Species coexistence in a mixed Mediterranean pine forestSpatio-temporal variability in trade-offs between facilitationand competition

Alicia Ledo auArr Isabel Cantildeellas a Ignacio Barbeito b Francisco Javier Gordo c Rafael A Calama aGuillermo Gea-Izquierdo d

a INIA-CIFOR Ctra La Coruntildea km 75 28040 Madrid Spainb INRA UMR1092 Laboratoire drsquoEtude des Ressources Forecirct Bois (LERFoB) Centre INRA de Nancy 54280 Champenoux Francec Forest Service of Junta Castilla y Leoacuten CDuque de la Victoria 5 47001 Valladolid Spaind CEREGE UMR 7330 CNRSAix-Marseille Universiteacute Europole de lrsquoArbois BP 8013545 Aix-en-Provence cedex 4 France

a r t i c l e i n f o

Article historyReceived 10 December 2013Received in revised form 19 February 2014Accepted 25 February 2014Available online xxxx

KeywordsGeneralized linear modelsGrowth modelsPinus pineaPinus pinasterPoint pattern analysisSpatial pattern

a b s t r a c t

Studying species coexistence is key to understanding the way in which forests will respond to climatechange We studied the patterns of mixed stands including two main Mediterranean pine species Pinuspinaster Ait and Pinus pinea L The spatial distribution of adult trees and saplings was studied via a pointpattern approach The effect of competition on growth of adult trees was investigated by comparing theperformance of several competition indexes for each pine species through generalized linear modelsAdult trees formed mixed clumps in which individuals of both species appeared together Part of the treegrowth variation was explained by tree size along with tree competition However the effect ofconspecific vs heterospecific competition on tree growth differed and reflected species-specificneighbor-asymmetric competition Facilitation was fundamental in the early stages for tree speciesdevelopment The spatial distribution of saplings was strongly related to the spatial distribution of adulttrees also being asymmetrically clustered and neighbor-species-dependent However the requiredfacilitation in early life-stage trees shifted to competition among trees in the adult stage Species mixturemay be desirable in terms of increasing and diversifying productivity although the conditions currentlypresent in the stand are likely to lead to future dominance of P pinea over P pinaster due not only to thegreater competition tolerance of the former but also to a greater ability to successfully recruit in the plotsforming clusters that may be in turn be impenetrable to P pinaster Therefore in order to maintain mixedstands it would be necessary to enforce adequate silvicultural management strategies which avoid futurestand dominance by P pinea

2014 Elsevier BV All rights reserved

1 Introduction

The spatial patterns observed in plant communities have beenexplained in terms of ecosystem self-organization resulting fromdifferent mechanisms and factors (both direct and indirect as wellas their interactions) acting on trees (Law et al 2001) Thesemechanisms generate the observed spatial structure in ecologicalcommunities (Callaway and Walker 1997 Koppel et al 2006)However the link between processes and patterns is not clearbecause many processes can create the same pattern (McIntireand Fajardo 2009 Perry et al 2006) Although examining a

pattern does not allow the specific mechanism that determinesthat pattern to be identified some ecological information cannonetheless be discerned (Brown et al 2011 Law et al 2009)A clumped distribution is an indication of species having similarecological requirements (Ruumlger et al 2009) facilitation amongindividuals (Barker et al 1997 Bever 2002) or dispersal limita-tions at larger scales (Burslem et al 2001) Segregation of speciesreflects different ecological requirements of species nichepartitioning or the existence of a mechanism that prevents thedevelopment of a particular species in the proximity of anotherspecies (Callaway and Walker 1997) A regular distribution patternarises when net competition dominates causing repulsionbetween individuals (Stoll and Newbery 2005)

In addition the factors which affect the pattern of a species inplant communities could change over the life of the individual

httpdxdoiorg101016jforeco2014020380378-1127 2014 Elsevier BV All rights reserved

uArr Corresponding author Tel +34 913476772E-mail address alicialedogmailcom (A Ledo)

Forest Ecology and Management xxx (2014) xxxndashxxx

Contents lists available at ScienceDirect

Forest Ecology and Management

journal homepage wwwelsevier comlocate foreco

Please cite this article in press as Ledo A et al Species coexistence in a mixed Mediterranean pine forest Spatio-temporal variability in trade-offsbetween facilitation and competition Forest Ecol Manage (2014) httpdxdoiorg101016jforeco201402038

given that many species appear to exhibit different ecological hab-itat preferences over the various life stages (Comita et al 2007)Moreover the sign of the interactions may not be constantthroughout the life of the individual (Cavard et al 2011) Further-more plantndashplant interactions can shift from competition to facil-itation when high levels of stress are present in herb and shrubcommunities (Garciacutea-Cervigoacuten et al 2013 Gea-Izquierdo andCantildeellas 2009 Brooker et al 2008) These notions tie in withthe postulated stress gradient hypothesis (Bertness and Callaway1994 Maestre et al 2009) which states that facilitativeplant-plant interactions are more important and more intense inhigh-stress environments such us arid ecosystems (Bertness andCallaway 1994) This hypothesis has been tested in Mediterraneanarid ecosystems although with differing conclusions (Lortie andCallaway 2006 Maestre et al 2009) hence the question has yetto be resolved

There is nevertheless some evidence of the existence of shiftsin intraspecific facilitationcompetition in Mediterranean shrubcommunities (Garciacutea-Cervigoacuten et al 2013 Verduacute et al 2010)and mixed Mediterranean tree stands (Pentildeuelas and Boada2003 Zavala et al 2000) Some species may also be better adaptedthan others to new environmental conditions such as intense dryperiods (Jump and Penuelas 2005 Pentildeuelas et al 2001)Differences in competition or facilitation responses in relation toclimatic variability along with differences in the spatial organiza-tion of the species in a mixed stand may lead to a reduction incoexistence certain species benefitting more than others increas-ing their dominance and thereby modifying the previously mixedstands

Research and management efforts in Mediterranean forestshave often focused on pure stands despite the fact that mixedMediterranean forest are complex ecosystems which exhibit great-er potential than pure stands in terms of specific and structuraldiversity (Barbeito et al 2009) as well as landscaping resistanceand resilience to biotic and abiotic hazards economic income opti-mal biomass production and stand stocking as well as and land-scaping and recreational use (Landeau and Landmann 2008) Inthe Northern Plateau of central Spain mixed stands of Pinus pineaL and Pinus pinaster Ait have traditionally been favoured (Gordoet al 2012) The economic importance of these multiple usespecies lies in the production of edible pine nuts in the case ofP pinea and good quality resin in the case of P pinaster Thesetwo species coexist in mixed stands partially due to the fact thatP pinea and P pinaster are quite similar in terms of ecologicalrequirements although they differ in certain aspects such us theprimary dispersal method (Del Peso et al 2012 Manso et al2012) These species coexist in the area in natural or seminaturalformations (Gordo et al 2012)

Recent research undertaken in mixed stands of these specieshas revealed that both P pinea and P pinaster have severe prob-lems for natural regeneration (Gordo et al 2012) In the case ofP pinaster the main limiting factor is seed survival during the sum-mer of the first year reducing recruitment almost to zero in thesesandy soils except in the area under the crown where highermoister levels are present (Del Peso et al 2012 Rodriacuteguez-Garciacuteaet al 2010 2011) The main problems for natural regeneration inP pinea are mast-year occurrence (Calama et al 2011) seeddispersal limitations and seed predation (Manso et al 2012)However under favorable conditions seedlings can establishand successfully develop under the crowns (Barbeito et al 2008)Mediterranean ecosystems are threatened due to their sensitiv-ity to new climatic conditions caused by climate change(Benito-Garzoacuten et al 2008 Gea-Izquierdo et al 2013 Lindnerand Calama 2012 Ruiz-Labourdette et al 2012) Hence it iscrucial to determine the main factors allowing tree coexistencein Mediterranean ecosystems

The aim of the present study is to investigate the arrangementof tree species in mixed P pinaster and P pinea stands in order togain an insight into the mechanisms which determine tree speciescoexistence Through the following analyses an attempt is made toclarify the spatial strategies of trees that allowed coexistence andcompetition among them (1) analysis of the spatial distributionof adult pine trees and whether or not DBH tree height crown sizeand slenderness are spatially structured (2) tree growth modelingcomparing different competition indices which express differentecological assumptions analyzing competition involving each spe-cies separately and (3) spatial distribution modeling of saplings inrelation to adult trees including alternative spatial covariates

2 Materials and methods

21 Study area and field sampling

Two plots of 1 ha (100 m 100 m) were installed in late fall2009 in two mixed P pineandashP pinaster stands located in theSpanish Northern Plateau Plots were selected in which the propor-tion of stemsha and basal area of the least represented species wasat least 30 In each plot all the adult trees of both species(DBH gt 75 cm) were positioned (coordinates x y) and breastheight diameter measured as the average of two perpendiculardiameters (DBH) was recorded along with total height (h) crowndiameter (cw) and height to crown base measured as the averageof two perpendicular diameters (hcb) Increment cores at 130 mwere extracted from all adult trees using a Pressler incrementborer All juvenile trees (h gt 130 m DBH lt 75 cm) within the plotwere positioned and marked (Fig 1) Core samples for 5 dominanttrees of each species in each plot were taken This information wasused to calculate the site-quality for both species in the plots Thetree-ring measurement of these trees indicated that they were allaged between 79 and 92 years The mature trees were naturallyregenerated in the area so the current spatial pattern reflects nat-ural conditions Some clear cutting may have been carried out butthere is no detailed information available in this regard In the Iacutescararea the management method employed is the floating periodicblock method However in the plots included in this analysis noregenerative cuttings have been undertaken and the currentregeneration is due to the natural dynamic of the stand

22 Characterization of the spatial pattern of adult trees

The spatial organization of adult trees was studied through apoint pattern analysis approach To test the null hypothesis of spa-tial independence among the individuals the Ripleyrsquos univariate Kcumulative function (Ripley 1977) was employed with thecomplete spatial randomness null model along with its non-cumulative o-ring function (Ripley 1981) The rejection of thishypothesis would imply the existence of either a cluster or a regu-lar pattern as reflected in Ripleyrsquos K function This function wascalculated including (i) all adult trees pooled together (ii) onlyP pinea trees (iii) only P pinaster trees To assess the spatialdependence between pairs of species the bivariate Krs functionwith the toroidal shift null model (Dale 1999 Van Lieshout andBaddeley 1999) was employed The toroidal shift null model sim-ulates independent distribution conditioned on the observed pointpattern of both classes keeping the position of the points of oneclass unchanged and shifting all the points of the other class bythe same random vector assuming that there is continuitybetween the upper and lower and between the right and leftboundaries of the plot The toroidal shift null model was chosenbecause we assumed that the spatial distribution of both species

2 A Ledo et al Forest Ecology and Management xxx (2014) xxxndashxxx

Please cite this article in press as Ledo A et al Species coexistence in a mixed Mediterranean pine forest Spatio-temporal variability in trade-offsbetween facilitation and competition Forest Ecol Manage (2014) httpdxdoiorg101016jforeco201402038

and the distribution of saplings and adults were a prioriindependent processes (Goreaud and Peacutelissier 2003)

To study the spatial structure of the dendrometric variables(DBH h cw slenderness and tree growth) the Stoyanrsquos mark cor-relation function was used (Stoyan and Stoyan 1994) to test spatialpositive negative or neutral correlation of the target variables Therandom labeling null model was also used because the size charac-teristics depend not only on age but also on competition so itmight be considered that different growth conditions involve ana posteriori process (Goreaud and Peacutelissier 2003)

In all cases 999 iterations were computed to create the nullmodel acceptance band The error probability in one-sided testingis a = 1(K + 1) where K is the number of simulations (Illian et al2008) A a = 0001 was considered The maximum distance of anal-ysis was 25 m which corresponds to 25 of the plot length Theisotropic edge (Ohser 1983 Ripley 1981) was used as an edge cor-rection for the aforementioned point pattern analysis functions

23 Modeling tree growth and competition

We used DBH increment over the past 10 years as an explana-tory variable to construct empirical models of tree growth Thisperiod is long enough to reduce the interanual climatic variabilityaffecting tree growth in these forests (Gea-Izquierdo and Cantildeellas2009 Calama and Montero 2005)

Tree growth is primarily related to tree age (or DBH as a surro-gate of age) site conditions and competition (Gea-Izquierdo et al2013 Pretzsch and Biber 2010 Biging and Dobbertin 1995)Weused the DBH of trees from ten years ago (to estimate future diam-eter growth instead of past diameter growth) as a surrogate of agealong with different competition indices although site index wasnot included We calculated the site quality curves for bothP pinaster (Bravo-Oviedo et al 2007) and P pinea (Calama et al2003) These curves gave similar results in both plots Thus itwas accepted that no differences in terms of site quality existed be-tween the two plots Nevertheless we checked whether P pinasterand P pinea diameter increment was significantly different andconfirmed that this was so (p lt 001 F-test)

Different neighborhood indices namely distance-independentand distance-dependent (Biging and Dobbertin 1995 Tome andBurkhart 1989) were compared to determine the most suitableexplanatory variable for competition (Table 1) These indices werecalculated with individuals of both species together and then foreach species separately in order to assess competition associatedwith conspecificheterospecific individuals considering the treeDBH as the initial tree size which was DBHt-10 as considered inthe growth model The indices were calculated considering an indi-vidual target tree for which the value for any competition indexwas calculated in a radius of 5 m and 10 m (for distance indepen-dent indices) and considering 3 and then 5 neighbors (for distance

Plot 1 Adult Saplings TotalP pinea 86 286 372

P pinaster 100 164 264Total 186 450 636

Plot 2 Adult Saplings TotalP pinea 190 1154 1344

P pinaster 97 110 207Total 287 1264 1551

PLOT 1 PLOT 2

15 25 35 45 55 65 75 85 95

PLOT 1

010

2030

4050

P pineaP pinaster

12 22 32 42 52 62 72

PLOT 2

020

4060

8010

012

014

0

P pineaP pinaster

(a)

(b)

(c)

Fig 1 Measured plots (a) map of the overall woody plant distribution where each adult tree is represented by a circle with a radius proportional to the tree diameter and thesaplings are represented by points P pinea is in dark grey and P pinaster in lighter grey (b) diameter distribution of adult trees and (c) number of saplings and adults trees ofeach species

A Ledo et al Forest Ecology and Management xxx (2014) xxxndashxxx 3

Please cite this article in press as Ledo A et al Species coexistence in a mixed Mediterranean pine forest Spatio-temporal variability in trade-offsbetween facilitation and competition Forest Ecol Manage (2014) httpdxdoiorg101016jforeco201402038

dependent indices) The averaged value of any competition indexwas the averaged value of the result for every tree The Donnellyedge correction (Donnelly 1978) was used for distance dependentindices

To define the growth models we fitted generalized linearmodels (GLMs) which took the form frac12Dfrac14DBHfrac14 f ethb0thornb1DBHthornb2Comp intraespecthornb3Comp interespecTHORN that is we included thecompetition associated with only conspecific and only heterospe-cific individuals separately A Gaussian error distribution wascompared to a Gamma distribution In both cases the canonicallink function was used (ie identity for Gaussian reciprocal forthe Gamma model) (McCullagh and Nelder 1989) The values ofthe b parameters of the model were estimated using maximumlikelihood methods obtained using iteratively reweighted leastsquares The different competition indices (Table 1) were includedalternatively (Pretzsch and Biber 2010 Schwinning and Weiner1998) and the resulting models were compared

To evaluate the model the residual distribution derived fromthe model was checked by analyzing the following plots to makesure that the desired results were obtained (i) residual vs fittedvalues to confirm error independence (ii) square root (stand resid-uals) vs fitted values to confirm the equal variance of the errors (iii)the QQ plot to confirm the normal distribution of the residuals

To validate the model a cross-validation was carried out bysplitting the data randomly into 20 groups of approximately equalsize The GLM was then fitted to the data omitting each group anda cost function was calculated between the observed responses ofthe group that were omitted from the fit and the prediction of thefitted models (Faraway 2005) Formerly t-tests were used todetermine whether the included explanatory variables were influ-ential in the model The efficiency of the fitted model was checkedand compared with nested models using the Akaike InformationCriterion (AIC)

24 Modeling spatial distribution of saplings

To study the factors influencing the spatial distribution of sap-lings we fitted different inhomogeneous Poisson process models(IPPM) for each species in each plot The key step in fitting Inhomo-geneous Poisson models is the estimation of k the intensity orpoint density (Illian et al 2008) which we considered to varyaccording to the explanatory variables included in the IPPM spatialmodel We included different variables in k estimation in order toidentify the best explanatory related to adult trees characteristics(DBH h cw growth slenderness adult tree density conspecific

adult tree density heterospecific adult tree density) with regardto the observed sapling distribution in the plots Nested modelswere compared with the AIC

We used a k exponential response khethx yTHORN frac14 expethh0 thorn h1m1ethx yTHORNTHORNwhere m1ethx yTHORN was the value of the spatial covariate at point (xy)and h were the parameters of the model We modeled k in the IPPMusing a GLM with a Gaussian error distribution and a canonical linkfunction For parameter estimation we used maximum likelihoodmethods using the approximate method proposed by Huang andOgata (1999) Point pattern models require edge correction sowe used the translation correction

However trees can display a clumped second-order distributiondue to the effects of dispersal (Wiegand et al 2007) Hence aspatial component was also included in the k estimation as apolynomial depending on the x and y spatial coordinates alongwith the explanatory variable included in a second approachkhethx yTHORN frac14 expethh0 thorn h1m1ethx yTHORN thorn h2ethx yTHORNTHORN

We used the KolmogorovndashSmirnov test to evaluate the good-ness-of-fit of the fitted IPPM models (Baddeley et al 2005) underthe null hypothesis that the model is true We also constructed thefour-panel plot according to Baddeley and Turner (2005) which in-cludes two graphs of residuals and the lurking variables against theCartesian coordinates

To include the aforementioned spatial variables in the IPPM itwas necessary to previously construct an image layer containingthe value of the covariate in the plots For this purpose we usedan ordinary kriging approach (Cressie 1993) to obtain the valuesfor growth diameter and height for the plot as a whole using a5 5 m grid As P pinaster and P pinea could have different growthpatterns the values for growth diameter and height were normal-ized for each species in order to obtain the growth diameter andheight value layers

25 Software

All the analyses were conducted using R statistical software (RDevelopment core Team 2005) We used the spatstats package(Baddeley and Turner 2005) for the spatial analysis to build thecompetition indices and to fit the point pattern models using theppm functions To fit the glm models and for the cross validationanalysis we used the boot package (Canty and Ripley 2013) andthe geoR package (Diggle and Ribeiro 2007) for the variogram fit-ting and ordinary kriging

3 Results

31 Characterization of the spatial pattern of adult trees

The spatial pattern of both species was similar Trees wereclustered in the plots forming clumps of 10ndash15 m in plot 1 and15ndash20 m in plot 2 (o-ring functions Appendix 1) In addition theclumps themselves clustered together resulting in a stronglyaggregated pattern in the plots (bK ethdTHORN Appendix 1) The clumpswere composed of a mixture of P pinaster and P pinea individualsin plot 1 whereas in plot 2 each species tended to remain separate(bK rsethdTHORN Appendix 1) As for the spatial correlation of dendrometricvariables the DBH was negatively correlated in the plots (bK mmethdTHORNAppendix 1) This implies repulsion between individuals of similarsize probably due to the effect of competition However P pinastershowed stronger negative correlation indicating that P pinea maybe more tolerant to the presence of neighbors (either conspecific orinterspecific individuals) than P pinaster in mixed stands As re-gards height and crown projection areas the results were similarto those for DBH although repulsion was less pronounced Stemslenderness showed positive spatial correlation as expected the

Table 1List of the different competition indices (CI) calculated and then included in thegrowth model fitting

Distance independentCI1 N5 N10 NCI2 BAL5 BAL10 BAL frac14

PN(1ifrac141 p=4DBH2

j when DBHi gt DBHjWykoff et al (1982)

CI3 BALmod5 BALmod10 BALmod frac14PN(1

ifrac141 1( 1( BALi=Gfrac12 =HISchroder and Gadow(1999)

CI4 SDI5 SDI10 SDI frac14PN(1

ifrac14i 10ethlog Nthorn1605 log DBH(1605THORN

REINEKE (1933)

Distance dependentCI5 CE3 CE5 CE frac14

PNifrac141ri=N

=eth1=2

ffiffiffiffiqp THORN N frac14 3 N frac14 5Clark and Evans (1954)

CI6 Hegyi index35 Hegyi frac14PN(1

ifrac141DBHj=DBHj

ri N frac14 3 N frac14 5

Hegyi (1974)

Five and ten indicated the diameter around the centre of each target tree in with thecompetition indices were calculated N is number of trees DBHi is the diameter atbreast height of tree j HI is the hart index HI frac14 100=H0

ffiffiffiffiNp

ri is the distance to thenearest tree q is the stand In all cases i ndash j

4 A Ledo et al Forest Ecology and Management xxx (2014) xxxndashxxx

Please cite this article in press as Ledo A et al Species coexistence in a mixed Mediterranean pine forest Spatio-temporal variability in trade-offsbetween facilitation and competition Forest Ecol Manage (2014) httpdxdoiorg101016jforeco201402038

greater the number of trees the stronger the positive correlation(bK mmethdTHORN Appendix 1) It should be noted that growth was notspatially structured (bK mmethdTHORN Appendix 1)

32 Modeling tree growth and competition

The competition indices calculated (Table 2) revealed thatP pinea tolerates a greater number of neighbors when consideringeither total trees or heterospecific individuals (F test p-valuelt 0001) However P pintaster tolerates a greater number ofconspecific individuals (F test p-value lt 0001)

321 P pinaster fitted modelIn the case of P pinaster the most accurate growth model

(lower AIC) was obtained using a Gamma error distribution

function with the canonical link including the Reineke SDI indexas a competition index considering an area of influence of 5 maround each tree

Dd10 frac14 f1=0297thorn 0016DBHthorn 0589SDIethpineaTHORN5g

With intercept (p lt 00001) DBH (p lt 00001) SDI(pinaster)5

(p = 02131) and SDI(pinea)5 (p = 00308) This model explained534 of the total growth variation with a RMSE = 0395 cm Ascan be seen in the model competition with P pinea was the mainfactor affecting P pinaster diameter increment

322 P pinea fitted modelsIn the case of P pinea the most accurate model was also ob-

tained using a model which included an error following a Gammadistribution with the canonical link again using the SDI competi-tion index considering a radius of 5 m The obtained model was

Dd10 frac14 f1=0337thorn 0004DBHthorn 3542SDIethpineaTHORN5g

With intercept (p lt 00001 t-test) DBH (p lt 00104)BAL(pinea)10 (p lt 00003) and BAL(pinaster)10 (p = 03560) Thismodel explained 206 of the total variation (Table 3) As can beseen the effect of P pinaster was not significant in the P pineadiameter increment model as might have been suspected fromthe competition index results

33 Modeling the spatial distribution of saplings

In the case of P pinaster the best IPPM model (in terms of lowerAIC) was obtained with a k that included the spatial density ofadult trees as an explanatory spatial variable both in plot 1 andplot 2 (Table 4) The models fitted in both plots were similar as ex-pected and the small differences in parameter h values may be dueto the difference in the number of trees in each plot P pinaster sap-ling distribution was found to be positively related to the distribu-tion of adult trees The seedling density distribution can be seen inFig 2a and the results of the models are presented in Fig 2c This

Table 3Comparison among the parameters of the different fitted empirical growth models for P pinaster and P pinea LL = logarithm of maximum likelihood estimate AIC = AkaikeInformation Criterion DEV = deviance and RMSE = root mean square error The most accurate models are in bold capitals

Gamma Gausian

(2LL AIC DEV RMSE (2LL AIC DEV RMSE

P pinasterN5 (146 302 4936 042 (190 3901 4557 064N10 (146 302 4936 042 (1912 3924 4492 064BAL5 (146 302 4937 042 (1913 3925 4489 064BAL10 (1485 307 4809 043 (1939 3978 4337 065BALmod5 (1398 2897 5079 042 (1771 3641 4926 062BALmod10 (1396 2893 5224 041 (1831 3762 4851 062ISD5 (536 1173 5414 040 (681 1462 4645 069ISD10 (1157 2415 5354 040 (1509 3118 4801 063CE5 (1493 3085 4769 043 (1944 3989 4307 065CE10 (1492 3085 4771 043 (1943 3986 4315 065Hegyi5 (1485 3071 4807 043 (1947 3994 4291 065Hegyi10 (1464 3029 4913 042 (1918 3937 4456 064

P pineaN5 (3089 6279 2403 033 (2827 5755 2865 068N10 (3108 6317 2299 033 (2854 5807 2727 069BAL5 (3196 6493 1798 035 (2979 6057 2032 072BAL10 (3009 6117 283 032 (2783 5665 3094 067BALmod5 (3249 6598 1439 035 (3043 6185 1613 074BALmod10 (3244 6587 1516 035 (3037 6173 1687 073ISD5 (99 2081 2068 031 (949 1999 2015 067ISD10 (2195 4491 2667 030 (2086 4273 2833 067CE5 (3126 6351 2204 034 (2857 5813 2711 069CE10 (3121 6343 2228 034 (2852 5805 2733 069Hegyi5 (3237 6573 1559 035 (3018 6137 1797 073Hegyi10 (3216 6531 1684 035 (2986 6072 199 072

Table 2Results of the different competition indices averaged for P pinea and P pinaster TheTotal column is the average competition index results for all trees of P pinaster and Ppinea respectively the conspecific and heterospecific columns are the average resultswhen only conspecific and heterospecific trees were included respectively

Total Conspecifics Heterospecifics

P pinaster P pinea P pinaster P pinea P pinaster P pinea

Distance independent indicesN5 265 371 143 284 122 087N10 995 1351 506 1000 409 351BAL5 006 011 505 504 001 007BAL10 031 048 025 018 006 030BALmod5 083 090 036 021 306 433BALmod10 084 094 059 046 136 162ISD5 002 003 036 021 306 433ISD10 084 094 059 046 136 162

Distance dependent indicesCE3 144 132 202 166 236 306CE5 181 167 258 209 287 366He3 018 024 013 027 017 009He5 006 005 007 017 007 004

A Ledo et al Forest Ecology and Management xxx (2014) xxxndashxxx 5

Please cite this article in press as Ledo A et al Species coexistence in a mixed Mediterranean pine forest Spatio-temporal variability in trade-offsbetween facilitation and competition Forest Ecol Manage (2014) httpdxdoiorg101016jforeco201402038

Table 4Results of the fitted IPPM models for P pinaster and P pinea saplings Z(xy) is the spatial covariate included in the model to estimate k the intensity of the point process modelkZ(x y) is the intensity of the IPPM model KS p-val is the probability-value of the KolmogorovndashSmirnov test to check the validity of the model and AIC is the Akaike InformationCriterion of the fitted model The AIC of the most accurate model is in bold

Z(xy) Plot P pinaster P pinea

kZ(xy) KS p-val AIC kZ(xy) KS p-val AIC

DBH 1 exp((399 to 061(xy)) 00143 1660 Not existing model 08802 26072 Not existing model 01127 1174 exp((222 + 015(xy)) 0000 7288

h 1 exp((403 to 045(xy)) 00996 1670 Not existing model 02782 26092 exp((509 + 096(xy)) 00833 1160 exp((220 + 011(xy)) 0000 7293

Growth 1 Not existing model 03687 1664 exp((355 to 001(xy)) 006242 26092 exp((477 to 081(xy)) 00089 1186 exp((219 to 019(xy)) 0000 7285

Density of adult trees 1 exp((531 + 586(xy)) 00846 1643 Not existing model 0452 26082 exp((621 + 444(xy)) 00067 1104 exp((208 to 253(xy)) 0000 7298

Density of conspecific adult trees 1 Not existing model 06918 1677 Not existing model 03458 25742 Not existing model 01397 1104 exp((244 + 167(xy)) 0000 7258

Density of heterospecific adult trees 1 Not existing model 07366 1543 exp((328 to 292(xy)) 008871 26012 Not existing model 01397 1104 exp((149 to 902(xy)) 0000 6998

Saplings of heterospecific individuals 1 Not existing model 07366 1543 Not existing model 05407 25932 exp((617 + 8594(xy)) 00127 1036 exp((1971862(xy)) 0000 7249

Actual density saplings IPM λ= f(Adult densityxy) IPM λ= f(Adult density)

P pinaster

Plo

t 1

Plo

t 1

Plo

t 2

Plo

t 2

P pinea

001

003

00

010

03

001

002

00

020

04

00

020

04

001

003

001

003

005

001

003

005

001

50

030

04

01

02

03

01

02

03

005

015

Fig 2 Results of the IPPM fitted models for the distribution of P pinaster (upper) and P pinea (lower) saplings in plots 1 and 2 (a) actual distribution of sapling density (b) arealization of the resultant IPPM including the spatial explanatory variable plus a function of the sapling coordinates (xy) in k estimation and (c) a realization of the resultantIPPM including the spatial explanatory variable in k estimation The explanatory variables were density of adult trees and density of heterospecific adult trees for P pinasterand P pinea respectively

6 A Ledo et al Forest Ecology and Management xxx (2014) xxxndashxxx

Please cite this article in press as Ledo A et al Species coexistence in a mixed Mediterranean pine forest Spatio-temporal variability in trade-offsbetween facilitation and competition Forest Ecol Manage (2014) httpdxdoiorg101016jforeco201402038

model provided a good estimation of sapling density However theinclusion of the spatial variation along with the spatial covariate ink estimation resulted in greater accuracy and explanatory capacityof the models (Fig 2b Table 4) Hence the inclusion of second or-der variation in sapling distribution improves the explanatorycapacity of the model to describe sapling density in the plots

In the case of P pinea an existing IPPM was obtained with a kthat included the spatial density of heterospecific adult trees onlyas an explanatory spatial variable Once again the models in bothplots were similar (Table 4) The inclusion of the spatial variationalong with the spatial covariate in k estimation resulted in moreaccurate and explanatory models (Fig 2 Table 4) For P pineathe spatial density of saplings depended not only on the positionof adult trees but also on the species unlike the distribution of Ppinaster saplings The fitted models revealed that P pinea saplingswere distributed such that they avoid proximity to adult P pinasterindividuals hence P pinea saplings may be more likely to formclusters of conspecific individuals or mixed clusters rather thanclusters of heterospecific individuals

4 Discussion

Our study reveals that coexistence in the studied Mediterraneanmixed pine forests is the result not only of silvicultural manage-ment but also of the interaction between similar spatial strategiesand trade-offs between competition and facilitation which changeover the lifespan of the trees sapling distribution was found to bepositively related with that of adult trees reflecting that facilita-tion is the driving process at that stage whereas we foundrepulsion between individuals during the adult stage when com-petition is the main factor influencing distribution Competitiontolerance differed between species and was also found to beneighbor-species-dependent Furthermore a mixture of species isdesirable in terms of productivity in order to reduce competitionMixed stands appear to be more efficient in terms of growth thanpure stands However our results indicate that a change from amixed stand to a pure P pinea stand has been occurring naturallyin recent decades

41 Spatial arrangement and competition patterns of adult trees

The results from the spatial analyses revealed that the treesarise in mixed clusters of P pinaster and P pinea individuals indi-cating a similar spatial strategy in both species Furthermore theclumps themselves also displayed a clustered pattern This com-plex clustered distribution has been also reported in other studiesin tropical ecosystems (Ledo et al 2012 Wiegand et al 2007) andhas been attributed to the effect of two different overlappingmechanisms underlying primary and secondary dispersion(Wiegand et al 2007) However competition between trees withinclusters was evident with diameter size reflecting spatial repul-sion as expressed by the Kmm(d) (Appendix 1) We found thatP pinea is capable of tolerating higher levels of competition thanP pinaster as reflected by the competition indices (Table 2)P pinea being surrounded by a greater number of individuals Treegrowth in both P pinaster and P pinea is more influenced (limited)by P pinea neighbors (Fig 2) This finding reflects differences interms of tree response to the nature of conspecifics indicatingneighbor-asymmetric competition This may indicate that P pineaoffers a greater competition value perhaps due to its morphology[as previously reported for temperate and tropical forests (Coateset al 2009 Harja et al 2012)] andor greater suitability to the re-gion Currently drought is the main problem in the area (Gordoet al 2012) and the roots of P pinea may be capable of reaching

the deeper humid regions of the soil which P pinaster is not ableto reach

In the present study in which we calculated competition treeper tree instead of per plot the most explanatory competition in-dex for both pine species was the Reineke ISD index consideringa 5 m radius around each tree This short distance indicates treeto tree competition This index considered both the number oftrees and the DBH of the trees in a given radius around each treeAccording to our results this combination provides the mostexplanatory competition value in the growth models For bothP pinaster and P pinea the use of distance-independent indicesresulted in a more accurate growth model than when distance-dependent indices are used The growth models revealed thatP pinea tolerates competition better in mixed than in pure clustersso mixed stands may be preferable in terms of competition balanceand increased productivity as long as competition remains belowcertain levels

42 Spatial distribution of saplings

Saplings of both species were spatially associated with adultindividuals as confirmed by the fitted IPPM models (Table 4) indi-cating that the aforementioned stand clumps are composed of bothadult trees and saplings It is known that both species require somedegree of shelter to develop after recruitment as the extremetemperatures reached on bare soil can cause seedling death(Rodriacuteguez-Garciacutea et al 2011 Barbeito et al 2008) Hencelsquonursery conditionsrsquo are necessary if seedling mortality is to beavoided and seedlings are to develop in the studied Mediterraneanstand (Castro et al 2004) The protection offered by tree crowns isa necessary factor for seedling development as reported byRodriacuteguez-Garciacutea et al (2011) for P pinaster seedling recruitmentand by Calama et al (2012) and Manso et al (2012) for P pineaseedling recruitment in pure pine stands However in the presentstudy an important difference in effective recruitment niche pref-erences between the species was identified through the IPPM anal-yses We assumed that the spatial pattern of saplings is limited byand related to the initial spatial distribution of seedlings followingthe effective recruitment process Hence the spatial pattern of sap-lings might be partially explained by the dispersion strategy of thespecies P pinaster occurs in close proximity to adult individualsregardless of the species (Table 4 Fig 2) P pinaster is mainly winddispersed so the seed shadow tends to display a uniform spatialdistribution (Del Peso et al 2012) with the possibility of becomingestablished under P pinea or P pinaster crowns However P pineais not found under heterospecific trees (Table 4 Fig 2) Further-more the position of P pinea saplings was better explained whenthe square coordinates of tree position were included in the anal-ysis (Fig 2) The fact that in the models the coordinates are squaredindicates the existence of secondary dispersion (Wiegand et al2007) Hence what is more likely to be occurring in the case ofP pinea is that some saplings found around the parent trees arethose derived from seedlings that were gravitationally dispersedwhereas other seedlings derived from secondary dispersal arespread throughout the plot although these latter seedlings avoidproximity to P pinaster This also indicates greater effectivenessof secondary dispersal in the case of P pinea compared to P pinas-ter As regards adult trees P pinea is a facilitation agent for saplingsof both species whereas P pinaster only facilitates conspecific indi-viduals (Table 4)

43 Implications of the results for species coexistence and standmaintenance

Competition was the main factor affecting adult tree growth(Fig 2) However the most suitable areas for regeneration are

A Ledo et al Forest Ecology and Management xxx (2014) xxxndashxxx 7

Please cite this article in press as Ledo A et al Species coexistence in a mixed Mediterranean pine forest Spatio-temporal variability in trade-offsbetween facilitation and competition Forest Ecol Manage (2014) httpdxdoiorg101016jforeco201402038

those areas occupied by the clumps of trees where saplings andseedlings can survive and develop thanks to the protectionprovided by the adults A similar shift from facilitation in theearly stages to competition in the adult stages has also beenobserved between closely related shrub taxa in arid ecosystems(Garciacutea-Cervigoacuten et al 2013 Valiente-Banuet and Verduacute 2008)This facilitation in the early stages of tree development in the aridMediterranean ecosystem is also in accordance with the stress gra-dient hypothesis (Bertness and Callaway 1994)

In the studied stand which can be considered a xeric site as aconsequence of low precipitation and sandy soil (causing moresevere soil drought) there is an area of coexistence of P pineaand P pinaster that has been favoured by silvicultural manage-ment However under different scenarios and in the absence ofhuman intervention it is likely that either P pinaster or P pineawould recruit and develop more effectively than the other speciesso that the stand would be dominated by the more successful ofthe two species to the detriment of the mixed composition Ouranalysis revealed that P pinea is increasing its density becausethe saplings of P pinea are far more abundant than the saplings ofP pinaster P pinea tolerates a greater level of competition (Table 2)and due to its gravitational dispersion method often forms densestands (Fig 1) which may become impenetrable to P pinaster

Moreover because of their greater mass both the seeds andseedlings of P pinea can be better adapted to survive the severedrought periods likely to occur in the area due to climate change(Castro et al 2004) The adaptive advantage of P pinea in the con-text of climate change (Benito-Garzoacuten et al 2008) along with itsgreater tolerance to competition may lead to a change in speciesdominance in the areas where the species currently coexist P pin-aster however has a seed which is more capable of colonizing newareas and can lsquomoversquo further in order to survive in the region

Human intervention through management can play an essen-tial role in tipping this balance of coexistence Both species areadapted to Mediterranean climatic conditions (Benito-Garzoacutenet al 2008) although human intervention is a key factor intraditionally managed pine stands Moreover in this case suchintervention can be used to address the differences in ecologicalpreferences and the possible dominance of one species over theother so that the current species coexistence is maintained partic-ularly given that both P pinea and P pinaster are important specieswith regard to the traditional economy of the area (Gordo et al2012)

5 Conclusions

P pinaster and P pinea coexist in mixed stands with a similarspatial arrangement The species occur in mixed clusters whichaccording to our models optimize production since competitionbetween conspecific neighboring adult trees is greater than be-tween heterospecific neighbors

To the best of our knowledge our study is the first to demon-strate the importance of facilitation (by tree crowns) in the earlystages for tree species coexistence in Mediterranean environmentsFurthermore a shift from facilitation in the sapling stage to compe-tition in the adult stage occurs so a trade-off between facilitationand competition are essential to tree coexistence Species mixturemay be desirable in terms of increasing and diversifying productiv-ity although under current conditions P pinea may dominate thestand in the future because its recruitment is more abundantand tends to form clusters which may be impenetrable to Ppinaster Besides P pinea tolerates greater competition in the adultstage from other P pinea trees contrary to P pinaster This may bedue to the deeper root system capable of reaching more waterHence adequate silvicultural management strategies need to be

enforced to maintain the mixed stands which optimize productionas well as offering other economic advantages Otherwise if thecurrent conditions continue future stands may be dominated byP pinea

Acknowledgements

This study was supported by the projects S2009AMB-1668REGENFOR AT 2010007 and AGL2010-211530001 projects Wewant to thank Aacutengel Bachiller Estrella Viscasillas and DanielMoreno for their technical support We also thank the owners ofthe studied forest lsquolsquoComunidad de Villa y Tierra de Iacutescarrsquorsquo andCastilla y Leoacuten Forest Service especially to the dedicate forestrangers for allowing plot installation and maintenance Thanksto the anonymous reviewers for their valuable comments whichnot only serve to improve the manuscript but also to enhancemy knowledge

Appendix A Supplementary material

Supplementary data associated with this article can be found inthe online version at httpdxdoiorg101016jforeco201402038

References

Baddeley A Turner R 2005 Spatstat an R package for analyzing spatial pointpatterns J Stat Softw 12 (6) 1ndash42

Baddeley A Turner R Moller J Hazelton M 2005 Residual analysis for spatialpoint processes J R Stat Soc B 67 617ndash666

Barbeito I Pardos M Calama R Cantildeellas I 2008 Effect of stand structure onstone pine (Pinus pinea L) regeneration dynamics Forestry 81 (5) 617ndash629

Barbeito I Fortin MJ Montes F Canellas I 2009 Response of pine naturalregeneration to small-scale spatial variation in a managed Mediterraneanmountain forest Appl Vege Sci 12 488ndash503

Barker MG Press MC Brown ND 1997 Photosynthetic characteristics ofdipterocarp seedlings in three tropical rain forest light environments a basis forniche partitioning Oecologia 112 453ndash463

Benito-Garzoacuten M Saacutenchez de Dios R Sainz-Ollero H 2008 Effects of climatechange on the distribution of Iberian tree species Appl Vege Sci 11 (2) 169ndash178

Bertness MD Callaway R 1994 Positive interactions in communities TrendsEcol Evol 9 191ndash193

Bever J 2002 Host-specificity of AM fungal population growth rates can generatefeedback on plant growth Plant Soil 244 281ndash290

Biging GS Dobbertin M 1995 Evaluation of competition indices in individualtree growth models For Sci 41 (2) 360ndash377

Bravo-Oviedo A del Riacuteo M Montero G 2007 Geographic variation andparameter assessment in generalized algebraic difference site indexmodelling For Ecol Manage 247 (1) 107ndash119

Brooker RW Maestre FT Callaway RM Lortie CL Cavieres LA Kunstler GLiancourt P Tielborger K Travis JMJ Anthelme F Armas C Coll LCorcket E Delzon S Forey E Kikvidze Z Olofsson J Pugnaire F QuirozCL Saccone P Schiffers K Seifan M Touzard B Michalet R 2008Facilitation in plant communities the past the present and the future J Ecol96 (1) 18ndash34

Brown C Law R Illian JB Burslem DFRP 2011 Linking ecological processeswith spatial and non-spatial patterns in plant communities J Ecol 99 1402ndash1414

Burslem DFRP Garwood NC Thomas SC 2001 Tropical forest diversitymdashtheplot thickens Science 292 606ndash607

Calama R Montero G 2005 Multilevel linear mixed model for tree diameterincrement in stone pine (Pinus pinea) a calibrating approach Silva Fennica 3937ndash54

Calama R Cantildeadas N Montero G 2003 Inter-regional variability in site indexmodels for even-aged stands of stone pine (Pinus pinea L) in Spain Ann For Sci60 (3) 259ndash269

Calama R Mutke S Tomeacute JA Gordo FJ Montero G Tomeacute M 2011 Modellingspatial and temporal variability in a zero-inflated variable the case of stonepine (Pinus pinea L) cone production Ecol Model 222 606ndash618

Calama R Madrigal G Manso R Garriga E Gordo FJ Pardos M 2012Germinacioacuten emergencia y supervivencia de regenerado en Pinus pinea L InGordo FJ Calama R Pardos M Bravo F Montero G (Eds) La regeneracioacutennatural de los pinares en los arenales de la meseta castellana IUIGFS-Uva-INIApp 115ndash129

Callaway RM Walker LR 1997 Competition and facilitation a syntheticapproach to interactions in plant communities Ecology 78 1958ndash1965

8 A Ledo et al Forest Ecology and Management xxx (2014) xxxndashxxx

Please cite this article in press as Ledo A et al Species coexistence in a mixed Mediterranean pine forest Spatio-temporal variability in trade-offsbetween facilitation and competition Forest Ecol Manage (2014) httpdxdoiorg101016jforeco201402038

Canty A Ripley B 2013 Boot Bootstrap R (S-Plus) Functions R Package Version13-9

Castro J Zamora R Hoacutedar JA Goacutemez JM 2004 Seedling establishment of a borealtree species (Pinus sylvestris) at its southernmost distribution limit consequencesof being in a marginal Mediterranean habitat J Ecol 92 (2) 266ndash277

Cavard X Bergeron Y Chen HY Pareacute D Laganiegravere J Brassard B 2011Competition and facilitation between tree species change with standdevelopment Oikos 120 (11) 1683ndash1695

Clark PJ Evans FC 1954 Distance to nearest neighbor as a measure of spatialrelationships in populations Ecology 445ndash453

Coates KD Canham CD LePage PT 2009 Above-versus below-groundcompetitive effects and responses of a guild of temperate tree species J Ecol97 118ndash130

Comita LS Condit R Hubbell SP 2007 Developmental changes in habitatassociations of tropical trees J Ecol 95 (3) 482ndash492

Cressie NAC 1993 Statistics for Spatial Data Wiley New YorkDale MRT 1999 Spatial Pattern Analysis in Plant Ecology Cambridge University

PressDel Peso C Bravo F Ruano I Pando V 2012 Patrones de diseminacioacuten y

nascencia de Pinus pinaster Ait En la Meseta castellana In Gordo FJ CalamaR Pardos M Bravo F Montero G (Eds) La regeneracioacuten natural de lospinares en los arenales de la meseta castellana IUIGFS-Uva-INIA pp 161ndash174

Diggle PJ Ribeiro Jr PJ 2007 Model Based Geostatistics Springer New YorkDonnelly K 1978 Simulations to determine the variance and edge-effect of total

nearest neighbour distance In Hodder I (Ed) Simulation Studies inArchaeology Cambridge University Press Cambridge New York pp 91ndash95

Faraway JJ 2005 Extending the Linear Model with R Generalized Linear MixedEffects and Nonparametric Regression Models CRC Press

Garciacutea-Cervigoacuten AI Gazol A Sanz V Camarero JJ Olano JM 2013Intraspecific competition replaces interspecific facilitation as abiotic stressdecreases the shifting nature of plantndashplant interactions Perspect Plant EcolEvol Syst 15 (4) 226ndash236

Gea-Izquierdo G Cantildeellas I 2009 Analysis of holm oak intraspecific competitionusing gamma regression For Sci 55 310ndash322

Gea-Izquierdo G Fernaacutendez-de-Untildea L Cantildeellas I 2013 In Growth projectionsreveal local vulnerability of Mediterranean oaks with rising temperatures ForEcol Manage 305 282ndash293

Gordo FJ Montero G Gil L 2012 La ploblemaacutetica de la regeneracioacuten natural delos pinares en los arenales de la Meseta Castellana In Gordo J Calama RPardos M Bravo F Montero G (Eds) La regeneracioacuten natural de los pinaresen los arenales de la meseta castellana Instituto Universitarios de Investigacioacutenen Gestioacuten Forestal Sostenible Valladolid

Goreaud F Peacutelissier R 2003 Avoiding misinterpretation of biotic interactionswith the intertype K12-function population independence vs random labellinghypotheses J Vege Sci 14 681ndash692

Harja D Vincent G Mulia R Noordwijk M 2012 Tree shape plasticity in relationto crown exposure Trees 26 1275ndash1285

Hegyi F 1974 A simulation model for managing jack-pine stands In Growthmodels for tree and stand simulation 30 74ndash90

Huang F Ogata Y 1999 Improvements of the maximum pseudo-likelihoodestimators in various spatial statistical models J Comput Graph Stat 8 510ndash530

Illian J Penttinen A Stoyan H Stoyan D 2008 Statistical Analysis and Modellingof Spatial Point Patterns Wiley

Jump AS Penuelas J 2005 Running to stand still adaptation and the response ofplants to rapid climate change Ecol Lett 8 (9) 1010ndash1020

Koppel Jvd Altieri AH Sillima BR Bruno JF Bertness MD 2006 Scale-dependent interactions and community structure on cobble beaches Ecol Lett9 45ndash50

Landeau S Landmann G 2008 Les peuplements forestieres meacutelanges Rev ForFranccedilaise L 99ndash106

Law R Purves DW Murrell DJ Dieckmann U 2001 Causes and effects of small-scale spatial structure in plant populations In Special Publication-BritishEcological Society 14 21ndash44

Law R Illian J Burslem DFRP Gratzer G Gunatilleke CVS GunatillekeIAUN 2009 Ecological information from spatial patterns of plants insightsfrom point process theory J Ecol 97 616ndash628

Ledo A Condeacutes S Montes F 2012 Different spatial organization strategies ofwoody plant species in a montane cloud forest Acta Oecol 38 49ndash57

Lindner M Calama R 2012 Climate change and the need for adaptation inMediterranean forests In Lucas-Borja ME (Ed) Mediterranean Forest

Management under the new context of Climate Change Nova SciencePublishers

Lortie CJ Callaway RM 2006 Re-analysis of meta-analysis support for thestress-gradient hypothesis J Ecol 94 (1) 7ndash16

Maestre FT Callaway RM Valladares F Lortie CJ 2009 Refining the stress-gradient hypothesis for competition and facilitation in plant communities JEcol 97 199ndash205

Manso R Pardos M Keyes C Calama R 2012 Modelling the spatio-temporalpattern of primary dispersal in stone pine (Pinus pinea L) stands in the NorthernPlateau (Spain) Ecol Model 226 11ndash21

McCullagh P Nelder JA 1989 Generalized Linear Models Monographs onStatistics and Applied Probability 37 Chapman Hall London

McIntire EJ Fajardo A 2009 Beyond description the active and effective way toinfer processes from spatial patterns Ecology 90 (1) 46ndash56

Ohser J 1983 On estimators for the reduced second moment measure of pointprocesses Math Oper Forschung Stat Ser Statist 14 63ndash71

Pentildeuelas J Boada M 2003 A global change-induced biome shift in the Montsenymountains (NE Spain) Glob Change Biol 9 131ndash140

Pentildeuelas J Lloret F Montoya R 2001 Severe drought effects on Mediterraneanwoody flora in Spain For Sci 47 (2) 214ndash218

Perry GL Miller BP Enright NJ 2006 A comparison of methods for thestatistical analysis of spatial point patterns in plant ecology Plant Ecol 187 (1)59ndash82

Pretzsch H Biber P 2010 Size-symmetric versus size-asymmetric competitionand growth partitioning among trees in forest stands along an ecologicalgradient in central Europe Can J For Res 40 (2) 370ndash384

Ripley BD 1977 Modelling spatial patterns (with discussion) J R Stat Soc B 39172ndash212

Ripley BD 1981 Statistical Inference for Spatial Processes Cambridge UniversityPress

Rodriacuteguez-Garciacutea E Juez L Bravo F 2010 Environmental influences on post-harvest natural regeneration of Pinus pinaster Ait in Mediterranean foreststands submitted to the seed-tree selection method Eur J For Res 129 (6)1119ndash1128

Rodriacuteguez-Garciacutea E Gratzer F Bravo F 2011 Climatic variability and other sitefactor influences on natural regeneration of Pinus pinaster Ait Mediterraneanforests Ann For Sci 68 (4) 811ndash823

Ruumlger N Huth A Hubbell SP Condit RS 2009 Response of recruitment to lightavailability across a tropical lowland rain forest community J Ecol 97 1360ndash1368

Ruiz-Labourdette D Nogueacutes-Bravo D Ollero HS Schmitz MF Pineda FD2012 Forest composition in Mediterranean mountains is projected to shiftalong the entire elevational gradient under climate change J Biogeogr 39 (1)162ndash176

Schroumlder J Gadow KV 1999 Testing a new competition index for Maritime pinein northwestern Spain Can J For Res 29 (2) 280ndash283

Schwinning S Weiner J 1998 Mechanisms determining the degree of sizeasymmetry in competition among plants Oecologia 113 447ndash455

Stoll P Newbery DM 2005 Evidence of species-specific neighborhood effects inthe dipterocarpaceae of a bornean rain forestunibasch America Ecology 863048ndash3062

Stoyan D Stoyan H 1994 Fractals Random Shapes and Point Fields New YorkTeam R Developement Core 2005 R A Language and Environment for Statistical

Computing R Foundation for Statistical Computing Vienna Austria 2013lthttpwww R-project orggt (ISBN 3-900051-07-0)

Tome M Burkhart HE 1989 Distance-dependent competition measures forpredicting growth of individual trees For Sci 35 816ndash831

Valiente-Banuet A Verduacute M 2008 Temporal shifts from facilitation tocompetition occur between closely related taxa J Ecol 96 489ndash494

Van Lieshout MNM Baddeley AJ 1999 Indices of dependence between types inmultivariate point patterns Scand J Stat 26 511ndash532

Verduacute M Jordano P Valiente-Banuet A 2010 The phylogenetic structure of plantfacilitation networks changes with competition J Ecol 98 1454ndash1461

Wiegand T Gunatilleke S Gunatilleke N Okuda T 2007 Analyzing the spatialstructure of a Sri Lankan tree species with multiple scales of clustering Ecology88 3088ndash3102

Wykoff WR Cookston NL Stage AR 1982 Userrsquos guide to the stand prognosismodel U S For Serv Gen Tech Rep INT-133

Zavala MA Espelta JM Retana J 2000 Constraints and trade-offs inMediterranean plant communities the case of Holm oak-Aleppo pine forestsBot Rev 66 119ndash149

A Ledo et al Forest Ecology and Management xxx (2014) xxxndashxxx 9

Please cite this article in press as Ledo A et al Species coexistence in a mixed Mediterranean pine forest Spatio-temporal variability in trade-offsbetween facilitation and competition Forest Ecol Manage (2014) httpdxdoiorg101016jforeco201402038