Mortality, recruitment and growth of the tree communities in three forest formations at the Panga Ecological Station over ten years (1997-2007)

The area evaluated in this study was a continuous stretch comprising three vegetation formations: gallery forest, semideciduous seasonal forest and cerradão (woodland savanna). The aim of this study was to examine the tree community dynamics in a forest gradient—from gallery forest to cerradão—at Panga Ecological Station, in the city of Uberlandia, located in the state of Minas Gerais, Brazil. The study was based on data from a previous inventory of the continuous forest conducted in 211 permanent 10 × 10 m sample plots in eight parallel transect running perpendicular to Panga Creek. Trees with a diameter at breast height ≥ 4.77 cm were sampled in 1997, 2002 and 2007. With the exception of the cerradão, there was a net reduction in tree density over the studied period of ten years, because mortality rates were higher than the recruitment rates. The basal area increased during the period of the study, especially at cerradão. The mean mortality rate in the studied area was 2.64%.yr−1 and 3.36%.yr−1 for the 1997-2002 and 2002-2007 periods, respectively, whereas the mean recruitment rate was 1.76%.yr−1 and 1.97%.yr−1, respectively. In general, mortality rates and recruitment rates have increased during the two successive periods of measurement and showed an imbalance in favor of mortality for the semideciduous seasonal forest and the gallery forest. This fact, added to the low density and high basal area, suggest that there was a process of thinning in the tree community. However, at cerradão, there was an imbalance in favor of recruitment, with a consequent increase in density and basal area, indicating that the cerradão is in a construction phase, which was further favored by a decrease in the occurrence of fire and other anthropogenic disturbances. When the turnover rates are taken into consideration, the global dynamics of the study area over the ten years evaluated can be expressed as cerradão > semideciduous seasonal forest > gallery forest.


Introduction
Long-term studies on temporal changes in tropical forest remnants are necessary to make the distinction between natural dynamic processes and changes resulting from human activity (Korning & Balslev 1994a).Such studies can contribute to the prediction of forest growth and productivity, facilitating the implementation of management programs, as well as promoting the rational use and recovery of tropical forests (Carey et al. 1994).The changes that occur over time in tropical forests, in terms of their composition and structure, are the product of internal processes, such as competition for light (Harcombe et al. 2002), and external processes, such as climatic, geological and anthropogenic changes (Condit et al. 1992).
Forest communities that are free of anthropogenic disturbance display a dynamic equilibrium.Changes occur continuously over time at the individual and population levels, due to a balance among growth, recruitment and mortality (Hartshorn 1980;Swaine et al. 1987;Felfili 1995b), as well as local extinctions, immigration of new species and ecological drift (Condit et al. 1992).Therefore, these communities do not constitute a single equilibrium stage but a mosaic of successional stages, with arrays of species and individuals in different phases of regeneration, subjected to recurrent disturbances at varying frequencies (Hartshorn 1980).
Studies of the dynamics of tropical forests have been fundamental to the characterization and understanding of the diversity and complexity of plant populations and communities (Condit et al. 1992).With knowledge of the dynamic processes that modify the structure of a forest community, it is possible to infer the tendencies of a given ecosystem (Clark & Clark 1999) regarding changes in basal area, biomass and carbon balance (Lewis et al. 2009).
In tropical seasonal forests and in gallery forests subjected to annual climate seasonality, the rates of mortality and recruitment are higher than 3% per year (Paiva et al. 2007;Oliveira & Felfili 2008;Silva & Araújo 2009;Carvalho & Felfili 2011), whereas in wet forests these rates are lower, not exceeding 2% per year (Lang & Knight 1983;Manokaran & Kochummen 1987;Korning & Balslev 1994a;Rolim et al. 1999;Marques et al. 2009).Therefore, because of seasonality, together with functional and successional ecological processes (Murphy & Lugo 1986;Swaine et al. 1990), the community dynamics are more accelerated in seasonal and gallery forests than in wet forests.
The aim of this study was to detect the patterns of mortality, recruitment and growth of the tree community of a forest gradient-from gallery forest to semideciduous seasonal forest to cerradão (savanna woodland)-in two successive five-year intervals (1997-2002 and 2002-2007).These forest formations are subjected to pronounced seasonality in the rainfall distribution, with a long period of low precipitation (Oliveira-Filho & Ratter 2002).We tested the following hypotheses: because the study site has a seasonal climate, it displays accelerated dynamics, with high mortality and recruitment rates; and the dynamic processes are most intense in the cerradão, followed by the seasonal forest and the gallery forest.

Study site
The study site is located at the Panga Ecological Station (PES), near the city of Uberlândia, which is in the state of Minas Gerais, Brazil, and comprises three continuous forest formations: gallery forest (surrounding the Panga Creek), semideciduous seasonal forest and cerradão (Fig. 1).Until 1984, the area now occupied by the PES was extensively used as agricultural land (Cardoso & Schiavini 2002).The PES was acquired in 1986 by the Federal University of Uberlândia.It has a total area of 403.85 ha and is located approximately 30 km from Uberlândia (Schiavini & Araújo 1989).The vegetation of the PES consists of several vegetation formations found in the Cerrado (Savanna) Biome (sensu Ribeiro & Walter 2008), including forests, savannas and fields (Schiavini & Araújo 1989).
Within the PES, the gallery forest is distributed mainly along Panga Creek, which constitutes the northern border of the station, and, to a lesser extent, along the drainage channels and small streams in its southeast portion (Schiavini & Araújo 1989).Although the semideciduous seasonal forest is physiognomically similar to the gallery forest, it is located on an inclined relief, as an extension of the gallery forest, and possesses a characteristic flora (Schiavini & Araújo 1989).The cerradão is adjacent to the semideciduous seasonal forest, occupying a higher topographic position (Schiavini & Araújo 1989).
According to the Köppen classification system (Köppen 1948), the climate of the region is predominantly Aw type, with hot, humid summers and cold, dry winters (Alves & Rosa 2008).The mean annual precipitation is 1590 mm (± 247 mm), and the mean temperature is relatively uniform throughout the year.According to the Uberlândia Station of the 5th Meteorology District of the Brazilian Ministry of Agriculture and to the Climatology Station of the Federal University of Uberlândia, the wet season is from October to March and the dry season is from April to September.
The soil of all three forest formations has been classified as being of average texture and average acidity (pH 5.0-5.9), with low aluminum content in that of the semideciduous seasonal forest (Moreno & Schiavini 2001).According to Moreno & Schiavini (2001), the gradient of fertility for the study site, represented by the values for base saturation and sum of bases, follows the sequence semideciduous forest > gallery forest > cerradão.The study site is located between elevations of 739 m and 773 m, thereabout, and the creek shows a mean declivity of approximately 0.18% (Cardoso & Schiavini 2002).

Continuous survey and data analysis
The survey was conducted in 211 permanent, 10 × 10 m sampling plots, in eight parallel transects running perpendicular to Panga Creek (Fig. 1).The transects end at the border between cerradão and cerrado sensu stricto (savanna).All individuals with a circumference at breast height ≥ 15 cm were sampled in 1997.The circumference data were transformed to diameter for the analyses employed in this study.In 2002 and 2007 the community was surveyed again, to record the growth (change in diameter), mortality and recruitment.The dataset from the surveys was used in order to calculate the demographic rates for both sampling periods (1997-2002 and 2002-2007).
The parameters of dynamics (mortality, recruitment and growth) were described for the study site as a whole, for each forest formation (gallery forest, semideciduous seasonal forest and cerradão) and for each vertical stratum (understory, midstory, overstory and emergent layer).The Shannon index (H') at the natural base was used as a measure of diversity and the Pielou index (J') was used as a measure of evenness, following Magurran (2004).The structure of the community was described in terms of number of individuals, basal area and distribution by diameter.
The classification of species according to their position in the vertical strata of the community was performed separately for each forest formation.The analysis of vertical stratification, adapted from Vale et al. (2009), was based on the height data for each species, with the aim to represent the true vertical position of the species (i.e., which stratum the species occupies when it reaches its maximum size) within the community.Only species with a minimum of ten individuals in the community in the 1997 survey were included in the analysis.The classification of strata considered the following criteria and intervals: • understory where Q3s and Q3c are the third quartiles of the heights of individuals, within a species and within the community, respectively; Mc is the median height of the individuals sampled in the community; and D9c is the ninth decile of the heights of the individuals sampled in the community.The mean annual rates of mortality and recruitment were based on the logarithmic model (Swaine & Lieberman 1987).Because the basal area is also involved in the processes of mortality and recruitment, the mean annual rates of loss and gain in basal area of the individuals were estimated, as described by Werneck & Franceschinelli (2004).To express the global dynamics, the rate of turnover in number of individuals was calculated from the mean rates of mortality and recruitment, and the rate of turnover in basal area was calculated from the rates of loss and gain in basal area (Oliveira-Filho et al. 2007).
Growth is usually expressed in terms of diameter and basal area.Diameter is the main determinant of basal area, which justifies the preference for expressing growth in terms of diameter (Braga & Rezende 2007).The annual diameter increment (ADI) expresses the annual growth of an individual and is based on the difference in diameter between two periods, divided by the time in years between the two measurements (Finger 1992).The mean increments were calculated for the surviving trees between the periods of 1997-2002 and 2002-2007.In calculating the mean ADI, we converted each of the negative ADIs recorded for some individuals to zero, thus assuming that there was no growth in the period, as adopted by Felfili (1995b).The negative ADIs represented less than 0.01% of the ADIs analyzed in the community.
The differences in the rates of mortality and recruitment among the vertical strata for each period were determined by the Kruskal-Wallis test, followed by the median test for multiple comparisons among the strata (Zar 2010).The median test quantifies the number of times each stratum appears above or below the median and compares this with the chi-square distribution in a contingency table of expected and observed values.The Wilcoxon nonparametric test was used in order to compare the two measurement periods (1997-2002 and 2002-2007) in terms of the mean rates of mortality and recruitment for each stratum (Zar 2010).The statistical analyses were performed at the 5% significance level with the Systat program, version 10.2 (Wilkinson 1990).

Results
In 1997, the forest gradient showed a total of 3797 individuals, with a basal area of 23.95 m 2 /ha (Table 1).In general, the results suggest a net reduction in density over the studied period of ten years (Tab.1), because mortality rates were higher than were recruitment rates (Tab.2), the only exception being the cerradão.The basal area of the gradient increased during the study period, especially for the cerradão (Tab.1).As for the floristic composition, the number of species declined over the study period.The fluctuation in the number of species in the community is due to the entry and exit of rare species (n < 2), which did not alter the estimated values of diversity and evenness (Tab.1).The same was observed for the botanical families.Among the rare species removed from the floristic list in the gradient are those whose occurrence was restricted to the cerradão and those that are typical of the adjacent area of cerrado sensu stricto (Appendix 1).
2002-2007, respectively.The recruitment rates for the same period were 1.76%.year−1 and 1.97%.year−1 , respectively.Between the two periods, the mortality and recruitment rates increased for all of the vegetation formations studied, except for the gallery forest, which showed a decrease in the recruitment rate (Tab.2).All formations except the cerradão showed an imbalance in favor of mortality.Recruitment rates were higher in the cerradão: approximately 3% per year.The semideciduous seasonal forest showed the greatest reduction in basal area for both measurement periods, due to the death of large trees, whereas the greatest gain, due to high recruitment, was in the cerradão (Tab.2).When the turnover rates are taken into consideration, the global dynamics of the study area over the ten years evaluated can be expressed as cerradão > semideciduous seasonal forest > gallery forest.
The species evaluated in the analysis of the vertical structure represented 92% and 41% of the number of individuals and the number of species of the tree community, respectively.From 1997 to 2007, the species of the overstory and midstory dominated the gradient, followed by the species of understory and emergent layer (Tab.3).There was no change in the proportional distribution of species among the strata over the course of the study period.
In general, only the species of the emergent layer showed mortality rates below the mean of the forest gradient and lower than those of the remaining strata (Fig. 2).However, the difference was significant only for the 2002-2007 period (Kruskal-Wallis, H 3,67 = 8.66; p = 0.03), the emergent layer showing lower mortality than did the understory (median test, z = 2.87; p = 0.02).Only the understory showed mean annual recruitment rates above those found for the forest community in both measurement periods (Fig. 2).A significant difference among the strata was observed only for recruitment rates in the 2002-2007 period (Kruskal-Wallis, H 3,67 = 10.70;p = 0.01), the understory showing higher recruitment than did the overstory (median test, z = 2.81; p = 0.03).
Only the overstory showed a significant increase in mortality rates from the 1997-2002 period to the 2002-2007 period (Wilcoxon, Z = 3.42; p = 0.001).The overstory was also the only stratum to show a significant decrease in recruitment rates along these same periods (Wilcoxon, Z = 2.25; p = 0.002).1997-2002 2002-2007 1997-2002 2002-2007 1997-2002 2002-2007 1997-2002 2002-2007 M (%.yearThe mean ADIs of the forest gradient in the periods of 1997-2002 and 2002-2007 were 0.15 and 0.13 cm.year −1 , respectively.The variation between the two measurement periods, in terms of the ADIs calculated (Tab.4), was high: 131% and 150% in the in the 1997-2002 and 2002-2007 periods, respectively.The median ADIs were much lower than were the means, showing a higher uniformity among the vegetation formations, except for the cerradão in the

Discussion
Over the ten-year period of monitoring in the gradient studied, we observed a reduction in tree density and an increase in basal area for the semideciduous forest and gallery forest.Similar patterns have been documented in other studies of gallery forests and semideciduous seasonal forests, as well as in those of areas of tropical forest, Atlantic Forest and deciduous forest (Tab.6).According to Finegan (1996) and Guariguata & Ostertag (2001), over the course of the successional process, density tends to decrease in parallel with the increasing age of the forest, a pattern cited as one of the main characteristics of forest succession.The process of increasing basal area and decreasing tree density, also known as self-thinning (Oliveira-Filho et al. 1997), consists of the growth and accumulation of biomass by a few trees that survive intraspecific and interspecific competition, coupled with the mortality of various individuals, most with small diameters (Machado & Oliveira-Filho 2010).Thus, the forest gradient tends to change from a community with high density of individuals to a community characterized by low tree density and increase in biomass by the individuals with larger size.This tendency, as observed here, is explained by the fact that the forest gradient is inserted in a protected area, with no occurrence of disturbances in the last 30 years.
The concomitant increase in basal area and density observed in the cerradão has also been reported for other areas of cerrado sensu stricto in Brazil (Henriques & Hay 2002;Libano & Felfili 2006;Roitman et al. 2008).The structure and composition of cerrado areas protected from fire have changed significantly over the years (Moreira 2000;Durigan & Ratter 2006;Cardoso et al. 2009), as, for example, the increases in density, basal area and number of species (19%, 15% and 35%, respectively) reported for a cerrado in southeastern Brazil that was protected from fire (Roitman et al.

2008
).The pattern of increase in both density and basal area recorded in the cerradão studied here seems to be related to a decrease in the incidence of fire and of anthropogenic disturbances at the Panga Ecological Station.The Shannon diversity indices found for all of the vegetation formations studied here (H' > 3.5) are within the range reported for other semideciduous forests in Brazil (Silva et al. 2004;Vale et al. 2009;Dias Neto et al. 2009;Prado-Júnior et al. 2010), as well as for other areas of cerradão (Pereira- Silva et al. 2004;Fina & Monteiro 2009).The final value calculated for the forest gradient as a whole (H' > 4.10) is close to that observed for tropical forests (Knight 1975) and reflects the environmental heterogeneity of the forest gradient, which is related to the edaphic gradient (Moreno & Schiavini 2001) and to the topographic gradient (Cardoso & Schiavini 2002).Between the two measurement periods, there were minor changes in the floristic composition, related to the exit of rare species, often restricted to the cerradão, which forms the border with an area of cerrado sensu stricto.
In general, the entry and exit of species in forest formations is observed for species with low abundance (Kellman et al. 1998;Nascimento et al. 1999;Pinto & Hay 2005).
In general, the rates of mortality and recruitment increased between the two successive measurement periods evaluated in the present study.For the semideciduous seasonal forest and the gallery forest, the imbalance in favor of mortality, the decrease in density and the increase in basal area suggest a process of self-thinning in the tree community.In the cerradão, however, the positive balance in favor of recruitment and the increases in density and basal area indicate that this vegetation formation is in the construction phase, which is favored by the decrease in the occurrence of fire and other anthropogenic disturbances (Cardoso et al. 2009), resulting in the expansion of the cerradão into areas of cerrado sensu stricto and in some areas of cerradão being supplanted by semideciduous forest.
There were no significant differences among the strata in terms of the mean annual rates of recruitment and mortality.We can highlight only the emergent layer, which had a lower mortality rate than did the understory, and the understory, which had a higher recruitment rate than did the overstory.The tendency for the mortality rate to be lower in the emergent layer than in the understory has also been found in studies of tropical forests.Manokaran & Kochummen (1987) reported differences in the mortality rates among groups of species in forests of dipterocarps in the sequence emergent layer < overstory < understory < shade-intolerant species.In tropical forests, the mean annual mortality rate has also been shown to decrease in parallel with increasing tree height, understory species showing the highest rate, followed by those of the midstory and overstory (Lang & Knight 1983;Korning & Balslev 1994b).According to Turner (2001), understory species show high mortality due to the risk of large trees falling and to competition; therefore, these species should recruit more individuals than do overstory species, which grow faster and are more efficient in escaping the limiting conditions of the understory.
The ADIs recorded for the forest gradient in the two measurement periods were comparable to those obtained for some tropical wet forests in the Amazon (Uhl et al. 1988;Laurance et al. 1998) and lower than those obtained for tropical forests (Lieberman et al. 1985), gallery forests (Felfili 1995b;Braga & Rezende 2007;Oliveira & Felfili 2008) and deciduous forests (Carvalho & Felfili 2011).These values are typical of undisturbed forests, values > 0.3 cm.year −1 being found in disturbed environments, which are subjected to timber exploitation (Silva et al. 1995) or display an edge effect with a predominance of pioneer and heliophytic species (Pulz 1998).
The forest gradient studied here, especially the cerradão, showed high rates of mortality and recruitment, confirming the pattern found in the forest formations subjected to climatic seasonality and having had no recent anthropogenic disturbances, because they are located in a protected area.These results suggest that the forest gradient studied is in the self-thinning phase, especially the semideciduous seasonal forest and the gallery forest, which showed a decrease in density and an increase in basal area.The cerradão, however, showed higher ADIs than did the other vegetation formations studied, probably because of the higher recruitment, which, in addition to the increase in basal area and tree density, suggests that this vegetation formation is in the construction phase.Panga Ecological Station over ten years (1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007) Acta bot.bras.28(2): 234-248.2014.
Appendix 1. List of the tree species within the forest gradient studied at the Panga Ecological Station in Brazil.The species are shown in alphabetical order by botanical family (53 families comprising 165 species), followed by their respective occurrences in the vegetation formations during the interval from 1997 to 2007.
1997-2002 period (Tab.4).However, in the 2002-2007 period, the difference in the ADIs was altered in the sequence cerradão = gallery forest > semideciduous forest.The results of the Kruskal-Wallis test (H) and of the median test (z) for multiple comparisons among the vegetation formations are shown in Tab. 5.

Figure 2 .
Figure 2. Mean annual rates of mortality and recruitment of the tree species in the vertical strata of the forest gradient at the Panga Ecological Station, in the city of Uberlândia, state of Minas Gerais, Brazil, at the periods of 1997-2002 and 2002-2007.Solid and dashed lines represent the annual rates of the forest gradient in the 1997-2002 and 2002-2007 periods, respectively.

Table 1 .
Characteristics of the forest gradient and its vegetation formations in 1997, 2002 and 2007 at the Panga Ecological Station, in the city of Uberlândia, state of Minas Gerais, Brazil.

Table 2 .
Parameters of the dynamics in the forest gradient and its vegetation formations over a period of ten years at the Panga Ecological Station, in the city of Uberlândia, state of Minas Gerais, Brazil.

Table 3 .
Changes in the distribution of species and individuals with higher density (n > 10) over a period of ten years (1997 to 2007), for each stratum of the forest gradient at the Panga Ecological Station, in the city of Uberlândia, state of Minas Gerais, Brazil.Numbers between parentheses are percentages in the strata.

Table 6 .
Structural and demographic parameters of successive monitoring studies in permanent plots in tropical forest formations.

Table 4 .
Mean and median of the annual diameter increment of the forest gradient and its vegetation formations at the Panga Ecological Station, in the city of Uberlândia, state of Minas Gerais, Brazil, for the periods of1997-2002 and 2002-2007.

Table 5 .
Results of the median test (z) for multiple comparisons of the annual diameter increment among the vegetation formations for the periods of 1997-2002 (H 3,6656 = 41.38;p = 0.001) and 2002-2007 (H 3,6218 = 37.31; p = 0.001) at the Panga Ecological Station, in the city of Uberlândia, state of Minas Gerais, Brazil.Values with asterisk show a significant difference (p < 0.05).