Effects ofplant physiognomies, season and year on soil N and P availability
Our results confirmed the hypothesis that open plant physiognomies would have less soil N and P availability than plant physiognomies with higher aboveground cover. Except in the dry season, the woodland savanna consistently showed the highest values of soil N availability throughout the two years of our study. These results align with the idea that plant-soil interactions should determine and maintain the plant physiognomies specificities within the Cerrado (Neri et al. 2012; Neri et al. 2013). In this biome, litter production increases as the plant layer becomes denser, which results in higher inputs of organic matter and nutrients to the soil, higher decomposition and mineralization rates (Table 1), consequently should result in higher N availability (Hossain et al. 1995; Li et al. 2003; Pellegrini et al. 2014). Further, specific characteristics of woodland savanna soils, such as higher soil depths and more clayey textures, favor soil water and nutrient retention and storage, potentially increasing nutrient availability and favoring the tree species establishment and growth (Neufeldt et al. 2002). Of particular interest are the low values of NO3-N concerning those of NH4-N found in all plant physiognomies. This low NO3-N availability could be related to low nitrification rates linked to the known low abundance of nitrifying bacteria in these soils (Bustamante et al. 2006), but also to a great demand for nitrate from soil microbiota and plants (Parron et al. 2003). In any case, NH4-N as the dominant form of inorganic N can be interpreted as a sign of a particularly conservative N cycle in the Cerrado (Pinto et al. 2002).
The lowest levels of soil P availability observed in the open savanna could be explained by several mechanisms. First, latosols with lower clay contents, such as those of the more open plant physiognomies, are known to have lower total soil P (Resende et al. 2011; Neri et al. 2013). Second, the lower litter production and soil organic matter content of the open savanna compared to those of the other plant physiognomies could also be behind its lower P availability (Lardy et al. 2002; Eberhardt et al. 2008; Bezerra et al. 2015). Third, litter quality (i.e. nutrient content) may also help to explain these differences. Different plant physiognomies have different species composition with different litter qualities (Ribeiro and Walter 1998; Neri et al. 2012). Specifically, the amount of P in the litter of trees from the denser plant physiognomies tends to be higher than that of more open plant physiognomies, as a result of a higher resorption efficiency linked to lower soil P availability (Ratnam et al. 2008; Vourlitis et al. 2014). Finally, higher Al3+ contents in the open savanna (Neri et al. 2013) might occlude soil organic P, hindering its mineralization and, therefore, decreasing its availability (Eberhardt et al. 2008).
As we hypothesized, soil N (NH4-N and NO3-N) availability was significantly lower in the dry season than in the other seasons, particularly in the woodland savanna. Soil moisture has been recognized as one of the main drivers of N mineralization in savannas (Bustamante et al. 2006). Even in gallery forests inside the Cerrado biome (under the same seasonal climate as the studied areas), N mineralization shows seasonal dynamics, with reductions in the driest period (Parron et al. 2003, 2011; Bustamante et al. 2006). Our results reflect the soil moisture dependence of N mineralization and availability. However, the fact that this seasonal pattern was not significant in the intermediate and open savanna supports the hypothesis that different plant physiognomies have different sensitivities to the typical seasonality of this biome. More importantly, they also suggest that different plant physiognomies might respond differently to forecasted changes in soil moisture due to climate change (see below). This differential response may be at least partially related to differences in soil N content among these three plant physiognomies (Table 1). The lower soil N contents in the intermediate and open savanna could limit plant and microbial growth throughout the year, resulting in a faster N uptake and cycling when it becomes available (Nardoto & Bustamante, 2003). Thus, although N mineralization and nitrification rates could be also higher during the rainy season in these two plant physiognomies, a superior N demand may rapidly decrease this new available N, being more difficult to capture the positive effect of the rainy season.
The seasonal pattern of soil P availability was not as clear as that of soil N availability. Phosphorus is the most limiting nutrient in savannas worldwide (Abrahão et al. 2019; Grace et al. 2006; Pellegrini, 2016; Resende et al. 2011). Consequently, plants and microorganisms are likely to strongly compete for it during the most active season (rainy), therefore decreasing its availability (Abrahão et al. 2019; Bustamante et al. 2012, 2006; Nardoto et al. 2013; Resende et al. 2011). However, when we considered the plant physiognomies separately, the intermediate savanna did not follow this seasonal pattern, confirming the different sensitivities among plant physiognomies to seasonality. In addition, available P dropped dramatically after the second late-dry season of the study. We cannot explain all these different temporal patterns of soil P availability, as they were related neither to soil water content nor to any of the plant physiognomies-related variables we studied. Different forms of soil P are differently related to several biotic (e.g. plants, microorganisms) and abiotic (e.g. minerals, organic matter, climate) factors, which makes particularly challenging to assess both its fate and main drivers (Ruttenberg, 2003). Also, the Cerrado biome is known for a strong dependence on the organic P turnover (Resende et al. 2011). As a result, the availability of P in the Cerrado could likely be more controlled by biotic processes (such as plant uptake and microbial mineralization, immobilization, and solubilization) than by seasonal changes in soil water content (Bezerra et al. 2015; Cross and Schlesinger, 1995; Pavinato and Rosolen, 2008). In any case, this study adds evidence to the inherent complexity of the P cycle, highlighting the need for more studies assessing its main mechanistic drivers and its potential responses to the ongoing environmental changes (Garcia-Velazquez et al. 2020).
Effects of rainfall reduction on soil N and P availability
Similar to what we found for the effect of seasonality on soil N and P availability, our rainfall reduction structures influenced N and P availability in opposite ways, tending to decrease N availability but to increase that of P, with exception of the intermediate savanna, where P availability also tended to decrease. These antagonistic effects of soil moisture reduction on soil N and P availability can influence organisms and ecosystem structure and function directly, but also indirectly via alterations in the N:P stoichiometry (Peñuelas et al. 2012, 2013; Yuan and Chen 2015). In a study carried out in a wide climatic gradient, Delgado-baquerizo et al. (2018) found that aridity is negatively correlated with available N but positively correlated with available P in soil, which could lead to a decoupling of these two elements. Here we add experimental evidence that changes in water availability could affect in opposite directions soil N and P availability, which could influence ecosystem functioning in the medium term and promote long-term changes in the stoichiometry of the whole ecosystem (Peñuelas et al. 2012; Delgado-baquerizo et al. 2018). Our results are the first indication that forecasted reductions in precipitation are likely to alter the N and P dynamics between plants, soil microorganisms, and soil N transformations in the Cerrado vegetation (see Engelhardt et al. 2021, Wang et al. 2021; Lin et al. 2022).
As we observed with seasonality, the woodland savanna was the most responsive plant physiognomies to the rainfall reduction in terms of soil N availability, despite having lower rainfall reduction values (4–8%) than the open savanna (10–32%). Differences in the nutritional status of both plant physiognomies (see above) could be behind their different response to changes in soil water content. Interestingly, the woodland savanna showed higher soil moisture values in the second than in the first year (0.129 and 0.110 m3 m− 3, respectively), whereas in the intermediate savanna this difference in soil moisture between years was slighter (0.111 and 0.106 m3 m− 3, respectively) and in the open savanna was the opposite (0.185 and 0.203 m3 m− 3, respectively). This could explain why we observed a higher N availability and a higher response of this nutrient to the rainfall reduction treatment in the second than in the first year only in the woodland savanna. All our plant physiognomies are located on the same site with the same climatic conditions, so observing or not differences in soil moisture among years is likely to be dependent on plant physiognomies-specific plant-soil interactions. Thus, these results indicate that particular plant-soil features of the woodland savanna could make soil nutrient availability in this plant physiognomies more sensible to short- (season and annual) and long-term (climate change) changes in soil water availability than the other two plant physiognomies, with important implications in the ecosystem functioning and composition.
Regarding the response of P availability to the rainfall reduction treatment, the intermediate savanna did not follow the general pattern observed in the other two plant physiognomies either. This particular response of the intermediate savanna could be due to the particularly low efficiency of the rainfall reduction structures in this plant physiognomies (Table 3). In the denser plant physiognomies (i.e. woodland and intermediate savanna) a significant fraction of the rain may be intercepted by the canopy and flow through the branches and trunk towards the soil (Klaassen et al. 1996; Yan et al. 2021), which would reduce the efficiency of our structures in excluding part of the rainwater. On the other hand, contrary to what we would have expected in terms of soil characteristics (e.g. soil organic matter content, clay content), the soil moisture registered in the open savanna was much higher than that in the intermediate and woodland savanna (Table 3, Figure S2). The high levels of silt in this plant physiognomies may be behind this result as silt (fine) particles account for greater water retention (Ferrari et al. 2016). In addition, in grasslands, there may be more moisture in the soil surface layer due to the entry of moisture from nocturnal dew (Ferrari et al. 2016), which is common in the study area. All these facts suggest that using soil moisture sensors in the most superficial soil layers may have not been the best solution for evaluating soil water availability and the efficiency of our rainfall exclusion structures. These insights should be considered in future studies that seek to simulate climate change in the different Cerrado plant physiognomies. However, the consistently different responses of P availability in the intermediate savanna and the highest responses of N in the woodland savanna to both seasonality and the rainfall reduction treatment, provide evidence that soil nutrient availability of the different Cerrado plant physiognomies may be affected in different ways by forecasted changes in climate, thus reinforcing the need of considering this particular characteristic of the Cerrado in future studies.