Can an invasive African grass affect carbon and nitrogen stocks in open habitats of the Brazilian Cerrado?

Considering the emergence of ecosystems dominated by invasive species, there is growing interest in estimating the effect of biological invasions in ecosystem processes and provision of services. African grasses are the most invasive plants in the Cerrado (Brazilian savanna), but their impact on C and N stocks is poorly known. We compared patterns of C and N stocks in the aboveground biomass, root biomass and soil in open Cerrado (campo sujo) sites, both uninvaded and invaded by the African grass Urochloa decumbens. In both sites we estimated the aboveground biomass of U. decumbens and native grasses, as well as the root biomass up to 50 cm. We obtained C and N contents in the soil, as well as C and N stocks, up to 1 m depth, and variation in soil δ 13 C and δ 15 N. Although invasion did not affect the aboveground biomass, it did affect belowground biomass, leading to higher C stock in ne roots and soil N content close to soil surface, as well as higher C content along the soil prole. C and N soil stocks, soil δ 13 C and δ 15 N values did not signicantly differ between invaded and uninvaded site. Even a relatively low level of invasion by U. decumbens changed the root distribution pattern and increased C and N contents in the upper soil, which may promote ecosystem changes by altering nutrient dynamics. Although still preliminary, our study shows that dominance by U. decumbens can have severe effects in the Cerrado belowground environment.


Introduction
Invasive exotic species may cause a wide variety of impacts on invaded environments, which often lead to biodiversity loss and changes in ecosystem processes (Simberloff et al. 2013;Downey and Richardson 2016). While some biological invasions produce easily detectable patterns, others become noticeable only after causing severe negative effects, such as the loss of regulating ecosystem services (Castro-Diez et al. 2021). Invasive plant species often show high photosynthetic and growth rates, and many of them make associations with N-xing organisms (Ehrenfeld 2003(Ehrenfeld , 2004Simberloff et al. 2013;Vitousek and Walker 2014). As a result, they may change both above and belowground nutrient stocks -especially carbon and nitrogen -and affect biogeochemical cycles and balances, as well as other ecosystem processes and services (Ehrenfeld 2003;Vilà et al. 2011). By stocking, producing or emitting C and N through their metabolic pathways, invasive plant species can synergistically affect the balance of atmospheric gases, thus increasing local C and N pools (Litton et al. 2008;Rossiter-Rachor et al. 2009;Macdougall and Wilson 2011) and changing patterns of primary productivity and nutrient cycling ( Ehrenfeld 2003( Ehrenfeld , 2010Standish 2004;Liao et al. 2007;Greaver et al. 2016).
Savannas and grasslands are very important C sinks all over the world (Scurlock and Hall 1998;Grace et al. 2006), as they store carbon mainly in the root system and soil (Grace et al. 2006). The Cerrado is the richest savanna in the world (Silva and Bates 2002), with more than 13,100 vascular species, about 80% of which belong to the herbaceous-shrub layer (Mendonça et al. 2008;Overbeck et al. 2015). However, intensive cattle foraging based on exotic species as fodder (Ratter et al. 1997;Klink and Machado 2005;Silva et al. 2006) has promoted the invasion of many African C4 grasses in the Cerrado, which pose a great threat to this biodiversity hotspot (Pivello et al. 1999a, b;Klink and Machado 2005;Damasceno et al. 2018). Grasses of the genus Urochloa (brachiarias) have shown great capacity to invade and rapidly dominate the herbaceous layer of open Neotropical savannas (Damasceno et al. 2018) thanks to their high e ciency in producing biomass (Williams and Baruch 2000), favoured by effective N acquisition associated with symbiotic organisms (Boddey and Victoria 1986;Okumura et al. 2013), as well as an effective seed regeneration strategy (Xavier et al. 2021) and the release of allelopathic substances to inhibit other species germination (Gorgone- Barbosa et al. 2008).
In a scenario of climate change, warmer climates with altered precipitation patterns are expected to intensify (Hoegh-Guldberg et al. 2018), probably facilitating the establishment and spread of invasive species (Barnosky et al. 2012;Chown et al. 2015), and specially C4 grasses which are well adapted to such conditions (Edwards and Still 2008). Therefore, it is essential to predict the impacts of grass invasions on ecosystem processes and services in the Cerrado, and hence support effective management. However, to date no study has investigated the effect of invasive species on C and N stocks and uxes in the Cerrado.
In this study we compared patterns of C and N stocks in the aboveground biomass, root biomass, and soil in open Cerrado sites, both in uninvaded and invaded by Urochloa decumbens Stapf.. We speci cally aimed to answer: i) Does invasion by U. decumbens increase C stocks in the aboveground biomass? (Hypothesis: Invasive species are highly productive, leading to high C assimilation and high production of aboveground biomass.); ii) Are there differences in C stocks in the root biomass between uninvaded and invaded sites along the 0-50 cm soil pro le? (Hypothesis: Highly productive invasive species demand higher nutrient uptake from the upper soil layers and a greater production of roots than native species); iii) Are the soil contents and stocks of C and N different between uninvaded and invaded sites throughout the soil pro le? (Hypothesis: Different patterns of root production and decomposition between native and invasive species, as well as the association of U. decumbens with N-xing microorganisms modify distribution patterns of elements throughout the soil pro le.); iv) Does the soil δ 13 C and δ 15 N differ between invaded and uninvaded sites? (Hypothesis: Invaded sites with a more homogenous grassy layer and N-xing species show higher δ 13 C and lower δ 15 N soil values).

Study Area
The study was conducted at Itirapina Ecological Station (IES, 22°11'-15' S; 47°51'-55' W), a 2,300 ha Cerrado protected area in Southeast Brazil . Regional climate is subtropical with hot summers and mild winters (Koppen's Cwa), and average temperatures range from 17.8° C in June to 24.9° C in January. Average annual rainfall is 331 mm in the dry season (April to September), and 1128 mm in the rainy season (October to March)  Crop plantations (mostly sugarcane), forestry (Eucalyptus spp. and Pinus spp.) and brachiaria (Urochloa spp.) pastures surround the IES. The brachiaria was introduced to the reserve by the 1980's and has spread and become dominant in parts of the IES, but open physiognomies still show high plant diversity (Tannus and Assis 2004;Xavier et al. 2019).This study was conducted in a well-drained open savanna (campo sujo) where trees or shrubs have remained sparse for at least 40 years (Leite et al. 2018). We delimited an uninvaded (visually free of U. decumbens, ca. 4000 m 2 ) and an invaded (where U. decumbens was one the most abundant species, ca. 2500 m 2 ) area, where we collected soil and vegetation samples (respectively six and eight samples in each area) during the rainy season. The soil in this site is an Entisol with 96% sand and 4% clay; pH is 4.2 and the amount of organic matter is around 11 g dm − 3 (Xavier et al. 2019, Suppl. Material); the minimum annual water table depth is 2.7 m and there is less than a month of less than-optimal soil water availability close to the soil surface (20 to 30 cm) (Xavier et al. 2017). Previous studies found little spatial variation in the hydrological regime and soil properties within this area (Leite et al. 2018, Xavier et al, 2017, Xavier et al. 2019 suggesting that soil moisture and fertility conditions are similar in the uninvaded and invaded site.

Carbon stocks in the biomass
We visually estimated the percent cover of U. decumbens and native herbaceous vegetation in both uninvaded and invaded sites in ten 1 x 1 m plots at least 15 m apart using a wooden frame divided by a 10 x 10 cm string grid. In the same plots we collected aboveground live and dead biomass, separated live biomass into U. decumbens and native species, and oven-dried the material at 60° C until constant weight. We thus estimated the total aboveground biomass (live + dead, dry weight), and the aboveground biomass of native species and U. decumbens for both sites.
At eight sampling points in both uninvaded and invaded sites, at least 15 m apart from each other and 1 m apart from the aboveground biomass sampling plots, we sampled the roots contained in 20 x 20 x 10 soil blocks collected at 0-10, 10-20, 20-30, and 40-50 cm soil depth. Samples were air dried and treated with pyrophosphate decahydrate sodium solution (0.27%) to release the roots (Böhm, 1979), which were then washed in water and separated into ne roots (diameter ≤ 2 mm) i.e., those responsible for most of the water and nutrient uptake (Jackson et al. 1996), and coarse roots (diameter > 2 mm). Roots were oven-dried at 60° C for 48h and weighed. Both the aboveground and belowground biomass were smapled in the wet season (November-January).
We determined carbon stocks in the aboveground biomass and roots assuming 45% of their dry weight was carbon (Paruelo et al. 2010), for both dead and live biomass (Schlesinger 1977).

Carbon and nitrogen contents and stocks in the soil
To estimate soil C and N stocks we collected soil samples at six random locations at least 15 m apart in both the uninvaded and invaded sites. As the amounts of C and N in the soil are expected to decrease and show less variation as the depth increases (Jobbágy and Jackson 2000;Donovan 2012), soil samples were taken at eight layers with smaller intervals close to the surface: 0-5, 5-10, 10-20, 20-30, 30-40, 40-50, 70-80 and 90-100 cm. The soil samples were dried at 40°C for 48h, homogenized and sieved through a 0.2 mm mesh. Fractions of 15-30 mg of pounded soil from each sample were sealed in tin capsules and loaded into a ThermoQuest-Finnigan Delta Plus isotope ratio mass spectrometer (Finnigan-MAT; CA, USA) interfaced with an Elemental Analyzer (Carlo Erba model 1110; Milan, Italy) at the Laboratory of Isotope Ecology (CENA-USP, Brazil). The gas from the combustion was puri ed by gas chromatography column and introduced directly into the mass spectrometer for analysis of total C and N contents. Stable isotope ratios were measured according to internationally recognized standards (Atropine, Yeast and LECO) which were included in every run. Stable isotope values are reported in "delta" notation as δ values in parts per thousand (‰), so that δ ‰ = (R sample / R standard -1) x 1000, where R is the molar ratio of the rare to abundant isotope (δ 15 N/ δ 14 N or δ 13 C/ δ 12 C) in the sample and the standard. Atmospheric air was used as nitrogen standard. The precision of isotope ratio measurements was ± 0.3‰ for δ 15 N and precisions of N and C concentrations were 0.15 % and 0.01%, respectively.
In four of the eight soil sampling locations of both the uninvaded and invaded sites we collected undeformed soil samples with a bucket-auger from six soil depth layers (0-5, 5-10, 10-20, 30-40, 40-50, 70-80 cm); the soil bulk density of each sample was obtained by dividing the soil dry weight by the sample volume. Soil C and N stocks (Mg ha − 1 ) were then calculated based on the concentration of total carbon and nitrogen in the soil, soil bulk density and the thickness of the sampled layer (Fernandes and Fernandes 2008):

Data analysis
The effect of invasion by U. decumbens on the total aboveground C stock was estimated using generalized least square models with Restricted Maximum Likelihood estimation, considering invasion as a categorical xed effect. We used linear mixed models including each sampling location as random effect to assess the in uence of invasion, soil depth, and their interaction on: (1) the C stock in the root biomass (four soil layers), (2) soil C content, soil N content, C:N content ratio, 13 C content and soil 15 N content (eight soil layers), (3) C and N stocks in the soil (six soil layers). Models for the root biomass were implemented separately for ne (< 2mm) and coarse (> 2mm) roots, and all root classes pooled together.
We plotted model residuals for checking violations of distribution assumptions and removed outliers associated with very high C stock on the root biomass, as well as very high C and N stocks in the soil, to improve model estimation. Residual variation in models for root and soil stocks was often greater in the invaded than in the uninvaded site, and hence we implemented models allowing for different variances for each site. Likewise, residual variation in models for the C:N content ratio was much lower in the 0-50 cm soil layer than in the 50-100 cm layer, and in this case we allowed a distinct variance for each of these layers. Experimental variograms showed no association between distance of sampling locations and residual variance. We used maximum likelihood ratio tests to compare models that included or not these variance structures and interaction terms. All analyses were implemented in the R package nlme (Pinheiro et al. 2018). When the best models included a signi cant effect of soil depth or interations we performed all combinations of a posteriori comparison between factor levels in the R package "emmeans" (Lenth et al. 2018). All analyses were implemented in the R statistical environment (R Core Team 2021).

Carbon stocks in the aboveground biomass
Estimated average aboveground C stocks in the uninvaded and invaded site were respectively 3.26 Mg.ha − 1 (± 0.66) and 3.29 Mg.ha − 1 (± 0.98). Although the invasive grass U. decumbens accounted for approximately 24% of the ground plant cover and 25% of the aboveground C stock in the invaded site, the total C stock in the aboveground biomass (native + U. decumbens + dead) did not differ between the invaded and the uninvaded site (t = 0.36, p = 0.727; Fig. 1a); as expected, the C stock of native species was higher in the uninvaded site (t = 2.39, p = 0.033, Fig. 1b).

C stocks in the root biomass
The C stock in the root biomass at a depth up to 50 cm was generally higher in the invaded than in the uninvaded area but this effect was variable depending on root diameter and soil depth ( Table 1, Table 1S). The C stock in all roots (F = 2.89, p = 0.111) and in coarse roots (F = 32.03, p = 0.176) did not differ between the uninvaded and invaded sites (Table 1) but it varied according to soil layers (Table 1S); post-hoc comparisons showed that at the 0 to 50 soil pro le the C stock in coarse roots and in all roots was high at 10 to 20 cm depth (Fig. 2). By contrast, the best model for the C stock in ne roots included signi cant effects of invasion, depth, and a marginally signi cant (F = 2.36, p = 0.086) interaction between invasion and soil layer (Table 1). Post-hoc comparisons showed that in the invaded site the biomass of ne roots was much higher at 0 to 10 cm than in deeper soil layers (p < 0.001), whereas in the uninvaded site the C stock at 0 to 10 cm was only marginally higher than at 10 to 20 cm (t = 2.643, p = 0.054) (Fig. 2); in addition, the C stock in ne roots in the upper soil layer (0-10 cm) was 30% higher in the invaded than in the uninvaded site (Fig. 2). 3.3. δ 13 C, δ1 5 N, C and N concentration of soil samples Consistent with C stock patterns in the ne root biomass, the soil C content was higher in the invaded than in the uninvaded site in all soil layers (F = 10.02, p = 0.010, Table 2, Fig. 3a). In both sites the soil C content decreased approximately 50% from the soil surface to the 20-30 cm soil layer, but showed smaller variation in deeper soil layers (Fig. 3a, Table 2S). The best model for the variation in N soil content and C:N ratio along the soil pro le included, respectively, a marginally signi cant (F = 1.93, p = 0.077) and a signi cant interaction (F = 2.40, p = 0.290) between invasion and soil depth layer (Table 2). Post-hoc comparisons showed that the soil N content was over 20% higher in the invaded than in the uninvaded site in the top 5 cm layer (t = 3.54, p = 0.005); in both sites the soil N content decreased from the soil surface to the 20 to 30 cm soil layer (p < 0.05) and exhibited less variation in deeper layers, although that decrease was greater in the invaded than in the uninvaded site (Fig. 3b). In contrast, the C:N ratio was higher in the invaded site at the 20 to 30 cm soil layer (t = 3.81, p = 0.003); in both sites values increased along the soil pro le from 10 to 40 cm (p < 0.05), but there were no consistent differences in deeper layers (Fig. 3c). There were no signi cant differences in δ 13 C values between the uninvaded (-17.9 to -18.1‰) and invaded (-17.4 to -18.0‰) sites (F = 1.93, p = 0.194, Table 2). The δ 15 N values also did not differ between sites (F = 1.88, p = 0.200) and varied from 2.4 to 9.3‰ and from 1.8 to 8.1‰ at the rst 100 cm of soil depth in uninvaded and invaded areas, respectively. In both sites δ 15 N values increased (~ 7‰) along the soil pro le (Fig. 4).

Soil bulk density, C and N stocks
The invasion by U. decumbens did not signi cantly affect soil bulk density (F = 2.69, p = 0.152, Table 3); as expected, in both sites values of soil bulk density were lower very close to the soil surface (0 to 10 cm) than in soil layers deeper than 30 cm (Fig. 1S). Although C soil stock values were slightly higher in the invaded than in the uninvaded site, differences were not signi cant (F = 1.07, p = 0.340; Fig. 5a, Table 3). In both sites C soil stock values were initially lower, peaking at a depth of 10 to 20 cm, and then remaining at intermediate levels ( Fig. 5a). Also, N soil stocks were higher in the invaded than in the uninvaded site in the rst soil layers and tended to be lower below 20 to 30 cm, but differences were not signi cant (F = 1.03, p = 0.349; Fig. 5b, Table 3).

Discussion
Estimates of C stocks in the Cerrado, especially belowground, are scarce and highly variable, across different Cerrado physiognomies and soil depths (Paiva and Faria 2007;Maquère et al. 2008;Neto et al. 2010;Fidelis et al. 2013;Oliveras et al. 2013;Brito et al. 2019). Similarly, there are few estimates of N stocks in Cerrado ecosystems, which point to a much higher concentration in the rst 50 cm soil layer (Bustamante et al. 2006;López-Poma et al. 2020). However, the effects caused by biological invasions in the Cerrado C and N stocks have not been reported to this moment, and this is a vital issue especially in times of global warming and changes in the natural ecological cycles.We compared C stocks aboveground, and C and N stocks belowground along the soil pro le between an uninvaded open cerrado and a site invaded by the African grass U. decumbens. Invasion did not affect the overall aboveground C stock, soil C and N stocks and the distribution of C and N along the soil pro le, but the C stock in ne roots and the C and N contents in the upper soil layers were higher in the invaded site.

Aboveground C stock
The overall aboveground C stock in the invaded site (3.29 Mg.ha − 1 ±0.978) was very similar to that of the uninvaded open cerrado (3.26 ± 0.658 Mg.ha − 1 ). This nding was unexpected since U. decumbens often maintains high productivity even in the typically infertile soils of Neotropical savannas (Braz et al. 2013;Gómez et al. 2013), which is pointed as one of the main reasons of its success as invasive in the Cerrado (Pivello et al. 1999a;Klink and Machado 2005;Forzza et al. 2012). In fact, in a review study Liao et al. (2008) found that 83% of the invaded communities signi cantly increased productivity. As in our study U. decumbens accounted for only 24% of the ground cover in the invaded site it is possible that at such level of invasion the biomass produced by U. decumbens possibly was just compensating the biomass of displaced native species. Alternatively, low soil fertility in our study site may be limiting the dominance and primary productivity of U. decumbens, thus contributing to the lack of difference in aboveground biomass between the invaded and uninvaded sites even after about three decades of invasion.

Root biomass C stock
We observed larger C stock due to ne roots in the upper layer (0-10 cm) in the invaded site. This indicates that U. decumbens produces more ne roots than Cerrado native grasses in the rst 10 cm of soil, increasing by 30% the C stock in that layer; this difference tended to be lower at greater soil depths. There was also a tendency of more coarse roots at lower soil layers in the uninvaded site. Higher biomass of coarse roots at greater depths (also found by Kauffman 1998, andAbdala et al. 1998) may relate to the presence of sparse shrubs in the campo sujo physiognomy, which produce thicker and deeper roots. This may also have contributed to the high variation in the C stock values of coarse roots in both the uninvaded and invaded site. The general pattern of root distribution along the soil pro le observed in both uninvaded and invaded sites -where ne roots are more plentiful in upper soil layers to uptake resources, and the coarse roots concentrate in layers just below -was expected and found in other Cerrado areas (Abdala et al. 1998;Castro and Kauffman 1998;Oliveira et al. 2005;Paiva and Faria 2007;Oliveras et al. 2013;Morais et al. 2017). This high investment on the production of ne roots-is consistent with the idea that plants need to expand their absorption area in infertile soils to meet their nutritional demands (Bustamante et al. 2012). African grasses are highly effective on soil nutrient uptake (Williams and Baruch 2000; Guenni et al. 2002), which is a major reason for their high productivity and dominance over native grasses in invaded Neotropical savannas (Williams and Baruch 2000). A high production of roots in the rst 50 cm of soil by U. decumbens was reported by Guenni et al. (2002).

Soil C stock
Even though the soil C content was signi cantly higher in the invaded than in the uninvaded site along the soil pro le, there were no statistical differences in the soil C stocks between sites. Other authors reported higher soil carbon contents and stocks in grassy ecosystems invaded by exotic species, especially in the upper layers, (Liao et al. 2007;Zhang et al. 2018, Haubensak andParker 2004). In contrast, Litton et al. (2008) and MacDougall & Wilson (2011) found no difference in the soil organic carbon stocks at different depths between native areas in North America and ecosystems invaded by African grasses.
Interestingly, MacDougall & Wilson (2011) found that although the invaded sites had twice as many roots than uninvaded areas there were no differences in the amounts of soil carbon; the authors attributed this fact to high decomposition rates in the invaded area that resulted in no C accumulation in the soil. We found higher ne root C stock at the upper soil layer and higher soil C content along the soil pro le in the site invaded by U. decumbens than in the uninvaded site, but differences in C stock along the soil pro le were small and statistically non-signi cant. Since we obtained C stock values for each soil layer by multiplying soil C content and soil bulk density, this discrepancy is likely related to the fact that soil bulk density tended to be lower in the invaded sites, especially in soil layers up to 50 cm (Fig. 1S). Such lower values of soil bulk density could be associated with the high production of ne roots by U. decumbens, which thus may increase soil aeration and promote a better environment for soil decomposing organisms (Swift 1979). Also, the increase of soil organic matter is directly related to soil water retention, which is an important soil hydraulic property (Yang et al. 2014), and thus is a major driver of ecological functions. Further studies are needed to compare soil microbial activity and water retention in both uninvaded Cerrado sites and invaded by exotic grasses.
Some studies have shown that especially shrubs from open Cerrado physiognomies have very deep roots, ranging from 2 to 18 m (Rawistcher 1948, Castro andKauffman 1998;Franco 2005). Therefore, the C stock in the total underground biomass of a native campo sujo must be much greater than that obtained in this study, where our estimates refer to the top 80 cm of soil and mainly based on ne roots. Also, the low cover and density of woody individuals in our study site (Leite et al. 2018) possibly contributed to low underground C stocks. In a campo sujo converted to Urochloa brizantha pasture, Rodin (2004) estimated a reduction in the ne-roots C stock of about 15% up to 3m soil depth, and around 70% due to coarse-roots at a depth of 4 m. Thus, considering higher soil depths we expect the dominance by U. decumbens and consequent elimination of native species will decrease the soil C stock.

Soil N stock and C:N ratio
Several authors have reported greater concentration of N in the soil due to biological invasions, as well as increased nitri cation and N-mineralization processes, usually associated to N-xing organisms (Ehrenfeld et al. 2001;Haubensak and Parker 2004;Liao et al. 2007Liao et al. , 2008Parker and Schimel 2010). We also found higher N levels in the rst 5 cm of soil in the invaded than in the uninvaded site, which we attribute mainly to the association of Urochloa species with N-xing microorganisms (Boddey and Victoria 1986). Several authors have found a diversity of symbiotic N-xing bacteria colonizing the roots of Urochloa species or in the rhizosphere (Reis et al. 2001;Okumura et al. 2013;Silva et al. 2013). The association of plants with N-xing microorganisms often increases the amount of N in the leaves and, consequently, in the soil after their decomposition (Vitousek and Walker 2014). Still, the quality of N-rich organic material is generally better, thus stimulating microbial activity and decomposition (Liao et al. 2008).The Cerrado soils are typically N-limited (Bustamante et al. 2006;Haridasan 2008), so that the presence of exotic species that increase N amounts in the soil may lead to imbalance in the system. Native Cerrado species are adapted to low soil fertility (Furley and Ratter 1988) and the input of N caused by the African grasses -which are more N-demanding -may create a positive feedback mechanism that favors their dominance over native species. The existence of this feedback in sites invaded by U. decumbens is consistent with the high production of ne roots in the top 5 cm soil layer, as this could allow the species to e ciently uptake the available soil N. In addition, the soil biota may be signi cantly modi ed under this fertilization condition, and the new micro ora can also bene t the exotic species to the detriment of the natives (Ehrenfeld et al. 2001;Parker and Schimel 2010). These feedbacks as a result of changes in N dynamics have been demonstrated in other ecosystems subjected to grass invasions (Rossiter-Rachor et al. 2009;Lee et al. 2012).
In our study, the high N content in the most super cial soil layer of the invaded site contrasted with a rapid decline in soil N content along the soil pro le compared to the uninvaded site. Below that thin layer the C:N ratio increased steadily along the soil depth up to a 20-30 cm soil depth, where the addition of C was much higher than that of N, and the C:N ratio was higher in the invaded site.Therefore, it seems that better conditions of soil for microbial activity and decomposition in the site invaded by U. decumbens were restricted to the top 5 cm soil layer. Previous studies also found higher C:N ratio in ecosystems invaded by exotic grasses (Drenovsky and Batten 2007;Yang et al. 2013), whereas others reported a small decrease (Haubensak and Parker 2004). 4.5. δ 13 C and δ 15 N of soil sample We found little variation in soil δ 13 C and δ 15 N between the invaded and uninvaded sites along the soil pro le. The soil δ 13 C concentration in savannas tends to decrease as the woody cover increase (Boutton et al. 1998), so that invasion by C 4 African grasses would lead to enrichment on δ 13 C if these invaders replaced either herbaceous or woody species with C 3 metabolism. Lack of δ 13 C variation in our study is possibly associated with the prevalence of C 4 photosynthesis among Cerrado grasses (Klink and Joly 1989) and the low woody cover in our study site (Leite et al. 2018). Likewise, variation in leaf and soil δ 15 N in savannas is often associated with distinct N acquisition strategies (Schmidt and Stewart 2003;Bustamante et al. 2004), and hence soil δ 15 N values would be expected, for example, as a result of dominance by invasive N-xing species and deposition of 15 N depleted leaf material (Haubensak and Parker 2004). Although U. decumbens has a high N-xing potential and this may be re ected in δ 15 N signatures (Boddey and Victoria 1986), leaf δ 15 N values are often similar among grass species (Wang et al. 2013). Therefore, even though N acquisition strategies among native Cerrado species are largely unknown, it is possible that N xing ability is also common among abundant native Cerrado grasses.

Conclusion
Even a relatively low level of invasion by the African grass Urochloa decumbens in an open Cerrado caused signi cant changes in the belowground environment, as the most super cial soil layer in the invaded site showed higher ne root stock and soil C and N contents compared to the uninvaded site.
Such inputs may modify the soil microbiota and decomposition dynamics, thus affecting C and N cycles. Besides, the increase of N in the Cerrado soils would create a positive feedback mechanism favoring invasive grasses, promoting further invasions and hindering the reestablishment of native vegetation. Further studies based on replicated plots along a gradient of invasion would detect dominance thresholds after which U. decumbens may shift underground stocks and nutrient cycling in the Cerrado. Although still limited to a single site, this study already shows serious effects of the Cerrado invasion by U. decumbens in the soil compartment, which is much less visible to managers.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. SupplementaryInformationFINAL.docx