Changes in the structure of a savanna forest over a six-year period in the Amazon-Cerrado transition, Mato Grosso state, Brazil

Universidade de Brasília UnB, Programa de Pós-graduação em Botânica, C.P. 4457, 70.904-970, Brasília, DF, Brasil. 2 Universidade do Estado de Mato Grosso UNEMAT, Programa de Pós-graduação em Ecologia e Conservação, C. P. 08, 78.690-000, Nova Xavantina, MT, Brasil. 3 Universidade de Brasília, Programa de Pós-graduação em Ciências Florestais, C. P. 4357, 70.910-900, Brasília, DF, Brasil. 4 Autor para correspondência: biamarimon@hotmail.com. Abstract Vegetation changes in transition zones are still poorly studied. Changes in the vegetation structure of a savanna forest (cerradão) were assessed in the Amazon-Cerrado transition (14o42’2.3”S; 52o21’2.6”W), eastern Mato Grosso, within a period of six years (2002, 2005 and 2008). In 2002, fifty plots of 10 × 10 m were set up, where all trees with DSH 30 ≥ 5 cm were measured; in 2005 and 2008 the plots were re-inventoried. In 2008, 84 species from 70 genera and 37 families were sampled; absolute density was 1,998 individuals/ha and basal area was 25.95 m.ha. On the one hand, the absolute density of live individuals decreased from 2005 to 2008 (2,066 individuals/ha); on the other hand, the basal area increased in 2008 compared to 2005 (23.56 m.ha) and 2002 (1,884 individuals/ha and 21.38 m.ha). The species with the highest importance value in the period were Hirtella glandulosa, Tachigali vulgaris and Xylopia aromatica. Except for these three species, all other species underwent hierarchic changes in the importance value, indicating that most species frequently alternate. Community structure exhibited changes throughout the period; hence, we suggest investigations on the role of T. vulgaris in these changes, since environmental conditions caused by gap opening from the fall of senile individuals of this pioneer species with a short life cycle may contribute to community dynamics.


Introduction
The Cerrado is seen as one of the 34 biodiversity hotspots of the world (Mittermeier et al. 2005); it is the largest Neotropical savanna in the world and has the second largest area of all biomes in South America (Oliveira & Marquis 2002).Among several threatened physiognomies of this biome, the savanna forest ("cerradão") stands out.It is denominated 'mesophyllous sclerophyllous forest' (Rizzini 1979) and is characterized by the presence of species from savanna and forest environments.In general, this physiognomy is not tolerant of anthropic disturbance and occupies small areas; exactly the areas that are most frequently used for agriculture and livestock, since these areas are usually on soils with higher availability of exchangeable cations (Ratter 1971;Eiten 1979;Oliveira-Filho et al. 1994), i.e. more humid and with clayey texture (Marimon-Junior & Haridasan 2005).
In the eastern portion of Mato Grosso state there are patches of savanna forest in the transition zone between Cerrado and the Amazon biomes (Marimon-Junior & Haridasan 2005;Marimon et al. 2006).Two types of vegetation were recorded in the area: the savanna forest of Magonia pubescens A. St.-Hil.and Callisthene fasciculata Mart., which occurs on mesotrophic soils; and the savanna forest of Hirtella glandulosa, which occurs on dystrophic soils (Ratter 1971;Ratter et al. 1973).These two savanna forest types are threatened, mainly because they are located in a region known as the 'deforestation arch', where the advance of cultivated areas represents an important threat to native vegetation (Nogueira et al. 2008).
Most studies carried out in savanna forests are based on information collected at a point in time (Costa & Araújo 2001;Marimon & Lima 2001;Gomes et al. 2004;Marimon-Junior & Haridasan 2005;Guilherme & Nakajima 2007;Kunz et al. 2009).There is still a huge need for studies on the dynamics of this vegetation.Therefore, studies on long-term vegetation changes are essential to understand the mechanisms and processes that maintain the community in a steady state (Aquino et al. 2007).A lot of information obtained from nativeforest functioning can be used for its management.
Hence, important subsidies to practices of conservation, management and restoration of degraded areas, for example, may come from studies on the remnants of native vegetation in Mato Grosso state.Therefore, the objective of this study was to assess changes in the structure of the woody vegetation of a savanna forest in eastern Mato Grosso, in the transition region between Cerrado and the Amazon, within a period of six years (2002 and 2008).The transition between Cerrado and the Amazon extends for over 4,500 km (Ackerly et al. 1989) and it is dynamic: studies show that forests are advancing over savannas (Marimon et al. 2006).In this context, the present study will also loot at whether floristic and structural changes in the savanna forest led this physiognomy to become a denser forest.

Material and Methods
The study was carried out in a savanna forest (14º42'2.3"S;52º21'2.6"W),Bacaba Municipal Park, Nova Xavantina, state of Mato Grosso, centralwestern Brazil.According to Marimon-Junior & Haridasan (2005), the park is located in a transition region between the Cerrado and Amazon biomes, where the predominant 'cerrado sensu stricto' vegetation (open savanna) is in contact with forests and savanna forests, in acid and dystrophic soils, with high levels of exchangeable aluminum and clayey texture.According to Köppen's classification, the regional climate is type Aw (Silva et al. 2008), characterized by two well-defined seasons: one dry and cold (April to September) and the other hot and rainy (October to March).
The transition between Cerrado and the Amazon is a zone of ecological tension that exhibits a mosaic of savannas and forests (Ratter et al. 1973;Ackerly et al. 1989;Ivanauskas et al. 2004;Marimon et al. 2006). Based on IBGE (2004), our study area is about 150 km from this zone of ecological tension.However, this zone is not regular; there are larger or smaller intrusions or fringes (Ratter et al. 1973;Marimon et al. 2006), which are currently fragmented due to the conversion of native vegetation into agricultural lands (Nogueira et al. 2008).Although our study area is located in a vegetation matrix dominated by savanna (Marimon-Junior & Haridasan 2005), in nearby areas (less than 10 km) there are fragments and intrusions of contact between savanna and seasonal forest (IBGE 2004;Marimon 2005).Records taken out in 1943 by members of the Roncador-Xingu Expedition confirmed that up to 40 km to the south of our study area there would have existed 'a dense vegetation, where to open the way they needed to cut down colossal trees' (Carpentieri 2008); this report characterizes the study region as a zone of ecological tension or a zone of transition between Cerrado and the Amazon.
Changes in the structure of a savanna forest.
In 2002, fifty 10 × 10 m permanent plots were set up (Marimon-Junior & Haridasan 2005), where woody species (except lianas) that had DSH 30 (diameter at soil height, measured at 30cm) e ≥ 5cm were sampled.At the occasion, all individuals were tagged with numbered aluminum plates, and were recorded and identified.Species were identified by comparison with herbaria vouchers (NX and UB) and by consulting specialists.The collected material was deposited in the NX Herbarium, UNEMAT -Nova Xavantina Campus, Mato Grosso state.In 2005 and in 2008, all surviving individuals were measured again and recruits (individuals that reached the minimal inclusion criterion) were tagged, measured, recorded and identified.Sampling and species identification followed the same procedures used in the first inventory.The classification system used for families was APG III ( 2009) and the revision of taxa names followed Forzza et al. (2010) in the list of Angiosperm species of the Brazilian flora.
Based on the number of individuals sampled, mortality and recruitment rates were calculated for each plot (Sheil et al. 1995(Sheil et al. , 2000) ) and comparisons were made (2002-2005 and 2005-2008) using paired t-tests.The average number of individuals and the basal area in each year sampled were compared by analysis of variance and Tukey's test at 5% probability (Zar 1999).

Results and Discussion
In 2008, the savanna forest studied had 84 plant species from 70 genera and 37 families (Tab.1), with an absolute density of 1,998 individuals/ha and a basal area of 25.95 m 2 .ha - .On the one hand, the absolute density of live individuals decreased compared to 2005 (2,066 ind.ha -1 ); on the other hand, the basal area increased compared to 2002 and 2005 (Tab.2).In 2002, 77 species from 65 genera and 36 families were recorded (Marimon-Junior & Haridasan 2005) and in 2005, 87 species from 71 genera and 38 families.The increase in the basal area in two consecutive periods (2005 and 2008) in the savanna forest is consistent with Phillips et al. (2002), Baker et al. (2004) and Lewis et al. (2009), who observed that in the last century nearly all terrestrial ecosystems have been under the influence of atmospheric and climatic changes.An increase in dynamics, biomass and carbon stock in tropical forests was recorded, probably due to an increase of CO 2 levels in the atmosphere.Further detailed long-term studies in the savanna forest studied here are essential to verify if the increase in biomass of this community is related to the increase in atmospheric CO 2 levels.However, the increase in biomass could also be explained by climatic changes that have been occurring since the early Holocene, when a drier climate was replaced by a warmer and more humid climate (Ledru et al. 1996).In a study carried out by Marimon et al. (2006) in a nearby area, 30-year records showed that the Amazon advanced 7 km into the Cerrado, reinforcing the expansion of forests over savannas in the region.
Recruitment and mortality rates were higher in the first inventory period, between 2002 and 2005 (Tab.2).If we consider intervals as well as the whole period (2002 to 2008), the values of the savanna forest of Bacaba Park were higher than the values observed in other studies carried out in forests of South and Central America, which varied from 0.5 to 2.8%.year - for mortality (Lieberman et al. 1985;Swaine et al. 1987;Condit et al. 1995;Felfili 1995) and from 2 to 4%.year -1 for recruitment (Oliveira-Filho et al. 1997;Higuchi et al. 2008;Silva & Araújo 2009;Miguel et al. 2011).According to Felfili (1995), mortality rates around 3.5%.year - are typical of areas that underwent disturbances.Oliveira & Felfili (2008) observed that high mortality and recruitment rates lead to a high turnover, confirming the dynamic aspect of the community, which even without undertaking direct disturbances (fire and cutting) exhibited high mortality and recruitment.Considering the whole period (2002)(2003)(2004)(2005)(2006)(2007)(2008), recruitment compensated mortality (t= -2.95, P= 0.0024).This compensation can be related to a 'construction' phase of the sylvigenetic cycle of the community, as proposed by Hallé et al. (1978), which is usually recorded in forests recovering from a disturbance (Oliveira-Filho et al. 1997;Chagas et al. 2001), suggesting that periods of higher mortality might have previously occurred (Felfili 1995), of which there is no record from the memory of local residents.
The reduction in density and the increase in basal area recorded in the present study (Tab.2) are consistent with a self-thinning pattern, as observed by Felfili (1995) and Werneck et al. (2000).Changes in the structure of a savanna forest.
In a riparian forest in Distrito Federal, Oliveira & Felfili (2005) recorded intense reduction in total density and increase in basal area; they suggested that greater shading of the area would hinder the growth of heliophilous species.This could also be happening in our study area, since all species that entered the community between 2002 and 2008, such as Diospyros sericea, Ficus enormis and Pouteria gardneri, are typical of forests and the ones that left the community, such as Dimorphandra mollis, Cordiera elliptica and Simarouba versicolor, are typical of savanna formations and open environments (Tab. 1, Marimon-Junior & Haridasan 2005).Besides, of the eight species that exhibited higher recruitment (> 10 individuals) in the period from 2002 to 2008 (Tab.3), at least six are typical of forests; and all species that exhibited higher mortality (> 10 individuals, Tab. 3) are typical of savanna and field vegetation (Ratter et al. 1973;Pott & Pott 1994;Oliveira-Filho & Ratter 1995;IBGE 2002;Durigan et al. 2004;Mendonça et al. 2008).
The ten most important species in 2008 represented c. 54% of the total importance value (IV) and of the total number of individuals sampled.In 2002 they represented 53% of IV and 51.7% of the total number of individuals sampled, and in 2005, 52% of IV and 51.5% of the total number of individuals (Table 1; Marimon-Junior & Haridasan 2005).In a riparian forest in Distrito Federal, Felfili (1993) observed that the ten most important species might be considered to be the ones that exhibit higher success exploiting resources of the habitat.Inventories carried out in forests and savannas of the Cerrado biome (Costa & Araújo 2001;Marimon et al. 2006;Kunz et al. 2009) reported that the species that have higher importance value also have higher number of individuals, as recorded in the present study.
The most important species (IV) in all three sampling periods was Hirtella glandulosa, which contributed with approximately 12% of the total number of individuals in 2002, 2005 and 2008, confirming the area as a savanna forest of Hirtella glandulosa, as described by Ratter (1971) and Ratter et al. (1973).Tachigali vulgaris (=Sclerolobium paniculatum) was the second most important species in all inventories, with 6.5% of the total number of individuals in 2002 (Marimon-Junior & Haridasan 2005) The species mentioned were also among the ten most important species in other savanna forest areas (Gomes et al. 2004;Pereira-Silva et al. 2004;Marimon et al. 2006;Guilherme & Nakajima 2007;Kunz et al. 2009) and in a dense savanna (Andrade et al. 2002), evidencing their broad distribution and high importance in different forests of the Cerrado biome.
Tachigali vulgaris has a short life cycle (< 20 years) and rapid growth (Felfili et al. 1999).In this case, it is suggested that the mortality of this species and the resulting fall of large-sized senile individuals, such as recorded by Franczak (2009) (Felfili 1993;Miguel & Marimon 2008), the temporal changes recorded in the savanna forest studied may have a biotic origin, led by Tachigali vulgaris, which might be a keystone species in the dynamics of this savanna forest.Besides, considering that this species maintained itself in the same hierarchical position during the study period, possibly its adult and senile individuals, after falling, opened new gaps, maintaining the possibility of regeneration and growth of the species in a type of positive feedback or virtual circle (Miguel et al. 2011).
Based on this assumption, the contribution of Tachigali vulgaris may be important to several ecosystem processes that affect community structure, as for instance the microclimate, since microclimatic factors such as light, humidity and soil and air temperature depend on canopy characteristics, especially regarding the dynamics Franczak, D.D. et al.

Table 1 -
Phytosociological parameters of species sampled in a cerradão in the Cerrado-Amazon Forest transition, in 2005 and 2008, Nova Xavantina-MT.Nº Herb.= registration number in Herbarium NX, N= number of individuals, DR= relative density (%), FR= relative frequency (%), DoR= relative dominance (%), and VI= importance value.Species listed in order of decreasing VI.

Table 2 -
Parameters of dynamics in a woody community of a cerradão in BacabaPark, Nova Xavantina, Mato  Grosso, between 2002 and 2005, 2005 and 2008, and 2002 and 2008.Where: t = paired t-test, F = result of analysis of variance.Values in parentheses correspond to standard deviation.Different letters indicate differences at 5% significance level.forests subjected to abiotic environmental changes, such as the case of riparian forests submitted to the seasonal flood of rivers and to an intense edge effect

Table 3 -
Number of dead and recruited individuals in the intervals between the years 2002 and 2005 (02-05), 2005 and 2008 (05-08), and 2002 and 2008 (02-08).Cerradão of the Bacaba Park, Nova Xavantina, MT.Species listed in descending order of number of individuals recruited between 2002 and 2008.Were considered only those species that presented at least five dead or recruited individuals at least in one interval.