Seedling density according to structure , dominance and understory cover in old-growth forest stands of the evergreen forest type in the coastal range of Chile

Securing timely regeneration is essential in maintaining the long-term ecological or silvicultural functions and values of forests. Its establishment, in turn, depends on many factors, including the structure and composition of the forest itself. Available information shows that seedling density varies greatly across the evergreen forest type in Chile. Yet stand variables that may affect the establishment of advance regeneration have not been studied. To that end, we evaluated seven stands of the coastal range, within the northern part of the evergreen forest type (39°14’-40°16’ S). We documented understory cover, tree density and dominance, and stand structure, and used the information to assess their effects over seedling density. Findings indicate that Laurelia philippiana was the dominant canopy and regenerating species in these stands. Also, seedling density was significantly greater in stands at lower elevations where shade-tolerant Aextoxicon punctatum was important. Chusquea spp. and Lophosoria quadripinnata, both understory species, had a significant negative effect on seedling density. Basal area and canopy cover, per se, showed little relationship with seedling density. Vertical structure, evaluated through a crown index, had a significant relationship with seedling density, but the direction depended on the species (e.g., L. philippiana and A. punctatum) and the diameter structure within our plots. Fitted models that included these variables were highly significant, and in most cases their significance increased considerably (14 to 26 %) when we accounted for the diameter structures of the plots.


INTRODUCTION
Forest regeneration is the key process that ensures successive generations of trees (Barnes et al. 1998), and a primary requirement for maintaining the ecological or silvicultural functions of a forest.Yet tree seedling establishment depends on many biophysical factors, including the structure and composition of a stand (Crow & Metzger 1987, Nyland 2002), the abundance and distribution of canopy gaps (Veblen 1992, and references therein), and the presence of microsites suited to tree seedling establishment (Lusk 1995, Szewczyk & Szwagrzyk 1996, Chistie & Armesto 2003).Within old-growth forests, tree seedling density also seems related to the availability of propagules of species that can grow in shade or beneath small canopy openings (sensu Oliver & Larson 1996), such as species adapted to the continuous and gapphase regeneration modes (Veblen 1992).In these forest stands, seedling establishment varies with the degree of shade cast by larger trees (e.g., Leak & Graber 1976, Crow & Metzger 1987, Clark et al. 1999), and by understory species that might oppress tree seedling development by reducing light availability near the ground.In forests with great vertical structure diversity, light levels in the understory are relatively higher than beneath stands with a poorer vertical stratification (Brown & Parker 1994,Van Pelt & Franklin 2000), and this may result in greater understory tree density in areas outside canopy gaps (Van Pelt & Franklin 2000).
Aside from these general findings, research has not clearly identified the stand variables that affect the characteristics of advance regeneration within mature and old-growth forest stands of the evergreen forest type.To address this question, we sampled seven stands (> 250 yrs old each) along the coastal range in the Province of Valdivia.We sought to describe differences in abundance and composition of regeneration with reference to overstory and understory characteristics.We hypothesized that tree seedling density would be lower in stands having greater understory and overstory cover, and more abundant in stands having a high degree of vertical structural complexity.

STUDY AREA AND STAND CHARACTERISTICS
Sample stands were located in the Provinces of Valdivia and Cautín, along the Coastal Range and its extension to the Central Depression of Chile, between 39 o 14' and 40 o 16' S (Table 1).This region has a temperate maritime climate with high annual precipitation, and an average annual temperature of 11-12 °C over the last 37 yr (Pezoa 2003).It is within a broader macroclimatic region lacking a dry period (Schlatter et al. 1995), except for the stand at Los Boldos on the western slope (Table 1).That area has one dry month per year (Schlatter et al. 1997).Soils on the eastern side of the Coastal Range (Table 1) have old volcanic ash deposited over the mica-schist metamorphic bedrock, are deep (> 60 cm), and have moderate to good drainage.The western side of the coastal range (Table 1) has soils developed in situ over the mica-schist bedrock.These have poor drainage, are shallow (< 46 cm) (Schlatter et al. 1995, Schlatter et al. 1997), more acidic, and have a higher bulk density (CIREN 1999).
We located four stands on the eastern side of the Coastal Range (two at 330 m, and two at 630 m elevation), and three along the western side (at 500-540 m).Each covered at least 3 ha, with original basal areas ranging from 71 to 111 m 2 ha -1 .Of the seven stands, three had been selectively cut 5-7 yr earlier.In another stand, groups of trees were removed 16 yr earlier, with no cutting between the groups (Donoso 1989).Three stands had no cutting history (Table 1).All together, the partial cuttings removed 16 to 26 % of the basal area, and about 10 % of the trees, leaving an average stocking of 64 to 98 m 2 ha -1 .The quadratic stand diameters (QSD, diameter of the tree of mean basal area) ranged from 33.6 to 47.1 cm.
None of the cuttings affected the overall composition or the diameter and vertical structures of the stands (Donoso 2002).They were characterized by an upper canopy layer of E. cordifolia and Weinmannia trichosperma, a subcanopy layer dominated by L. philippiana, and an understory of Myrtaceae species with scattered Raphythamnus spinosum and Lomatia ferruginea.The Llancacura stands also had an emergent layer of scattered Nothofagus dombeyi trees.At La Montaña, Aextoxicon punctatum shared all diameter classes with L. philippiana.In stands on the western slopes, Podocarpaceae species (particularly Saxegothaea conspicua) where abundant (Donoso 2002).Overall, the stands correspond to the evergreen-with-shade-intolerantemergents forest subtype as defined by Donoso (1981).

Seedlings, saplings, and undergrowth
In each stand, we systematically located five to six square 1,000 m 2 plots, each separated by 50 m.In these plots we measured the diameter at breast height (dbh) of all trees ≥ 5 cm.Each plot was then subdivided into four equal 250 m 2 squares, each having four 2-m 2 circular subplots located at 7.45 m from the corners.This provided a total of 16 subplots (32 m 2 ) in each 1,000 m 2 plot.In the subplots, tree seedlings were recorded by height class (5-< 50 cm, 50-< 100 cm, and 100-200 cm) and species.The cover (%) of herbs, shrubs, and ferns was also estimated by four height classes (< 50 cm, 50-< 100 cm, 100-< 200 cm, and ≥ 200 cm).Seedlings < 5 cm tall were not tallied because most of them are ephemeral.Saplings (trees < 5 cm dbh and taller than 2 m) were also counted in the 2 m 2 subplots.This systematic sampling allowed us to estimate seedling abundance over average microsite conditions, without bias towards any particular set of conditions within the stands.

Overstory variables
Among overstory variables that could potentially affect regeneration, we evaluated basal area, canopy cover, and vertical structure.Although small gaps were observed in portions of the sample stands, the high average basal areas (50 to 90 m 2 ha -1 ) indicate that the plots To estimate the actual canopy cover, we measured the dbh and canopy area of all trees ≥ 5 cm dbh in one 250 m 2 subplot within each 1,000 m 2 plot.The widest crown diameter and the one perpendicular to it were recorded for each of these trees.From these measurements, we modelled the relationship between dbh and crown area (based on an ellipse) (Donoso 2002), and used the models to calculate canopy cover across the plots.
To analyze vertical structure, we assigned each plot to a specific structure class based on a modification of the Triangle of Structures (Fig. 1) developed by the French Society of Franche-Comté for irregular (multistrata) forests (SFFC 2000).Their classification recognizes seven possible structures, depending on the proportion of trees in small (10-25 cm dbh), medium (> 25-50 cm dbh), and large (> 50 cm dbh) dbh classes, as follows: S = stand or plot dominated by small trees; SM = stand or plot dominated by a mixture of small and medium trees; SL = stand or plot dominated by a mixture of small and large trees; M = stand or plot dominated by medium trees; L = stand or plot dominated by large trees; ML = stand or plot dominated by a mixture of medium and large trees; IR = stand or plot dominated by a mixture of small, medium, and large trees.Vertical structure was also represented by a Crown Index (CI) (Donoso 2002), based on canopy cover for four broad classes of trees: upper main canopy (UMC; >50 cm), lower main canopy (LMC; 30-50 cm), below main canopy (BMC; 15-< 30 cm), and overtopped (OT; < 15 cm).We calculated CI as: CI = (UMC/UMC+UMC/LMC+ UMC/BMC+UMC/OT) / 4 Values close to 1.0 indicate a balance in the distribution of crown area among the different tree size classes.Values greater than 1.0 reflect a disproportionate crown area among large trees.A CI of < 1.0 identifies stands with higher concentrations of crown area in small trees.

Analyses
Cover of understory species, and seedling and sapling density by species, was compared among the stands.Data for canopy and noncanopy species were analyzed separately.Variation in seedling density was also compared between plots with different structures: plots where small trees were dominant (S) or codominant (SM and SL) were analysed separately from plots with a more balanced mixture of tree sizes (IR) (see Fig. 1).Tree dominance was evaluated through basal area and canopy cover.Understory variables considered were the overall cover of Lophosoria quadripinnata and Chusquea spp., and the cover of these two species in the tallest height classes (> 50 cm for L. quadripinnata, and >200 cm for Chusquea spp.).were selected.Transformations were conducted when necessary to improve the models.Possible collinearity between regressor variables were addressed with the COLLIN option in SAS.Individual independent variables were kept in the regression analysis when significant at α ≤ 0.05 and were not collinear with other variables in the model.Analysis of variance and means separation by the least square difference (LSD) test (α = 0.05) were used to compare seedling densities among stands.

Differences in species composition
Twenty-two species of trees (≥ 5 cm dbh) were sampled across the seven sites.Stands on the eastern slopes had between 10 and 12 tree species each, while those in the west had from 12 to 16 per stand (Table 2).However, in all stands only two or three species dominated, as reflected by their importance values (IV) (Table 2; Cottam & Curtis 1956, Donoso 1993).We identified more than 60 species in the understory of the seven stands (Donoso 2002).Only 10 species were common to all stands, or had a relatively high cover in at least some stands (Table 3).Chusquea spp.(a bamboo) had the highest cover in all stands except for Llancacura Uncut and Llancacura Cut.There L. quadripinnata, a 1 to 2 m tall fern, had the highest cover.Chusquea spp.accounted for > 42 % cover at Hueicolla, 30 % at Los Boldos, 12 % at La Montaña, and < 5% at Llancahue and the Llancacura stands.Other species with relatively high cover in most stands were Luzuriaga spp., Nertera granadensis, and the vine Mitraria coccinea (Table 3).
Llancahue had the least understory cover (12.6 %) and the lowest average cover in each of the three height categories.Llancacura Uncut had 20 %, and Llancacura Cut 25 % cover.The two cut stands on the western slope (Hueicolla Cut and Los Boldos) had the highest covers (64 and 63 %).Hueicolla Uncut and La Montaña had 52 and 53 % cover (Table 3).For the two tallest understory height classes, Hueicolla Cut had 56 % cover, Hueicolla Uncut 43 %, and Los Boldos 31 % cover.La Montaña and Llancacura Cut ranked intermediate (14.8 % and 15.6 %), and Llancacura Uncut (10.2 %) and Llancahue (5.6 %) had the lowest covers.The relatively high tall understory cover along the western slope was due to Chusquea spp.

Seedling density of canopy species
Stands on the eastern slope had more seedlings.In all stands, the 5-< 50 cm height class had the greatest numbers of seedlings per ha (Table 4), and we could distinguish three groups of stands based on stocking in this seedling height class: (1) La Montaña, with 24,583 seedlings ha -1 and significantly different from all other stands; (2) Llancahue, Llancacura Uncut, Llancacura Cut, and Los Boldos, with 13,332 seedlings ha -1 at Llancahue and 6,000-9,000 seedlings ha -1 beneath the other stands; and (3) Hueicolla Cut and Hueicolla Uncut, with the fewest seedlings (3,500 and 1,668 seedlings ha -1 , respectively).
For the 50-< 100 cm class, Llancahue, La Montaña, and Llancacura Uncut had > 3,000 seedlings ha -1 .Those stands differed significantly from Hueicolla Uncut (968 seedlings ha -1 ).Llancacura Cut, Hueicolla Cut, and Los Boldos ranked intermediate (1,200 to 2,300 seedlings ha -1 ).For seedlings 100-< 200  cm tall, La Montaña, Llancacura Uncut, and Llancahue had > 3,000 seedlings ha -1 , and Hueicolla Uncut had the fewest (563 seedlings ha -1 ).The remaining stands had from 1,438 to 2,396 seedlings ha -1 in this height class.Considering all height classes, La Montaña had significantly more seedlings than the other stands.Llancahue ranked second, and although it had > 20 thousand seedlings/ha, it did not differ significantly from the Llancacura stands that had 14,479 (Cut) and 12,708 (Uncut) seedlings ha -1 .Los Boldos, Hueicolla Cut, and Hueicolla Cut, all on the western slope of the Coastal Range, had the fewest seedlings.These stands were significantly different from La Montaña and Llancahue.The high seedling density at La Montaña was due to the abundance of A. punctatum in the 5-< 50 cm class, with numbers significantly greater than all other stands.Seedlings at Llancahue were more evenly distributed among A. punctatum, E. cordifolia, and L. philippiana (Table 4).E. cordifolia was most abundant at Llancahue.At Llancacura, L. philippiana was the most abundant species, and accounted for 68 % (Llancacura Uncut) and 60 % (Llancacura Cut) of all seedlings.
The density of saplings was also greater in stands on the eastern slope, with significantly greater numbers at Llancahue than elsewhere (2,656 versus an average of < 1,000 saplings ha -1 for the other stands), except for La Montaña.Although Hueicolla Uncut and Hueicolla Cut had only limited numbers of seedlings and saplings of upper canopy species, they had the lowest ratio of numbers in the 5-< 50 cm class to those of sapling size, suggesting a better rate of survival for the few seedlings that became established.By stand, these ratios were: Hueicolla Uncut = 3.4,Hueicolla Cut = 4.3, Llancahue = 5.0, Los Boldos = 5.8, Llancacura Uncut = 6.3,Llancacura Cut = 17.3,La Montaña = 22.1.quadripinnata, was also negatively related to seedling density in one model (S plots for all canopy species).For both species, in many cases the tall cover in L. quadripinnata (> 50 cm) and Chusquea spp.(> 200 cm) had a more significant effect than the cover with all height classes combined.
Basal area and canopy cover were not significant explanatory variables in any model.Among structural variables, QSD was negatively related to seedling density in the model for all plots and canopy species.CI was positively related to seedling density in the models for all plots, for all canopy species and L. philippiana, and for S plots and L. philippiana.It was negatively related to A. punctatum in S plots.
Models that were fitted for plots according to their structure (as reflected by proportions of large, medium, and small trees) had, in all cases, greater R 2 values than those without this separation.Except for the model for S plots based on all canopy species, increases in R 2 values were from 14 to 26 %.

Effects of diameter structure over seedling densities
Although all plots were dominated by similar species and had a multistoried canopy, those with S or SL structures had greater numbers of seedlings.Stands with an IR structure had intermediate seedling densities (Table 7).Stands with diameter distributions having few large trees, medium numbers of intermediatesized ones, and a discontinuous upper canopy layer apparently favoured the development of advance regeneration.By contrast, stands with a dense middle canopy layer had the lowest density of regeneration, possibly due to heavy shading from the many intermediate-sized trees of medium heights.For L. philippiana, seedling density was greatest among plots having similar numbers of small and large trees (SL plots, Table 7), compared to plots dominated by small trees or a combination of small and medium trees.This suggests that L. philippiana regenerates best in plots that have a limited middle canopy layer, and important proportions of large trees.In stands with an important degree of middle story development (i.e., IR and SM plots), seedling densities were lower (Table 7).As with L. philippiana, seedling density of A. punctatum was lower in IR and SM plots (Table 7).

Effects of understory species over seedling density
Chusquea spp. was the most influential understory species affecting seedling density.These species characterize the understory component of Chilean temperate forests located from 37 to 43° S. Their negative effects on regeneration success have been reported extensively for other temperate forests within Chile, and elsewhere (Veblen 1992, Donoso Average and standard deviation of seedling densities of all canopy species, L. philippiana, and A. punctatum in plots with differing proportions of trees in broad diameter classes, according to Fig. 1 El promedio de desviación normal de densidades de plántulas de todas las especies de dosel, L. philippiana y A punctatum en las parcelas con diferentes proporciones de árboles en las clases del diámetro anchas, según Fig.The fern L. quadripinnata had a relatively important effect on tree seedling densities.In fact, tall cover (> 50 cm) of this species was a significant explanatory variable in one of the regeneration models.Similarly, the tall cover of Chusquea spp.(> 200 cm), rather than total cover of that species, was a significant explanatory variable of tree seedling density within several regeneration models.
Available information suggests that interference may be caused by rhizomatous ferns (e.g., reported in North America by Horsley 1993, andin New Zealand by Beveridge 1973), by Chusquea spp.(Veblen 1992, Donoso 1993, Lusk 1995, Lusk 1996, González et al. 2002; see also references in these sources), or by a generally dense cover of other tall understory vegetation (Lorimer et al. 1994, Clinton 2003).Under appropriate conditions, bamboos and ferns rapidly dominate the understory, preempt resources, and create an unfavourable understory microclimate for other species (Clinton 2003), including trees.This seems the case among our sample stands having the highest cover of L. quadripinnata and Chusquea spp., which were also the stands with the lowest number of vascular plants and at an elevation > 500 m (C.machrostachya at Hueicolla Uncut, Hueicolla Cut, and Los Boldos, and L. quadripinnata at Llancacura Uncut and Llancacura Cut).The two lower elevation stands (Llancahue and La Montaña) seem to presently have better regeneration of canopy species.Based on the diameter distributions in these stands, with the interspersion of trees having different diameters and heights (Donoso 2002), this might also have been the situation in the past.

Effects of tree dominance and vertical structure over seedling density
Basal area and canopy cover were used as measures of dominance in this study.In the past, canopy cover has been regarded as a more reliable measure of dominance, because basal area continues to increase at a linear rate even after crown area has approached an asymptote (Cole 1991).Both high basal area (e.g., Leak & Graber 1976, Crow & Metzger 1987) and a dense canopy cover (Clark et al. 1999) reportedly have negative effects on regeneration success.Yet those variables did not explain differences in seedling density in our study.Similarly, Berg & Van Lear (2004) also found little relationship between these factors and regeneration density in other temperate hardwood forests.
We used crown index (CI) and quadratic stand diameter (QSD) as surrogates for stand structure, and entered them as variables in the regression models.QSD was only significant in the model for all plots and all species, and had a negative relationship to seedling density.CI was a significant explanatory variable in several models.While all the sample plots had high vertical structure (sensu Brokaw & Lent 2001), CI reflected the degree of balance among the different tree strata within a plot.
For all canopy species, CI was positively related to seedling density when we lumped all plots.When we separated the plots into Sdominated or IR plots, CI was not a significant explanatory variable.The positive relationship of CI and seedling density for L. philippiana in the model for combined plots was very weak (Table 6).The relationship greatly improved when only S-dominated plots were considered.For these plots CI was the only explanatory variable for seedling density.For plots having more balanced numbers of trees across the diameter classes (IR plots) we observed no relationship between seedling density and CI for L. philippiana.Understory cover rather than vertical structure seems to more profoundly affect tree seedling density in those cases (Table 6).
With A. punctatum we found no relationship between seedling density and CI for all plots combined, or among S-dominated plots.For the latter condition, density was negatively affected by cover of Chusquea spp.Density was sensitive to CI only within IR plots, and seedling numbers declined with higher CI values (i.e., plots with a relative high proportion of crown area among large trees).
Although both species seem sensitive to shade by Chusquea spp., A. punctatum seems more tolerant than L. philippiana to conditions beneath a continuous canopy in stands having more balanced numbers of trees across the diameter classes (IR plots).Laurelia philippiana appears especially sensitive to shade cast by a middle canopy layer and regenerates better beneath stands with a broken vertical structure.Usually both species have been classified as being very shade tolerant (Donoso 1993, Figueroa & Lusk 2001).Yet, in addition to the present findings that relate the regeneration of these species according to cover and vertical structure, little has been reported on their habitat choices in regions where both species grow and coexist.For regions where A. punctatum does not grow, it has been reported that for L. philippiana numbers of seedlings established on logs are similar or greater than those established on litter (Lusk 1995, Christie & Armesto 2003).On the contrary, in low-land forests of Valdivia, Neumann (2001) reported that both L. philippiana and A. punctatum had a small proportion of seedlings growing on logs, although the former had a greater proportion (18 versus 3 %).Thus, in addition to possible microsite preferences, these two species (widely known as shade tolerant) also seem to respond differently to light availability.Analyses of both variables combined could provide a better explanation for plant recruitment, especially considering that regeneration on coarse woody debris may help these tree species to escape understory species interference.
This observational study revealed a wide range of seedling densities among stands of the evergreen forest type in the coastal range of Chile.Findings indicate that within the levels of basal area studied (only one plot having < 40 m 2 ha -1 ), understory cover of Chusquea spp.and L. quadripinnata, and stand vertical structure, were the best indicators of seedling density for canopy species in general, as well as that of the two major shade-tolerant species.Neither the degree of canopy cover nor the total stand basal area, alone or in combination, appeared to significantly affect seedling density in stands of this forest community type.Silvicultural experiments that test alternate levels of residual basal area might identify thresholds where those two stand attributes have greater effects on the establishment and development of advance regeneration, and where less shade-tolerant species like C. quila begin to importantly interfere with seedling establishment within low elevation stands.

Fig. 1 :
Fig.1: Triangle of structures for irregular forests from the Society of Foresters of the French-Comté in France (SFFC 2000) for classification of irregular forests into one of seven possible diameter structures (see text): T = timber; S = small; M = medium; L = large; IR = irregular.

TABLE 1
General locations of the seven stands studied in evergreen forests of Chile Ubicación general de los siete rodales estudiados en bosques siempreverdes

TABLE 3
Mean cover a and standard errors for main understory vascular non-tree species in the five or six plots sampled in each stand

TABLE 4
Seedling densities (number ha -1 ) of canopy species in the seven stands studied in Valdivian forests