie ad ple t lns esd rm the from leased lands to applied and limited

host species species host

o h ery 1970s early the to 7

might strong that the the that A challenge facing . This ). -

scale feedbacks of the forest on budworm on forest the of feedbacks scale

(Heinselman 1996) er lo be also we during Wilderness

gr enabled - effect of forest landscape forest of effect cld ouain yais t the at dynamics population scaled

eater eater nw that know at earlier time periods time earlier at is more dynamic than the otherwise the than dynamic more is landscape earlier the outes s iey due likely is robustness later Further, . osqec of consequence

any had a lower disturbancelowerrate a had cycles. small by

. Without access to high to access Without . approach

structure . accumulated launch of of launch Forestryinoperations -

7 sized

h erir Landsat earlier the The availability of availability The ) provided , we cannot we ,

such ours as n spatially and n budworm on in the region the in increasingly structure on structure the the

over -

Landsat specific a more a o the to forest time yet is -

iety upr h nto ta dcdos otn afce otra bhvo,a upiig result surprising a behavior, outbreak affected content deciduous that notion the support directly outbreaks. budworm spruce on structure landscape forest of contributions identifyingthe ofchallenge whichhighlightsgradient,theclimatic a weakresponsesto with and structure landscape forest between Relationships Liebhold 2000) controversial 2005) Russo is outbreaks budworm of behavior cycling true is situation reverse the demonstrating the over outbreaks budworm spruce of severity increasing that hypothesis” “silvicultural the with consistent host where areas and frequency, high and intensity low synchrony. influence configuration spatial W Conclusion budworm cycling spruce budworm affecting process key a was renewal forest that hypothesized which dynamics, budworm 1975)(Baskerville ago relationship decades hypothesized feedback as scale, landscape reciprocal the at a occurs that forest is and budworm the there between that implies study our configuration, spatial forest chang significant for responsible is disturbances, replacing host in muted be may these cycles cyclic, generally are outbreaks budworm while that view synthetic the support results Our outbreakinsect dynamics over broad spatial scales. b attention more that requires ecosystems forest of stewardship sustainable Continued processes. ecological spatial on BLL. land the consequent of conditions forest mixed the given e

Accepted Articlegenerally found

at etr rsle fo ps mngmn activities management past from resulted century last ht pta vrain n oet adcp srcue icuig pce cmoiin and composition species including structure, landscape forest in variation spatial that In areas with reduced fir abundance and connectivity, and abundance fir reduced with areas In

oe accepted more for defoliators for n wsen Europe western and . s

use changes at large scales has the potential to have significant unintended effects unintended significant have to potential the has scales large at changes use . This This . wa (Jones 1977) (Jones ‘This article is protected by copyright. All rights reserved.’ e given to spatial legacies in forest landscape structure as these legacies affect legacies these as structure landscape forest in legacies spatial to given e s

oe bnat a i te eaiey undisturbed relatively the in as abundant, more pr ovides - for such as the s the as such deficient forests. deficient . Even if, as hypothesized by hypothesized as if, Even . s ak beetle bark considerable qualitative validation to the earliest models of spruce of modelsearliest the considerablequalitativeto validation

pue uwr otra cce frequ cycle outbreak budworm spruce Tmel e a. 2013) al. et (Temperli were . This paradigm This . pruce budworm pruce –

not well synchronized. The opposite The synchronized. well not s i.e., a reduction in landscape in reduction a i.e., ystems To the extent that the budworm itself, through stand through itself,budworm the that extent the To

Never

n etr Nrh America North western in outbreak of strong of hls, non ln mngmn and management land ongoing theless, Mle ad unc 93 Mzk and Muzika 1993, Rusnock and (Miller bt ni nw a be considered been has now until but , s n oet pce cmoiin and composition species forest in es Royama Royama

ht nacd fir, enhanced that atrs per o e intertwined be to appear patterns

f orest outbreaks budworm spruce ( 1984) - Wilderness insect reciprocal feedback reciprocal insect - Our results also did not did also results Our level host mitigates the mitigates host level ny itniy and intensity, ency, ,

reciprocal feed reciprocal Wieed and (Whitehead was n hs ae by case this in is result This . observed in observed

unique back had -

aoce o ter ep n h fed n i te laboratory. the in Boulanger, and two anonymous reviewers for helpful comments onthe manuscript. and field the in help their for Larouche BRS. supporting Plan Fire National the and LER, support to E2014 # Project (www.usendowment.org; B DK, to Canada of Network Management Forest Sustainable the from grant a MJF, and BJC, DK, PAT, (2005 program Ecosystems Managed (CSREES) Service Extension and was project This Acknowledgements universal a is feature of forest insect enemies forest how forest of influence the quantify better the patterning. outbreak that in complexity appears it behaviour, cycling outbreak of budworm source primary a is enemies natural with Accepted Article - host relationship in other defoliator systems to determine if reciprocal if determine to systems defoliator other in relationship host S ad J, n truh gat rm h U Edwet o Frs ad Communities and Forest for Endowment US the from grant a through and BJC, and RS, .

Additionally, we think there would there think we Additionally, - oet eircl edak ant e ue ot s sgiiat opnn o te full the of component significant a as out ruled be cannot feedback reciprocal forest landscape uddtruha rn rmte SACoeaie tt Rsac, Education Research, State Cooperative USDA the from grant a through funded ‘This article is protected by copyright. All rights reserved.’

tutr my niety fet top affect indirectly may structure

defoliator

A dtoa suy n different in study dditional landscape

outbreak - 0260) to BRS, BJC, and PMAJ, all of which were used were which of all PMAJ, and BJC, BRS, to 0260)

be structure on budworm outbreak cycling outbreak budworm on structure enormous systems.

value in examining the strength of the of strength the examining in value

We thank Nate Aspelin and Aspelin Nate thank We e lo hn Ptik oi, Yan Tobin, Patrick thank also We ecological contexts ecological - on oto b natural by control down - 35101 forest

64) o BRS, to 16342) - insect s needed is

as well as well as

feedback François to northwestern Ontario. Ontario. northwestern bu spruce of recurrence The 1954. R. J. Blais, functions. covariance nonparametric spatial ncf:

Can. J. For. Res. 34: 1035 Res.34: For. J. Can. 2319 87: Ecology Accepted Canada. ineastern outbreaks budworm ofspruce severity and extent, frequency, in the Trends 1983. J. lais, Article - le, G. L. 1975. Spruce budworm: super silviculturist. silviculturist. super Spruce budworm: L. G. 1975. le,

Can. J. For. Res.25: 1375 For. J. Can. -

Can. Entomol. 97: 129 97: Entomol. Can. -

Can. J. For. Res. Special Res. For. Can.J. et al. 2006. Forest dynamics after successive spruce budworm outbreaks in mixedwood forests. forests. mixedwood in outbreaks budworm spruce successive after dynamics Forest 2006. al. et – - 346. project.org/package=ncf>. - –

U of Minnesota Press. ofMinnesota U 2329.

Ecology 35: 62 35: Ecology – 547.

– ion cycles: the case for trophic interactions. interactions. trophic for case the ioncycles: 1043.

– – 136. 1384. - p atr cnrbt t large to contribute factors up

–

Issue:1 71. Lob trees in the wilderness: the human and natural history of the the of history natural and human the wilderness: the in trees Lob

–

497.

- dworm infestations in the past century in the Lac Seul area of of area Seul Lac the in century past the in infestations dworm

– In: Numerical Ecology with R. Springer, pp. 153 pp. withR.Springer, Ecology Numerical In:

For. Chron. 51: 138 Chron. 51: For. - scale synchrony in spruce budworm budworm spruce in synchrony scale -

Oxford university press. university Oxford pcae eso 1.1 version package R – 140.

– 225.

In: In: - 3, 3, -

pp. 487 pp. Press. Academic Elsevier K. Miyanishi, and E.A. Johnson, response. the and process the ecology: disturbance ecosystems. forest northern in disturbances periodic as defoliators Insect 2007. al. et J. B. Cooke, ring tree standardization. to approach analysis A series times R.1985. E. Cook, Canada. Brunswick, inNew control Chang, 236 114: Oecologia landscape. patchy a in rate parasitism and abundance impact, budworm Spruce 1998. al. et N. Cappuccino, Ontario. in change J. Candau, conditions. bioclimatic to relation J. Candau, forest l surrounding of content hardwood with the varies 2008. (Choristoneura al. et M. E. Campbell, budworm spruce of reconstruction Dendrochronological 2012. fumife al. et Y. Boulanger, Accepted easternCanada. in budworm spruce the of characteristics outbreak and climate between relationship The 2008. R. D. Gray, 325 65: Monogr. Ecol. southern Minnesota a in succession and patterns Spatial 1995. B. P. Reich, and E. L. Frelich, USA. Maine, northern in outbreaks budworm spruce of reconstruction Dendrochronological 2007. al. et S. Fraver, webs. Flu 2007. al. et S. E. Eveleigh, Forests. Canadian 2010. al. et C. C. Dymond, S. Dray, Article - - rana) outbreaks in southern Quebec for the last 400 years. 400 thefor last Quebec southern in outbreaks rana)

– Can. J. For. Res. For. Can.J. Proc. Natl. Acad. Sci. 104: 16976 104: Sci. Acad. Natl. Proc. W.

525. et al. 2007. Packfor: forward selection with permutation, R package version 0.0 version R package withpermutation, selection forward Packfor: 2007. al. et - - - N. and Fleming, R. A. 2005. Landscape A. 2005. R. Fleming, and N. N. and Fleming, R. A. 2011. F 2011. A. R. Fleming, and N. Y. et al. 2012. 2012. al. et Y.

Clim. Change 87: 361 87: Change Clim. - -

–

37: 523 37:

Economic impacts of forest pests: a case study of spruce budworm outbreaks and and outbreaks budworm spruce of study case a pests: forest of impacts Economic

The severity of budworm of severity The Future Spruce Budworm Outbreak May Create a Carbon Source in Eastern Eastern in Source Carbon a Create May Outbreak Budworm Spruce Future ctuations in density of an outbreak species drive diversity cascades in food food in cascades diversity drive species outbreak an of density in ctuations – 529. -

Can. J. For. Res. 35: 2218 Res.35: For. Can.J. -

Can. J. For. Res. 42: 490 Res.42: For. J. Can.

– – 383. 931. – 16981. orecasting the response of spruce budworm defoliation to climate climate to defoliation budworm spruce of response the orecasting

– 1960.

- scale spatial distribution of spruce budworm defoliation in in defoliation budworm spruce of distribution spatial scale

- andscapes. andscapes. caused growth reductions in balsam fir/spruce stands stands fir/spruce balsam in reductions growth caused – – 2232. 505. -

Can. J. For. Res. 42: 1264 Res.42: For. J. Can. -

For. Sci. 54: 195 54: Sci. For. -

Univ. Ar Univ. – 205.

– 7. iz. in press. iz.

– 1276. - boreal forest. boreal

In: Plant Plant In: - -

oms R L 19. edohoooy rga lbay Lbrtr o tree of Arizona,Tucson. Laboratory library. program Dendrochronology 1992. L. R. Holmes, systems. ecological stability and of Resilience 1973. C.S. Holling, spruce. black 2008. al. et R. C. Hennigar, wildernessecosystem. waters boundary The L. 1996. M. Heinselman, 329 Res.3: M Area, Canoe Waters Boundary the of forests virgin the in Fire 1973. L. M. Heinselman, intensity. outbreak moth gypsy of oscillation harmonic of 249 159: Oecologia analysis Spatial 2009. al. et J. K. Haynes, 377 38: Math.Biol. inf budworm spruce for model site Jones’s of analysis mathematical A 1999. al. et C. D. Hassell, aircraft. and radar, fromobservations, canopy understanding budw Spruce 1980. al. et O. D. Greenbank, Canada. eastern budworm in spruce the of characteristics outbreak on climate and composition forest of influence The 2013. R. D. Gray, Accepted auQuébec. l’épinette de bourgeons des tordeuse la de épidémies des synchronisme et Périodicité 2003. al. et Y. Jardon, in ignition fire of risk the affects defoliation Canada. Ontario, budworm spruce cumulative Lagged 2017. al. et P. James, Lakes landscape. Two 2011. al. et M. P. James, dynamics? synchronous sub Do 2004. al. et A. R. Ims, program for manual user a measuring: and crossdating of control LaboratoryofTree COFECHA. Quality 1986. al. et L. R. Holmes, Arizona. Arizona,Tucson, of University R Holmes, Article . and Swetnam, T. 1996. Program OUTBREAK user manual. user OUTBREAK Program 1996. T. Swetnam, and . – - 382.

Can. J. For. Res. 33: 1947 Res.33: For. Can.J. -

For. Sci. 54: 158 54: Sci. For.

- -

J. Anim. Ecol. 73: 1129 Anim. 73: J. Ecol. Differences in spruce budworm defoliation among balsam fir and white, red, and and red, white, and fir balsam among defoliation budworm spruce in Differences -

- Can. J. For. Res. 43: 1181 Res.43: For. Can.J. dimensional dimensional - – Ring Research. University of Arizona, Tucson, Arizona. Tucson, ofArizona, University Research. Ring 166. - Arctic winter moth populations in coastal birch forest exhibit spatially spatially exhibit forest birch coastal in populations moth winter Arctic

– – 5 –

1961. 44. 2209. wavelet analysis of spruce budworm host basal area in the Border Border the in area basal host budworm spruce of analysis wavelet

orm (Lepidoptera: Tortricidae) moth flight and dispersal: new new dispersal: and flight moth Tortricidae) (Lepidoptera: orm

– 1136.

– 1195. -

Mem. Entomol. Soc. Can. 112: 1 112: Can. Soc. Mem.Entomol.

Annu. Rev. Ecol. Syst. 4: 1 4: Syst. Rev.Ecol. Annu. -

University of Minnesota Press. Minnesota of University -

Laboratory of Tree of Laboratory - ring research, University of of University research, ring

innesota. innesota. - Ring Research. Ring – – 49. estations. estations. 23.

Quat. Quat. -

J. -

maps. University of Massachusetts, Amherst. Massachusetts, of University maps. progra analysis pattern spatial v4: FRAGSTATS 2012. al. et K. McGarigal, discussion. and review fir of Vulnerability 1980. A. D. MacLean, 315 Ecol.: Anim. fores and budworm spruce the systems: outbreak insect of analysis Qualitative 1978. al. et D. Ludwig, Numerical ecology. 2012. F. L. Legendre, and P. Legendre, Germany. Park, National Spatio 2013. al. et A. Lausch, 987 change. climate to feedback carbon forest and beetle pine Mountain 2008. al. et A. W. Kurz, demonstration. survey a benchmarks: mining data series time for need the On 2003. S. Kasetty, and E. Keogh, systems. ecological to theoryapplied Catastrophe 1977. D. D. Jones, LBM. in the waves travelling and geometry 2004. al. et M. D. Johnson, Can. Entomol. 148: S33 148: Can.Entomol. conifer of ecology Comparative 2016. G. V. Nealis, defoliating managing to approaches silvicultural in of critique A 2000. M. A. Liebhold, and M. R. Muzika, Lynx cycle. Canadian the of statistical analysis A. The 1953. P. P. Moran, and Scanner Multispectral Landsat north in vegetation forest of Classification 1990. E. M. Bauer, and M. M. Moore, hypoth silvicultural the of fall and rise ( The 1993. P. Rusnock, and A. Miller, Canada. for models Customizedclimate spatial 2007. al. et D. McKenney, Choristoneura fumiferana Choristoneura Accepted America. North in sects Article – 990.

– For. Chron. 56: 213 Chron. 56: For. S57. - -

) management in eastern Canada. Canada. eastern management in ) Ecol. Indic. 31: 31: 73 Indic. Ecol.

Agric. For. Entomol. 2: 97 2: Entomol. For. Agric.

adcp goer ad rvlig ae i te ac bdoh Landsca budmoth: larch the in waves travelling and geometry Landscape - eprl netto pten o Is typog Ips of patterns infestation temporal Thematic Mapper data. data. Mapper Thematic - - spruce stands during uncontrolled spruce budworm outbreaks: a outbreaks: budworm spruce uncontrolled during stands spruce –

– Ecol. Lett. 967 7: Ecol. 221. 81.

– 371. - – feeding spruce budworms (Lepidoptera: Tortricidae). Tortricidae). (Lepidoptera: budworms spruce feeding 105.

- For. Sci. 36: 330 36: Sci. For. -

Elsevier. For. Ecol. Manag. 61: 171 Manag.61: Ecol. For. – 974.

Simulation 29: 1 29: Simulation - raphus (L.) in the Bavarian Forest Forest Bavarian the in (L.) raphus -

Aust. J. Zool. 1: 291 1: Zool. J. Aust. Frontline Tech. Note i Note Tech. Frontline m for categorical and continuous continuous and categorical for m – 342.

esis in spruce budworm budworm spruce in esis - central Minnesota using using Minnesota central –

– 15. 189.

– - and empirical empirical and

298. Nature 452: 452: Nature 108

t. -

pe pe J. -

Régnière, J. and Lysyk, T. 1995. Population dynamics of the spruce budworm, Choristoneura fumiferana. fumiferana. Choristoneura budworm, spruce the of dynamics Population 1995. T. Lysyk, and J. Régnière, eruptions. beetle bark dynamics of Cross 2008. al. et F. K. Raffa, minutum. Trichogramma by eggs fumiferana Choristoneura host composition, Forest 2003. al. et D. Quayle, Debate. Aof Century Ecology: 2016. al. et S. D. Pureswaran, 1984. al. W.et P. Price, fractions. Peres 2.4 version packge R dispersal. “vegan”. 200 al. et M. Peltonen, Package 2017. al. et project.org/package=vegan>. J. Oksanen, Host Tortricidae). (Lepidoptera: pinus Choristoneura budworm, 1994. V. P. Lomic, and G. V. Nealis, Accepted 25 93: Oecologia Large 1993. J. Roland, fumiferana L. Robert, Timber 1997. Evaluation. Landscape al. et S. R. Rempel, example. as an budworm spruce J Régnière, the of budworm. spruce outbreaks during change population of mechanisms Ecological 2007. G. V. Nealis, and J. Régnière, 95 Ont.pp. in pests insect Forest In: Article - Neto, P. R. et al. 2006. 2006. al. et R. P. Neto, - -

– Ecology 87: 2 87: Ecology . et al. 2012. Predicting insect distributions under climate change from physiological responses: responses: physiological from change climate under distributions insect Predicting 2012. al. et . Ecology 83: 3120 83: Ecology ) outbreaks. outbreaks. ) E. et al. 2012. Effects of forest management legacies on spruce budworm ( ( budworm spruce on legacies management forest of Effects 2012. al. et E. 105.

– - 30.

Ecol. Entomol. 32: 461 32: Entomol. Ecol.

A new ecology: Novel approaches to interactive systems. systems. interactive to approaches A Novel new ecology: 2. 2. 517. Manag.Wildl.61: J. - 614 Can. J. For. Res. 42: 463 For.Res. 42: J. Can. Canada. Edited by JA Armstrong and WGH Ives. Canadian Forest Service, Ottawa, Ottawa, Service, Forest Canadian Ives. WGH and Armstrong JA by Edited Canada. scale forest fragmentation fragmentation forest scale Spatial synchrony in forest insect outbreaks: roles of regional stochasticity and and stochasticity regional of roles outbreaks: insect forest in synchrony Spatial

– – 2625. - 3129. Variation partitioning of species data matrices: estimation and comparison of of comparison and estimation matrices: data species of partitioning Variation scale drivers of drivers scale Paradigms in Eastern Spruce Budworm (Lepidoptera: Tortricidae) Population Tortricidae) (Lepidoptera: Budworm Spruce Eastern in Paradigms -

- Environ. Entomol. Environ.

Biol. Invasions 14: 1571 Invasions14: Biol. -

Bioscience 58: 501 58: Bioscience - aaeet n Natural and Management – 477.

- natural disturbances prone to anthropogenic amplification: the the amplification: anthropogenic to prone disturbances natural ln ifune n the on influence plant

–

475. - 45: 1333 45: population density, and parasitism of spruce budworm budworm spruce of parasitism and density, population increases the duration of tent caterpillar outbreak. outbreak. caterpillar tent of duration the increases –

517. – -

1586. - Ecol. Ecol. –

Entomol. Exp. Appl. 107: 215 107: Appl. Exp. Entomol. 1342.

Entomol. 19: 367 19: Entomol. - Disturbance Effects on Moose Habitat: Habitat: Moose on Effects Disturbance

ouain clg o te ak pine jack the of ecology population -

Wiley. – 373. - ,

– 227. Choristoneura Choristoneura

- - -

outbreaks. outbreaks. Swe 1620 Res.26: For. budworm. spruce by defoliation fir balsam on content hardwood of influence The 1996. al. et Q. Su, budworm. landscape forested across disturbance insect Modeling 2015. al. et R. B. Sturtevant, (Canada). Ontario and (USA) Lakesof Minnesota region Border 2014. al. et R. B. Sturtevant, connectivity. habitat predict to Using indices landscape 1996. H. N. Schumaker, processes. population nonlinear on effect Moran 2005. T. Royama, 429 54: fumiferana. Choristoneura budworm spruce the of dynamics Population 1984. T. Royama, spatial different at structure forest scales. to respond species parasitoid Insect 1997. D. P. Taylor, and J. Roland, time spatially in apparent dynamics cyclic in variation spatial of seeds” the“ Are 2005. J. Roland, AcceptedBusinessMedia. animals. of the abundance and nitrogen environment: inadequate The C.2012. T. White, offorest landscape. structure by affected outbreak d L. brumata Operophtera moth winter by defoliation Tree 2006. P. Rowiński, and T. Wesołowski, northwestern in behaviour Wisconsin. population budworm pine Jack 1994. G. D. McCullough, and A. J. W. Volney, Press. synthesis. theoretical/empirical a dynamics: population Complex 2003. P. Turchin, modelingapproach. landscape a Temperli Article tnam, T. W. and Lynch, A. M. 1993. Multicentury, regional Multicentury, 1993. M. A. Lynch, and W. T. tnam, - series? An example from the forest tent caterpillar. caterpillar. the forest tent from An example series?

- –

Nature 386: 710 386: Nature 462. , C. et al. 2013. Cross 2013. al. et C. , - -

- In: Simulation Modeling of Forest Landscape Disturbances. Springer, pp. 93 pp. Springer, Disturbances. Landscape ofForest Modeling In: Simulation Ecol. Monogr. 63: 399 63: Monogr. Ecol.

Can. J. For. Res. 24: 502 Res.24: For. Can.J.

713.

- oet eoey atrs n epne o iegn dsubne eie i the in regimes disturbance divergent to response in patterns recovery Forest scale interactions among bark beetles, climate change, and wind disturbances: disturbances: wind and change, climate beetles, bark among interactions scale -

Ecol. Monogr. 83: 383 83: Monogr. Ecol. – 424. – 510.

For. Ecol. Manag. 221: 299 Manag.221: Ecol. For. - –

402. Ann. Zoo Ann.

- For. Ecol. Manag. 313: 199 Manag.313: Ecol. For. scale patterns of western spruce budworm budworm spruce western of patterns scale - l. Fenn.: 397 Fenn.: l.

Ecol. Monogr. 75: 277 75: Monogr. Ecol. – – - 305.

407. Ecology 77: 1210 77: Ecology s: insights from the spruce spruce the from insights s:

– Princeton Princeton - 134. Springer Science & Science Springer – 293. - - –

replicated short replicated 211. Ecol. Mon Ecol.

University University

– 1225. uring an an uring -

Can. J. Can. ogr. ogr.

Accepted ArticleMedia. R. with ecology in extensions and models effects Mixed 2009. al. et A. Zuur, an classification cover land Lakes Region Border Greater al.et 2012. T. P. Wolter, Ontario. and Minnesota inNorthern P. Wolter, America. North eastern in outbreaks budworm spruce of synchrony Spatial 2000. M. A. Liebhold, and W. D. Williams, beetle. pine mountain from damage have less Beetle 2005. L. G. Russo, and J. R. Whitehead, Council. Resources Euro Cha 2003. E. G. Host, and A. M. White, - mrcn o eteet o rsn i North in Present to Settlement to American

et al. 2008. Remote sensing of the distribution and abundance of host species for spruce budworm budworm spruce for species host of abundance and distribution the of sensing Remote 2008. al. et Ecology 81: 2753 81: Ecology

Remote Sens. Environ. 112: 3971 112: Environ. RemoteSens. nges in Disturbance Frequency, Age and Patch Structure from Pre from Structure Patch and Age Frequency, Disturbance in nges -

Pacific Forestry Centre. Forestry Pacific - proofed lodgepole pine lodgepole proofed - central and Northeastern Minnesota. Minnesota. Northeastern and central

– 3982. stands in interior British Columbia Columbia British interior in stands

Spr d change detection. change d inger Science & Business Business & Science inger

Minnesota Forest Forest Minnesota

- “ ( Zone Management = Letter 1 labels: Subarea analysis. of unit subarea the by grouped are that outbreaks of reconstructions and (Canada) 1: Figure FIGURE LEGEND

Accepted Articleentral”, S h Bre lks adcp (L) td ae lctd t h bre bten Ontario between border the at located area study (BLL) landscape lakes Border The outh, Minnesota (United States). Points represent sampling sites for dendrochronological for sites sampling represent Points States). (United Minnesota N orth); 3 ‘This article is protected by copyright. All rights reserved.’ S

Letter =Letter Subarea ( O

ntario, W ilderness, N orth, M S outh). neoa; 2 innesota);

etr A = Letter

e ( rea E ast, W est,

Accepted Articlearrows. (I cycles major four with associated peaks Cycle based. is line) black trees; affected (percent chronology disturbance the which upon trees of number the to refers line) (red (ie., unit sample the of level the at size sample low to due analyses formal 2 Figure

subarea). Sample depth Sample subarea). - IV) are indicated with indicated are IV)

in the 16 subareas. B. The four clusters overlaid in geographical space. See Table 1 for subarea for 1 acronyms. Table See space. geographical in overlaid clusters four The B. subareas. 16 the in 3 Figure Accepted Article :

A. Results A. ‘This article is protected by copyright. All rights reserved.’ of the cluster analysis (n = 4) of the percentage of trees affectedbydefoliationtrees percentage theof of 4) = analysis (n cluster theof

Cluster: Cluster 4& 3: 33years, Cluster 2: 12 subarea raw for 2Appendix and acronyms, subarea for 1 Table See pattern. temporal BLL broader the from cluster analysis. any in applied not threshold 25% the indicates 2012) line al. et (Robert horizontal study prior in Dottedoutbreaks defining 1). (Fig. area study entire the for affected the for affected trees = Red cluster. the intervalsconfidencefor 95% of = grayDashed cluster. percentage average = Black (top). northeast the to (bottom) southwest 4 Figure Accepted Article : Combined chronologies for each of the four clusters shown in Figure 3, listed from the from listed 3, Figure in shown clusters four the of each for chronologies Combined - level chronologies. Spectral peak from spectral analysis is the following for e for following the is analysis spectral from peak Spectral chronologies. level ‘This article is protected by copyright. All rights reserved.’ ifrne ewe rd n bak uvs hw te eito o a of deviation the shows curves black and red between Difference

- 15 years, Cluster 441: years –

but this line is used for reference only and only reference for used is line this but average percentage of trees of averagepercentage

ach entire BLLentire (red). the for calculated SNCF the with overlaid and lines) (dashed interval confidence 95% their with se on individuallyrepresented is (black) managementzone Each BLL. the in 5 Figure Accepted Article : Spatial non parametric covariance functions (SNCF) for the different management zones managementdifferent the for (SNCF) functionscovariance parametric non Spatial

parate graph (A, B, C) B, (A, graph parate

(Fig. 7). se with contrasted is chronology full the to applied is RDA The diagram). bubble the in variables of class the to corresponds label classofbyexplainedeach predictor variables 2 for variable (see Tabledescriptions, where color the for 1 Table See climate. (3) diagramsInsetblackbubblesubarea acronymsfont. indicate independenin the and composition; forest (2) configuration; forest (1) hypotheses: trees percent 6 Figure Accepted Article

: Redundancy Analysis (RDA) and variance partitioning relating dissimilarity the in annual affected to independent variables chosen by forward selection for each of our three our of each for selection forward by chosen variables independent to affected ‘This article is protected by copyright. All rights reserved.’

parate RDAs applied to each of the three major outbreak periods outbreak major three the of each to applied RDAs parate t and shared varianceshared and t

the the three major outbreak periods (A of each to applied is RDA The diagram). bubble the in variables of class the to corresponds label classofbyexplainedeach variables (see T for 1 Table See climate. (3) diagramsInsetblackbubblesubarea acronymsfont. indicatevariance independentin the shared and and composition; forest (2) configuration; forest (1) hypotheses: for selection forward by chosen variables independent to affected trees percent 7 Figure Accepted Article : Redundancy Analysis (RDA) and variance partitioning relating dissimilarity the in annual ‘This article is protected by copyright. All rights reserved.’

- C).

able 2 for predictorable2 for variabledescriptions, where color the

each of our three our of each

the programthe COFECHA. 1 Table TABLE LEGENDS Accepted Article Wilderness Ontario Zone Minnesota : Summary statistics of the dendr Summarythe of statistics

Western Western Eastern Eastern Center Center Western Western South North Eastern Eastern Western Western Eastern Eastern Area

South North South North South North South North South North South North South North South North Subarea

MWS MWN MES MEN MCS MCN WWS WWN WSS WNN WES WEN OWS OWN OES OEN Label

ochronologicalbysubareasproducedreconstruction all for

16 17 19 22 18 trees of No. 22 22 22 22 23 21 15 21 24 20 13

1924 19 1895 1883 1849 TimeSpan 1920 1903 1872 1836 1928 1915 1880 1836 1897 1906 1902

26 ------2006 2005 2005 2005 2005 2005 2005 2005 2005 2005 2005 2006 2006 2005 2005 2006

0.279 0.272 0.28 0. 0.25 Mean Sensitivity 0.263 0.277 0.249 0.238 0.207 0.208 0.253 0.259 0.277 0.268 0.25 226

1.029 1.159 1.157 1.024 1.13 SD 1.508 1.363 1.075 1.403 1.413 1.351 0.826 0.92 1.026 1.356 1.105

0.661 0.571 0.604 0.472 0.496 Correlation Inter 0.61 0.63 0.609 0.368 0.583 0.636 0.594 0.426 0.604 0.643 0.506

tree tree

Composition Category 20, 30km) on centered each subarea. 10, 5, 1, (0.5, radii increasing of buffers extracted multiple using at calculated were scales spatial time the indica to Predictors available. refer was predictor this Dates which for details). periods for text see 1; (Fig. subarea sample each at attributes outbreak 2 Table AcceptedConfiguration Article

: umr o ptnil n slce peitr aibe t dsrb vrain n SBW in variation describe to variables predictor selected and potential of Summary

forest of Proportion disturbance forest of size patch weighted Area forestold of Proportion forest of Proportion BasalArea Deciduous BasalArea Fir & Spruce BasalArea Fir Balsam Variable opening forest of Proportion openings of size patch weighted Area disturbance - - Candidate Landscape Factors affecting budworm dynamics affecting budworm Factors Landscape Candidate

% Mean % % Mean Mean Mean Unit % Mean

POpn[SpatialScale] AWMOpn[SpatialScale] PDist[SpatialScale] AwmPtch[SpatialScale] POld[SpatialScale][Year] PFor[SpatialScale][Year] Decid[ Hst[SpatialScale][Year] Fir[SpatialScale][Year] Abbreviation SpatialScale][Year]

ted as being calculated at multiple at calculated being as ted 1985 & 2002 1985 & 2002 1985 & Period Time 5yr 5yr 1980 of Average increment 5yr 1980 of Average increment 5yr 1980 of Average increment 5yr 1980 of Average 2000 1975 2002 1985 & - - - -

2000, 2000, 2000, 2000,

Yes Yes Scale Multi Yes Yes Yes Yes Yes Yes Yes

- Landsat Landsat TM Landsat TM Landsat Source Data Landsat Landsat and MSS Landsat TM Landsat and MSS Landsat TM Landsat and MSS Landsat TM Landsat and MSS Landsat TM Landsat TM Landsat TM

Climate Accepted Article

temperature summer Min. temperature spring Min. temperature winter Min. Index Growth Budworm Spruce patches spruce of COHESION fir patches of COHESION ‘This article is protected by copyright. All rights reserved.’ - fir

Mean Mean Mean Index Mean Mean

SumTmp SpTmp WintTmp ClimInd CoheHst[SpatialScale][Year] CoheFir[SpatialScale][Year]

Average of of Average 1971 of Average 2002 1985 & 2002 1985 & increment 1900 of Average 1900 of Average 1900 - - - - 2000 2000 2000 2000

No Yes Yes No No No

McKenney McKenney BIOSIM TM Landsat and MSS Landsat TM Landsat and MSS Landsat TM 2007 others and McKenney 2007 others and McKenney 2007 others and

Accepted Articlemultiple values separated by “/” character. the by indicated variable given a for selected scales spatial and/or periods time Multiple window. time 3 Table : umr fvrals eetd o nlso ntefnlRAmdl a ah outbreak each at models RDA final the in inclusion for selected variables of Summary 1983 1958 1957 1928 Full Window Temporal 2005 1984

- - -

Spruce Budworm Budworm Spruce disturbance of forest Proportion forest old of Proportion Area Basal Fir Balsam Selected Variable rprin of forest Proportion forest old of Proportion of forest Proportion Area Fir Basal disturbance of forest Proportion Forest Old of Proportion of forest Proportion Area Basal Fir & Spruce disturbance of forest size patch Area Index Growth -

weighted weighted

Time Average Time 2000 / 2002 / Period Time Average Time 2000 1975 2002 Average Time 2000 1975 2002 Average Time Average - - - - -

serie serie serie serie serie

s s s s s

Subarea Subarea km 30 / km 5 km 20 / km 10 5 km 1 km/ Spatial Scale km km 1 30k 0.5 km 1 km 5 km 5 km / 5km km 0.5 30 km Average Subarea Average

30 30

Min. summer summer temperature Min. spring temperature Min. patches fir of COHESION disturbance

Average Time Average Time Average Time - - - serie serie serie

s s s

Average Subarea Average Subarea 1 km