Introduction

Forest fragments are small and separate remnants of reduced continuous forest areas, which probably have limited ecosystem functionality compared to the original forest (Young and Boyle 2000). The first consequence of forest fragmentation is demographic (Lande 1993; Hanski and Ovaskainen 2000; Jaquiéry et al. 2008), but fragmentation has also genetic consequences, which contributes to risks of extinction (Frankham 1995; Lande 1995; Reed 2005). Forest fragmentation has shown to produce an immediate loss of alleles that is associated with a reduction in population size (White et al. 1999; Cascante et al. 2002). Theoretical studies (Barrett and Kohn 1991) predict that severe fluctuations in the number of individuals or continually small population sizes result in genetic drift, which in turn leads to the fixation of alleles and consequently to a loss of genetic variation. Another consequence of small population size is susceptibility to inbreeding, which reduces heterozygosity and the performance of various fitness-related traits and thereby substantially increases the probability of extinction (Reed and Frankham 2003; Matthies et al. 2004; Reed 2005).

Moreover, forest fragmentation may disrupt the mating system and modify gene flow, resulting in significantly elevated levels of inbreeding, population divergence and a reduction in genetic diversity within populations (Hanson et al. 2008). Inbreeding is a concern to outcrossing populations where highly deleterious mutations and polymorphisms for overdominant alleles may contribute to inbreeding depression (Cheptou and Schoen 2007). However, the negative effects predicted in the genetic theory of small populations may be absent in forest fragments if they are experiencing an effective seed and pollen dispersal over long distances (Hamrick 2004; Kramer et al. 2008).

The study of genetic structure and diversity permits knowledge of the organization and distribution of the genetic variability among and within natural populations (Hamrick et al. 1992; Young et al. 1996; Dick et al. 2008). These understandings have helped to provide important assumptions regarding conservation and the elaboration of management strategies (Bawa 1992; Álvarez-Buylla et al. 1996; Reis et al. 2000; Sebbenn et al. 2008). In addition, the knowledge of genetic variation in conservation strategies secures populations a high adaptive potential to withstand environmental stress such as climatic change, pollution and diseases (Geburek 2000).

Studies have established the fact that most of the genetic variation is distributed within populations (Auler et al. 2002; Gaiotto et al. 2003; Conte et al. 2008), and in many cases the distribution within population is not random, due to the limited dispersal of pollen and seeds or selection pressures or historical events (Heywood 1991; Vekemans and Hardy 2004; Hardy et al. 2006; Dick 2008). This therefore promotes spatial genetic structure (SGS).

Many studies on plants have been devoted to characterizing SGS (Heywood 1991; Epperson et al. 1999; Mantovani et al. 2006; Hardy et al. 2006; Born et al. 2008), because the presence of SGS within populations has important implications in conservation programs as it can affect estimates of genetic parameters (Young and Boyle 2000; Jones and Hubbell 2006), such as the outcrossing rate in natural populations (Ennos and Clegg 1982; Bittencourt and Sebbenn 2007), influencing the sampling strategies for improvement of conservation projects ex situ and it aids management of populations, where exploitation patterns can affect the levels of genetic diversity (Doligez and Joly 1997; Sebbenn et al. 2008).

The species Ocotea catharinensis Mez. (Lauraceae) is a native tree of the Atlantic rain forest (ARF), with natural occurrence in the south and southeast of Brazil. O. catharinensis was the most common and typical tree in the ARF of Santa Catarina State in southern Brazil. It accounted for a third of all tree volume in a hectare (Reitz et al. 1978). Because of its economic value due to its high quality wood and essential oils, this species has intensively been exploited since 1940 (Carvalho 1994). Intensive exploitation along with expansion of agricultural frontiers have caused forest fragmentation and losses of the numbers of individuals and populations. The remaining ARF fragments in Santa Catarina State that may house O. catharinensis populations represent 14.46% (13,693.34 km2) of its original area (IESB 2007). Today, O. catharinensis is threatened of extinction in its natural habitat and also appears in the International Union of Conservation of Nature and Natural Resources Red List of endangered species as vulnerable to extinction (Varty and Guadagnin 1999).

To conserve natural populations of O. catharinensis in the ARF more effectively, demographic and genetic information about this species needs to be obtained. However, there are no studies yet that evaluate the demography, genetic diversity and fine-scale spatial genetic structure in natural populations of O. catharinensis. Thus, the objective of this study was to analyze the demography, genetic diversity and structure and intrapopulation spatial genetic structure of four natural populations of O. catharinensis using allozymic markers. Since we studied two populations inside conservation units in continuous forest and other two inside small, isolated and fragmented forests in private proprieties, the following questions were raised: (1) Do the populations in small, isolated and fragmented forests have equivalent levels of genetic diversity as the ones in the continuous forest? (2) How is it possible to enhance reproductive success and survival of O. catharinensis in the remaining forest fragments? (3) Since seeds of O. catharinensis are animal- and gravity-dispersed (Moraes and Paoli 1995, 1999), should spatial genetic structure be expected in all the studied populations?

Materials and methods

Species biology

Ocotea catharinensis Mez. (Lauraceae) is a long-living, evergreen and dominant canopy species, reaching a height of 45 m and 150 cm in diameter at breast height. The species is indigenous to south and southeast of Brazil. The natural distribution of the species in Brazil ranges from 21°20′ S (Minas Gerais State) to 30°15′ S (Rio Grande do Sul State) and it is found exclusively in the Atlantic rain forest, between altitudes of 30 and 1,400 m (Carvalho 1994). O. catharinensis’ small and white hermaphrodite flowers are pollinated by diverse insects such as bees, wasps, flies and beetles (Reitz et al. 1978). Its seeds are animal- and gravity-dispersed. Animal seed dispersal is carried out by primates such as Cebus apella, Alouatta fusca and Brachyteles arachnoids (Moraes and Paoli 1995, 1999) and birds, e.g., Pipile jacutinga (Galetti et al. 1997). Seedlings of O. catharinensis establish in shade, regardless of the presence of gaps for their development (Moraes and Paoli 1999).

Field collections

Studies were carried out in four natural populations of O. catharinensis in the cities of Ilhota [conservation unit—Parque Botânico do Morro Baú (PMB); 26°47′56″ S, 48°55′49″ W], Santo Amaro da Imperatriz [conservation unit—Parque Estadual da Serra do Tabuleiro (PEST); 27°44′43″ S, 48°49′13″ W], Corupá [Private property of MOBASA/BATISTTELLA (CP); 26°25′06″ S, 49°22′02″ W] and Grão Pará [Private property of Mr. Antônio Alberton (GP) 28°14′17″ S, 49°17′40″ W], all located in the state of Santa Catarina (SC) in the Brazilian ARF (Fig. 1). The distance between populations ranged from 60.7 (CP to MB) to 202.1 km (GP to CP), with an average of 122.6 km. These populations were separated from each other as a result of forest fragmentation.

Fig. 1
figure 1

Map of the study sites (CP, PMB, PEST and GP) in Santa Catarina State, Brazil

Full size image

In all study areas, we georeferenced and numbered all the reproductive individuals of O. catharinensis with aluminum plates (Fig. 2). We collected demographic data [density (D), diameter at breast height (DBH), Clark and Evans (1954) aggregation index (R), mean distance to the nearest neighbor (MD) and local declivities (LD)] and leaf samples from all the reproductive trees in the study areas: PMB (n = 51; 12.5 ha; 500 m × 250 m; LD = 13%), PEST (n = 35; 7.5 ha; 500 m × 150 m; LD = 17%), CP (n = 49; 3.00 ha; 200 m × 150 m; LD = 12%) and GP (n = 53; 14.0 ha; 350 m × 400 m; LD = 13%). Samples were put into plastic bags in ice containers and taken to the laboratory, where they were stored at 5°C until needed.

Fig. 2
figure 2

Spatial distribution of O. catharinensis in PMB, PEST, CP and GP

Full size image

Laboratory procedures

The populations were analyzed by horizontal starch-gel electrophoresis (13%). The enzymes glutamate dehydrogenase (locus Gtdh-1; EC 1.1.1.2), diaphorase (Dia-1 and Dia-2; EC 1.8.1.4), glutamate oxaloacetate transaminase (Got-1 and Got-2; EC 2.6.1.1), isocitrate dehydrogenase (Idh-1; EC 1.1.1.42), phosphoglucomutase (Pgm-1; EC 5.4.2.2), superoxide dismutase (Sod-1; EC 1.15.1.1), acid phosphatase (Acp-1; EC 3.1.3.2), shikimate dehydrogenase (Skdh-1; EC 1.1.1.25), esterase (Est-1,Est-2 and Est-3; EC 3.1.1.1), peroxidase (Po-1 and Po-2; EC 1.11.1.7), malate dehydrogenase (Mdh-1 and Mdh-2; EC 1.1.1.37) and malic enzyme (Me-1; EC 1.1.1.40) were assayed using a continuous, number 10, Tris–citrate buffer system (pH 7.5) (Alfenas and Brune 1998).

Data analysis

For demographic characterization we analyzed Clark and Evans (1954) mean distances between neighbors (MD) and aggregation index (R). In addition, a Z-test was applied to determine the statistical level (P-value) at which the observed pattern deviates from an expected random pattern, when R = 1 the spatial pattern is random, R < 1 clumping occurs (0 = minimum) and R > 1 when a uniform distribution pattern exists (2.149 = maximum). All calculations were accomplished through SGS program (Degen et al. 2001).

For genetic characterization we identified the genotypes and estimated the allelic frequencies for the studied populations from the interpretation of the zymograms. We used the GDA computer program (Lewis and Zaykin 2001) to estimate the percentage of polymorphic loci (P), the mean number of alleles per locus (A), the mean observed (H o) and expected (H e) heterozygosities, and the fixation indices for each population (f i). P was obtained considering the number of loci that presented the most frequent allele with an occurrence of less than 95% in relation to the total number of loci. A was estimated by dividing the total number of alleles by the number of loci. H o was obtained by counting the heterozygote. H e was obtained by Nei’s (1978) formula \( \hat{H}_{\text{e}} = 1 - \sum {p_{i}^{2} } \), where p i is the mean frequency of the allele i in the loci. 95% bootstrap confidence intervals were estimated on f i by resampling loci 10,000 times using the GDA program (Lewis and Zaykin 2001).

Wright’s (1951, 1965) F-statistics (F IS, F IT and F ST) were estimated for the populations with BIOSYS-2 program (Swofford and Selander 1997), using Nei’s (1987) corrections for small sample size. In addition, genetic structure was quantified using Weir and Cockerham’s (1984) F-statistics (f = F IS, F = F IT and θ P = F ST) implemented in GDA program (Lewis and Zaykin 2001). 95% bootstrap confidence intervals (CI) were estimated from F, f and θ P by re-sampling loci 10,000 times using the GDA program (Lewis and Zaykin 2001).

The apparent outcrossing rate \( \left(\hat{t}_{\text{a}} = \left( {1 - \hat{f}} \right)/\left( {1 + \hat{f}} \right)\right) \) was estimated from f as suggested by Carlini-Garcia et al. (2001). And the apparent gene flow \( \left(\hat{N}m = \left( {1/4\alpha } \right)[(1/\hat{F}_{\text{ST}} ) - 1]\right) \) among populations was estimated indirectly, using the island model proposed by Crow and Aoki (1984), which corrects the analyses for a small number of populations. In this expression the genetic divergence among populations (F ST) was substituted by θ P, on account of it is less biased as suggested by Cockerham and Weir (1993), and the correction of the population number (n) was calculated as α = [n/(n-1)]2.

Spatial genetic structure in the four populations was assessed using Moran’s I coefficient (Moran 1950; Sokal and Oden 1978) for each locus. Moran’s I was calculated using alleles with a frequency higher than 5% and for each of seven consecutive distance classes from 0–20 m (class 1) to 120–140 m (class 7) with 20 m distance intervals. 20 m was the minimum distance class size that could be tested in order to have 30 pairs of data points in each distance class, the minimum statistical sample size requirement for Moran’s I (Degen et al. 2001). An average Moran’s I coefficient was also calculated for all loci as a summary statistic. A permutation procedure using Monte-Carlo simulations was applied to test the significance of deviations from randomness of the spatial distribution for each calculated measure (Manly 1997). Each permutation consisted of a random redistribution of multilocus genotypes over the spatial coordinate of the sampled trees. For each distance class, the observed values were compared with 95% confidence intervals constructed for each of the parameters from the distributions obtained after 1,000 permutations. The expected value for the absence of spatial autocorrelation was −1/(n−1) (Sokal and Wartenberg 1983), n was the population sample size. These calculations were performed through the SGS program (Degen et al. 2001). Additionally, Wright’s (1943) neighborhood sizes were estimated for the populations by using the observed average Moran’s I values for the first distance class, and comparing these to theoretical values in Table 4 of Epperson et al. (1999).

Results

Demographic characterization

Population CP presented the highest tree density with an average distance among trees of 8.22 m (Table 1). PEST, PMB and GP presented an equally moderate tree density with an average distance among trees of 17.06, 16.98 and 18.98 m, respectively (Table 1). Based on Clark and Evans (1954) aggregation index (R), this value indicates a rejection of null hypothesis of a strictly random pattern of distribution for the trees within the four populations, yet it suggests a much greater tendency toward randomness than clumping (Table 1). Pooling the four populations, the DBH average of 188 trees was 45.51 ± 23.06 cm, ranging from 6.85 to 148.30 cm. Population PEST presented the lowest DBH, and PMB, CP and GP presented significantly higher values of DBH compared to PEST (Table 1). DBH values for CP and GP did not differ significantly.

Table 1 Demographic characterization in four natural populations of O. catharinensis in the ARF of Santa Catarina state
Full size table

Genetic diversity

Among the eighteen loci resolved into the 12 enzymatic systems, two loci were monomorphic (Got-1and Gtdh-1), and the other sixteen were polymorphic. About 38 alleles were found in population GP, 37 alleles were found in populations CP and PMB, and 36 were found in PEST. The maximum number of alleles per locus was three and the mean number of alleles per locus for the species was 2.2 for all loci (Table 2). The observed and expected heterozygosity values obtained for PEST and PMB deviated from Hardy–Weinberg equilibrium, indicating excess of heterozygote. CP also deviated from Hardy–Weinberg equilibrium, endogamy was detected. Unlike GP that was in Hardy–Weinberg equilibrium (Table 2).

Table 2 Summary of genetic diversity measures and fixation indices (f i ) for four natural populations of O. catharinensis in the ARF
Full size table

The estimate of the fixation index (\( \hat{F}_{\text{IT}} = 0.0 9 7 ;\,\hat{F} = 0. 1 3 3 \), CI = 0.128–0.136) indicated a significant deficit of heterozygosity for the total sample, but the subpopulation fixation index indicated excess of heterozygosity (\( \hat{F}_{\text{IS}} = - 0.0 2 3;\,\hat{f} = - 0.0 1 1 \), CI = −0.016 to −0.009). The genetic divergence among the four populations (\( \hat{F}_{\text{ST}} = 0. 1 1 8;\,\hat{\theta }_{P} = 0. 1 4 3 \), CI = 0.138–0.143) was significant. In agreement with this, the number of migrants per generation (\( \hat{N}m \)) using \( \hat{\theta }_{P} \) was 0.842. The \( \hat{\theta }_{P} \) values calculated among the closest geographical populations (CP-PMB = 0.206; PMB-PEST = 0.126; PEST-GP = 0.102) were significant. In this study, the populations showed a discontinuous distribution, like in the case of the stepping stone model.

Spatial genetic structure

There was significant fine-scale spatial genetic structure within the four populations of O. catharinensis, indicating that the trees share common alleles with a frequency greater in comparison with that expected at random (Fig. 3). Significant Moran’s I values were found among trees separated by less than 40 m in CP and less than 60 m in PEST and GP (Fig. 3). For PMB, significant Moran’s I values were found among trees separated by less than 80 m, except for trees found in the distance class from 20 to 40 m (Fig. 3). The average Moran’s I values for all pairwise tree comparisons for CP, PMB, PEST and GP, within the first distance class (0–20 m) were estimated to be 0.072, 0.166, 0.118 and 0.191, respectively (Fig. 3). In concordance, neighborhood size was approximately 83 individuals in the case of CP and about 25 individual for PMB, PEST and GP.

Fig. 3
figure 3

Correlograms for the four studied populations of O. catharinensis using expected value for the absence of spatial autocorrelation (- -), Moran’s I observed (♦) and 95% bootstrap confidence intervals (…) values

Full size image

Discussion

Demographic patterns

The deviation from an expected random spatial pattern found in the four populations of O. catharinensis is a trend of most tropical tree species (Hubbell 1979; He et al. 1997; Condit et al. 2000; Plotkin et al. 2000, 2002). The tendency toward randomness could be related to the age of the sampled trees, where spatial pattern tends to random distribution as age increases (Itoh et al. 1997; Condit et al. 2000; Plotkin et al. 2000). Spatial aggregation of tree species in tropical forests results from many abiotic and biotic factors, such as edaphic factors, limited seed dispersal and habitat specialization (Ng et al. 2006; Gardner and Engelhardt 2008; Muller-Landau et al. 2008). As the four sampled areas are mature forest and have a similar declivity, differences in spatial distribution of individuals could be more probably related to biotic dispersion of seed. Therefore, O. catharinensis animal- and gravity-dispersed seeds are contributing to the aggregated spatial distribution of adult individuals and each local frugivore community that interacts with the species causes variation in the spatial pattern and dispersal distance.

Diameter at breast height values revealed a large amount of reproductive trees in all the study areas. These trees through seed dispersal can help the species reclaim its original area. Nevertheless, seedlings of O. catharinensis establish in shade, requiring mature or advanced successional stage forest for their development (Carvalho 1994; Moraes and Paoli 1999). To enhance reproductive success and survival of O. catharinensis the remaining mature forest fragments must be protected and landscape connectivity reestablished. To achieve this goal, Harvey et al. (2008) and Perfecto and Vandermeer (2008) provide conservation plans appropriate for Santa Catarina State by reconciling farming and biodiversity conservation. As Santa Catarina State is dominated by agricultural landscapes and conservation units respond for less than 10% of its areas, owners of rural properties must be treated as stakeholders in conserving biodiversity and incentives for sustainable development of the landscape must be provided.

Brazil has very strict laws for forest conservation, the national environment policy (Brazilian federal law no. 6938/1981), the forestry law (Brazilian federal law no. 4771/1965) both updated until the year 2006, and ARF conservation law (Brazilian federal law no. 11428/2006), already treat rural property owners as stakeholders in conserving biodiversity, but there is a lack of policies to encourage conservation. The law in Brazil requires that 20% of the rural property area be reserved to ARF conservation; however in most cases this law is not respected. With proper law enforcement and incentives, in situ conservation of the ARF in rural properties in Brazil may overcome conservation units in area and achieve the minimum requirement established in the Convention on Biological Diversity (CBD 2003) for ecosystem conservation.

As O. catharinensis seedlings need mature or advanced successional stage forest to establish and these forests account for less than 14.46% of ARF original area in Santa Catarina (IESB 2007), a viable solution for expanding O. catharinensis populations and reconnecting forest fragments is the use of agroforestry systems. On the one hand, agroforestry systems may offer shade for the establishment of O. catharinensis seedlings and on the other hand, it can be integrated into biodiversity corridors for a variety of uses, such as timber and non-timber forest products, thereby minimizing the exploitation of protected areas (Huang et al. 2002) and ameliorating the environmental characteristics that produce the edge effect (Perfecto et al. 2007). Therefore, with sustainable agriculture activity and proper law enforcements, the rural landscape could be integrated in the conservation of O. catharinensis and maintain the forest ecosystem functionality.

Genetic diversity

The heterozygosity found in this study (\( \hat{H}_{T} = 0. 4 40 \)) is above the mean heterozygosity found for woody perennial outcrossing species (\( \hat{H}_{T} = 0. 2 8 3 \); Hamrick and Godt 1996; Hamrick 2004). The expected genetic diversity (\( \hat{H}_{e} = 0. 4 2 7 \)), the mean number of alleles per locus (\( \hat{A} = 2. 2 \)) and the mean percentage of polymorphic loci (\( \hat{P} = 8 3\% \)) of O. catharinensis were above the mean values found in Lauraceae (Table 3). The \( \hat{A} \) and \( \hat{P} \) values found for O. catharinensis were closer to those found for other two Ocotea species from the ARF, O. porosa (\( \hat{A} = 2.0 7 \) and \( \hat{P} = 7 1. 4\% \); Silva et al. 2003) and O. odorifera (\( \hat{A} = 2. 3 6 \) and \( \hat{P} = 6 7. 8\% \); Kageyama et al. 2003) than to those for Cryptocarya aschersoniana (\( \hat{A} = 2. 9 4 ;\,\hat{P} = 9 5. 1\% \); Moraes and Derbyshires 2002) another Lauraceae species from the ARF, probably due to the phylogenetic relationships inside Lauraceae.

Table 3 Genetic diversity in different species of Lauraceae
Full size table

The high genetic diversity values detected for O. catharinensis demonstrate that all the studied populations have genetic potential for conservation. Moreover, the high genetic diversity observed might be related to the wide geographical distribution in the ARF and to the mating system, apparently outcross pollination (\( \hat{t}_{\text{a}} = 1.023 \)). Perennial species with a long life cycle, sexual reproduction, predominant outcrossing pollination and wide distribution, accumulate a great genetic variability within the populations (Hamrick et al. 1992; Leimu et al. 2006).

The studied populations occurring inside conservation units in continuous forest and inside small, isolated and fragmented forests in private proprieties presented equally moderate levels of genetic diversity, probably due to the fact that the sampled individuals were present before the forest fragmentation events initiated in the mid-1900. This can be inferred because O. catharinensis is a long-living tropical tree species that may live for more than 300 years, concerning the high dimensions that the trees can reach (DBH > 150 cm) and the very slow growth rate (Carvalho 1994).

Our F-statistics demonstrated population genetic structure and a low migration rate, approximately one migrant per generation. The first event as a determinant of gene flow among populations is an effective pollen dispersal (Hamrick and Nason 2000). A lower divergence among populations can be achieved if the pollen disperser travels over long distances (Kramer et al. 2008). As O. catharinensis is pollinated by various small insects such as wasps, bees, flies and beetles (Reitz et al. 1978) and reproductive trees are located in closed-canopy forest with an average distance among trees of 15 m, probably its pollen dispersers are traveling over short distances, promoting high genetic divergence among the populations. Although long-distance pollination mediated by small insects is common for tropical tree species (Kramer et al. 2008), short pollination distances appear when tree density is high and in closed-canopy forest (Dick et al. 2008; Hanson et al. 2008).

As gene flow is a two stage event and depends on seedling establishment (Hamrick and Nason 2000) we suggest that the dynamic of allele movement in O. catharinensis is strongly related to its restricted seed dispersal carried out by gravity and animals, promoting spatial genetic structure within the populations and high divergences among the populations. Dick et al. (2008) reviewed 42 tropical tree species and found high genetic differentiation among populations (\( \hat{F}_{ST} = 0. 1 7 7 \)) related to low tree density, restricted seed dispersion and biogeographic barriers. An evidence of a biogeographic barrier affecting genetic differentiation among populations in O. catharinensis is the Jaraguá Montain located between populations CP and PMB, where \( \theta_{P} \) presented the highest value (0.206) among the closest populations.

Our results showed that the realized gene flow (\( \hat{N}m = 0. 8 4 2 \)) is not sufficient to stop the increase in the genetic differentiation among population due to genetic drift and local selection, threatening the future viability of O. catharinensis populations. The rule of one migrant per generation (OMPG rule) has been proposed as the required minimum to protect populations from the accumulation of deleterious mutations (Mace and Lande 1991; Hedrick 1995; Mills and Allendorf 1996). Nevertheless, two conditions need to be fulfilled for the OMPG rule: large population sizes and high growth rates (Couvet 2002; Bouchy et al. 2005). Moreover, when the condition of having less than one migrant per generation associated with local adaptations could lead to an increase of the probability of population extinction (Couvet 2002). To reduce O. catharinensis probability of extinction, urgent efforts towards conservation must be intensified in these four different populations. The remaining long-living trees should be used as seed-trees to help the species reclaim its original area and as genetic connectors among fragmented populations.

Spatial genetic structure

The analysis of fine-scale spatial genetic structure within the populations detected SGS in adult trees separated by less than 80 m. In this context, genetic similarity is higher among neighboring individuals than among more distant ones, which is in agreement with the expected pattern of SGS at drift-dispersal equilibrium predicted by the theory of genetic isolation by distance (Vekemans and Hardy 2004). SGS within tropical tree populations has been found for several studies (e.g., Degen et al. 2001; Luna et al. 2005; Hardy et al. 2006; Born et al. 2008) and can be a result of selection or genetic drift, due to environmental heterogeneity or genetic isolation by distance, respectively (Heywood 1991; Epperson 1993).

The distances of SGS in the studied populations were similar to those found for others insect-pollinated and gravity- and animal- seed dispersed Brazilian Lauraceae tree species, for instance, Ocotea odorifera (Kageyama et al. 2003) with significant SGS found in trees separated by less than 74 m (Moran’s I = 0.05) and shorter than Cryptocarya moschata (Moraes et al. 2004) with significant SGS found in trees separated by less than 150 m (Moran’s I = 0.05). Moreover, when our SGS results are combined with the findings of Kageyama et al. (2003) and Moraes et al. (2004) and compared with other Asian Lauraceae species that are insect-pollinated, whose seeds are bird-dispersed and lack SGS (e.g., Neolitsea sericea, Chung et al. 2000; Cinnamomum insularimontanum: Chung et al. 2003), we demonstrate how biotic seed dispersal influences the SGS. Therefore, we should expect genotypes to be randomly distributed within populations with bird-seed dispersal (Chung et al. 2003) and should expect SGS within local populations caused by gravity- and mammal-seed dispersal (Moraes et al. 2004; Hardy et al. 2006).

The observed Moran’s I values for the first distance class corresponds approximately to a neighborhood size of 83 individuals in the case of CP and about 25 individuals for the other populations. Under different densities, the area of O. catharinensis demes corresponds approximately to 5 ha for CP and 6 ha for PMB, PEST and GP. The lowest SGS patterns found for population CP may probably reflect the sampling strategy that only surveyed 3 ha and 49 individuals, less than the expected deme area and size, respectively. These results also alert us that small forest fragments with at least 5 ha may contain exclusive alleles with high adaptive potential to withstand environmental stress. Therefore, the conservation of genetic diversity across an entire population and in small fragments must be considered in conserving the species evolutionary potential (Chung et al. 2001; Cruse-Sanders and Hamrick 2004).

Conservation implications

The high genetic diversity in the studied populations illustrates the importance of preserving them for the conservation of the species evolutionary potential. Our results also proved that the realized gene flow is not enough to reduce the genetic differentiation among populations due to genetic drift and local selection. As illegal deforestation continues in the Brazilian ARF, the removal of small forest patches, up to 6 ha, can eliminate an entire O. catharinensis deme and consequently, eliminate alleles within these demes. Moreover, forest fragmentation may increase, reducing the connectivity between populations of O. catharinensis and hence, the SGS and genetic differentiation could also increase. However, connectivity between fragments could be reestablished to maintain gene flow among populations if rural landscape is integrated to conservation planning through agroforestry systems.

The analyses of demographic patterns and SGS are powerful tools for developing conservation strategies and pointing out undisturbed populations (Chung and Nason 2007). As the four studied populations revealed spatial aggregation of trees and SGS, these features may represent a natural state in O. catharinensis undisturbed populations. Therefore, we can use our results as a pattern to alert conservation managers that increased SGS values may indicate forest fragmentation effects, or even the lack of a frugivore community that disperses O. catharinensis seeds. On the other hand, the absence of SGS may indicate anthropogenic disturbance, such as harvesting. Accordingly, an efficient restoration plan within forest fragments, which would provide a balanced family structure, should involve the entry of new alleles through seeds. These seeds could come from gene flow corridors or even by direct human intervention. Although, a special regard should be taken when using these seeds, they must come from nearby populations to avoid maladaptation problems (e.g., McKay et al. 2005; Bower and Aitken 2008). Concerning in situ and ex situ genetic conservation, the seeds must be collected from trees at least 80 m apart reducing genetic similarity among seed-trees to obtain a representative sample.