Nutrient cycling between soil and leaf litter in the Cerrado ( Brazilian savanna ) on eutrophic and dystrophic Neosols

We evaluated nutrient cycling by measuring leaf litter production and decomposition in the Brazilian savanna (cerrado sensu stricto) on eutrophic (calcareous) and dystrophic (arenitic) Neosols over the same spatiotemporal scale. We installed three transects in the Cerrado with plots in each Neosol type. Areas along the transects were randomly chosen for determining soil chemical properties, and accumulated leaf litter was collected monthly for a year for analysis. We used litter bags to test for a correlation between loss of dry mass and nutrient release. Th e calcareous cerrado had higher levels of macronutrients, higher pH, and lower Al compared to the arenitic cerrado. Th e estimated annual release of leaf litter nutrients in kg ha-1 yr-1 was higher in the calcareous cerrado. Nutrient concentration was signifi cantly higher in diff erent months in the calcareous cerrado, except for S and N. In general, the loss of dry mass was not correlated with nutrient release. Calcareous Neosol is unique to the cerrado sensu stricto, and retains more nutrients and facilitates higher leaf litter chemical quality than cerrado of the same physiognomy on dystrophic soils. Nutrient release is not always related to loss of dry mass, regardless of the amount of leaf litter.


Introduction
Some of the most important scientifi c questions in terrestrial ecosystem ecology have been concerned with understanding nutrient cycling dynamics in the atmosphere, vegetation biomass, and soil (D'Odorico & Porporato 2006;Sutherland et al. 2006).Nutrient release occurs through various mechanisms that depend on the type of litter (Berg & McClaugherty 2008).Chemical elements, especially macronutrients such as nitrogen, sulfur, and phosphorus, sustain trophic chains (Moss 2010) and ecosystem function (Raff aelli & Frid 2010).
Nutrient cycling is a process that occurs primarily in the edaphic environment, and involves litter accumulation and subsequent decomposition (Berg & McClaugherty 2008), followed by the transfer of nutrients to the soil (Schulze et al. 2002;Raff aelli & Frid 2010).When soil fertility is high, this can be refl ected in the concentrations of foliar nutrients (Rossatto et al. 2015), and consequently, litter with higher chemical quality may stimulate greater performance by decomposing microorganisms (Seastadt 1984).Th e Brazilian savanna, known as cerrado sensu stricto (Paiva et al. 2015), predominantly occurs on dystrophic, alic and deep soils (Latosols) (Eiten 1972;Gottsberger & Silberbauer-Gottsberger 2006).Several authors partially attribute these dystrophic soils to high aluminum saturation (Goodland 1971;Larcher 2000;Gottsberger & Silberbauer-Gottsberger 2006), which contributes to lower soil pH, decreasing nutrient solubility for plants (Eiten 1972).However, there is currently some debate on whether nutritional disorders in plants are directly or indirectly affected by Al (Souza et al. 2017).
Most of the studies involving aspects of nutrient cycling in the cerrado have been conducted in areas comprised entirely of Latosols (Peres et al. 1983;Schiavini 1983;Nardoto et al. 2006;Kozovits et al. 2007;Valenti et al. 2008;Villalobos-Vega et al. 2011;Paiva et al. 2015;Valadão et al. 2016).There is also a smaller portion of the Cerrado with shallow soils on rocky outcrops (Neosols), which can be mineral-rich (e.g., basalt, diabase and gabbro) or mineral-poor (e.g., arenite, quartz and shale) (Eiten 1972;Reatto et al. 1998).This portion of the Cerrado (~7.3 %) is dominated by campos rupestres, and there is no description to date of cerrado sensu stricto on calcareous outcrops.Instead, calcareous descriptions are limited to forest formations of the Cerrado, mainly dry deciduous forests (Reatto et al. 1998).However, the occurrence of calcareous soil was identified in cerrado sensu stricto vegetation in the municipality of Ituiutaba, Minas Gerais State, Brazil, (Souza et al. 2017).In addition, our field observations (white coloration) combined with chemical analysis (excess calcium) indicate specifically the occurrence of Neosol with rocky calcareous outcrop in the cerrado sensu stricto.Since plants extract nutrients mainly from soils, (Cianciaruso et al. 2006) and the storage of these nutrients occurs in the litter (Campos et al. 2008), it is expected that nutrient concentration in the soil reflects, to some degree, nutrient concentrations in the litter.
The calcareous presence in soil increases the pH and cation exchange capacity in the edaphic milieu (Vitti et al. 2015), increasing nutrient availability for plants, which is a factor that may favor biomass productivity (Haridasan 2000;2008), except with calcifuge species, which may suffer impairment in nutrient uptake (e.g., P and K) (Lee 1998) and growth (Haridasan & Araújo 1988;Lee 1998).In Neotropical ecosystems, such as the Cerrado, higher soil fertility is also associated with higher leaf litter nutrient concentration (Rossatto et al. 2015), which enables greater performance by decomposing microorganisms (Seastadt 1984).In addition to the potential differences in decomposition rate due to differences in leaf chemical quality, the proportion of leaf mass also declines exponentially over time (Villalobos-Vega et al. 2011), and the loss of leaf dry mass is directly related to the release of leaf nutrients through the action of decomposers (Berg & McClaugherty 2008).
In this context of nutrient dynamics between soil and vegetation, there is a shortage of experimental approaches on the same spatiotemporal scale, which allows a reasonable control of climatic variables, such as moisture, that may modulate nutrient dynamics between soil and plants (Goodland 1971;Peres et al. 1983;Coutinho 2002).Thus, we evaluated whether nutrient cycling differs in the same physiognomy of savanna in an apparently different soil (calcareous influence) and in an apparently typical soil (dystrophic -arenite influence) in the same region.Since the edaphic differences between areas have been confirmed experimentally, we expected higher soil fertility and lower acidity on calcareous soil.Also, considering that leaf nutrients would respond to soil fertility (Fyllas et al. 2009), we expected (1) a greater contribution of nutrients (especially Ca, Mg, N and S) in both leaf litter deposited and leaf litter undergoing decomposition in the cerrado with calcareous influence and (2) a decrease in nutrient content in the leaf litter with decomposition time in the cerrado, regardless of whether the soil was eutrophic or dystrophic.

Study areas
We conducted the study on two contiguous remnants of cerrado sensu stricto (s.s) on a mountain range within the limits of Ituiutaba, Minas Gerais (Fig. 1).The range hosts areas of seasonal forest or cerrado on the slopes, in addition to springs of streams (Pinheiro, pers.obs.).Ituiutaba is located in the mesoregion of the Triângulo Mineiro and Alto Paranaíba regions, where the predominant edaphic class is Red Latosol (Martins & Costa 2014).The total area encompassing both remnants is approximately 50 ha, and it is situated along a cuesta plateau (IBGE 1995).The regional climate is characterized as tropical savanna climate (Aw), with rainy summers and dry winters (Alvares et al. 2014), where dystrophic Red Latosol is the primary soil type (IBGE 2001).
Among the study areas are dispersed rocky outcrops that can be defined as Neosols (Benedetti et al. 2011;IBGE 2007), originating from the Bauru Group, Marília Formation, which correspond to entisol according to the American classification (USDA 1999) and leptosol according to the international classification (FAO 1998).The rock types were identified by observations in loco (the analysis being performed by Dr. Alice Bosco).From rock coloration (whitish) and effervescence reaction with dilute hydrochloric acid, we identified the mineral calcite in calcareous rocky outcrop, sparsely present in area 1, in accordance with Souza et al. (2017), who conducted their study close to this outcrop.While in Area 2, we identified the scattered occurrence of arenitic rocky outcrop (brownish).In theory, these two rock types would provide different soil fertility between sites.For simplification purposes, we referred the cerrado of Area 1 with calcareous Neosol as "calcareous cerrado" and the cerrado of Area 2 with arenitic Neosol as "arenite cerrado."The calcareous cerrado has an altitude ranging between 680 and 700 m, and it is referenced by a calcite point with the coordinates 19°03'438"S 49°26'422"W.The arenitic cerrado has an altitude ranging from 705 to 710 m, and it is referenced by an arenite point with the coordinates 19°03'633"S 49°26'075"W.
Both areas contain abundant herbaceous stratum, and according to a phytosociological survey in progress, the shrub-tree community frequently includes cerrado species.To date, the most common plant species in the calcareous cerrado are Callisthene fasciculata Mart.(Vochysiaceae), Luehea paniculata Mart.(Malvaceae) and Terminalia argentea Mart.& Zucc (Combretaceae).In the arenite cerrado, the most common species are Miconia albicans (Sw.)Steud.(Melastomataceae), Ouratea hexasperma (A.St.-Hil.)Baill (Ochnaceae) and Roupala montana Aubl.(Proteaceae).Th e two areas do not form a closed canopy, but the apparent thickening of the shrubby-arboreal stratum diff ers between them; the calcareous cerrado is a dense cerrado s.s. with predominantly arboreal stratum, especially near the fi rst transect, while the arenite cerrado area is a typical cerrado s.s. with shrubby-arboreal stratum that is less predominant (see Fig. S1A-B in supplementary material for general appearance of the vegetations).

Soil collection and analysis
We installed three 100-m transects per area (i.e., calcareous cerrado and arenite cerrado), with a minimum distance of 50 m in between.Twenty contiguous plots (10 m 2 ) were marked along each transect, for a total of 60 plots per area.We then randomly chose 20 plots for soil collection and chemical characterization.We obtained soil (depth = 0.2 m) from 15 diff erent sample points, which were randomly determined.Each sample was placed in an individual, labeled plastic bag, and the chemical analysis was later performed in the laboratory.
Soil layers from the 15 collection points within a single plot were combined and homogenized (N= 20 plots in each area).Soil pH was determined in water using a soil-liquid ratio of 1:2.5.Exchangeable divalent cations (Ca 2+ and Mg 2+ ) and exchangeable acidity (Al 3+ ) were extracted with 1 mol L -1 KCl, and concentrations were determined by titration (Embrapa 1997).Exchangeable potassium was determined using the photometric method directly in the soil stratum, obtained using Melhich-1 solution according to Raij et al. (2001).Organic matter (OM) content was determined using a colorimetric method with 5 mol L -1 sulfuric acid (Cantarella et al. 2001).P was determined by spectrophotometry using the reaction of molybdate with ascorbic acid (Embrapa 2009).We also calculated the base saturation (V%) of all samples according to the following formula: V% = 100.S/T, where S = sum of bases and T = cation exchange capacity at pH 7. With the mean values obtained from base saturation (V%) of all samples of each area, we determined whether soils were eutrophic (≥ 50 %) or dystrophic (<50 %) according to Embrapa criteria (2006).
The differences in natural fertility between the two cerrado areas were confirmed as assumptions for choosing the sites and proceeding with the comparisons.The calcareous cerrado showed eutrophic characteristics, while the arenite cerrado displayed dystrophic characteristics.The mean base saturation values (V%) of all samples were 77 V% in the calcareous cerrado and 29 V% in arenite cerrado.The P, K, Ca and Mg contents were significantly higher in the calcareous cerrado, as were OM and pH.Al was very low (0.0) in calcareous cerrado Neosol according to the quantitative method used, while mean Al content in the arenite cerrado was 1.28 cmol dm -3 (Tab.1).The calcareous and arenite cerrado areas were then characterized as eutrophic and dystrophic, respectively, according to the mean base saturation values (V%) and Embrapa criteria (2006).

Leaf litter collection and macronutrient analysis of accumulated and decomposed leaf litter
From July 2015 to June 2016, we collected monthlyaccumulated litter in 20 litter traps (50 x 50 cm) for evaluation of production, since twenty was considered a sufficient number for sample significance (Proctor 1983).Litter traps were composed of 2 mm nylon screens situated 20 cm above the ground.We separated only the leaf component for chemical analysis in the laboratory.
To evaluate leaf litter decomposition in the two areas, we randomly selected 10 litter traps to place in litter bags on the soil, totaling 100 litter bags per area.In each litter bag, 10 g of leaf litter were allocated.We removed 20 litter bags per area, at approximately 15, 30, 60, 90 and 120 days after the beginning of the rainy season (between October and April) (Bustamante et al. 2012).The plant material with both methods was placed in individual plastic bags for sorting, drying, and weighing.Chemical analysis was performed to quantify the nutritional condition of the leaf litter produced (accumulation in litter traps) and decomposed (fixed quantity in the litter bags).When possible, we compared the mean values obtained in our chemical analysis with the mean values obtained in other cerrado areas and in deciduous forest remnants.Such forest formations in Brazil typically occur on shallow soils with calcareous concretions (Silva et al. 2013a).
For both areas, the monthly-accumulated leaf litter (representing production) and the litter bag content (representing decomposition) were ground, dried (at 85 ºC for 48 h), and subjected to chemical analysis for quantification of macronutrient content.Nitrogen was determined using the Kjeldahl method (Tedesco et al. 1985), while P and K were determined by nitric-perchloric acid digestion (Bataglia et al. 1983), and P also colorimetrically (Embrapa 1997).Ca and Mg were determined by atomic absorption spectrophotometry (Bataglia et al. 1983), and S by turbidimetry (Tedesco et al. 1985).The mean macronutrient contents of accumulated leaf litter were multiplied by the corresponding dry mass and presented in kg ha -1 at annual and monthly levels.The mean macronutrient content of leaf litter from decomposition bags was related to the dry mass of the samples over the chosen time intervals (sampling months).

Statistical analyses
When evaluating differences in macronutrient content of leaf litter between the two areas, we were not able to group data for all months in a temporal analysis due to field incidents (e.g., seasonal interference by wild animals or by severe rain and wind), and the subsequent lack of sampling balance between the areas.Thus, to determine whether the leaf litter nutrient content differed between the areas, we performed independent t-tests from month to month.For months in which the data (original or transformed) did not show a normal distribution and/or homoscedastic variance, we performed Mann-Whitney tests.
We also determined whether there were correlations between the amount of dry mass in leaf litter (in litter bags) and concentrations of each macronutrient over the course of decomposition.Therefore, we performed a Spearman's correlation test for all macronutrients, since the data were not parametric (normal distribution and homoscedastic variance).
Normality and homoscedasticity in the data distribution were tested by the Kolmogorov-Smirnov and Levene tests, respectively, using P<0.05 as threshold criterion.For all analysis, we adopted α = 0.05 as our threshold for statistical significance.

Macronutrients in accumulated leaf litter
Over the year, macronutrients stored by leaf litter in the calcareous cerrado (Area 1) followed the descending order: Ca > N > K > Mg > P > S, while in the arenite cerrado (area 2), macronutrients stored by leaf litter followed almost the same descending order: N > Ca > K > Mg > P > S (Tab.2).The contents of all macronutrients oscillated over the year, following the leaf litter production seasonality in both areas, which was concentrated in the dry season.From comparisons between areas, the efficient use of most macronutrients (except N and S) was inversely proportional to their return in the areas' leaf litter (Tab.2).
Throughout the year, leaf litter in the calcareous cerrado displayed significantly higher Ca content than in the arenite cerrado (Fig. 2), with values proportionally ranging from 0.5 to 29 % of mean available dry mass in the areas.During most of the year, K, Mg, and P contents were also significantly higher in the calcareous cerrado (Figs.3-5), with values ranging from 0.3 to 22 % for K, from 0.1 to 5 % for Mg and from 0.01 to 2 % for P. In contrast, the S content was significantly higher in leaf litter of the arenite cerrado in the dry season (Fig. 6), with values ranging from 0.02 to 1 %.N content was significantly higher in the calcareous cerrado leaf litter in two of the rainy season months, but higher in leaf litter in the arenite cerrado in one of the dry season months (Fig. 7), with values ranging from 0.9 to 20 %.

Macronutrientes in leaf litter with decomposition
Among macronutrients, only variation in K correlated significantly and positively with the amount of leaf litter in both areas.That is, its concentration decreased as the dry mass decreased with decomposition (Fig. 8A-B) in the two areas, showing 70 to 75 % variation over the months.This effect did not occur with the other macronutrients -N, P, Ca, Mg and S -which oscillated irregularly during decomposition in the two areas, showing no significant correlation over months.

Soil fertility
From comparison between areas in relation to surface soil layers, where nutrient uptake mostly occurs (Cramer et al. 2009), we found that calcareous cerrado soil was less acidic, with negligible Al saturation, and retained higher contents of most macronutrients compared to arenite cerrado soil, probably due to alkaline calcareous rock and mineral richness (Reatto et al. 1998;Felfili et al. 2007).These differences between calcareous and typical (more acidic) soils are in accordance with other cerrado areas (Haridasan 1988;Souza et al. 2017) and forest communities of the Cerrado (Haridasan & Araújo 1988).Moreover, OM, which also was higher in Area 1, may be even important in providing nutrients for plants in the cerrado (Goodland & Ferri 1979), riparian forests of the Cerrado (Moreno & Schiavini 2001) and Australian savannas (Holt & Coventry 1990).Even in experiential situations of OM addition, the results usually indicate pH increase and decrease in Al saturation (Wong & Table 2. Leaf litter dry weight, macronutrient content (g kg -1 ) in leaf litter dry mass and macronutrient use efficiency estimated in kg ha -1 yr -1 in each study area.Notation: standard error (S.E.).* This attribute was estimated as the mean dry mass of leaf litter divided by the mean amount of the macronutrient in leaf litter during the year.In this estimate it is assumed that nutrient use efficiency does not vary with the amount of macronutrient circulated, according Vitousek (1982).*indicates significant differences between areas with α < 0.05 and **indicates significant differences between areas with α < 0.01.We show the raw data for all months.The data for all months were parametric (independent t-test; T values varied from 0.87 to 6.07; significant p values varied from 0.00001 to 0.015; non-significant p values varied from 0.089 to 0.392).Swift 2003), probably due to precipitation of soluble Al and the formation of Al-OM complexes (Bessho & Bell 1992).Th us, OM may play an important role in Al alleviation in acidic soils, where there are species non-tolerant to Al toxicity, mainly cultivated plants (Haridasan 2008).However, such role is not important for native plant species, which are adapted to high levels of acidity and Al (Watanabe & Osaki 2002), especially Al-accumulating species present in the families Rubiaceae, Melastomaceae and Vochysiaceae (Jansen et al. 2002a;b).Th e calcareous cerrado showed higher mean values for all soil chemical attributes except for acidity, compared to cerrado areas on Oxisols and a deciduous forest area on Basaltic Eutrophic Neosol (Tab.3).Unlike the calcareous and arenite cerrado over one year (mean ± S.E.).**indicates significant differences between areas with α < 0.01.We show the raw data for all months.The data for the months of July, August, October, January, February, March, April and May were parametric (independent t-test; T values varied from -4.88 to 1.37; significant p values varied from 0.0001 to 0.13; non-significant p values varied from 0.186 to 0.943).The data for the months of September, November, December and June were non-parametric (Mann-Whitney test; U values varied from 94.00 to 289.00; significant p value of indicated month was 0.001; non-significant p values varied from 0.186 to 0.943).and arenite cerrado over one year (mean ± S.E.).*indicates significant differences between areas with α < 0.05 and **indicates significant differences between areas with α < 0.01.We show the raw data for all months.The data for the months of July, September, November, December, January, February and April were parametric (independent t-test; T values varied from -1.11 to 2.98; significant p values varied from 0.005 to 0.032; non-significant p values varied from 0.123 to 0.986).The data for October were parametric after log transformation (independent t-test; T= 2.22 and p= 0.032).The data for the months of August, March, May and June were non-parametric (Mann-Whitney test; U values varied from 109.00 to 277.50; significant p value for indicated month was 0.035; non-significant p values varied from 0.241 to 0.282).cerrado area, these other areas are not influenced by calcareous rock, and according to Reatto et al. (1998), young soils with calcareous rock in rough reliefs (such as Area 1) tend to be diff erentially rich in bases.However, when compared to calcareous cerrado and a deciduous forest with the same edaphic classifi cation (i.e., calcareous Neosol), the former showed higher proportions of most of the chemical attributes present in the forest (Tab.3).Th ese diff erences may be due to the calcareous rocky outcrops of this forest that are very developed, forming extensive and thick blocks along the area (Rossatto et al. 2015).
Th e dystrophic arenite cerrado showed intermediate mean macronutrient values compared to areas of dystrophic cerrado (Latosols) and deciduous eutrophic forests (Tab.3).The highest nutrient content of the arenite cerrado compared to cerrado on Oxisols can be attributed to drainage diff erences between more or less developed soils.At least in Atlantic forests, shallow soils are generally less leached and less acidic than well-developed soils (Guilherme et al. 2012), which may favor nutrient retention in the surface layer.In addition, Rossatto et al. (2015) assert that young soils are rich in macronutrients, particularly Ca, Mg, and P, showing higher base saturation.In addition, the fi nding that our study areas, savannas, contained higher mean values of nutrients than a deciduous eutrophic forest partially refutes the hypothesis of Bond (2010) that savannas do not contain suffi cient nutrients to sustain forests.

Macronutrient concentration in accumulated leaf litter
Throughout the year, the patterns for estimated macronutrient content in dry leaf mass were similar between areas, since the macronutrients were measured in the seasonal leaf fall, which in both areas were concentrated in the dry season (especially July to October), according to others studies in the cerrado (Nardoto et al. 2006;Kozovits et al. 2007).The similarity between the two areas concerning the decreasing order of mean macronutrient returned by leaf litter indicated that the proportionality of leaf macronutrients in the cerrado does not depend solely on soil fertility.The fact that Ca in leaf litter of Area 1 was the macronutrient returned in greatest magnitude was probably due to the influence of high Ca levels in the calcareous rock, which may consist of about 99 % calcium carbonate (Braun-Blanquet et al. 1983).In Area 2, the decreasing order in annual mean macronutrients returned by leaf litter was similar to that of a cerrado on Latosol (Nardoto et al. 2006), which also began with N and only differed in sequence from the arenite cerrado due to inversion of S and P in the last position.The return of all macronutrients was higher in calcareous cerrado leaf litter, and it may be explained by the fact that areas with higher natural fertility reflect higher nutrient contents in leaves (Haridasan 2000).
Other information at the annual level about areas includes nutrient concentration stoichiometry, especially Ca:Mg and N:P, which have intrinsic relationships that may be unbalanced due to small changes in soil conditions (Rossatto et al. 2015).The annual mean Ca:Mg ratio was higher in Area 1 (4.38 g kg -1 ) than in Area 2 (3.57g kg -1 ), probably due to an antagonistic relationship between them (Baker & Pilbeam 2006), which can increase with very abundant Ca in the system.In relation to mean ratio of N:P, Aerts & Chapin (2000) defined values above 16 as a limitation of P. The annual mean N:P ratio was higher in Area 2 (23.06 g kg -1 ) than in Area 1 (17.00 g kg -1 ), which was associated with low P availability in soil analyses.Such P limitation is in accordance with that commonly expected for a large number of Neotropical systems (Bustamante et al. 2006;Araújo & Haridasan 2007;Rossatto et al. 2015).
The higher estimated annual return of Ca, K, Mg and P in the calcareous cerrado compared to arenite cerrado reflected concentrations of these macronutrients, which were significantly higher in Area 1 than Area 2 during the whole year or most months.Such results suggested that cerrado species on eutrophic soils show a less conservative strategy for most macronutrients.Additionally, the use efficiency of Ca, K, Mg and P between the areas was also the inverse of their concentrations in leaf litter, in accordance with Vitousek (1982).According to Parron et al. (2011), cerrado plants in soils with higher natural fertility tend to deposit more nutrients per unit of leaf litter biomass compared to plants in soils with lower natural fertility.We can then deduce from this that calcareous cerrado plants on eutrophic soils lose more nutrients, while plants on dystrophic arenitic soils are able to re-absorb and conserve nutrients.This is in accordance with the hypothesis that nutrient resorption efficiency increases as soil fertility decreases (Vourlitis et al. 2014).
Despite that the estimated annual return of S was higher in Area 1, this was not reflected in the S concentrations nor in S use efficiency.In the dry months, the S concentrations in leaf litter were significantly higher in Area 2, where Ca content was much lower.It is known that high Ca content may cause a decrease in other minerals, such as S (Ferri 1979).Additionally, about 90 % of S occurs in the organic form (Ferri 1979) and it can be drained along with water to other sites, or exported from the soil milieu by transformation into the gas phase (Begon et al. 2007).These factors can lower S content in leaf litter, making it difficult to determine differences in concentrations between areas during the rainy season.
The monthly N concentrations in accumulated litter leaf showed high variability between the areas, besides S use efficiency being relatively close between the areas.On the one hand, eutrophic soil conditions with presence of calcareous outcrops contributed to a significantly higher N concentration in Area 1 for a few months.In addition, such differences may arise from the fertility effect on the increase in species richness (Gentry 1988), which increases the likelihood of finding nitrogen-fixing species (Silva et al. 2013b).Such relation may be supported at the end of the phytosociological survey in progress, in which the proportion of Fabaceae species and individuals in the areas will be recorded and analyzed.On the other hand, N is a primary food source for decomposers, being present in proteins and vitamin compounds (Ferri 1979), and is one of the main elements that determines decomposition rate (i.e., based on the C:N ratio) (Larcher 2000).Considering that some decomposers can access the leaves accumulated in the litter traps throughout the month, and that decomposition of nutrients renders them unavailable to leaves (Schiavini 1983), this may have contributed to the lack of difference in leaf N content between the areas.Additionally, one of the litter traps in the arenite cerrado showed much higher N concentrations in some months.This particular litter trap contained mostly leaves of Ouratea hexasperma as there was one plant near the litter trap with branches located directly above.However, the association of a high N content with leaf litter of this species is in contrast to a study by Araújo & Haridasan (2007), who demonstrated that this typical cerrado species is an evergreen and has relatively low N content in mature leaves.Thus, such process in the current study deserves further investigation.
Comparisons between study areas and cerrado areas on Oxisols showed that, with the exception of S, leaf macronutrient concentrations (Tab.4) suggested that calcareous soil increases the nutritional quality of cerrado leaf litter, which tends to accelerate decomposition rates (Dahlgren et al. 2003) with subsequent mineralization.When comparing concentrations of leaf macronutrients of the arenite cerrado with areas of cerrado on Oxisols, results were highly variable (Tab.4).Since cerrado soil is also dystrophic, it may be that the differences between soils are due to phytosociological differences between areas, for example, by one area having higher abundance of leguminous species.
When comparing dystrophic cerrado to deciduous forests, we found that foliar concentrations of N, Ca, Mg, and P were invariably higher in the latter (Tab.4), in agreement with Paiva et al. (2015) who showed that forest litter has higher nutrients than savanna litter.However, foliar K concentrations were higher in most cerrado areas compared to deciduous forest areas (Tab.4), possibly due to differences in soil leaching rates.Considering only the comparison between the calcareous cerrado and eutrophic deciduous forests, we noted that Mg concentration was higher in Area 1.A plausible explanation for this difference is that excess Ca may exert an antagonistic effect on Mg uptake, competing for binding sites on plant cell membranes (Baker & Pilbeam 2006) and reducing Mg content.We noticed that results from other studies of deciduous seasonal forests may be due to differences in the methods used.For example, in a study by Rossato et al. (2015), nutrient concentrations were measured in photosynthetically active mature leaves that were mechanically extracted, which probably have more nutrients than dry naturally fallen leaves (Miranda et al. 2002); in a study by Viera et al. (2010), the litter used for the analysis was not only composed of the leaf component but also of twigs.

Decomposition of macronutrients in leaf litter
The loss of dry leaf mass is directly related, among other factors, to the loss (release) of nutrients during the decomposition process (Schiavini 1983;Berg & McClaugherty 2008).This relation was found only for K, in which its concentration decreased as a function of the loss of dry mass of the leaf litter, according a significant positive relationship between percent weight remaining and K concentration in residual material, as observed in the leaf litter of some dry deciduous forest species (Bargali et al. 2015).These results indicate that K is subject to extensive leaching from decomposing leaf litter; however, K is not the only mobile macronutrient, and explanations for the absence of significant correlations between dry mass loss and release of other macronutrients are needed with further investigations.
Regardless of the significance of the correlations, the similarity of the results between the areas showed that the relationship between the loss of dry mass and macronutrient release is not only altered due differences in the chemical quality of the leaf litter.The absence of correlations between leaf dry mass and most of the macronutrients measured suggests that other factors also modulate decomposition, such as lignin content of the species in the litter bags (Gholz et al. 2000), the diversity of decomposers between sample points or areas (Seastadt 1984;Berg & McClaugherty 2008), surface water availability for decomposers, which varies with soil texture (Coleman et al. 2004), microclimatic conditions (Gholz et al. 2000), or even external sources of macronutrients in the natural environment.

Conclusions
We confirmed that the cerrado in the study area occurs under eutrophic and dystrophic conditions on the same spatiotemporal scale.The cerrado under calcareous influence has a higher content of most macronutrients in accumulated   leaf litter, in comparison with cerrado under the influence of arenitic outcrops or on Oxisols.This demonstrates that Brazilian savannas on eutrophic soils usually exert a less conservative strategy with nutrients.In relation to decomposing leaf litter, the release of macronutrients in the cerrado does not differ according to leaf litter nutritional quality.Moreover, heterogeneity of the difference in macronutrients between distinct calcareous cerrado and some forests reinforces the notion that nutrient availability does not always act as one of the driving factors for savanna formation.

Figure 1 .
Figure 1.Map indicating study area locations.Area 1 is composed of calcareous cerrado, and Area 2 is composed of arenite cerrado in Ituiutaba municipality, situated in Minas Gerais State, Brazil.We used ArcGIS to adapt an image obtained from Google Earth.

Figure 2 .
Figure 2. Calcium content (Ca) in the leaf litter of calcareous and arenite cerrado over one year (mean ± S.E.).**indicates signifi cant diff erences between areas with α < 0.01.We show the raw data for all months.Th e data of the months July, August, December, March, April, May and June were parametric (independent t-test; T values varied from 2.81 to 5.76; signifi cant p values varied from 0.00001 to 0.007).Th e data for the months of September, October, November, January and February were parametric after log transformation (independent t-test; T values varied from 2.81 to 6.52; signifi cant p values varied from 0.00001 to 0.008).

Figure 3 .
Figure 3. Potassium content (K) in the leaf litter of calcareous and arenite cerrado over one year (mean ± S.E.).*indicates significant differences between areas with α < 0.05 and **indicates significant differences between areas with α < 0.01.We show the raw data for all months.The data for all months were parametric (independent t-test; T values varied from 0.87 to 6.07; significant p values varied from 0.00001 to 0.015; non-significant p values varied from 0.089 to 0.392).

Figure 4 .
Figure 4. Magnesium content (Mg) in the leaf litter of calcareousand arenite cerrado over one year (mean ± S.E.).**indicates signifi cant diff erences between areas with α < 0.01.We show the raw data for all months.Th e data for the months of July, September, October, November, December, January, February, March, April and June were parametric (independent t-test; T values varied from 2.96 to 7.62; signifi cant p values varied from 0.00001 to 0.007).Th e data for August were parametric after log+1 transformation (independent t-test; T= 4.12 and p= 0.0002).Th e data for May were non-parametric (Mann-Whitney test; U= 179.50 and p= 0.772).

Figure 5 .
Figure 5. Phosphorus content (P) in the leaf litter of calcareous and arenite cerrado over one year (mean ± S.E.).**indicates signifi cant diff erences between areas with α < 0.01.We show the raw data for all months.Th e data for the months of July, September and November were non-parametric (Mann-Whitney test; U values varied from 17.00 to 67.50; signifi cant p values varied from 0.0001 to 0.005).Th e data for the months of August, October, December, January, February, March, April, May and June were parametric (independent t-test; T values varied from -0.81 to 7.74; signifi cant p values varied from 0.0001 to 0.003; non-signifi cant p values varied from 0.204 to 0.503).

Figure 6 .
Figure 6.Sulfur content (S) in the leaf litter of calcareous

Figure 7 .
Figure 7. Nitrogen content (N) in the leaf litter of calcareous

Table 1 .
Soil chemical attributes in the edaphic layer of calcareous and arenite cerrado.Minimum, median and maximum values are presented for Mann-Whitney test.Means ± S.E. are presented for independent t-test.Degrees of freedom (d.f.) = 38.Notation: standard error (S.E.).

Table 3 .
Soil chemical attributes in the two study areas compared to other areas of cerrado sensu stricto (s.s.) and deciduous seasonal forest (d.s.).

Table 4 .
Mean total macronutrient content of leaf litter in calcareous vs. arenite cerrado in the current study compared with other areas of cerrado sensu stricto (s.s.) and deciduous forest.