modelling climate change-driven treeline...

Journal of Ecology

2004

92

241ndash252

copy 2004 British Ecological Society

Blackwell Publishing Ltd

Modelling climate change-driven treeline shifts relative effects of temperature increase dispersal and invasibility

STEFAN DULLINGER THOMAS DIRNBOumlCK and GEORG GRABHERR

Institute of Ecology and Conservation Biology University of Vienna Althanstraszlige 14 A-1090 Vienna Austria and

Federal Environment Agency Spittelauer Laumlnde 5 A-1090 Vienna Austria

Summary

1

Global warming will probably shift treelines upslope in alpine areas and towards thepole in arctic environments However responses of regional treelines to climatic trendsover the last century do not show any clear trends We hypothesize that these equivocalresponses may partly be caused by limitation of dispersal andor recruitment that isspecies-specific to particular trees with potentially expanding ranges

2

To test this hypothesis we established and parameterized a temporally and spatiallyexplicit model of plant spread and analysed its sensitivity to (a) variation in predictedclimatic trends (b) the spatial distribution of recruits around a seed source and (c)variation in the resistance of resident non-woody vegetation to invasion We used datafrom a high mountain landscape of the Northern Calcareous Alps in Austria where thetreeline is dominated by

Pinus mugo

Turra a shrubby pine

3

Low growth rates and long generation times together with considerable dispersal andrecruitment limitation resulted in an overall slow range expansion under various climate-warming scenarios

4

Running the model for 1000 years predicted that the area covered by pines willincrease from 10 to between 24 and 59 of the study landscape

5

The shape of the dispersal curve and spatial patterns of competitively controlledrecruitment suppression affect range size dynamics at least as severely as does variationin assumed future mean annual temperature (between 0

deg

C and 2

deg

C above the currentmean) Moreover invasibility and shape of the dispersal curve interact with each otherdue to the spatial patterns of vegetation cover in the region

6

Ambiguous transient responses of individual treeline systems may thus originate notonly from variation in regional climatic trends but also from differences in speciesrsquodispersal and recruitment behaviour and in the intensity and pattern of resistance ofresident alpine vegetation to invasion

Key-words

alpine treeline climate change dispersal European Alps invisibility plantspread model

Journal of Ecology

(2003)

92

241ndash252

Introduction

Predicted global warming will probably affect rangesizes and the geographical distribution of biota (egGrabherr

et al

1994 Parmesan 1996 Sturm

et al

2001Thomas

et al

2001 Walther

et al

2002 Parmesan amp Yohe2003) As climatically determined ecotones arctic andalpine treelines are assumed to be particularly sensitiveto altered temperature regimes and climate warming isexpected to drive treelines upslope and poleward at theexpense of alpine and arctic ecosystems respectively

(Kittel

et al

2000 Hansen

et al

2001 Theurillat amp Guisan2001) However field studies and remote sensinganalyses that have investigated recent treeline shiftsin response to increasing temperatures during the lastcentury have provided ambiguous results that spanthe whole gradient from rapid dynamics to apparentlycomplete inertia (eg Kullman 1993 Lavoie amp Payette1994 Szeicz amp MacDonald 1995 Hessl amp Baker 1997Meshinev

et al

2000 Cullen

et al

2001 Masek 2001Motta amp Nola 2001 Sturm

et al

2001 Klasner amp Fagre2002 Kullman 2002) Apart from variation in regionalclimatic trends these equivocal findings probablyresult from differences between individual treelinesystems As an analogy to alien plant invasions one may

Correspondence Stefan Dullinger (tel +43 01 427754379fax +43 01 42779542 e-mail dullpflaphypphunivieacat)

242

S Dullinger T Dirnboumlck amp G Grabherr


Journal of Ecology

92

241ndash252

hypothesize that dispersal capacity of tree species (egKot

et al

1996 Neubert amp Caswell 2000 Bullock

et al

2002 Shigesada amp Kawasaki 2002 Watkinson amp Gill2002) and competitive interactions with residentalpine vegetation (Richardson amp Bond 1991 Magee ampAntos 1992 Rochefort amp Peterson 1996 Moir

et al

1999) will be major sources of variation in treelineresponses to climate change

Throughout most of the north-eastern CalcareousAlps the treeline is currently dominated by a shrubbypine (

Pinus mugo

Turra) In an earlier study (Dirnboumlck

et al

2003) we have used an equilibrium-based staticmodelling approach (Guisan amp Zimmermann 2000) toevaluate effects of predicted climate warming on theregional distribution of this species in some mountainranges at the north-eastern fringe of the Alps Resultsof these models suggest a considerable range expansionof

P mugo

at only moderate levels of temperature in-crease However additional work on

P mugo

dynamicsindicated that the possible range shift might be ham-pered by restricted dispersal as well as by competitiveinhibition of recruitment in dense grassland layersPredicted potential distributions may thus not be achievedin the mid-term (Dullinger Dirnboumlck amp Grabherr 2003)

Unfortunately most existing simulation models thatexplore potential range shifts and forest dynamicsin response to climate change either do not representtransient dynamics (in the case of equilibrium-basedbioclimatic habitat distribution models eg Guisan ampZimmermann 2000) or they disregard processes thatoccur beyond the boundaries of the simulation plot(in the case of gap models eg Loehle amp LeBlanc 1996)Models that integrate both the spatial and temporaldimensions have until now primarily been applied to thespread of alien plants (eg Higgins amp Richardson 19961999 Neubert amp Caswell 2000 Shigesada amp Kawasaki2002) However they have also been used to reconstructHolocene vegetation dynamics (eg Clark 1998) and theirpotential for simulating possible responses of plantspecies to predicted climate change has been inferred fromsuch applications (Pitelka amp Plant Migration WorkshopGroup 1997 Clark 1998 Higgins amp Richardson 1999)

We use a spatially and temporarily explicit plant spreadsimulator to address the relative roles of climate warmingdispersal capacity and competitive interactions with estab-lished alpine vegetation in determining the range expan-sion of

P mugo

Focusing on a 1000-year time frame weexamine how predictions of pine shrub distribution areaffected by (i) the degree of temperature increase (ii)the shape of the dispersal curve and (iii) the degree ofinvasibility of the alpine vegetation Moreover we examinethe possible interactions among these driving processes

Methods

The study area covers the treeline ecotone and the alpinebelt of Mt Hochschwab (47

deg

34

prime

to 47

deg

38

prime

latitude and

15

deg

00

prime

to 15

deg

18

prime

longitude uppermost summit 2273 masl) which is part of the north-eastern CalcareousAlps of Austria The mountain range is characterizedby displaced plateaus of different altitudes with surfacesshaped by Pleistocene glaciation and karst landformdevelopment Soils are predominantly lithic and rendzicLeptosols as well as chromic Cambisols Climaticconditions are temperate humid Mean annual temper-ature near the summit is approximately 0ndash2

deg

C whereannual precipitation averages 2000ndash2500 mm with amarked peak during the growth period The alpine areasare covered by snow for approximately 6ndash8 months ofthe year (OctoberndashMay) There is much fine-scale varia-tion in the duration of snow cover due to the rugged reliefand strong winds

Summer pasturing (June to September) in the MtHochschwab region dates back at least to the 16thcentury At least some historical livestock grazing hasbeen documented over 30 of the study area Since themid-19th century grazing intensity has decreased andmuch former pasture land has been abandoned so thattoday only 75 of the study area remains as pasture forfree-ranging cattle at a density of about 05 animals perhectare (Dullinger Dirnboumlck Greimler amp Grabherr2003)

The dominant woody plant species of the uppersubalpine belt is prostrate pine (

P mugo

) The upperlimit of single

P mugo

individuals is currently at about1950 m asl In fact the current subalpine belt isa mosaic of woody and non-woody vegetation Non-woody vegetation below the treeline mainly consists ofdifferent kinds of pastures and natural grasslands withthe latter covering disturbed sites like avalanche pathsand exposed ridges Above the treeline natural grass-lands dominate with a gradual switch from

Carex sem-pervirens

Vill to

Carex firma

Mygind grasslands withincreasing altitude Additionally rock faces scree andsnowbeds are widespread from the valley bottoms tothe summits

Pinus mugo

is an obligatory prostrate pine with adultcanopy height varying between

c

03 and 25 m in thestudy area (for convenience we use the term treelinefor the upper altitudinal range margin of

P mugo

despiteits shrubby growth form) Seedling establishmentseems to be inhibited by low light availability (Hafen-scherer amp Mayer 1986) and deep litter layers (Michiels1993) Thus within-stand regeneration is entirelydependent on clonal propagation by means of layer-ing Intensive multidirectional layering makes clonespotentially immortal and inhibits gap-phase regenera-tion processes in established stands (Hafenscherer ampMayer 1986) although recruitment of seedlings iscommon in grasslands Seeds of

P mugo

are primarilywind dispersed and secondary redistribution of seedsby birds and small mammals has been observed(Muumlller-Schneider 1986)

243

Modelling climate change driven treeline shifts


Journal of Ecology

92

241ndash252

A digital elevation model (DEM Austrian MappingAgency) with a cell size of 20 m served as main inputfor the representation of abiotic habitat conditions (seeDirnboumlck

et al

2003 for details) The following variableswere calculated (cf Table 1)

1

Climatic conditions represented by annual degree days(= days with a mean daily temperature gt 0

deg

C DD) solarradiation income at the beginning (15 May SRM) inthe middle (15 July SRJ) and at the end (15 SeptemberSRS) of the growth period and site water balance inAugust (WBA)

2

Topography characterized by slope inclination(SLOPE) a topographical wetness index (WET) a topo-graphical soil erosion index (EROS) and an estimateof topographically modified near-surface wind velocityduring strong north-westerly winds (WSP)

Additionally we provide data sets on bedrockmineralogy (GEO) spatial distribution of chromicCambisols and current vegetation cover Bedrockmineralogy was derived from recently updated geologicalmaps (scale 1 50 000 Geological Survey of Austriaunpublished information) The distribution of chromicCambisols was extrapolated from 557 sample pointsin the study area and adjacent mountain ranges usinga binary classification tree procedure with topographicalvariables (altitude slope EROS WET and theirinteractions) as predictors (misclassification rate 13)Information on current vegetation cover including pineshrub distribution comes from a fine-scale vegetationmap (1 10 000 Dirnboumlck

et al

1999) and high-resolutionIR-orthophotographs (acquired on 23 July 1994 pixelresolution 25 cm) All these data sets were re-sampledto meet the resolution of the DEM

In accordance with environmental descriptors the modellandscape is represented by a two-dimensional grid witha cell size of 20 m Overall it spans about 53 km

2

(131 901

cells) Pine shrub dynamics across this landscape aretracked as the changing percentage of

P mugo

coverper individual grid cell (= site) over time These changesresult from the spatial distribution of recruits origin-ating from a site and from the growth and mortality ofindividual pine shrubs growing at the respective site

Model formulation is based on several simplifyingassumptions

1

Time passes in discrete steps Our parameterizationdata demonstrate very low recruitment rates (see below)We thus decided to use a rather long time step of 50 yearsMoreover using a long time step avoids bias in recruit-ment rate estimation due to variable seed productionand seedling survival (eg Clark

et al

1999 De Stevenamp Wright 2002) Although little is known about inter-annual patterns of seed production for

P mugo

mast-ing behaviour is widespread among treeline species ofthe northern Calcareous Alps and masting frequencyis very low at high elevations (eg gt 10 years for

Piceaabies

(L) Karsten Mayer 1976)

2

The canopy increment of an individual

P mugo

shrubper time step is adequately represented by a growingcircle While the canopy shape is irregular the multistemgrowth-form (Hafenscherer amp Mayer 1986) justifies thisgeometrical approximation

3

Due to intensive clonal propagation by multidirec-tional layering mortality does not have an age-dependentcomponent but is entirely due to catastrophic dis-turbances The most important disturbance regimes areavalanches and extreme weather events Such events donot cause mortality of just one individual but usuallykill the whole population at a site We thus assumemortality events to reset the canopy cover of a grid cellto 0

4

Recruitment and mortality of pines include astochastic component that is due to unpredictability ofannual snow fall and melting processes unusual weatherevents or spatial patterns of seed and seedling predationand secondary seed dispersal (eg Vander Wall 1992Greene amp Johnson 1997)

5

Pine shrub populations of already densely coveredcells invade neighbouring cells vegetatively For ease

Table 1 Variables representing abiotic habitat conditions and sources they were derived from DEM = digital elevation modelSOLARFLUX NUATMOS and TAPES-G are software packages for calculating solar radiation income topographicallymodified near-surface wind velocity and different topographical indices respectively

Variable Source Abbreviation

Degree days DEM climate station data DDSolar radiation in May July and September

DEM SOLARFLUX (Dubayah amp Rich 1996)SRM SRJ SRS

Water balance in August DEM climate station data SOLARFLUX (Dubayah amp Rich 1996) WBAWind speed DEM climate station data NUATMOS (Ross et al 1988

Bachmann 1998) WSPSlope inclination DEM SLOPESoil erosion potential DEM TAPES-G (Gallant amp Wilson 1996) EROSTopographic wetness index DEM TAPES-G (Gallant amp Willson 1996) WETBedrock mineralogy Geological map GEODistribution of chromic Cambisols DEM 573 sample points SOIL

244



Journal of Ecology

92

241ndash252

of computation the invading front is assumed to bea straight line with a width of one cell size We used athreshold cover of 90 to trigger this process of vegetativeinvasion into adjacent sites

6

Rock faces lack appropriate space and soil substrateto support a dense

P mugo

canopy According to our fieldexperience we set the maximum value of pine shrub coverin rock habitats to 10

7

Besides seed availability and climatic constraintspine shrub colonization of debris cones is mainly con-trolled by mechanical stress Successful invasion is thusonly possible if the site is positioned on a ridge or ifthere is a neighbouring cell above the focal site thatprovides shelter as a result of existing dense pine coverThese rules produce the typical colonization pattern ofpine shrubs on debris cones ie propagating downhillin a conical fashion

The model starts from the current distribution ofpine shrub populations across the landscape Initiallyall currently occupied sites are assigned a cover of 100and an age of 100 years The model first calculates thefecundity of the population at each site as a function ofthe age of the individuals and of environmental condi-tions Sites with populations above a certain thresholdof fecundity (see below) are defined as seed sourcesNext it determines the number of newly germinatingrecruits per individual site during one time step Yearof germination is chosen randomly The canopy coverof all individuals at a site increases by a site-specificgrowth rate independently of age (Michiels 1993) Theresulting increase in pine shrub cover per site is aug-mented by vegetative invasion from neighbouringcells The model proceeds by simulating catastrophicmortality events A random number generator (0ndash1uniform distribution) is used and for each cell thisnumber is compared with the site-specific probabilityof occurrence of such an event If the number is greaterthan this probability for a given site its pine shrubcover is reset to 0 Lastly the model re-calculates theage of each

P mugo

population at the end of the timestep Age of a population is represented as a cover-weighted mean of the age of all its individuals

Recruitment growth fecundity and mortality functionswere estimated using data collected at 140 plots (each20

times

20 metre) in the study area and on adjacent mountainranges Selection of plots was based on a stratified ran-dom sampling design (Dullinger Dirnboumlck amp Grabherr2003) For growth and fecundity parameters we addition-ally used 196 pine shrub individuals selected duringrandom walks in the same areas

Recruitment dispersal functions and invasibility

The site-specific 50-year recruitment rate was deter-mined by counting the number of individuals youngerthan 50 years on each plot (see Dullinger Dirnboumlck amp

Grabherr 2003 for methods of age determination) Wefitted a recruitment kernel to these data using thedistance (two-dimensional Euclidean distance) to thenearest pine shrub stand (= grid cell with a pine shrubcover gt 10 as determined from aerial photographs) aspredictor Two alternative statistical models were applieda negative exponential and a restricted cubic spline withfour knots Restricted cubic splines are third-orderpolynomials within intervals of the predictor forcedto be smooth at the joining points (= knots) and con-strained to be linear in the tails (Stone amp Koo 1985Harrell 2001)

In simulation runs stochastic variation in the numberof recruits per site was implemented by drawing a ran-dom number from an exponential distribution for eachsite with the site-specific recruitment rate (determinedfrom the recruitment kernel) as the respective meanWe used an exponential distribution to mimic the errorpattern in fitted recruitment kernels for plots 0 and20 m from the nearest seed source Subsequently thepredicted number of recruits per site and time step wasweighted by two alternative invasibility layers Theselayers were derived from the vegetation map The firstone assumed equal invasibility (weighting factor = 1)across all types of alpine vegetation (but holding forestsand snowbeds uninvasible) the second one assigned eachplant community a specific invasibility value These werecalculated as the ratio between observed recruitmentrates in individual plant communities and expectedrates under a null model of equal invasibility using thesame parameterization data set as for demographicvariables and vary between 01 and 21 for the plantcommunities of the study area (see Dullinger Dirnboumlck ampGrabherr 2003 for details)

Growth

The vegetative growth rate of 231 individuals was meas-ured as the mean length-increment of a randomly selectedmajor branch between 1996 and 1999 This growth ratewas regressed against abiotic environmental conditionsusing ordinary least-squares regression Both linear andnon-linear effects were considered Non-linear effectswere tested for by restricted cubic splines with four knotsMoreover we tested for all two-way interactions amongpredictors Model and predictor significances wereobtained from the Wald test statistic assuming a chi-squaredistribution with one degree of freedom (Harrell 2001)The full model was reduced by backward eliminationknot reduction and linearization respectively (threshold

P

-value 005) The final coefficient of determination wascorrected for possible overfit by means of bootstrapping(1000 re-samples with replacement)

Fecundity

Using the number of cones produced by an individualas an indicator each pine shrub was assigned an ordinalfecundity value on a four-level scale (0ndash3) Fecundity

245



Journal of Ecology

92

241ndash252

was regressed against environmental conditions age ofthe individual and occurrence of damage (see below) bymeans of a proportional odds model (= ordinal logisticregression model Harrell 2001) using the same procedureas for the growth function Instead of least-squares

R

2

we report Nagelkerkersquos

R

2

and Somerrsquos

D

xy

as a measureof concordance between regression model predictionsand data (Table 2)

In the simulation model the population of a site wasdefined a seed source if its predicted fecundity valuewas greater than 1

Mortality

We used the distribution of dead individuals across oursampling plots to estimate the site-specific risk of adultmortality by means of logistic regression The procedureof regression analysis was the same as for growth andfecundity functions

In the simulation model the site-specific probabilityof finding a dead individual was set as the site-specificadult mortality per time step This somewhat arbitrarydefinition was found to produce realistic spatio-temporalmortality patterns in exploratory simulation runs Aval-anche paths which are unlikely ever to support closed

P mugo

cover on account of frequent severe disturbancedo not become overgrown whereas the establishmentand persistence of closed populations at less exposedsites is not affected

Furthermore we adapted the concept usually appliedto mortality routines in forest gap model formulations

(Keane

et al

2001) and defined sites with predictedgrowth rates lower than the minimum observed in ourparameterization data set to be not suitable for perman-ent

P mugo

establishment In the simulation this wasrealized by resetting the population of all such sites to0 cover after each time step

For all individuals we also recorded occurrence of dam-age due to climatic constraints (frost desiccation snow-iceabrasion) on a four-level ordinal scale Occurrence ofdamage turned out to be a significant predictor in bothgrowth and fecundity models We thus introduced the site-specific probability of damage (estimated by means of aproportional odds model cf Table 2) to the growth andfecundity functions interpreting it as a weighted interac-tion term of individual age and abiotic site conditions

A factorial design was established combining four dif-ferent climatic scenarios with the two distance-dependentrecruitment functions (exponential and restricted cubicspline) and the two alternative invasibility patterns(homogenous and varied)

As

P mugo

performance in parameterization plotsdid not show any significant effect of water balance (cfTable 2) climatic scenarios only took account of tem-perature rise A baseline scenario assumed that currenttemperature conditions remain unchanged (overall mean

Table 2 Regression functions for parameterizing the demographic processes of the Pinus mugo spread model LS = least squaresregression LR = logistic regression PO = proportional odds regression For each multiple regression function only significantpredictors are listed Distance = distance to nearest pine shrub stand Damage = probability of climatic damage Age = age ofpine shrub individual For abbreviations of abiotic habitat variables see Table 1 P lt 0001 P lt 001 P lt 005 R 2- andSomerrsquos Dxy-values were corrected for possible overfit by bootstrapping (1000 resamples) exp = exponential recruitment kernelrcs = restricted cubic spline recruitment kernel Regression equations and coefficients are listed in Appendix S1 date used forestablishing regression models are given in Appendix S2 (see Supplementary Material)

Model type Predictors R 2-orig R 2-corr Dxy-orig Dxy-corr P

Recruitment_EXP LS Distance 062 059 ndash ndash lt 00001Recruitment_RCS LS Distance 088 085 ndash ndash lt 00001

Growth LS DD 049 044 ndash ndash lt 00001DamageSOILWETSRJ

Fecundity PO Age 057 054 080 079 lt 00001GEODamage SRMDD

Mortality LR EROS 027 024 069 066 lt 00001WETDDSLOPE

Occurrence of damage PO DD 04 038 07 068 lt 00001SRJAgeSLOPEWET

246



Journal of Ecology

92

241ndash252

annual temperature in the study area for 1995ndash98calculated by linear regression of data from nearbymeteorological stations against altitude and geo-graphical longitude 12

deg

C) Scenario 2 was derived fromsimulation outputs of the global circulation modelECHAM4 (Roecker

et al

1996) downscaled for Austriaby Lexer

et al

(2001) According to this scenario a065

deg

C increase in mean annual temperature might beexpected for the study area by the year 2050 (means for2035ndash65 vs 1961ndash95) but temperature is assumed toremain constant thereafter Scenario 3 is as scenario 2but with a further increase to +12

degC above current meantemperature in total by the year 2100 and scenario 4has a further increase to +2 degC by the year 2150 Thedifferent scenarios were incorporated into the modelby adapting the degree day-values of all sites accordingto the assumed temperature regime

Overall this 2 times 4 times 2 combination of fixed effectsyields 16 different scenarios in a fully orthogonaldesign Ten replicates of each combination of treat-ments were run For each time step of each replicaterun we recorded the total percentage of pine shrubcover the elevation of the uppermost population andthe maximum distance from an existing seed sourceat which a new recruit had established To focus ex-clusively on climate change effects and to exclude theconfounding influence of unpredictable changes in landuse all scenarios were run under the assumption thatcattle grazing and logging will cease instantaneouslyTo minimize edge effects we applied the model to thewhole landscape (131 901 sites 527 km2) but resultsare reported only for the area above 1700 m asl (84 827sites 339 km2) was used to test for main effectsand interactions of climate scenario recruitment kerneland invasibility on the recorded response variablesAssumptions of were evaluated using diagnosticplots (normal probability plots eg Ellison 2001) andcalculating Cochranrsquos C for homogeneity of variances

Results

With an average of about 00005 individuals mminus2 yearminus1recruitment of Pinus mugo is generally sparse even whereseed sources are available within a radius of 20 m anddeclines sharply to 76 times 10minus6 or 59 times 10minus5 individuals mminus2

yearminus1 at 100 m depending on the recruitment kernel usedThe main differences between the two recruitment func-tions are that the restricted cubic spline-model predictshigher recruitment in the immediate vicinity of a seedsource and has a longer tail whereas the exponential modelsimulates more recruits at intermediate distances (50ndash300 m see Fig 1) Both functions provide significant fitsto the data though residual variance is considerably lowerfor the more flexible restricted cubic spline kernel (seeTable 2)

Mean annual growth is extremely slow with a 4-yearaverage of only about 5 cm yearminus1 (minimum 15 cm yearminus1maximum 129 cm yearminus1) Of the abiotic site conditionsvariation in growth rate is most sensitive to temper-ature (see Table 2 Fig 2)

Fecundity levels were primarily controlled by ageof individuals pine shrubs usually do not start coneproduction until they are about 15ndash20 years old andhigh fecundity rates (ordinal fecundity value 3) arerarely achieved by individuals younger than 50 yearsFecundity becomes independent of age after 80 yearsand is affected only slightly by temperature (Table 2Fig 2)

Topography is the most important factor controllingadult mortality (Table 2) Average mortality is low (2ndash3) with high values spatially clustered on exposedridges steep slopes with high erosion potential and riskof avalanches and at the base of slopes where snowaccumulates Among climatic variables temperature isagain the most effective predictor However in contrastto growth and fecundity functions its effect is non-linear with lowest mortality (ie the highest survivalprobability) at intermediate DD-values (Fig 2)

Pine shrubs currently cover 10 of the study areaWithin the next 1000 years the model predicts P mugoto increase its cover to between 24 and 59 depend-ing on the degree of climatic warming the assumedshape of the recruitment kernel and the invasibilitymatrix used in simulation runs (Fig 3) Within-scenariovariance of predicted values is generally low

As expected the rate of pine shrub expansion increaseswith rising temperatures The effect of higher temper-ature is most pronounced when comparing the baseline

Fig 1 Recruitment kernel of Pinus mugo fitted to data from140 field plots Solid line restricted cubic spline function Dashedline negative exponential function Each site is 400 m2 thetime step is 50 years Distance = distance to the nearest sitewith P mugo cover gt 10 Points represent average recruit-ment intensities at each distance

247Modelling climate change driven treeline shifts

copy 2004 British Ecological Society Journal of Ecology 92 241ndash252

scenario with an assumed increase of 065 degC and lowestwhen switching from a +12 degC scenario to a +2 degCscenario (Fig 4a)

Assuming an exponential recruitment kernel resultsin considerable acceleration of pine shrub spreadAfter 1000 years simulation runs with an exponentialkernel predict that P mugo will cover an area nearly50 larger on average than simulations with a restrictedcubic spline-kernel (Fig 4b) Similarly spatially variedinvasibility of the resident vegetation facilitates pinespread (an additional increase of about 30 in the areacovered by P mugo after 1000 years compared with thehomogenous model Fig 4c)

results demonstrate that all main effects andall interactions significantly affect model predictions coefficients are highest for the +065 degC climatescenario and the recruitment function used (Fig 5) Theinteraction between shape of the recruitment kerneland invasibility pattern is especially pronouncedwhereas all others are of minor importance

The uppermost position of P mugo individualsshifts from the present value (1935 m asl) to between2076 m and the highest peak of the model area (2273 masl) depending on the scenario considered Maineffects and interactions change model predictions in asimilar way to that observed for the area covered bypine shrubs Again the most pronounced interaction isthat between recruitment kernel and invasibility pat-tern (Table 3)

Figure 6 illustrates why this latter interaction is sali-ent in both cases If an exponential kernel is used themaximum distance of a new recruit from the nearestseed source is only marginally influenced by the wayinvasibility is modelled In contrast when applying arestricted cubic spline kernel this maximum distance isstrongly affected by the invasibility pattern

Discussion

Results of this simulation study indicate that whilepine shrubs will invade and displace current alpinevegetation under future climate change scenarios theyare likely to do so rather slowly This inertia is due torecruitment and dispersal limitation as well as to long

Table 3 Effects of climate change scenario (ClimScen) shape of the recruitment kernel (RecrKern) spatial invasibility pattern(Invas) and their interaction terms on the area predicted to be covered by P mugo in 1000 years time (= Area) the uppermostposition of P mugo individuals after 1000 years (= Altitude) and the maximum distance from a seed source at which a new recruitis predicted to establish during one time step (= MaxDistance) Results of fixed-effects factorial df = degrees of freedom

Treatment df

Area Altitude Max Distance

F-value P F-value P F-value P

ClimScen 3 26150 lt 00001 530 lt 00001 53 0001RecrKern 1 32568 lt 00001 1817 lt 00001 02 065Invas 1 8020 lt 00001 929 lt 00001 1234 lt 00001ClimScen RecrKern 3 900 lt 00001 8 00005 32 002ClimScen Invas 3 181 lt 00001 23 lt 00001 35 002RecrKern Invas 1 4241 lt 00001 347 lt 00001 1124 lt 00001

Fig 2 Partial effects of temperature (number of days withmean temperature gt 0 degC per year) on growth fecundity andmortality of P mugo in multivariate regression models (cfTable 2) Dashed lines represent 95 confidence intervalsAdditional predictors are set to the following values Growthdamage = 0 SOIL = 1 SRJ = 271 MJ WET = 594 Fecundityage = 40 damage = 0 SRM = 258 MJ GEO = limestoneMortality WET = 59 EROS = minus027 SLOPE = 185deg Forabbreviations see Table 1

248S Dullinger T Dirnboumlck amp G Grabherr


generation times and slow growth rates Although ourestimates of growth rates and generation times agreewith values reported in Schroeter (1926) and Michiels(1993) comparable data on the intensity and spatialpattern of P mugo recruitment in subalpine and alpineenvironments are lacking In general recruitmentlimitation seems to be a widespread phenomenon intemperate forests (Clark et al 1998 1999) but the num-ber of recruits detected in our study is particularlylow Ribbens et al (1994) for example have definedrecruitment limitation as occurring where the expectednumber of recruits drops below one individual mminus2

yearminus1 Mayer (1976) reports 01ndash05 successful recruit-ment events mminus2 yearminus1 for Picea abies in gaps of sub-alpine spruce forests which are dominant below the pineshrub belt in our study area Moreover the initially lowrate declines sharply with distance from seed sourcesAs seed dispersal distance is closely linked to seedrelease height (eg Nathan amp Muller-Landau 2000) weassume that this sharp decline in recruitment is causedby narrow seed shadows of the low growing pine shrubs

Nevertheless despite such a slow response pineshrub cover of the model landscape is predicted toincrease considerably over the next 1000 years Withclimatic conditions unchanged (control scenario) mostof this range expansion is due to encroachment ontosummer farms both those abandoned during thelast 150 years and those set as abandoned for the

Fig 3 Predicted increase of the area covered by P mugo during the next 1000 years Maximum minimum and median of 160model runs Error bars represent standard deviations

Fig 4 Effects of climate change scenario shape of the recruit-ment kernel and spatial pattern of invasibility on the areapredicted to be covered by P mugo 1000 years from the presentday Boxes show the limits of the middle half of all simulationresults lines inside the boxes represent the median and lines tothe top and the bottom of each box highlight minima andmaxima predicted under the respective variable value exp =exponential recruitment kernel rcs = restricted cubic splinerecruitment kernel



simulation runs However even under a moderate levelof temperature rise the area predicted to be encroachedby pine shrubs by the year 3000 greatly exceeds theextent of both historically and currently used pastures

The rate of this range expansion is significantlyaffected by the climatic scenario assumed the recruit-ment kernel used and the supposed spatial pattern ofinvasibility As expected expansion rate increases withrising temperatures an effect that directly follows fromhigher growth rates and fecundity levels and an overalllower risk of mortality for P mugo For growth rates ofPinus cembra L another European treeline pine thestimulating effect of climate warming has been demon-strated to be due to higher sink activity rather than toincreased carbon assimilation (Hoch et al 2002) Pro-vided that these results are applicable to other pinespecies confounding effects of rising atmospheric CO2

concentrations linked with climate warming whichwere not considered in our simulations may thus be ofminor significance

The accelerating effect of using an exponentialdispersal kernel seems somewhat counter-intuitiveas this function approaches zero more rapidly thanthe restricted cubic spline kernel (cf Figure 1) and thelength of the kernel lsquotailrsquo has been repeatedly demon-strated to control the speed of migration and invasion(Kot et al 1996 Clark 1998 Higgins amp Richardson1999 Neubert amp Caswell 2000 Bullock et al 2002Shigesada amp Kawasaki 2002) However the probabilityof recruitment in the tail of the restricted cubic splinekernel is too low to create significant differences in themaximum distance of successful dispersal events pertime step (cf Table 3) Thus higher predicted recruit-ment at intermediate distances is responsible for thelsquokernel-effectrsquo on pine shrub expansion In accordancewith the simulation results of Shigesada amp Kawasaki(2002) more intense short distance dispersal as pre-dicted by the restricted cubic spline kernel has negli-gible effects on simulation results

We found a pronounced interaction between dis-persal and invasibility patterns (Fig 6 Table 3)The assumption of spatially differential invasibilityincreases the maximum distance of successful dispersalevents when using a restricted cubic spline kernel butinvasibility patterns hardly affect maximum dispersaldistances under an exponential recruitment kernelThis interaction is explained by peculiarities ofregional vegetation distribution The grassland com-munity that is most invasible for pine shrubs (swards ofCarex firma see Dullinger Dirnboumlck amp Grabherr 2003)is rare in subalpine regions but predominates above thecurrrent treeline in alpine areas Thus under assump-tions of differential invasibility expansion of pineshrubs into the alpine belt is facilitated The interactionof dispersal and invasibility patterns is caused by anincreased probability that seeds dispersed over longerdistances (ie into these alpine areas) will reach a habitat

Fig 5 coefficients for factor levels (a) area predictedto be covered by P mugo in 1000 years time (b) uppermostposition of P mugo individuals in 1000 years time ClimScen =climate change scenario Invas = spatial pattern of invasibilityRecrKern = shape of the recruitment kernel Colons symbolizeinteraction terms All main effects and interactions are signi-ficant (P lt 005)

Fig 6 Interacting effects of invasibility pattern and shape ofthe recruitment kernel on the maximum distance from a seedsource at which a new recruit of P mugo is predicted toestablish during one 50-year time step Hom = invasibilityassumed to be homogeneous across different alpine plantcommunities var = invasibility assumed to differ betweenplant communities Dashed line = recruitment kernel modelledwith a restricted cubic spline function solid line = recruitmentkernel modelled with a negative exponential kernel



with favourable recruitment conditions Assuming dif-ferential invasibility strengthens the effect of the thinbut longer tail of the restricted cubic spline kernel dueto the region-specific spatial distribution of habitatssuitable for recruitment

Although invasibility of resident vegetation has longbeen recognized as a major factor controlling alienplant spread (Richardson amp Bond 1991 Wiser et al1998 Shea amp Chesson 2002) its possible impact onthe rate of altitudinal treeline shifts driven by climatechange has been largely ignored However there isevidence that resident vegetation may strongly affectrates of woody plant encroachment into non-forestsubalpine habitats (Magee amp Antos 1992 Rochefortamp Peterson 1996 Moir et al 1999 Dullinger Dirnboumlckamp Grabherr 2003) Competition with herbaceousvegetation was also invoked as a possible explanationfor regionally observed unexpectedly slow expansionof boreal forests into arctic tundra in recent decades(Masek 2001) Such encroachment processes prob-ably resemble the invasion of exotic pine species intonative non-forest vegetation (Richardson amp Higgins1998 Higgins et al 2001) Our results confirm this func-tional similarity Like the interaction of dispersal andinvasibility patterns the accelerating effect of spati-ally varied invasibility is due to the particular spatialarrangement of plant communities easily invasiblecommunities dominate at and above the current treeline(Dirnboumlck et al 1999) Under predicted climate warm-ing scenarios it is exactly this altitudinal belt wheremost of the terrain that can be newly colonized by P mugois concentrated

Our simulation model does not consider possibleclimate change effects on recruitment Pine shrub spreadis likely to be driven mainly by exceptionally favourableyears for recruitment (Kullman 1993) the frequency ofwhich may increase with climate warming This couldconsiderably accelerate range expansion driven byclimate change However our parameterization data setprovides little evidence for major effects of temperatureon local-scale recruitment intensity during the last50 years (Dullinger Dirnboumlck amp Grabherr 2003)When controlling for distance to seed sources differ-ences in recruitment success between lower subalpineand treeline populations were only marginally signi-ficant during this time span Nevertheless we cannotexclude the possibility that exceptionally extendedperiods of warmer summers may produce thresholdeffects on recruitment intensity at and above the cur-rent treeline that may cause our simulation results forP mugo to be underestimates However even enhancedrecruitment of P mugo under climate warming wouldnot diminish the codominant effect of dispersal andcompetition on its spread

In conclusion the results of this modelling studysuggest that patterns and rates of transient treelinedynamics that are driven by climate change may beidiosyncratic Species will not only vary in their responseto increasing temperatures in terms of their growth rates

as they may also possess different dispersal capacitiesand competitive abilities during their recruitment phaseOur study demonstrates that these species-specific traitsmay affect future range dynamics as much as if notmore than variation in regional climatic trends More-over they interact with each other and with spatialvegetation patterns This further complicates the taskof accurately predicting transient treeline dynamicsdriven by climate change

Acknowledgements

We are grateful to M Steinkellner G Mandl G Brydaand R Tscheliesnig for meteorological geological andland use data to J Greimler I Schmidsberger NSauberer and T Englisch for field data collection andto N Zimmermann and A Bachmann for GIS softwareD Moser M Abensperg-Traun L Haddon and twoanonymous referees provided helpful comments onearlier versions of the manuscript The study was fundedby the Austrian Federal Ministry for Education Scienceand Culture and the Water Department of the Vienneseadministration

Supplementary material

The following material is available from httpwwwblackwellpublishingcomproducts journalssuppmatJECJEC872JEC872smhtm

Appendix S1 Parameter values and regression equa-tions for models of recruitment growth fecunditymortality and probability of damage of Pinus mugo

Appendix S2 Data used for regression models of growthfecundity mortality and probability of damage ofPinus mugo

References

Bachmann A (1998) Coupling NUATMOS with the GISARCInfo Final Report for MINERVE 2 Department ofGeography of the ETH Zuumlrich Division of Spatial DataHandling

Bullock JB Moy IL Pywell RF Coulson SJ NolanAM amp Caswell H (2002) Plant dispersal and colonizationprocesses at local and landscape scales Dispersal Ecology(eds JM Bullock RE Kenward amp RS Hails) pp 279ndash302 Blackwell Scientific Oxford

Clark JS (1998) Why trees migrate so fast Confronting the-ory with dispersal biology and the paleorecord AmericanNaturalist 152 204ndash224

Clark JS Beckage B Camill P Cleveland B HilleRis-Lambers J Lichter et al (1999) Interpreting recruitmentlimitation in forests American Journal of Botany 86 1ndash16

Clark JS Macklin E amp Wood L (1998) Stages and spatialscales of recruitment limitation in southern Appalachianforests Ecological Monographs 68 213ndash235

Cullen L Stewart GH Duncan RP amp Palmer G (2001)Disturbance and climate warming influences on New Zea-land Nothofagus tree-line population dynamics Journal ofEcology 89 1061ndash1071



De Steven D amp Wright J (2002) Consequences of variablereproduction for seedling recruitment in three neotropicaltree species Ecology 83 2315ndash2317

Dirnboumlck T Dullinger S Gottfried M amp Grabherr G(1999) Die Vegetation des Hochschwab (Steiermark) ndashAlpine und Subalpine Stufe Mitteilungen Des Naturwis-senschaftlichen Vereins der Steiermark 129 111ndash251

Dirnboumlck T Dullinger S amp Grabherr G (2003) A regionalimpact assessment of climate and land use change on alpinevegetation Journal of Biogeography 30 401ndash418

Dubayah R amp Rich PM (1996) GIS-based solar radiationmodeling GIS and Environmental Modeling (eds MFGoodchild LT Steyaert amp BO Parks) pp 129ndash134 JohnWiley amp Sons New York

Dullinger S Dirnboumlck T amp Grabherr G (2003) Patternsof shrub invasion into high mountain grasslands of theNorthern Calcareous Alps (Austria) Arctic Antarctic andAlpine Research 35 434ndash441

Dullinger S Dirnboumlck T Greimler J amp Grabherr G(2003) A resampling approach for evaluating effects ofpasture abandonment on subalpine plant species diversityJournal of Vegetation Science 14 243ndash252

Ellison AM (2001) Exploratory data analysis and graphicdisplay Design and Analysis of Ecological Experiments2nd edn (eds SM Scheiner amp J Gurevitch) pp 37ndash62Oxford University Press Oxford

Gallant JC amp Wilson JP (1996) Tapes-G a grid-basedterrain analysis program for the environmental scienceComputers and Geoscience 22 713ndash722

Grabherr G Gottfried M amp Pauli H (1994) Climate effectson mountain plants Nature 369 448

Greene DF amp Johnson EA (1997) Secondary dispersal oftree seeds on snow Journal of Ecology 85 329ndash340

Guisan A amp Zimmermann NE (2000) Predictive habitatdistribution models in ecology Ecological Modelling 135147ndash186

Hafenscherer J amp Mayer H (1986) Standort AufbauEntwicklungsdynamik und Verjuumlngung von Latschen-bestaumlnden im Karwendeltal Tirol Schweizer Zeitschriftfuumlr Forstwesen 137 177ndash203

Hansen AJ Neilson RP Dale VH Flather CH Iverson LRCurrie DJ et al (2001) Global change in forests responseof species communities and biomes Bioscience 51 765ndash779

Harrell FE (2001) Regression Modeling Strategies SpringerNew York

Hessl AE amp Baker WL (1997) Spruce and fir regenerationand climate in the forest-tundra ecotone of Rocky Moun-tains National Park Colorado USA Arctic and AlpineResearch 29 173ndash183

Higgins SI amp Richardson DM (1996) Modeling invasiveplant spread the role of plantndashenvironment interactionsand model structure Ecology 77 2043ndash2054

Higgins SI amp Richardson DM (1999) Predicting plantmigration rates in a changing world the role of long-distance dispersal American Naturalist 153 464ndash475

Higgins SI Richardson DM amp Cowling RM (2001)Validation of a spatial simulation model of a spreading alienplant population Journal of Applied Ecology 38 571ndash584

Hoch G Popp M amp Koumlrner C (2002) Altitudinal increaseof mobile carbon pools in Pinus cembra suggests sinklimitation of growth at the Swiss treeline Oikos 98 361ndash374

Keane R Austin M Field C Huth A Lexer MJPeters D et al (2001) Tree mortality in gap models applica-tion to climate change Climatic Change 51 509ndash540

Kittel TGF Steffen WL amp Chapin FS III (2000) Globaland regional modelling of arctic-boreal vegetation distri-bution and its sensitivity to altered forces Global ChangeBiology 6 1ndash18

Klasner FL amp Fagre DB (2002) A half century of change inalpine treeline patterns at Glacier National Park MontanaUSA Arctic Antarctic and Alpine Research 34 49ndash56

Kot M Lewis MA amp van den Driessche P (1996) Disper-sal data and the spread of invading organisms Ecology 772027ndash2042

Kullman L (1993) Pine (Pinus sylvetris L) tree-limit surveil-lance during recent decades Central Sweden Arctic andAlpine Research 25 24ndash31

Kullman L (2002) Rapid recent range-margin rise of tree andshrub species in the Swedish Scandes Journal of Ecology90 68ndash77

Lavoie C amp Payette S (1994) Recent fluctuations of thelichen-spruce forest limit in subarctic Quebec Journal ofEcology 82 725ndash734

Lexer MJ Houmlnninger K Scheifinger H Matulla ChGroll N Kromp-Kolb H et al (2001) The sensitivity ofthe Austrian forests to scenarios of climatic change Mono-graphien des Umweltbundesamt Wien 132 1ndash132

Loehle C amp LeBlanc D (1996) Model-based assessments ofclimate change effects on forests a critical review Ecolo-gical Modelling 90 1ndash31

Magee TK amp Antos JA (1992) Tree invasion into a mountain-top meadow in the Oregon Coast Range USA Journal ofVegetation Science 3 485ndash494

Masek JG (2001) Stability of boreal forest stands duringrecent climate change evidence from Landsat satelliteimagery Journal of Biogeography 28 967ndash976

Mayer H (1976) Gebirgswaldbau ndash Schutzwaldpflege FischerStuttgart

Meshinev T Apostolova I amp Koleva ES (2000) Influenceof warming on timberline rising a case study on Pinuspeuce Griseb in Bulgaria Phytocoenologia 30 431ndash438

Michiels HG (1993) Die Stellung einiger Baum- und Strauchar-ten in der Struktur und Dynamik der Vegetation im Bereichder hochmontanen und subalpinen Waldstufe der BayrischenKalkalpen Forstliche Forschungsberichte Muumlnchen 135 1ndash300

Moir WH Rochelle SG amp Schoettle AW (1999) Micro-scale patterns of tree establishment near upper treeline SnowyRange Wyoming USA Arctic Antarctic and Alpine Re-search 31 379ndash388

Motta R amp Nola P (2001) Growth trends and dynamics insub-alpine forest stands in the Varaita Valley (PiedmontItaly) and their relationships with human activities andglobal change Journal of Vegetation Science 12 219ndash230

Muumlller-Schneider P (1986) Verbreitungsbiologie der Bluumltenp-flanzen Graubuumlndens Veroumlffentlichungen des GeobotanischenInstituts der ETH Stiftung Ruumlbel 85 1ndash263

Nathan R amp Muller-Landau C (2000) Spatial patterns ofseed dispersal their determinants and consequences forrecruitment Trends in Ecology and Evolution 15 278ndash285

Neubert MG amp Caswell H (2000) Demography and dis-persal calculation and sensitivity analysis of invasion speedfor structured populations Ecology 81 1613ndash1628

Parmesan C (1996) Climate and speciesrsquo range Nature 382765ndash766

Parmesan C amp Yohe G (2003) A globally coherent finger-print of climate change impacts across natural systemsNature 421 37ndash42

Pitelka LF amp Plant Migration Workshop Group (1997)Plant migration and climate change American Scientist85 464ndash473

Ribbens E Silander JA amp Pacala SW (1994) Seedlingrecruitment in forests calibrating models to predict patternsof tree seedling dispersion Ecology 75 1794ndash1806

Richardson DM amp Bond WJ (1991) Determinants of plantdistribution evidence from pine invasions American Nat-uralist 137 639ndash668

Richardson DM amp Higgins SI (1998) Pines as invaders inthe southern hemishpere Ecology and Biogeography ofPinus (ed DM Richardson) pp 450ndash473 CambridgeUniversity Press Cambridge



Rochefort RM amp Peterson DL (1996) Temporal and spatialdistribution of trees in subalpine meadows of Mount RainierNational Park Washington USA Arctic and Alpine Re-search 28 52ndash59

Roecker E Oberhuber JM Bacher A Christopch M ampKirchner I (1996) ENSO variability and atmosphericresponse in a global coupled atmosphere-ocean GCMClimate Dynamics 12 737ndash754

Ross DG Smith IN Manins PC amp Fox DG (1988)Diagnostic wind field modeling for complex terrain modeldevelopment and testing Journal of Applied Meteorology27 785ndash796

Schroeter C (1926) Das Pflanzenleben der Alpen AlbertRaustein Zuumlrich

Shea K amp Chesson P (2002) Community ecology theory asa framework for biological invasions Trends in Ecology andEvolution 17 170ndash176

Shigesada N amp Kawasaki K (2002) Invasion and the rangeexpansion of species effects of long-distance dispersalDispersal Ecology (eds JM Bullock RE Kenward ampRS Hails) pp 350ndash373 Blackwell Scientific Oxford

Stone CJ amp Koo CY (1985) Additive Splines in StatisticsProceedings of the Statistical Computing Section ASA 45ndash48 American Statistical Association Washington DC

Sturm M Racine C amp Tape K (2001) Increasing shrubabundance in the Arctic Nature 411 546ndash547

Szeicz JM amp MacDonald GM (1995) Recent white sprucedynamics at the subarctic alpine treeline of north-westernCanada Journal of Ecology 83 873ndash885

Theurillat J-P amp Guisan A (2001) Potential impact ofclimate change on vegetation in the European Alps a reviewClimatic Change 50 77ndash109

Thomas CD Bodsworth EJ Wilson RJ SimmonsAD Davies ZG Musche M et al (2001) Ecological andevolutionary processes at expanding range marginsNature 411 577ndash581

Vander Wall SB (1992) The role of animals in dispersing alsquowind-dispersedrsquo pine Ecology 73 614ndash621

Walther G-R Post E Convey P Menzel A ParmesanC Beebee TJC et al (2002) Ecological responses torecent climate change Nature 416 389ndash395

Watkinson AR amp Gill JA (2002) Climate change anddispersal Dispersal Ecology (eds JM Bullock RE Ken-ward amp RS Hails) pp 410ndash430 Blackwell ScientificOxford

Wiser SK Allen RB Clinton PW amp Platt KH (1998)Community structure and forest invasion by an exotic herbover 23 years Ecology 76 2071ndash2081

Received 22 May 2003 revision accepted 4 November 2003 Handling Editor Malcolm Press

242



Journal of Ecology

92

241ndash252

hypothesize that dispersal capacity of tree species (egKot

et al

1996 Neubert amp Caswell 2000 Bullock

et al

2002 Shigesada amp Kawasaki 2002 Watkinson amp Gill2002) and competitive interactions with residentalpine vegetation (Richardson amp Bond 1991 Magee ampAntos 1992 Rochefort amp Peterson 1996 Moir

et al

1999) will be major sources of variation in treelineresponses to climate change

Throughout most of the north-eastern CalcareousAlps the treeline is currently dominated by a shrubbypine (

Pinus mugo

Turra) In an earlier study (Dirnboumlck

et al

2003) we have used an equilibrium-based staticmodelling approach (Guisan amp Zimmermann 2000) toevaluate effects of predicted climate warming on theregional distribution of this species in some mountainranges at the north-eastern fringe of the Alps Resultsof these models suggest a considerable range expansionof

P mugo

at only moderate levels of temperature in-crease However additional work on

P mugo

dynamicsindicated that the possible range shift might be ham-pered by restricted dispersal as well as by competitiveinhibition of recruitment in dense grassland layersPredicted potential distributions may thus not be achievedin the mid-term (Dullinger Dirnboumlck amp Grabherr 2003)

Unfortunately most existing simulation models thatexplore potential range shifts and forest dynamicsin response to climate change either do not representtransient dynamics (in the case of equilibrium-basedbioclimatic habitat distribution models eg Guisan ampZimmermann 2000) or they disregard processes thatoccur beyond the boundaries of the simulation plot(in the case of gap models eg Loehle amp LeBlanc 1996)Models that integrate both the spatial and temporaldimensions have until now primarily been applied to thespread of alien plants (eg Higgins amp Richardson 19961999 Neubert amp Caswell 2000 Shigesada amp Kawasaki2002) However they have also been used to reconstructHolocene vegetation dynamics (eg Clark 1998) and theirpotential for simulating possible responses of plantspecies to predicted climate change has been inferred fromsuch applications (Pitelka amp Plant Migration WorkshopGroup 1997 Clark 1998 Higgins amp Richardson 1999)

We use a spatially and temporarily explicit plant spreadsimulator to address the relative roles of climate warmingdispersal capacity and competitive interactions with estab-lished alpine vegetation in determining the range expan-sion of

P mugo

Focusing on a 1000-year time frame weexamine how predictions of pine shrub distribution areaffected by (i) the degree of temperature increase (ii)the shape of the dispersal curve and (iii) the degree ofinvasibility of the alpine vegetation Moreover we examinethe possible interactions among these driving processes

Methods

The study area covers the treeline ecotone and the alpinebelt of Mt Hochschwab (47

deg

34

prime

to 47

deg

38

prime

latitude and

15

deg

00

prime

to 15

deg

18

prime

longitude uppermost summit 2273 masl) which is part of the north-eastern CalcareousAlps of Austria The mountain range is characterizedby displaced plateaus of different altitudes with surfacesshaped by Pleistocene glaciation and karst landformdevelopment Soils are predominantly lithic and rendzicLeptosols as well as chromic Cambisols Climaticconditions are temperate humid Mean annual temper-ature near the summit is approximately 0ndash2

deg

C whereannual precipitation averages 2000ndash2500 mm with amarked peak during the growth period The alpine areasare covered by snow for approximately 6ndash8 months ofthe year (OctoberndashMay) There is much fine-scale varia-tion in the duration of snow cover due to the rugged reliefand strong winds

Summer pasturing (June to September) in the MtHochschwab region dates back at least to the 16thcentury At least some historical livestock grazing hasbeen documented over 30 of the study area Since themid-19th century grazing intensity has decreased andmuch former pasture land has been abandoned so thattoday only 75 of the study area remains as pasture forfree-ranging cattle at a density of about 05 animals perhectare (Dullinger Dirnboumlck Greimler amp Grabherr2003)

The dominant woody plant species of the uppersubalpine belt is prostrate pine (

P mugo

) The upperlimit of single

P mugo

individuals is currently at about1950 m asl In fact the current subalpine belt isa mosaic of woody and non-woody vegetation Non-woody vegetation below the treeline mainly consists ofdifferent kinds of pastures and natural grasslands withthe latter covering disturbed sites like avalanche pathsand exposed ridges Above the treeline natural grass-lands dominate with a gradual switch from

Carex sem-pervirens

Vill to

Carex firma

Mygind grasslands withincreasing altitude Additionally rock faces scree andsnowbeds are widespread from the valley bottoms tothe summits

Pinus mugo

is an obligatory prostrate pine with adultcanopy height varying between

c

03 and 25 m in thestudy area (for convenience we use the term treelinefor the upper altitudinal range margin of

P mugo

despiteits shrubby growth form) Seedling establishmentseems to be inhibited by low light availability (Hafen-scherer amp Mayer 1986) and deep litter layers (Michiels1993) Thus within-stand regeneration is entirelydependent on clonal propagation by means of layer-ing Intensive multidirectional layering makes clonespotentially immortal and inhibits gap-phase regenera-tion processes in established stands (Hafenscherer ampMayer 1986) although recruitment of seedlings iscommon in grasslands Seeds of

P mugo

are primarilywind dispersed and secondary redistribution of seedsby birds and small mammals has been observed(Muumlller-Schneider 1986)

243



Journal of Ecology

92

241ndash252


et al


1


deg


2



et al



2

(131 901


P mugo



1


et al


P mugo


Piceaabies


2


P mugo


3


4


5








244



Journal of Ecology

92

241ndash252


6


P mugo


7



P mugo



times






Growth


P


Fecundity


245



Journal of Ecology

92

241ndash252


R

2


R

2

and Somerrsquos

D

xy



Mortality



P mugo



(Keane

et al


P mugo




As

P mugo









246



Journal of Ecology

92

241ndash252


deg


et al


et al


deg




Results
















Discussion



Treatment df




























Acknowledgements






References







































































243



Journal of Ecology

92

241ndash252


et al


1


deg


2



et al



2

(131 901


P mugo



1


et al


P mugo


Piceaabies


2


P mugo


3


4


5








244



Journal of Ecology

92

241ndash252


6


P mugo


7



P mugo



times






Growth


P


Fecundity


245



Journal of Ecology

92

241ndash252


R

2


R

2

and Somerrsquos

D

xy



Mortality



P mugo



(Keane

et al


P mugo




As

P mugo









246



Journal of Ecology

92

241ndash252


deg


et al


et al


deg




Results
















Discussion



Treatment df




























Acknowledgements






References







































































244



Journal of Ecology

92

241ndash252


6


P mugo


7



P mugo



times






Growth


P


Fecundity


245



Journal of Ecology

92

241ndash252


R

2


R

2

and Somerrsquos

D

xy



Mortality



P mugo



(Keane

et al


P mugo




As

P mugo









246



Journal of Ecology

92

241ndash252


deg


et al


et al


deg




Results
















Discussion



Treatment df




























Acknowledgements






References







































































245



Journal of Ecology

92

241ndash252


R

2


R

2

and Somerrsquos

D

xy



Mortality



P mugo



(Keane

et al


P mugo




As

P mugo









246



Journal of Ecology

92

241ndash252


deg


et al


et al


deg




Results
















Discussion



Treatment df




























Acknowledgements






References







































































246



Journal of Ecology

92

241ndash252


deg


et al


et al


deg




Results
















Discussion



Treatment df




























Acknowledgements






References














































































Discussion



Treatment df




























Acknowledgements






References






























































































Acknowledgements






References







































































modelling climate change-driven treeline...

Documents