Revista chilena de historia natural
versión impresa ISSN 0716-078X
Rev. chil. hist. nat. v.83 n.3 Santiago 2010
Revista Chilena de Historia Natural 83: 349-363, 2010
Spatial distribution and regeneration strategies of the main forest species on Robinson Crusoe Island
Distribución espacial y estrategias de regeneración de las principales especies forestales de la Isla Robinson Crusoe
RODRIGO VARGAS1,*, JAIME G. CUEVAS2, CARLOS LE QUESNE3, ALBERT REIF1 & JAN BANNISTER1
1 Institute of Silviculture, Albert-Ludwigs University, Freiburg, Germany, Tennenbacher str 4, Freiburg i.Br 79085, Germany
2 Instituto de Investigaciones Agropecuarias (INIA), Oficina Técnica INIA-Valdivia, Campus Isla Teja s/n, casilla 567, Valdivia, Chile
3 Facultad de Ciencias Forestales y Recursos Naturales, Universidad Austral de Chile, Instituto de Silvicultura, Campus Isla Teja s/n, Valdivia, Chile
* Corresponding author: firstname.lastname@example.org
We describe the structure of the endemic Robinson Crusoe Island forest. We analyse the regeneration strategies of the trees: Myrceugenia fernandeziana, Fagara mayu, and Drimys confertifolia. Inventories were taken on nearly intact montane forest remnants with emphasis on spatial patterns by using Ripley's K (t) function. We hypothesized that i) ornithochorous and barochorous species should show a clumped regeneration, as a consequence of seed rain; ii) nurse effect, if present, should also contribute to a clustered pattern; iii) clustering should disappear at more developed life-stages due to self-thinning; iv) if gaps influence tree mortality and regeneration, these will present grouped patterns. Myrceugenia fernandeziana was the most abundant tree species, followed by F. mayu and D. confertifolia. The diameter distribution of M. fernandeziana suggested a continuous regeneration. The diameter distribution of F. mayu was typical of species that require disturbances to regenerate, whereas the diametric pattern of D. confertifolia was intermediate compared with the previous species. M. fernandeziana demonstrated a spatially clumped regeneration given its dispersal mode (hypothesis i), but did not show association with the large tree individuals, discarding a possible nurse effect (hypothesis ii). The large specimens of M.fernandeziana were randomly distributed, likely due to self-thinning (hypothesis iii). Mortality of M. fernandeziana trees did not seem to be a consequence of local disturbances, rejecting hypothesis iv). Fagara mayu distributed randomly showing in cases some aggregation. Drimys confertifolia presented a clustered pattern. Regeneration of F. mayu and D. confertifolia occurred mainly in gaps, but also under canopy for the latter species. D. confertifolia tended to regenerate near larger trees of the same species (hypothesis ii). M. fernandeziana seems to exhibit shade tolerance, although it showed plasticity to use different regeneration environments. F. mayu showed characteristics of a shade intolerant species, possibly requiring gaps, or larger disturbances to regenerate. D. confertifolia seems to be a shade semi-tolerant species, requiring areas with low canopy cover to establish.
Key words: forest regeneration, Juan Fernández Islands, Myrtisylva, Ripley's K-function, spatial point pattern analysis.
Se describe la estructura del bosque endémico de la Isla Robinson Crusoe. Se analizan las estrategias de regeneración de los árboles: Myrceugenia fernandeziana, Fagara mayu, y Drimys confertifolia. Se inventariaron remanentes de bosque montano, con énfasis en los patrones espaciales usando la función K(t) de Ripley. Se hipotetiza que: i) especies ornitócoras y barócoras debieran mostrar regeneración agrupada, a consecuencia de la lluvia de semillas; ii) si existe efecto nodriza, existirá también un patrón agregado de regeneración; iii) en estados de desarrollo avanzado la agregación se atenuará, a consecuencia del autorraleo; iv) si las aperturas de dosel influencian la regeneración y mortalidad, estas presentarán patrones agrupados. M. fernandeziana fue la especie arbórea más abundante, seguida por F. mayu y D. confertifolia. La distribución diamétrica de M. fernandeziana sugirió una regeneración continua. F. mayu exhibió una distribución típica de especies con regeneración esporádica. La distribución diamétrica de D. confertifolia fue intermedia respecto a las otras especies. La regeneración de M. fernandeziana fue espacialmente agrupada, dado su modo de dispersion (hipótesis i), pero no mostró asociación con individuos adultos, descartándose un posible efecto nodriza (hipótesis ii). Los individuos adultos de M. fernandeziana se distribuyeron aleatoriamente probablemente debido a autorraleo (hipótesis iii). La mortalidad de adultos de M. fernandeziana no pareció responder a alteraciones locales, descartándose la hipótesis iv. F. mayu se distribuyó aleatoriamente exhibiendo, en casos, leve agregación. D. confertifolia presentó un patrón agrupado. La regeneración de F. mayu y D. confertifolia ocurrió principalmente en claros, aunque esta última especie se presentó también bajo dosel. D. confertifolia tiende a regenerar cercana a ejemplares adultos de la propia especie (hipótesis ii). M. fernandeziana mostró tolerancia a la sombra, aunque es muy plástica, ocupando diferentes ambientes de regeneración. F. mayu mostró características de intolerante a la sombra; posiblemente requiera claros, o disturbios mayores para regenerar. D. confertifolia parece ser semitolerante a la sombra requiriendo claros, o áreas con menor cobertura para establecerse.
Palabras clave: análisis espacial, archipiélago Juan Fernández, K (t) de Ripley, Myrtisylva, regeneración forestal.
The Robinson Crusoe Island (RCI), part of the Juan Fernández Archipelago, is a recognised centre of plant endemism (61.1 %; Danton & Perrier 2006), that presents a precarious flora conservation status (Dirnbock et al. 2003). The tree species of the RCI forest are all endemic. They exhibit a wide range of dispersal strategies (Skottsberg 1928, Danton 2006) and rely on many different species-specific disturbance regimes (Castro et al. 1995, Cereceda et al. 1996). Few studies have considered either the structure of the natural forest, or auto-ecological aspects of the main tree species of the island (Bannister et al. 2006, Cuevas & Vargas 2006, Vargas et al. 2006). An understanding of the original structure and dynamics of the forests has fundamental importance when designing a management plan to contribute to their restoration and conservation (Donoso 1993). We studied nearly intact, forest remnants on RCI, which means that these data can serve as a reference for future conservation and restoration attempts.
The appearance of the vegetation in the community is an aspect regularly considered to describe a forest (Le., physiognomy, Mueller-Dombois & Ellenberg 1974). By analyzing the physiognomy and structure of a forest, it is possible to make conclusions about the ecological site conditions and the autoecological behaviour of the species (Donoso 1993).
Disturbance is considered to be a major factor driving plant dynamics, creating environmental heterogeneity for species coexistence (Pickett & White 1985). Alternative mechanisms for coexistence are life-history differentiation (Veblen 1992), resource partitioning and different regeneration niches for the species in competition (Grubb 1977). An approach for understanding regeneration dynamics is through spatial distribution patterns of trees (Donoso 1993).
Different stages and processes of plant life cycles may influence the distribution pattern of individuals. For instance, differences on seed/ fruit dispersal modes (Willson 1992, Bustamante 1995), seedling germination (Harper 1977), seedling survival (Janzen 1970) or a nurse effect of large trees upon regeneration can affect the spatial pattern of species (Fuentes et al. 1986). This could also be determined by self-thinning and competition at more advanced life stages, especially when the initial recruitment was clustered (Closset-Koop et al. 2007).
We analyse the structure of the forest in RCI, focusing on the main tree species: Myrceugenia fernandeziana (Hook. et Arn.) Johow, Fagara mayu (Bertero ex Colla) Engler and Drimys confertifolia Phil. We studied the spatial distribution of trees and regeneration, considering also standing dead individuals, in order to show that different species life histories, in addition to the disturbance regime that they face, result in different regeneration strategies, making possible species coexistence. We hypothesize the following: i) ornithochorous and barochorous species should have clumped seed rain, and thus, a clumped regeneration, following antecedents of grouped propagule dispersal by gravity or by directed bird dispersal (Bustamante 1995, Cuevas 2000, Willson 1992). ii) Nurse effect, if present, should also contribute to a clustered pattern (Oliver & Larson 1990, Fuentes et al. 1986). iii) Clustering should disappear with the self-thinning that is common on many species at older life stages, resulting in a random pattern (Fangliang et al. 1997). iv) Tree mortality and tree regeneration, if is triggered by gaps, should also show a clustered pattern due to the restricted area of these disturbances (Salas et al. 2006). Our four hypotheses are not mutually exclusive.
In addition to hypotheses testing, we aim i) to describe the structure of the endemic forest on RCI, and ii) to provide an explanation about the regeneration strategies of the main tree species.
Study sites and main tree species
The Juan Fernández Archipelago comprises three islands located between 680 and 860 km west of Chile. The Robinson Crusoe Island is the closest one to the continent (47.9 km2, 33°37' S, 78°51' W). The archipelago was designated National Park in 1935, and UNESCO World Biosphere Reserve in 1977. San Juan Bautista, on RCI, is the only town permanently inhabited of the archipelago (Fig. 1). The climate is Mediterranean with oceanic influence. Average annual temperature and precipitation are 15.2 °C and 957 mm, respectively (Hajek & Espinosa 1987).
Fig. 1: Robinson Crusoe Island and study site characteristics. The shading indicates the vegetation cover. Study sites: PLY = Plazoleta del Yunque, QDA = Quebrada de Damajuana, QVI = Quebrada de Villagra (based on maps prepared by Greimler et al. 2002 and Universidad Mayor 2004).
Isla Robinson Crusoe y características de los sitios de estudio. La zona sombreada indica la cobertura vegetacional. Sitios de estudio: PLY = Plazoleta del Yunque, QDA = Quebrada de Damajuana, QVI = Quebrada de Villagra (basado en mapas preparados por Greimler et al. 2002 y Universidad Mayor 2004).
The forests on RCI are closely related to the Valdivian rainforests of southern Chile, and exhibit common characteristics with the Cleistocalyx communities of the subtropical islands in the New Zealand region, as well as with the Hawaiian Metrosideros-forest (Skottsberg 1953, Mueller-Dombois & Fosberg 1998). Skottsberg (1953) and Greimler et al. (2002) identified several vegetation units on the island, including two forest assemblages: 'endemic upper mountain forest' and 'endemic lower mountain forest'. Danton (2006) proposed a specific ñame for these forest communities, 'Myrtisylva (-ae)', based on their unique characteristics, and the dominance of the endemic myrtle tree, M. fernandeziana.
With no human influence before 1574, disturbances like fires, selective cuttings, and introduction of exotic animáis and plants have affected severely the RCI, since human occupation (Woodward 1969, Dirnbock et al. 2003). Natural abiotic disturbances as erosión, landslides, rock falls, torrential rains (Castro et al. 1995), strong winds (Cereceda et al. 1996), earthquakes, volcanic eruptions and even tsunamis (Skottsberg 1920-1956, p. 403) have also shaped the landscape of the island.
We selected the study sites considering absence or negligible biotic, non-native disturbances impeding natural forest regeneration, in particular herbivory by rabbits, rats, domestic cattle and wild goats; the presence of exotic species in the undergrowth (mainly Aristotelia chilensis (Mol.) Stunz and/or Rubus ulmifolius Schott.), and previous anthropogenic alterations. Plazoleta el Yunque (PLY), Quebrada de Damajuana (QDA) and Quebrada de Villagra (QVI) sites were selected (Fig. 1). These areas were located on middle elevation sectors, and form part of the endemic lower and upper mountain forests (Greimler et al. 2002).
From January to February (2003), we studied the three most abundant endemic trees in the RCI forests (Cuevas 2002): (i) Myrceugenia fernandeziana (Myrtaceae) that reaches 28 m height, and 1.1 m of diameter at breast height (DBH). Its fruit is a red round berry of 7-8 mm in diameter that contains 2-3 seeds of 5 mm length (Rodríguez et al. 1983), which may be dispersed by birds or gravity (Skottsberg 1953). (ii) Fagara mayu (Rutaceae), the biggest tree on RCI, reaches more than 30 m height and 200 cm DBH. Its fruit is a hard capsule of ca 6-7 mm in diameter, which contains a seed of 5 mm, dispersed by gravity. It grows up to 600 masl. (Rodríguez et al. 1983). (iii) Drimys confertifolia (Winteraceae) grows between 200 and 900 masl. It can reach 15 m height and 50 cm DBH (Skottsberg 1953). Its fruit is a round berry (0.8-1 cm diameter) that contains 14-20 angular seeds of 2.5-3 mm, dispersed by birds or gravity (Rodríguez et al. 1983, Skottsberg 1953).
Forest structural aspects
A 2000 m2 sample plot (40 m x 50 m) was established at each study site (Fig. 1). In each plot, we registered all trees > 5 cm DBH, considering: species, sociological position (sensu Donoso 1993), DBH, total height, and crown radii projection lengths. Heights were measured with a Suunto® hypsometer. Tree individuals < 5 cm DBH were considered regeneration and classified as sexual (i.e., single seedling or sapling) or vegetative renewal (i.e., clustered, originated from a thicker root or stem base). The coverage of strata was estimated using the average crown radius, assuming circular areas (Donoso 1993).
There is scant age data for RCI trees. Johow (1896) referred to the slow growth of M. fernandeziana, pointing out that the annual rings would be about 3 mm wide. Honeyman et al. (2005), using a small sample of cores, found a direct relationship between DBH and age (r = 0.86, P = 0.000, N = 9 for M. fernandeziana; r2 = 0.98, P = 0.008, N = 4 for F. mayu; and r2 = 0.92, P = 0.009, N = 5 for D. confertifolia). Sample sizes were necessarily low due to the endangered status of the archipelago flora, and its protection in a National Park (Honeyman et al. 2005). In spite of these restrictions, our interpretation of diameter structures assumed a good correlation with age, taking into account that stand size structures, combined with spatial information, can provide good insights into forest regeneration processes and disturbance histories (Veblen 1992, Salas et al. 2006).
The diameter distributions were tested against three common models in forest dynamics: a negative exponential model (F = a* e-b*DBH), which applies to a population with constant recruitment and constant mortality rates. A power function model (F = a* DBH-b), which assumes a stable recruitment, but allows a decreasing mortality rate (Hett & Loucks 1976). And a normal distribution model, that applies to a population with episodic recruitment (equation 1) typical for species that need disturbances to regenerate (Veblen 1992, Donoso 1993). The Statistica 6.0 software was used to fit the models through the Levenberg-Marquardt algorithm.
F is the frequency in trees per hectare, DBH is the diameter class measured at breast height (in cm), e is the base of the natural logarithms, π is the mathematical constant, a and b are constant values > 0; σ and μ are the parametric standard deviation and mean, respectively.
Tree spatial distribution
We divided the 2000 m2 plot in 80 squares (5 x 5 m) to facilitate the register of individuals. Every seedling, sapling and tree was recorded to the nearest centimetre using an X, Y coordinate system. We studied the forest spatial arrangement using Ripley's K(t) function (Ripley 1977). This method calculates the distance between all pair of points (plants), and permit to analyse, at different spatial scales, the association within a class of points (e.g., species), and between separate point categories (e.g., between species 1 and 2) (Moeur 1993, Salas et al. 2006). An univariate analysis (null hypothesis of randomness) was conducted, calculating the function L(t), a transformation of Ripley's K(t) function that allows easier interpretations (Salas et al. 2006):
If the tree distribution is random (Poisson distribution), then K(t) ≈ πt2 and L (t) = 0 (Duncan 1991). If the L(t) value< 0, it indicates an uniform or regular horizontal distribution, and if the L(t) value > 0 it indicates that the distribution is clustered or aggregated (Duncan 1991).
We considered the spatial distributions of each tree species independently (M. fernandeziana, D. confertifolia and F. mayu). For all spatial analysis, groups with less than 8 individuals were not considered (Salas et al. 2006). Due to the major abundance of M. fernandeziana, its spatial distribution could be analysed for different sizes (regeneration < 5 cm; small > 5 < 10 cm; intermediate > 10 < 30 cm; and large individuals > 30 cm DBH), and for standing dead individuals.
To further analyse the species regeneration, we carried out a bivariate spatial relationship analysis. We compared the distribution pattern of the large individuals (i.e., tree species independently, > 30 cm DBH), with the regeneration individuals of M. fernandeziana and D. confertifolia (< 5 cm DBH). Due to sample size restrictions, F. mayu regeneration was not analysed. We used the L12(t) function, which is a transformation of the K12(t) function (Salas et al. 2006) to test the spatial independence hypothesis (Duncan 1991). If the L12(t) function value equals zero, the groups are spatially independent; a value > 0 indicates attraction between individuals, and a value < 0 indicates repulsion (Salas et al. 2006). We used Duncan's (1990) spatial distribution programmes for the estimation of K(t). Significant differences were determined between the observed and expected values for a random (univariate) and an independent (bivariate) distribution, generating 99 Monte Carlo simulations to establish the upper and lower confidence envelopes at 99 % (Duncan 1991). The t distances attained a maximum of 20 m (half of the shortest plot side).
Regeneration in gaps
We considered as gaps all openings larger than 25 m, where there was no tree coverage and no direct impediment to solar radiation reaching the forest floor (Veblen 1992). Gap percentage at the landscape level was determined by using three 100 m long transects starting in the centre of each plot with a defined cardinal direction (i.e., adjacent transects were perpendicular between them). Along these transects, the distance below tree coverage was quantified and the occurrence of gaps along the transects recorded. In each study site, we selected one gap visually similar in size and shape to the average gaps observed within the study area (i.e., a 'representative' gap). Inside the expanded gap we recorded the regeneration of trees by direct counting, registering also the main potential successors (i.e., trees > 5 cm DBH, Rebertus & Veblen 1993).
Structural aspects ofthe Myrtisylva
The montane forest of RCI exhibited a M. fernandeziana dominant stratum (16.3 ± 3.5 m height; mean values from all sites ± standard deviation), with emergent individuals of F. mayu (24.6 ± 0.9 m). In the intermediate stratum (8.2 ± 2.2 m), M. fernandeziana and D. confertifolia individuals were found. On occasions, the latter species occupied the codominant stratum (12.3 ± 2.3 m). M. fernandeziana also grew in the suppressed stratum (5.2 ± 0.1 m), accompanied in low numbers by Rhaphithamnus venustus. Boehmeria excelsa and Coprosma pyrifolia individuals were also found, mostly in the dominant or codominant strata, but these were rare. Coprosma oliveri was occasionally observed, present at very low densities in the intermedíate stratum of the upper forest.
The most important species in the Myrtisylva forest was M. fernandeziana (Table 1), which contributed most basal area (> 65 %), and stocking density (90 %). F. mayu was less frequent but represented > 18 % of the stand basal area (Table 1, Fig. 2). D. confertifolia presented between 20 and 60 individuals ha-1, contributing 0.6-8.8 % to the basal area. Less important was R. venustus. Coprosma oliveri, C. pyrifolia, and B. excelsa occurrences were rare (Table 1).
Fig. 2: Diameter distributions of the tree species studied at three locations on Robinson Crusoe Island. The correlation coefficient (r) and the associated probability value (P) for the best fit models are shown (tested models: exponential(e), normal(n), power(p)). See Fig. 1 for information about the sites.
Distribución diamétrica de las especies arbóreas estudiadas en los tres sitios de la isla Robinson Crusoe. Se muestra el coeficiente de correlación (r) y la probabilidad asociada (P) para los modelos de mejor ajuste (modelos probados: exponencial(e), normal(n), potencia(p)). Ver Fig. 1 para información sobre los sitios.
The diameter distributions of the cumulated and single Myrtisylva species followed a negative power curve at PLY (inverse J) (Fig. 2). The diameter trend of F. mayu and D. confertifolia revealed in all stands an absence of individuals in certain diameter classes. In QDA, the negative exponential model was the best fit for the whole stand and M. fernandeziana trees, while D. confertifolia fit better the power function, although not significantly (Fig. 2). Fagara mayu showed a normal distribution with a reduced fit (r = 0.36). The best models in QVI were the exponential for the whole stand and M. fernandeziana individuals; the normal one for F. mayu, and the power function for D. confertifolia, which was significant (Fig. 2).
M. fernandeziana dominated all sociological strata (Table 2). The only exception was the emergent stratum in PLY, where F. mayu prevailed. The relative importance of F. mayu increased in the upper sociological positions (dominant, emergent). The QVI site did not present emergents and had less stratification.
Gaps were formed mainly as result of small scale disturbances (tree falls), representing almost one quarter of the forest area in PLY and QDA, but only 8 % in QVI (Table 3).
Tree spatial distribution
The regeneration of M. fernandeziana (< 5 cm DBH) showed a clustered distribution (Fig. 3). The small individuals of this species (5-10 cm DBH) presented an inconsistent trend among sectors, tending to aggregation only in PLY. The intermediate individuals of M. fernandeziana (> 10-30 cm DBH) distributed randomly up to 4 m, tending to aggregate at longer distances in PLY. The larger M. fernandeziana specimens (> 30 cm DBH) distributed randomly; only in QVI they were clustered over 5 m. The mortality pattern of M. fernandeziana was generally random, although it showed a slight aggregation at short distances in the PLY and QVI sites (Fig. 3). Fagara mayu distributed mostly random, showing some aggregation at short distances in QDA and QVI sites (Fig. 4). D. confertifolia horizontal distribution presented an aggregated pattern, especially in QDA (Fig. 4). The spatial association between large individuals (> 30 cm DBH) and the regeneration of M. fernandeziana (< 5 cm DBH) was mostly independent in PLY and QDA sectors. In QVI, these groups repelled between 4 and 12 m, the opposite occurred in PLY at larger distances (Fig. 5).
Fig. 3: Spatial distribution (function L(t)) oí Myrceugenia fernandeziana trees of different diameter categories. The dashed lines represent the confidence envelopes for a random distribution (a = 99 96). See Fig. 1 for ¡nformation about the sites.
Patrón espacial (función L(t)) de individuos de Myrceugenia fernandeziana para diferentes clases diamétricas. Las líneas segmentadas representan las envolturas de confianza para una distribución al azar (a = 99 96). Ver Fig. 1 para información sobre los sitios.
Fig. 4: Spatial pattern (function L(t)) for all individuals oí Fagara mayu and Drimys confertifolia. The dashed lines represent the confidence envelopes for a random distribution (a = 0.99). See Fig. 1 for information about the sites.
Patrón espacial (función L(t)) para todos los individuos de Fagara mayu and Drimys confertifolia. Las líneas segmentadas representan las envolturas de confianza para una distribución al azar (a = 0.99). Ver Fig. 1 para información sobre los sitios.
Fig. 5: Association between regeneration (< 5 cm DBH) and large individuals (> 30 cm DBH) of several tree species. The dashed lines represent the confidence envelopes (a = 99 96) for an independent distribution. See Fig. 1 for information about the sites.
Asociación entre la regeneración (< 5 cm DAP) y los individuos grandes (> 30 cm DAP) de varias especies arbóreas. Las líneas segmentadas representan los intervalos de confianza (a = 99 96) para una distribución independiente. Ver Fig. 1 para información sobre los sitios.
Large individuals of M. fernandeziana were, in general, not related with the regeneration of D. confertifolia, although slightly in QDA at large distances (Fig. 5). Large trees of D. confertifolia were spatially related with the regeneration of M. fernandeziana in PLY, but repelled it in QDA and QVI beyond 8 m. The pattern of D. confertifolia with respect to its own regeneration showed always a positive relationship between 3-7 m.
The large trees of F. mayu showed a slight repulsion towards the regeneration of M. fernandeziana between 6-8 m in QVI (Fig. 5). A similar pattern presented the regeneration of D. confertifolia at 8-12 m in QDA.
The composition of the tree regeneration below canopy was similar in all sites (Table 3). Myrceugenia fernandeziana was the most common species, followed by D. confertifolia and R. venustus. Fagara mayu regenerated below canopy only in QDA and QVI. There was no recruitment of B. excelsa and Coprosma spp. The QVI site presented in average 11 times less regeneration than the other study sectors (Table 3).
Below canopy, the regeneration of M. fernandeziana was higher than in gaps at PLY and QDA. Fagara mayu regeneration was absent below canopy in PLY, even though in gaps, exceeded 1000 individuals ha-1. Fagara mayu did not show regeneration in the gaps of the other sites, and recruited scarcely below canopy (10 to 15 plants ha-1).
Drimys confertifolia presented low recruitment below coverage in PLY, inferior to that observed in gaps. In the other study sites, D. confertifolia showed an intermedíate regeneration below canopy (90-25 individuals ha-1), with no recruitment in gaps.
Regeneration developed principally from seeds below canopy, particularly that of M. fernandeziana (Table 3). F. mayu showed 40 % of vegetative plants. Most regeneration of D. confertifolia originated asexually below canopy (54 %). R. venustus regenerated exclusively from seeds. In gaps, M. fernandeziana, F. mayu and D. confertifolia tended to regenerate sexually (Table 3).
Within the studied gaps, the main potential successors in QDA and QVI belonged to M. fernandeziana. F. mayu was the only potential replacement species in the PLY gap (Table 1). D. confertifolia did not feature as potential successor.
Structural aspects ofthe Myrtisylva
Our findings agree with previous reports that give the following sequence of tree heights and diameters: F. mayu > M. fernandeziana > D. confertifolia (Johow 1896, Skottsberg 1953, Rodríguez et al. 1983). Specimens of M. fernandeziana growing > 110 cm DBH have not been reported earlier. Similar stand features were obtained by Cuevas (2001, 2002), based on the analysis of 17 forest areas on RCI. Tree species richness is poorer than that of analogous Pacific island forests such as Lord Howe and Norfolk Islands (Mueller-Dombois & Fosberg 1998) or conifer-broadleaved forests, lowland conifer, and warm-temperate Nothofagus forests of New Zealand (Wardle J. 1984, Wardle P. 1991). However, the forest physiognomy does not seem to differ greatly from them (Mueller-Dombois & Fosberg 1998). The gap proportions are in the range of warm temperate forests (Yamamoto 1992). Gaps on RCI forest were formed by the individual trees falling, sometimes causing a small-scale domino-effect, but not as stand-level mortality like the synchronous forest dieback reported for subtropical Metrosideros-forests oí Hawaii (Hart 2010) or températe Nothofagus forests of New Zealand (Wardle J. 1984).
Myrceugenia fernandeziana diameters always fitted the negative function model, either the exponential or the power functions. Both models usually fit on uneven aged forests that present continuous regeneration (Veblen 1992, Donoso 1993). Although F. mayu adjusted to the power function in PLY, generally the distribution can be considered discontinuous given the lack of some diameter classes. The best adjustment to the normal model in QDA and QVI is consistent with a sporadic regeneration strategy that may be triggered by fine scale gaps or larger disturbances such as landslides and blowdowns, which are common on RCI (Castro et al. 1995). The diametric distribution of D. confertifolia, fitted significantly the power model in two sites. The recurrent lack of diametric classes suggests that D. confertifolia also presents discontinuous recruitment, possibly requiring the occurrence of small gaps.
Tree spatial distribution
The clustered distribution exhibited by the smallest individuals of M. fernandeziana may be the result of the species dispersion strategy (hypothesis i). Vargas et al. (2006) stated that M. fernandeziana berries are dispersed by either endozoochory or barochory. The only native frugivore on the RCI is the Austral Thrush (Turdus falcklandii magellanicus), and given the abundance of fleshy-fruited species, there is probably an excess of food supply for these birds (Skottsberg 1928). M. fernandeziana seeds usually germinate inside the fruits (Ricci 1998), and it is unknown whether they can survive the digestive tract of the thrush. Consequently, barochore dispersal of fruits is a probable explanation for the clustering of small M. fernandeziana individuals at distances of about 5 m (hypotesis i). We cannot rule out that this pattern could also be the result of distribution of sites suitable for its development. The spatial association between the regeneration of M. fernadeziana and the large trees of the same species was generally independent. The regeneration of M. fernandeziana showed in several cases a spatial repulsión against the large trees beyond 8 m. Thus, our data does not support a possible nurse effect on the regeneration of M. fernandeziana (hypothesis ii). The clustering of M. fernandeziana disappeared progressively with increasing diameters, with the exception of the QVI site. This is consistent with results from other forest communities such as the Malaysian rainforest, where a self-thinning effect has been reported (Fangliang et al. 1997) (hypothesis iii). The standing dead individuals of M. fernandeziana did not show a considerable deviation from randomness, thus we discard the premise of localized mortality under canopy for this species (hypothesis iv).
Fagara mayu generally showed a random spatial pattern. At two study sites, the individuals of this species clustered slightly at short distances. Barochore dispersión may explain this pattern, given the fruit characters not eaten by birds (Bannister et al. 2006) (hypothesis i). This aggregated pattern of F. mayu could also be explained by local disturbances as gaps (Salas et al. 2006), but we cannot support this premise based only on the spatial data of the species (hypothesis iv).
Considering the fruit characters of D. confertifolia, its spatial grouped pattern might be indicative of zoochory. Drimys confertifolia seeds would be strong enough to survive to the thrush's digestive tract (Skottsberg 1928). Nevertheless, it is easy to find numerous fruits laying on the forest litter (personal observations), indicating that birds do not consume all fruits. Drimys confertifolia presents germination difficulties without seed stratification (Cuevas & Figueroa 2007). Accordingly, birds could improve dispersal, and probably germination chances through endozoocory (hypothesis i). The vegetative reproduction was recurrent for D. confertifolia, and may also contribute to the clump generation (Oliver & Larson 1990). Drimys confertifolia regeneration was spatially and positively related with the large trees of the same species, which suggests an intra-species nurse effect (hypothesis ii). The clumped pattern of D. confertifolia could also be explained by the limited area of sites suitable for its recruitment, like canopy gaps (hypothesis iv). In continental Chile, Drimys winteri seems to present a similar behaviour in primary forests, where the species requires space for its development, but present dispersal limitations that links its recruitment with mother trees (Donoso et al. 1984, Lusk 1996).
Tree regeneration strategies
M. fernandeziana regeneration was considerably higher below canopy than in gaps, with one exception at the QVI site. Thus, M. fernandeziana seems to show a plástic behaviour, from tolerant to shade to shade intolerant. The shade tolerance is consistent with the continuous regeneration suggested by the diameter structure of the species, which reflected common features of multilayered, uneven aged forests (Donoso 1993). The good germination of this species, at least under laboratory conditions (Ricci 1998), and its regeneration plasticity might explain its higher abundance in the Myrtisylva. In continental Chile, Myrceugenia planipes possess a high degree of shade tolerance, although it may colonise open fields (Donoso 2006a), and it exhibits an inverted J-shaped diameter distribution (Donoso et al. 1985). Myrceugenia exsucca (Donoso 2006b) and M. ovata (Donoso 2006c) show similar characteristics, and the latter species establishes also in gaps (Armesto & Fuentes 1988). Then, the regenerative behaviour of M. fernandeziana looks similar with its continental relatives.
There was no regeneration of F. mayu below canopy in PLY, whereas in gaps it exceeded 1000 individuals ha-1. F. mayu did not regenerate in the gaps present at the other sites (smaller than the PLY gap), but there was scant recruitment of this species below the canopy. These antecedents, in addition to F. mayu's diameter distribution, are characteristics of species with sporadic regeneration (Veblen 1992, Donoso 1993). The lack of individuals in some diameter classes is likely due to the absence of canopy openings during a certain period. Seed cycles and restrictions in dispersal and germination may also impose additional regeneration discontinuities. Seeds require acid scarification to germinate (Ricci 1998), and they are not dispersed by birds (Bannister et al. 2006). The relative importance of F. mayu increased towards the upper sociological strata (dominant, emergent). This indicates the species' shade intolerant nature. The superficial root system (Rodríguez et al. 1983, Bannister et al. 2006) and shade intolerance are also typical characteristics of pioneer species (Oliver & Larson 1990). The species' dependence on open areas might be responsible in part for its minor abundance relative to M. fernandeziana, where Myceugenia does not need to "wait" for the occurrence of gaps in order to recruit. Studies of other Fagara species worldwide (Le., F. ailanthoides in Japan) have revealed a direct relationship between gaps and its occurrence, with F. ailanthoides considered to be a pioneer species requiring large disturbances to regenerate (Tanouchi & Yamamoto 1995).
D. confertifolia regeneration in PLY was considerably lower below the canopy than in gaps. In QDA and QVI, D. confertifolia exhibited a reduced self-replacement under canopy, with no regeneration in gaps. The species presented a semi-continuous diameter distribution. Drimys confertifolia individuals were observed growing in areas where the canopy coverage was lower (semi-shade). Like M. fernandeziana, the species would also benefit from the protection granted by other higher plants at early stages (Cuevas & Vargas 2006). In continental Chile, Drimys winteri var. chilensis and var. punctata (Donoso et al. 2006) exhibit pioneer behaviour in open forests (Corvalán et al. 1987), requiring semi-shade to establish (Lusk 1995). The punctata variety regenerates below the canopy and responds rapidly to gap formation (Rebertus & Veblen 1993). Thus, D.winteri and D. confertifolia match, both exhibiting a dual mode of regeneration, in semi-shade and in gaps, which is indicative of a semi-shade tolerant species.
We studied the spatial patterns of the main endemic trees of RCI. We related their spatial patterns with their regeneration strategies considering: seed dispersal, nurse effect, self-thinning and localized disturbances. Barochory and endozoochory reflected particularly in the clumped pattern of M. fernandeziana and D. confertifolia. Intra-specific nurse effect was suggested in the case of D. confertifolia. Self-thinning was suggested for M. fernandeziana. Localized disturbances like gaps appear to determine the occurrence of F. mayu, and may influence positively the recruitment of D. confertifolia and M. fernandeziana.
This information acquires interest considering the conservation problems affecting the Myrtisylva of RCI. Currently the action of alien invasive species (principally R. ulmifolius and A. chilensis) leads actively a process of novel forest formations (Dirnbock et al. 2003). Structural parameters of original forest remnants should be used as a reference to guide recovery activities. Detailed ecological studies considering light, soil, micro-site characters and the influence of invasive species over the natural regeneration of this forest seem urgent to further develop effective conservation measures.
We are grateful to the CONAF functionaries: Javiera Meza (Viña del Mar), and especially to the National Park Administrator, Iván Leiva and his wife Silvia Moreno. Philippe Danton and Christophe Perrier (Robinsonia Association, France), helped us on the field and on many further conversations. Thanks to Miguel Álvarez (Bonn University) for commenting the manuscript, especially to Úrsula Eggert, David Butler-Manning, David Forrester and Stefanie Gartner (Freiburg University) for the English checking and comments. We appreciate the great contribution of two anonymous reviewers. This work is dedicated to the victims of the tsunami of February 27th 2010, and to all Juan Fernández community on these difficult moments after this event that destroyed the lower part of San Juan Bautista town.
ARMESTO J & E FUENTES (1988) Tree species regeneration in a mid-elevation, temperate rain forest in Isla de Chiloé, Chile. Plant Ecology 74: 151-159. [ Links ]
BANNISTER J, R VARGAS & P DANTON (2006) Fagara mayu (Bertero ex Colla) Engler. In: Donoso C (ed) Las especies arbóreas de los bosques templados de Chile y Argentina. Autoecología: 599-605. Marisa Cuneo Ediciones, Valdivia, Chile. [ Links ]
BUSTAMANTE R (1995) Depredación de semillas en bosques templados de Chile. In: Armesto JJ, C Villagrán & MTK Arroyo (eds) Ecología de los bosques nativos de Chile: 265-278. Editorial Universitaria, Santiago, Chile. [ Links ]
CASTRO C, LA BRIGNARDELLO & P CERECEDA (1995) Determinación de áreas con riesgo morfodinámico en San Juan Bautista, Isla Robinson Crusoe, comuna de Juan Fernández, V Región. Revista Geográfica de Chile Térra Australis 40: 43-61. [ Links ]
CERECEDA P, C CASTRO, R SCHEMENAUER, R PONCE & P OSSES (1996) Comportamiento de las precipitaciones en cinco sectores de la isla Robinson Crusoe durante los inviernos de los años 1992 a 1995. Revista Geográfica de Chile Térra Australis 41: 37-48. [ Links ]
CLOSSET-KOPP D, O CHABRERIE, B VALENTÍN, H DELACHAPELLE & G DECOCQ (2007) When Oskar meets Alice: Does a lack of trade-off in r/ K-strategies make Prunus serótina a successful invader of European forests? Forest Ecology and Management 247: 120-130. [ Links ]
CORVALÁN P, L ARAYA & M JIMÉNEZ (1987) El Canelo: Una alternativa de desarrollo para la Décima Región. Vol III. Ministerio de Agricultura. Universidad de Chile, Santiago. [ Links ]
CUEVAS JG (2000) Tree recruitment at the Nothofagus pumilio alpine timberline in Tierra del Fuego, Chile. Journal of Ecology 88: 840-855. [ Links ]
CUEVAS JG (2001) Distribución y abundancia de la flora vascular del archipiélago de Juan Fernández. I Parte. Technical report N° 23. CONAF V Región, Viña del Mar, Chile. [ Links ]
CUEVAS JG (2002) Distribución y abundancia de la flora vascular del archipiélago de Juan Fernández. II Parte y Final. Technical report N° 29. CONAF V Región, Viña del Mar, Chile. [ Links ]
CUEVAS JG & R VARGAS (2006) Drimys confertifolia Phil. In: Donoso C (ed) Las especies arbóreas de los bosques templados de Chile y Argentina. Autoecología: 593-598. Marisa Cuneo Ediciones, Valdivia, Chile. [ Links ]
CUEVAS JG & J FIGUEROA (2007) Seed germination of species of the Juan Fernández archipelago under laboratory conditions. Gayana Botánica 64: 60-80. [ Links ]
DANTON P (2006) The "myrtisylva" of the Juan Fernandez Archipelago (Chile), a threatened forest. Acta Botánica Gallica 153: 179. [ Links ]
DANTON P & C PERRIER (2006) Nouveau catalogue de la flore vasculaire de l'archipel Juan Fernández (Chili). Acta Botánica Gallica 153: 399-587. [ Links ]
DIRNBOCK T, J GREIMLER, P LÓPEZ & T STUESSY (2003) Predicting future threats to the native vegetation of Robinson Crusoe Island, Juan Fernández Archipelago, Chile. Conservation Biology 17: 1650-1659. [ Links ]
DONOSO C (1993) Bosques templados de Chile y Argentina. Variación, estructura y dinámica. Editorial Universitaria, Santiago, Chile. [ Links ]
DONOSO C (2006a) Myrceugenia planipes (H et A.) Berg. In: Donoso C (ed) Las especies arbóreas de los bosques templados de Chile y Argentina. Autoecología: 385-389. Marisa Cuneo Ediciones, Valdivia, Chile. [ Links ]
DONOSO C (2006b) Myrceugenia exsucca (D.C.) Berg. In: Donoso C (ed) Las especies arbóreas de los bosques templados de Chile y Argentina. Autoecología: 375-380. Marisa Cuneo Ediciones, Valdivia, Chile. [ Links ]
DONOSO C (2006c) Myrceugenia ovata (Hooker et Arnott) Berg. In: Donoso C (ed) Las especies arbóreas de los bosques templados de Chile y Argentina. Autoecología : 381-384. Marisa Cuneo Ediciones, Valdivia, Chile. [ Links ]
DONOSO C, B ESCOBAR & J URRUTIA (1985) Estructura y estrategias regenerativas de un bosque virgen de Ulmo (Eucryphia cordifolia)-Tepa (Laurelia philippiana Phil. Looser) en Chiloé, Chile. Revista Chilena de Historia Natural 58: 171-186. [ Links ]
DONOSO C, R GREZ, B ESCOBAR & P REAL (1984) Estructura y dinámica de bosques del tipo forestal siempreverde en un sector de Chiloé insular. Bosque 5: 82-104. [ Links ]
DONOSO C, B ESCOBAR, P DONOSO & F UTRERAS (2006) Drimys winteri J.R. et G. Forster. In: Donoso C (ed) Las especies arbóreas de los bosques templados de Chile y Argentina. Autoecología: 220-232. Marisa Cuneo Ediciones, Valdivia, Chile. [ Links ]
DUNCAN R (1990) Spatial Analysis Programmes. School of Forestry, University of Canterbury, Christchurch, NZ. [ Links ]
DUNCAN R (1991) Competition and coexistence of species in a mixed podocarp stand. Journal of Ecology 79: 1073-1084. [ Links ]
FANGLIANG H, P LEGENDRE & JV LAFRANKIE (1997) Distribution patterns of tree species in a Malaysian tropical rain forest. Journal of Vegetation Science 8: 105-114. [ Links ]
FUENTES E, A HOFFMANN, A POIANI & MC ALLIENDE (1986) Vegetation change in large clearings: Patterns in the Chilean matorral. Oecologia 68: 358-366. [ Links ]
GREIMLER J, P LÓPEZ, TF STUESSY & T DIRNBOCK (2002) The vegetation of Robinson Crusoe Island (Isla Masatierra), Juan Fernandez Archipelago, Chile. Pacific Science 56: 263-284. [ Links ]
GRUBB PJ (1977) The maintenance of species richness in plant communities. The importance of the regeneration niche. Biological Reviews 52: 107-145. [ Links ]
HAJEK E & GA ESPINOZA (1987) Meteorología, climatología y bioclimatología de las islas oceánicas chilenas. In: Castilla J (ed) Islas oceánicas chilenas: Conocimiento científico y necesidades de investigación: 55-83. Ediciones Universidad Católica de Chile, Santiago, Chile. [ Links ]
HARPER JL (1977) Population biology of plants. Academic Press, London, UK. [ Links ]
HART P (2010) Tree growth and age in an ancient Hawaiian wet forest: Vegetation dynamics at two spatial scales. Journal of Tropical Ecology 26: 1-11. [ Links ]
HETT J & O LOUCKS (1976) Age structure models of balsam fir and eastern hemlock. Journal of Ecology 64: 1029-1044. [ Links ]
HONEYMAN P, P CRUZ & C SCHULZE (2005) Estudio de los factores bióticos asociados a la regeneración del bosque de Juan Fernández y propuesta de restauración. Informe final. Proyecto CONAF - Universidad Mayor, Santiago, Chile. [ Links ]
JANZEN DH (1970) Herbivores and the number of tree species in tropical forests. American Naturalist 104: 501-527. [ Links ]
JOHOW F (1896) Estudios sobre la flora de las islas de Juan Fernández. Imprenta Cervantes, Santiago, Chile. [ Links ]
LUSK C (1995) Seed size, establishment sites and species coexistence in a Chilean rain forest. Journal of Vegetation Science 6: 249-256. [ Links ]
LUSK C (1996) Stand dynamics of the shade-tolerant conifers Podocarpus nubigena and Saxegothaea conspicua in Chilean temperate rain forest. Journal of Vegetation Science 7: 549-558. [ Links ]
MOEUR M (1993) Characterizing spatial patterns of trees using stem-mapped data. Forest Science 39: 756-775. [ Links ]
MUELLER-DOMBOIS D & H ELLENBERG (1974) Aims and methods of vegetation ecology. J. Wiley and sons, NY, USA. [ Links ]
MUELLER-DOMBOIS D & F FOSBERG (1998) Vegetation of the tropical Pacific islands. Springer Verlag, NY, USA. [ Links ]
OLIVER CH & B LARSON (1990) Forest stand dynamics. Biological Resource Management Series. McGraw-Hill, Inc, NY, USA. [ Links ]
PICKETT STA & PS WHITE (1985) The Ecology of Natural Disturbances and Patch Dynamics. Academic Press, New York, USA. [ Links ]
REBERTUS AJ & TT VEBLEN (1993) Structure and tree-fall gap dynamics of oíd growth Nothofagus forests in Tierra del Fuego, Argentina. Journal of Vegetation Science 4: 641-654. [ Links ]
RICCI M (1998) Técnicas de propagación y viverización de algunas especies de la flora vascular endémica del Archipiélago de Juan Fernández. Technical handbook N° 1. CONAF V Región, Viña del Mar, Chile. [ Links ]
RIPLEY B (1977) Modelling spatial patterns. Journal of the Royal Statistical Society B 39: 172-212. [ Links ]
RODRÍGUEZ R, O MATTHEI & M QUEZADA (1983) Flora arbórea de Chile. Editorial de la Universidad de Concepción, Concepción, Chile. [ Links ]
SALAS C, V LEMAY, P NUÑEZ, P PACHECO & A ESPINOSA (2006) Spatial patterns in an old-growth Nothofagus obliqua forest in south-central Chile. Forest Ecology and Management 231: 38-46. [ Links ]
SKOTTSBERG C (1920-1956) The natural history of Juan Fernández and easter island. Almqvist & Wiksells Boktrickeri AB, Uppsala, Sweden. [ Links ]
SKOTTSBERG C (1928) Pollinationsbiologie und Samenverbreitung auf den Juan Fernández-Inseln. In: Skottsberg C (ed) The natural history of the Juan Fernández and easter islands: 503-547. Almqvist & Wiksells Boktryckeri AB, Uppsala, Sweden. [ Links ]
SKOTTSBERG C (1953) The vegetation of the Juan Fernández Island. In: Skottsberg C (ed) The natural history of Juan Fernández and easter island: 793-960. Almqvist & Wiksells Boktryckeri AB, Uppsala, Sweden. [ Links ]
TANOUCHI H & S YAMAMOTO (1995) Structure and regeneration of canopy species in an old-growth evergreen broad-leaved forest in Aya district, southwestern Japan. Plant Ecology 117: 51-60. [ Links ]
UNIVERSIDAD MAYOR (2004) Estudio interpretativo de imágenes satelitales y fotografías aéreas del Parque Nacional Archipiélago de Juan Fernández. Laboratorio de Ciencias de Información Geográfica, Universidad Mayor, Santiago, Chile. [ Links ]
VARGAS R, J BANNISTER & P DANTON (2006) Myrceugenia fernandeziana (Hook. & Arn.) Johow. In: Donoso C (ed) Las especies arbóreas de los bosques templados de Chile y Argentina: 610-616. Marisa Cuneo Ediciones, Valdivia, Chile. [ Links ]
VEBLEN TT (1992) Regeneration dynamics. In: Gleen-Lewin DC, RK Peet & TT Veblen (eds) Plant succession: Theory and predictions: 152-187. Population and community biology series 11, Chapman & Hall, London, UK. [ Links ]
WARDLE J (1984) The New Zealand beeches: Ecology, utilisation, and management. New Zealand Forest Service, Wellington, New Zealand. [ Links ]
WARDLE P (1991) Vegetation of New Zealand. Cambridge University Press, Cambridge, UK. [ Links ]
WILLSON MF (1992) The ecology of seed dispersal. In: Fenner M (ed) Seed ecology and regeneration in plant communities: 6-16. CAB International, Southampton, UK. [ Links ]
WOODWARD R (1969) Robinson Crusoe's Island: A history of the Juan Fernández Islands. University of North Carolina Press, Chapel Hill, North Carolina, USA. [ Links ]
YAMAMOTO S (1992) The gap theory in forest dynamics. Journal of Plant Research 105: 375-383. [ Links ]
Associate Editor: Javier Figueroa
Received September 1, 2009; accepted April 5, 2010