Clber Rodrigo de Souza · Jean Daniel Morel · Alisson Borges Miranda Santos ·Wilder Bento da Silva · Vincius Andrade Maia · Polyanne Aparecida Coelho ·Vanessa Leite Rezende · Rubens Manoel dos Santos

Abstract Our work aimed to test the hypothesis that soil microscale heterogeneity act as a community ecological driver, increasing diversity and promoting structural shifts on the Seasonally Dry Tropical Forest (SDTF) tree community. We evaluated the relationship between microscale edaphic variations and floristic—structural patterns of tree communities in a SDTF fragment located in the southern end of the Brazilian Caatinga domain. Vegetation and soil data were obtained through 27 sample units of 400 m2(20 m × 20 m), within each one we measured and identified at species level all arboreal individuals with Circumference at the Breast Height greater or equal to 10 cm,and also collected the soil samples. Through the data we evaluated soil variation influence on the tree community structural and floristic patterns trough generalized linear models. Soil explained the small-scale structural and floristic variations, contributing significantly to biomass,sprouting and to floristic relationships between sample units. It was also observed a possible relation of the result with the Caatinga domain biogeographic history, due the presence of Sedimentary Caatinga species, which are not expected for the study region. Soil plays an important role in driving small-scale complexity and diversity of SDTF,but we also suggest that Caatinga biogeographic events have influence on the high heterogeneity patterns.

Keywords Caatinga domain · Edaphic variables ·Sprouting · Sedimentary Caatinga

Introduction

The relevance of soil in explaining ecological patterns in tropical forests has been increasingly recognized in its range from broad to small scales, with important implications for the biogeography domains and for biomes characterization (Oliveira-Filho and Ratter 1995; Queiroz 2006; Arruda et al. 2017; Neves et al. 2017; Bueno et al.2018), and also for abundance and growth patterns at regional and community level (Siefert et al. 2012; Vleminckx et al. 2015; Krishnadas et al. 2016). At small-scales,the interaction between vegetation and soil are related to a broad spectrum of essential ecological processes in which conditions and resources may influence the communities attributes (Pausas and Austin 2001; Meinders and Van Breemen 2005; Putten et al. 2016). Through these interactions,biogeochemical and hydrological cycles associated with factors such as water availability,pH,growth essential nutrients (N, K, C, P, Mg, Ca, S) and possibly toxic elements (Al, Pb, Mn, among others) come into contact with plant populations(Pausas and Austin 2001;Chapin III et al.2009; Wang et al. 2017). The local occurrence and variability of these edaphic may produce some ecological filters and may determine how the resources can become available for plant survival and growth (Hutchings et al.2003; Hart and Marshall 2013; Cadotte and Tucker 2017).Through these influences, the soil may play a role in the species selection from the regional pool as well as in their patterns of establishment and growth, in addition to influence the successional process and functional/phylogenetic diversities through the selection of specific traits and lineages (Phillips et al. 2003; Bagousse-Pinguet et al. 2017;Pinho et al. 2017).

Soil effects at small-scales vegetation patterns was little studied in Seasonally Dry Tropical Forests (SDTF) (Pennington et al. 2000), although has been well known for tropical rainforests (Sollins 1998; Bohlman et al. 2008;Vleminckx et al. 2015). Soil is one of the main key attributes for this vegetation type jointly with climatic seasonality and deciduousness (Prado and Gibbs 1993;Pennington et al. 2000), and it is also related to the Seasonal Deciduous Forest patches associated to SDTF located inside other phytogeographic domains (Prado and Gibbs 1993; Oliveira-Filho and Ratter 1995; DRYFLOR 2016).Soil is also important in phytogeographic patterns inside the several SDTF nuclei, as in the Brazilian Caatinga domain which has its main subdivision attributed to the differentiation between crystalline and sedimentary soils supposedly occurred at the Quaternary Period by a pediplanation event (Queiroz 2006; Santos et al. 2012;Moro et al. 2016).

Edaphic attributes have been pointed out as relevant in SDTF communities patterns due to its association with water availability, which is the main factor in this forest type (Markesteijn et al. 2010; Pen~a-claros et al. 2012;Apgaua et al. 2014, 2015; Castellanos-Castro and Newton 2015). The soil is also important for the successional process due to its influence in species establishment and growth by selecting resistant traits to environmental conditions such as sprouting, which is the ability to emit new stems from vegetative shoots in response to stressful conditions such as water shortage, flooding or fire occurrence(Ceccon et al.2006;Becknell and Powers 2014;Pausas and Keeley 2014; Castellanos-Castro and Newton 2015). By influencing ecological processes through interaction with conditions and resources at various scales (climate,topography), the soil attributes have direct consequences on communities structure and floristic composition, thus being an important factor for tropical forests diversity patterns(Chesson 2000;Vleminckx et al.2015;Krishnadas et al. 2016). However, questions that have been already discussed for other vegetation types, such as the relationship between soil heterogeneity and communities attributes such as dominance, biomass, taxonomic, functional and phylogenetic diversity, remain unanswered for SDTF(Huston 1980; Harms et al. 2001; John et al. 2007; Sander and Wardell-Johnson 2011). Diversity and soil fertility have been found to show negative, positive and no relationships, varying with the vegetation type and location(Clinebell et al. 1995; Tuomisto et al. 2002; Poulsen et al.2006;Pea-Claros et al. 2012).In SDTF these patterns are less studied than in Rainforests, where few studies prospected soil vegetation relationships at this biome (Apgaua et al. 2014; Reis et al. 2017).

Besides the lack of basic knowledge about SDTF, the need for studies in these forests has been motivated by its high degradation level and low contemplation of these forest in conservation units, as well as by the increased recognition of their importance in environmental services maintenance (Santos et al. 2011; Sunderland et al. 2015;DRYFLOR 2016). The potential SDTF importance in climate change scenarios has been pointed out as an additional motivation, considering the global predictions of increased aridity and seasonality (IPCC 2014), its higher vulnerability to these changes (Seiler et al. 2015) and the soil importance in ecological processes and consequently in the forest responses (Santos et al. 2014; Allen et al.2017). The knowledge construction concerning the relationship between edaphic heterogeneity and vegetation is also important for ecosystems ecological restoration projects that may have their effectiveness increased with the knowledge of suitable conditions for certain species and successional stages(Ceccon et al.2006;Castellanos-Castro and Newton 2015; Pinho et al. 2017). Thus, we aimed to evaluate the relationship between tree communities floristic—structural patterns and small-scale edaphic variations in Seasonally Dry Tropical Forests, evaluating whether the soil variation can explain the community patterns. Our hypothesis is that the small-scale edaphic heterogeneity plays an important role and acts as a driver of Seasonally Dry Tropical Forests tree communities, thus being an important factor to explain its ecological patterns.

Methods

The studied area is a forest fragment of 70 ha located in Juvenlia, North of Minas Gerais state, Brazil(44°4′28.31′W;14°27′51.10′S)(Fig. 1).The site is located in a region of Koppen Aw/As climate(tropical dry winter)that presents average annual rainfall of 868 mm and average monthly temperature of 23 °C. The relief is flat to gently undulating and the average altitude is 445 m. The fragment is a Caatinga Decidual Seasonal Forest that it located in an ecotonal region between Caatinga and Cerrado domain, also near to Atlantic Domain enclaves(Santos et al. 2012).

Fig. 1 Location of the study region in South America and Brazil, as well as its proximity to the Brazilian phytogeographic domains

We allocated 27 sample units of 400 m2(20 m × 20 m)in the fragment that were arranged non-contiguously in an aleatory way. We measured and identified at the species level all the tree individuals with Circumference at Breast Height (CBH) ≥10 cm within each sample unit. We included individuals with multiple stems in the sampling when the root of their circumferences squares sum reached the individual inclusion criterion (Scolforo and Melo 1997). We have adopted the APG IV (Angiosperm Phylogeny Group-APG 2016) for families and the NeoTrop-Tree database (Oliveira-Filho 2017) for species and botanical synonyms correction. We collected the soil data by the collection of superficial samples of 0.5 L to 10 cm of depth in three points of each sample unit to form composite samples. Each composite sample was stored in a plastic bag and later sent to the Federal University of Lavras Laboratory of Soil Analysis. The following variables were analyzed: pH in water; (K), Phosphorus (P),Sodium (Na), Calcium (Ca), Magnesium (Mg), Aluminum(Al) and Organic Matter (OM); Potential Acidity(H + Al); Sum of Bases (SB); effective Cation Exchange Capacity (t); Cation Exchange Capacity at pH 7.0 (T);Bases Saturation (V); saturation by aluminum (m); and proportions of Sand, Silt and Clay.

With the vegetation data, we obtained the variables of(1) density (number of individuals/ha); (2) sprouting(number of stems/number of individuals); (3) aboveground-biomass(AGB)(t ha-1);(4)the number of species(richness)per sample unit.The AGB was obtained through the Chave et al. (2014) pantropical equation, using the biomass package(Rejou-Mechain et al.2017),based on the individual quadratic CBH(root of the sum of the squares of the individual stems), the species wood density or the genus or family means, and also based on the E constraint measure calculated on the site spatial coordinates, that works by reducing tree height in stressful climates. This equation has been developed and has been widely applied in rain Forests, with expectations of uncertainties when applied to other vegetation types.However,its use has still been pointed out as potentially informative by adding a lot of information to the classical Basal Area approach,which considers all species at all sites as equivalent in their contribution to plot biomass. The soil variables were synthesized in an explanatory summary variable represented by the value of each sample unit on the axis 1 of a Principal Component Analysis (PCA) (Jolliffe and Cadima 2016)performed with all the soil variables collected. We have performed this synthesis due to the high number of edaphic variables, which can impair the adjustments quality of by reducing the degrees of freedom. We have used only the axis 1 PCA values due to its high explanatory power of 71.28%,compared to axis 2 that has explained only 9.91%of the existing variation.

We have analyzed if the soil can explain the vegetation patterns using generalized linear models (GLM), using gaussian (Sprouting, AGB), poisson (Richness) and negative binomial (number of individuals), according to the criteria of normality,homoscedasticity for gaussian models and Pearson chi square goodness of fit test for non-Gaussian models.All the selected models were absent of spatial autocorrelation.Soil influence on each one of the structural response variables was analyzed in two ways.The first one was by using the PCA axis scores as an edaphic variation synthesis, thus assuming an approach of continuous variation. The second one used a categorical approach, based on the results of abrupt separation between sample units according to the soil characteristics in PCA, considering that the units apparently are separate in isolated groups and not in the form of continuous gradients. So, we separated the sample units into two categories related to their soil texture and fertility attributes, with 9 units being classified as areas of high sand and low fertility (Sandy-Less Fertile area) and 18 sample units as areas of low amount of sand and high fertility (Less Sandy-High Fertile area). This classification was used as explanatory variable in GLMs with the structural variables evaluated. Based on the model’s quality of each alternative, evaluated from the AICc (Akaike Information Criterion of Second Order), we analyze if the structural variations are better explained by a continuous approach (Axis 1 of the general PCA) or grouping (categories).

We also calculated floristic dissimilarity between sample units using Jaccard as a measure of distance and later have partitioned the values obtained in the components of betadiversity turnover and nestedness (Baselga 2010) by the beta.pair function of the betapart package (Baselga et al. 2013) to reduce distortions due to differences in richness.We then analyzed which of the two components is more representative of the floristic relationships between sample units through GLM using the distance measure as a response variable and the component type as an explanatory variable. The most explanatory component was used as a variable response representative of the floristic community to assess whether soil (variable synthesis) significantly explain the patterns found. For this, we used the Permutational Multivariate Analysis of Variance Using Distance Matrices(PERMANOVA)by the adonis function of the vegan package(Oksanen et al.2017),thus using the main component value as the response variable and the continuous variables (PCA Axis 1) and categorical(groups) as explanatory variables in isolated analyzes.Then we compared the R2of each alternative, in order to identify whether the possible variations are formed by edaphic gradients of continuous variation or by groupings.All the analysis were performed in the R program v. 3.3.1(2016).

Results

Sample units have differed in relation to soil characteristics along axis 1 of the principal component analysis (Fig. 2)due to soil texture and fertility variations. In the diagram left portion are located sample units associated to higher values of sand and of attributes associated with the aluminum presence (Sandy-Less Fertile area), as opposed to the sample units located to the diagram right side, which have presented more basic pH, less sandy texture and higher values of attributes related to fertility, such as Sum of Bases and Bases Saturation (Less Sandy-High fertile area). Thus, higher values in PCA axis 1 for sample units are associated with higher fertility and lower amounts of aluminum and sand (Supplementary Table S1). Although there are differences in the sand amount between the sample units, all of them present higher sand amounts in comparison to the clay and silt fractions, what classifies them as markedly sandy soils.

Fig.2 Principal Components Analysis(PCA)for the soil attributes of the sampled units of the Seasonally Dry Tropical Forest in the north of Minas Gerais,Brazil.Proportion of variation explained by the axes 1 and 2 are 71.28% and 9.91%, respectively. Note: pH: pH in water;P: Phosphorus; Na: Sodium; Al: Aluminum; OM: Organic Matter;H + Al: Potential Acidity; SB: Sum of Bases; t: Effective Cation Exchange Capacity;T:Cation Exchange Capacity at pH 7.0;V:Bases Saturation(V);m:Saturation by aluminum.Due to visual aspects,the variables K,Ca and Mg were not represented and are included in the Sum of Bases variable

In general, we found 1395 individuals of 82 species, 65 genera and 22 botanical families (Supplementary Table S2).Individuals total density is 1291.6 ind. ha-1,the sprouting is 2125.92 stems/ha and the total above ground biomass is 113.9 t ha-1. The general diversities of Shannon—Wiener (H′) and the Pielou’s Equability (J′) for the fragment are 3.46 nat ind-1and 0.76, respectively.

The floristic and structural patterns can be significantly explained by soil characteristics, with sample units with different edaphic characteristics tending to present distinct structure and species composition. For all the analyzed variables the approach using categorical variables presented lower AICc (structural variables) and higher R2(floristic) (Supplementary Table S3) in comparison to the approach with continuous values. Thus, the soil variations observed occur by two isolated groups that present distinct edaphic characteristics that influence the vegetation. The small-scale edaphic variations between groups significantly explain the variation in sprouting (Fig. 3) and biomass(Fig. 4), and do not explain the variation in number of individuals (Fig. 5) and species richness (Fig. 6). The sprouting was significantly higher in the Sandy-Less Fertile area (p = 0.01), while the biomass was higher in the Less Sandy-High Fertile area (p = 0.01). Thus, the sprouting tends to be higher in sites of lower fertility and with more aluminum and sand (lower values of the synthesis variable), and biomass tends to be higher in sites with higher fertility and with less sand and aluminum. The floristic relationships in the fragment were marked by high dissimilarity values (Fig. 7), represented mainly by the turnover component, significantly higher than the nestedness(estimate = 0.73; p ＜ 2e—16). These floristic relationships based on turnover are significantly explained by the edaphic attributes represented by the edaphic groups (F model = 8.8924; p = 0.001; df = 26), thus being the floristic variations associated with soil variations. In addition, most species occur in a few sample units (Table S3)associated with some particular condition in the edaphic spectrum.

Fig. 3 Density (number of individuals per ha) in relation to PCA 1 axe values (a) and between the two edaphic group considered (b) of the Seasonally Dry Tropical Forest in the north of Minas Gerais,Brazil. The estimate and the p value in the first figure corresponds to the results in the continuous analysis,while the p value in the second figure are the value obtained in the categorical approach. Note: LSHF: Less Sandy-High Fertile area; S-LF: Sandy-Less Fertile area

Fig. 4 Sprouting (number of stems/n of individuals) in relation to PCA 1 axe values(a)and between the two edaphic groups considered(b) of the Seasonally Dry Tropical Forest in the north of Minas Gerais, Brazil. The estimate and the p value in the first figure corresponds to the results in the continuous analysis, while the p value in the second figure are the value obtained in the categorical approach. Note: LS-HF: Less Sandy-High Fertile area;S-LF: Sandy-Less Fertile area

Fig. 5 Above ground biomass (Mg/ha) in relation to PCA 1 axe values (a) and between the two edaphic group considered (b) of the Seasonally Dry Tropical Forest in the north of Minas Gerais, Brazil.The estimate and the p value in the first figure corresponds to the results in the continuous analysis, while the p value in the second figure are the value obtained in the categorical approach. Note: LSHF: Less Sandy-High Fertile area; S-LF: Sandy-Less Fertile area

Fig. 6 Species richness in relation to PCA 1 axe values (a) and between the two edaphic group considered (b) of the Seasonally Dry Tropical Forest in the north of Minas Gerais,Brazil.The estimate and the p value in the first figure corresponds to the results in the continuous analysis, while the p value in the second figure are the value obtained in the categorical approach.Note:LS-HF:Less Sandy-High Fertile area; S-LF: Sandy-Less Fertile area

Fig. 7 Representation of the partitioning of Jaccard’s similarity (PlotTrix) (a) and boxplot (b) for the Jaccard’s dissimilarity partitioned into turnover and nestedness of the Seasonally Dry Tropical Forest in the north of Minas Gerais,Brazil

Discussion

Our results confirmed the hypothesis proposed, with the small-scale edaphic variations contributing significantly to explain the community ecological patterns. However, soil did not significantly influence all the attributes considered,being significant for sprouting, biomass and for floristic relationships; and not significant for density of individuals and species richness.

The result found of no significant soil influence for number of individuals and richness, but with influence on sprouting, possibly is related to the influence of restrictive factors of distinct scales on the regeneration patterns.Individual’s density probably has been conditioned by the macroecological SDTF climate characteristics that are common to the whole fragment, while the differentiation for sprouting has responded to small-scale environmental differences in habitat restrictiveness. The species richness occurs together with seed regeneration, the absence of significant soil influence on the species richness have occurred due to the absence of effects on the number of individuals. The fragment as a whole is subject to low precipitation, climatic seasonality, high temperatures and other SDTF characteristics (Pennington et al. 2009;DRYFLOR 2016), which act as constant ecological filters for seed regenerants in the development early stages, thus controlling the density and improving the sprouting adoption as an establishment strategy(Bond and Midgley 2001;Ceccon et al. 2004, 2006; Pausas et al. 2016). However,this sprouting adoption will vary due to ecological constraints associated with distinct microenvironmental conditions and soil variations(Bond and Midgley 2001;Pausas et al. 2016). These results also highlight the soil effects on density (regeneration) and species richness as being dependent on the environmental context,mainly associated with the presence of strong environmental filters such as seasonality(Ceccon et al.2006;Pausas et al.2016;Cadotte and Tucker 2017). Thus, although the soil influence on alpha diversity (richness) has been reported in several studies (Hutchings et al. 2003; Vleminckx et al. 2015;Krishnadas et al. 2016), it would be mediated by environmental factors (climate, topography) and biotic interactions, in addition to its interaction with soil elements, in which fertility effects interact with texture effects.

The adoption of regeneration by sprouting occurs when the survival of individuals from reproductive material is affected by some restrictive factors such as fire,pathogens,herbivory, wind, anthropic interventions, floods and soil toxicity (Bond and Midgley 2001, 2003; Zeppel et al.2015a, b; Pausas et al. 2016). In these situations, species with this ecological characteristic allocate resources for the emission of secondary shoots from vegetative material in order to maximize their establishment chances,often being accompanied by events of reduction of reproductive intensity (Bond and Midgley 2001; Pausas and Keeley 2014;Pausas et al.2016).The new stem has the advantages of being associated with a developed root and energy system that will make it less susceptible to environmental constraints and will enhance its permanence (Bond and Midgley 2001; Moreira and Tormo 2012; Pausas and Keeley 2014).In the studied fragment,such restrictions are associated with high acidity, high concentration of aluminum and high sand amount that can negatively interfere in the regenerating individuals development through toxicity and water deficit(Sollins 1998).These results emhaces the sprouting importance in tree communities structural characterization, in contrast to the historical neglect attributed to its relevance (Zeppel et al. 2015a, b; Pausas et al. 2016).

The result of a positive relationship between biomass,higher fertility and less sandy soil is expected, considering that plants in places with higher amounts of essential nutrients have a higher growth(Baker et al.2009;Becknell and Powers 2014).The plants are also benefited by greater soil water retention provided by the smaller sand amount(Jenny 1980; Markesteijn et al. 2010), considering the water importance in nutrients absorption and in the SDTF ecological processes (Pennington et al. 2009; Sunderland et al.2015;DRYFLOR 2016).Besides the direct influence on growth,the soil and the water availability can still act in the selection of species and functional traits that are related to the greater vegetal biomass accumulation, such as the presence of symbiotic associations (Lebrija-Trejos et al.2010; Xu et al. 2016). The relation between biomass, fertility and water availability is still pointed out in other few studies carried out with dry forests in which conclusive direct relationships between them have been presented(Becknell and Powers 2014; Becknell et al. 2012). Thus,species in places with more soil nutrients availability have greater chances of absorbing them by the possible presence of traits that may have improved this process.

The pattern of high floristic specificity found was significantly explained by edaphic variations, which through its effects on the conditions and resources has originated a microhabitats mosaic where the regional species pool has been distributed. Each point inside the edaphic variation presents a particular combination of both conditions and resources that is important in the composition differences due to offering different restrictions and possibilities to species in different stages of establishment and succession(John et al. 2007; Lai et al. 2009; Jones et al. 2011; Pinho et al. 2017). Thus, species tend to associate with microhabitats in which their success can be enhanced(Hutchings et al. 2003; Gómez-Aparicio et al. 2005; Vleminckx et al.2015). The composition partitioning along the microhabitats have reflected in the representativeness of turnover in relation to nestendness in the comparisons between sample units,as well as in the soil explanatory force in this pattern.The floristic specificity found may also have been influenced by the region where the study fragment is located,that is a central transition zone between the Caatinga and Cerrado domains that is also influenced by the Atlantic domain (Santos et al. 2012; Moro et al. 2016). This region can be considered an ecotone in macro scale that presents a diverse pool of species of different origins and ecological characteristics. In interaction to the environmental variation, this regional diversity of species may associate itself with some local micro-habitat,thus partitioning itself along the environments and improving the beta diversity.

Another additional point to be discussed is the relationship between edaphic attributes and floristic—structural characteristics in the study area, that suggest a specieshabitat association possibly also influenced by biogeographical factors related to the domain phytogeographic context in which the study area is located, the Brazilian Caatinga domain(Queiroz 2006;Santos et al.2012).In our study area species of the called Arboreal Caatinga, a Caatinga unit commonly located in the domain southern end,presents high representativeness in most of the sample units, with species common to this formation being abundant and structurally important (Santos et al. 2012; Moro et al. 2014, 2016). Besides this, in sample units associated to higher fertility (Low Sandy-High Fertile area) there are restricted elements considered as characteristics of Arboreal Caatinga such as Goniorrhachis marginata Taub.,Cavanillesia umbellata Ruiz & Pav. and Coccoloba schwackeana Lindau(Santos et al.2012;Moro et al.2014).However,in in the less fertile and more sandy sample units there the presence and structural importance of elements associated with Sedimentary Caatinga,an another Caatinga unit with disjoint occurrence in Brazilian Northeast that area not predicted for the region, such as Cenostigma gardneriana Tul., Byrsonima correifolia A. Juss., and Psidium salutare(Kunth)O.Berg(Santos et al.2012;Moro et al. 2014).

This species differentiation in response to textural and fertility variables,with Sedimentary Caatinga characteristic species associated to sandy and less fertile sites, suggests the Sedimentary Caatinga past influence on the region and contests the current spatial distribution pattern of the Caatinga domain flora (Queiroz 2006; Santos et al. 2012;Moro et al.2016).This influence may be associated to two non-contrasting hypotheses that may explain the existence of Sedimentary Caatinga species in the southern region of the domain where they are not predicted. The first one is associated to the biogeographic separation by a pediplanation event in the transition between Neogene and lower Quaternary Periods and the second associated to the Sa~o Francisco River retraction during Quaternary Periods as creator of sedimentary units. According to the first hypothesis, the pediplanation would have exposed pre-Cambrian crystalline rock and isolated the Sedimentary Caatinga units in small nuclei such as the study area,which presents a high amount of sand in relation to other sites studied in the region(Ab’Saber 1974;Queiroz 2006).Thus,the fragment studied would have been a former Sedimentary Caatinga remnant isolated in the middle of a crystalline matrix originated by the pediplanation event.However, over time this area would have been colonized by species of the close matrix (Arboreal Caatinga), thus restricting the Sedimentary Caatinga species to sandy and less fertile points where their competitive capacity is greater.

From the second perspective, the existence of some sedimentary units in domain is also associated to Sa~o Francisco River retractions during the Quaternary Period due to aridity increase periods that would have exposed large portions of sand (Ab’Saber 1974; Tricart 1985;Queiroz 2006).Considering the study area proximity to the Sa~o Francisco River course (20 km),the studied site could be relict of sand banks discovered in this retraction and have been related through the river marginal areas to sedimentary formations of similar origin, as for example the‘Dunes of the Sa~o Francisco’ (Velloso et al. 2002;Queiroz 2006; Moro et al. 2016). This contact would have allowed the Sedimentary Caatinga species occurrence that would have been associated to points where the conditions allow a greater competitiveness,such the sandy study area.In this scenario, the Sa~o Francisco River would play an essential role as a latitudinal way of contact of the domain floras, which would add complexity to the current spatial distribution model of its floristic units(Queiroz 2006;Moro et al. 2016), since there may be situations similar to those presented here along the river adjacent areas. The shared presence of Sedimentary Caatinga and Arboreal Caatinga species with small-scale structural variations associated with the soil make us consider the last hypothesis more plausible to explain the results found, considering that in this situation the sand patches would have more discrete extensions inside the Arboreal Caatinga,contrary to what is known for the larger disjoint nuclei of Sedimentary Caatinga that are originated from the pediplanation event.Thus,we consider that the San Francisco retraction effects would have mainly local influences and that fit more logically to the studied context. However, the absence of concrete biogeographical studies in the Caatinga domain compromises any statement that goes beyond what is presented.

Thus, the fragment small-scale edaphic heterogeneity has acted as a differentiating agent that have contributed together to biogeographic factors for the existence of structural floristic complexity. This result emphasizes the importance of both small-scale soil and environmental variations to Seasonally Dry Tropical Forest ecological patterns and to tropical diversity (Chesson 2000; Munoz et al.2014;Vleminckx et al.2015).The diversity of smallscale structural behaviors and of floristic composition still draws attention to the need to consider such variations in studies that seek to evaluate the vegetation response to environmental changes due to climatic changes, anthropic interventions and in projects of degraded areas recovery(Santos et al. 2014; Ribeiro-Neto et al. 2016; Allen et al.2017; Rito et al. 2017). Considering that formations are cohesive units of homogeneous behavior may imply inadequate knowledge construction to be used in probably inefficient conservation policies,with wasted resources and serious risk to biodiversity.

AcknowledgementsTo Federal University of Lavras,Foundation for the Support to the Researches in Minas Gerais(FAPEMIG),Brazilian National Council for Scientific and Technological Development(CNPq) and to Coordination for the Improvement of Higher Education Personnel (CAPES) for all the support.