Introduction

The species composition and individual species characteristics of savanna vegetation are shaped by fire (Williams and Cook 2001; Bond et al. 2005; Hoffmann et al. 2012), soil water availability (Cook et al. 2002; Scholes et al. 2004; Scott et al. 2009; Rossatto et al. 2012; Murphy et al. 2015), soil nutrient stocks and climate (Pinheiro and Monteiro 2010; Lehmann et al. 2011, 2014). Savannas around the world are frequently burnt by natural or anthropogenic means, and up to 75 % of some savanna areas burn annually (Hao et al. 1990), which affects plant recruitment (Rossiter-Rachor et al. 2008) and development (Beringer et al. 2015) and their phenological and functional events (Hoffmann 1998; Pausas et al. 2004; Alvarado et al. 2014). Despite the intense wet season from December to May, Australian savannas are the most frequently burnt vegetation in the world (Andersen et al. 2005; Chuvieco et al. 2008; Beringer et al. 2015). Across Australian savannas, half of the vegetation burns annually (Edwards et al. 2001; Andersen et al. 2005; Beringer et al. 2015) and in the Kakadu National Park (Northern Territory, Australia) few areas remain unburnt for more than 2–3 years (KBMPA 1999; Cook 2001).

Soils from savannas often have limited nutrient availability, affecting the leaf nutrient concentrations (mainly N and P) and leading to changes in the specific leaf area (SLA, ratio of leaf area per unit leaf dry mass) of some Australian species (Wright et al. 2001; Prior et al. 2003, 2005). Frequent incidence of fire over a long time frame may limit the available pool of soil nutrients, and fire may also diminish the soil organic matter concentration and accumulation (Andersson et al. 2004; Silva and Batalha 2008) when litter is incinerated (Oesterheld et al. 1999). In addition, frequent fires can result in savanna physiognomies with a conspicuous dominance of grasses, whereas trees and shrubs are more evident in savannas with lower fire incidence (Sankaran et al. 2005). Conversely, savanna vegetation protected from fire for long periods (>10 years) becomes less flammable as the density of trees increases over time, resulting in low sunlight interception and less vegetative biomass at the ground level, due to denser tree crowns, when compared to frequently burnt areas (Andersen et al. 2005; Beringer et al. 2015). Therefore, sunlight availability may also influence SLA in such a manner that in savannas high sunlight availability is associated with low SLA (Prior et al. 2003; Franco et al. 2005), while in low light environments, such as forests, high SLA is predominant (Givnish 1988; Prior et al. 2003; Habermann and Bressan 2011).

In this study, we examined the concentration of nutrients in leaves, the SLA and the soil fertility of two sites in a mesic savanna at the Territory Wildlife Park in the Northern Territory, Australia. One site has been protected from fire for the past 20 years, referred to here as the ‘unburnt site’, while the other site (‘burnt site’) has burnt 14 times during this same period. We screened and identified tree species unique to each site as well as those species common to both the burnt and unburnt sites. Due to the scarcity of sites protected from fire for more than 5 years within Australian savanna areas (KBMPA 1999; Cook 2001) we were unable to identify additional unburnt sites for this study. Thus, a single pair of burnt and unburnt sites ca. 30 km apart were assessed to understand the relationships between plant nutrition, soil nutrient stocks and SLA of mesic savanna vegetation under contrasting fire regimes. Plants within the unburnt site were expected to exhibit higher foliar nutrient concentrations and greater SLA compared to vegetation from the burnt site. This should contribute to the understanding of how fire and soil fertility affects the mineral nutrition and leaf scleromorphism of Australian savanna species.

Materials and methods

Leaf nutrient concentrations and SLA of the most representative species in two mesic savanna sites in the Northern Territory, Australia, were examined (Table 1). A savanna site in the Territory Wildlife Park (S12° 36′ 56.0″ E131° 00′ 45.3″), which has burnt 14 times during the last 20 years, was selected to represent the burnt savanna. Within the same region, a savanna remnant in Berrimah, a suburb of Darwin (S12° 24′ 43.7″ E130° 55′ 08.3″) that has been protected from fire for more than 20 years was selected as a representative unburnt savanna site (BOM 2013) (Fig. 1). The two sites are approximately 30 km apart; both sites are ~30 m above sea level and receive approximately 1700 mm annual rainfall (Prior et al. 2003; BOM 2013).

Table 1 List of plant species from the burnt and unburnt Australian savanna sites, Northern Territory, 2013
Full size table
Fig. 1
figure 1

Mesic Australian savanna, Darwin, Northern Territory, 22 February 2013. a Burnt site (14 fire events in 20 years), and b unburnt site (>20 years without any fire events)

Full size image

Plant species were identified and classified as being unique to each site (exclusive) or common to both sites (common). This procedure was used to avoid differences between nutritionally distinct groups of plants (e.g., leguminous species) and those species that may be influenced by fire frequency or soil fertility (Araújo and Haridasan 1988). The tree species that were found in both sites included eight common species; four species were exclusively found at the unburnt site and six were unique to the burnt site (Table 1). We recognize that this scenario (eight common species, four species from the unburnt site, and six species from the burnt site) may limit conclusions to a local plant community structure. However, studies examining the effects of fire on plant communities growing close to each other and with the same physiognomy are extremely rare (see Alvarado et al. 2014). In addition, this is a study of frequent long-term fire and fire protection effects on savanna vegetation, which is also uncommon (Cook 2001; KBMPA 1999; Edwards et al. 2001). Moreover, the dominant woody species in Australia are principally eucalypts (Eucalyptus and Corymbia spp.) (Burrows 2013), and this might be due to their fire resistance capacity, especially during the juvenile plant phase. Other woody species occurring in Australia belong to Acacia, Terminalia, Erythrophelum, Syzygium and Xanthostemon genera (Bond et al. 2012), and these genera were considered in the present study. Therefore, the plant communities from the two sites can be considered as representative tree species occurring in the region.

To determine the SLA and leaf nutrient concentration, we used four adult trees per species and sampled five fully expanded undamaged leaves per tree (therefore, 20 leaves/species) at the end of February 2013. In both sites, a mix of leaves was sampled under and outside the canopy. Due to the higher density of trees, we were not able to only sample sunlight leaves on the unburnt site. Ten leaf disks (6 mm in diameter) not including the midrib were obtained per leaf and oven-dried at 60 °C to constant mass. The SLA was calculated as the ratio between the leaf disk area (cm2) and its dry mass (g) (Habermann and Bressan 2011). The same leaf samples used for measuring SLA were washed in deionized water, oven-dried for 72 h at 60 °C, ground and digested in a 5:1 nitric:perchloric acid solution (Ratnam et al. 2008). After digestion, the P, K, Ca and Mg concentrations were assessed using inductively coupled plasma (ICP) spectrometry (Leman Labs, Hudson, MA, USA). Nitrogen (N) was determined by combustion, using a LECO CHN analyzer (LECO Corp., St Joseph, MI, USA).

At each site, four soil samples were collected at a depth of 10–20 cm, where the roots of most native plants are able to absorb nutrients (Wigley et al. 2013). These samples were analyzed to assess fertility parameters. Soil pH (in CaCl2), base (K, Ca and Mg) concentrations, cation exchange capacity (CEC), organic matter (OM), N, P and Al concentrations, as well as Al saturation (m %), were determined according to the international standard procedure for soil analysis (Robertson et al. 1999).

Plant species were grouped into exclusive, common and total (exclusive + common) species for statistical analysis. We used the Kolmogorov–Smirnov test to check the normality of data. A one fixed factor multivariate analysis of variance (MANOVA) was used to detect variations in leaf traits (N, P, K, Ca, Mg and SLA) of exclusive, common and total species between the burnt and unburnt sites. If MANOVA was significant, individual univariate ANOVAs were performed as post hoc test (α = 0.05) to determine which response variable differed between sites (Zar 2010). The variations on soil fertility between burnt and unburnt sites were determined using a Student t test (Zar 2010). The statistical procedures were performed using R (R Development Core Team 2012).

Results and discussion

Fire disrupts the nutrient cycle in savannas because it incinerates the biomass, volatilizes nutrients from plants and litter, and moves ash due to convection during the fire (Oesterheld et al. 1999). In Australian savannas, a single fire event promotes the loss of up 94 % N, 54 % P and 82 % K from plant biomass (Cook 1994). Consistent with previous studies (Cook 1994), the concentrations of N, P, K, Ca and Mg in the soil from the burnt and unburnt sites differed significantly (p < 0.05), being 63, 33, 46, 80 and 80 % lower, respectively, in the burnt than in the unburnt site. In addition, soil from the burnt site contained 42 % less organic matter when compared to soil from the unburnt site, possibly indicating a reduction in litter accumulation as a result of frequent fire, as also observed by Mills and Fey (2004). The soil fertility was lower at the frequently burnt than at the unburnt site (Table 2), which was reflected in the lower leaf nutrient concentrations observed in plants from the burnt site when compared to those in plants from the unburnt site (Fig. 2).

Table 2 Chemical properties of the soils collected from the burnt and unburnt sites of a mesic Australian savanna, Northern Territory, 2013
Full size table
Fig. 2
figure 2

Leaf macronutrient concentrations of the total (common + exclusive), common and exclusive species from the burnt and unburnt sites of a mesic Australian savanna. The box extends from the 25th to 75th percentiles, the continuous line within the box shows the median, and error bars represent 5th and 95th percentiles (n = 4 individuals per species). (ns not significant, *p < 0.05, **p < 0.01 and ***p < 0.001 using ANOVA)

Full size image

Leaf nutrient concentration and SLA significantly differed between the burnt and unburnt sites (MANOVA: Wilks’ λ = 0.23039; F = 10.5). Across all plant species (common + exclusive), plants from the burnt site exhibited lower N, P and Mg leaf concentrations as compared to leaves of plants from the unburnt site. Common species growing at the burnt site had lower leaf concentrations of N and K, and species exclusive to the burnt site contained less N, P, and Mg (Fig. 2). The exclusive, common and total (common + exclusive) species from the unburnt site exhibited greater SLA than those from the burnt site (Fig. 3). A long period (>10 years) of fire exclusion may result in vegetative systems with high tree density (Andersen et al. 2005; Durigan and Ratter 2006). Thus, the higher tree density on the unburnt site may have resulted in lower irradiances (high SLA), because we were not able to only sample sunlit leaves.

Fig. 3
figure 3

Specific leaf area (SLA) of the total (common + exclusive), common and exclusive species from the burnt and unburnt sites of a mesic Australian savanna. The box extends from the 25th to 75th percentiles, the continuous line within the box shows the median, and error bars represent 5th and 95th percentiles (n = 4 individuals per species). (ns not significant, *p < 0.05, **p < 0.01 and ***p < 0.001 using ANOVA)

Full size image

Related studies (Wright et al. 2001; Prior et al. 2005; Delgado et al. 2013) have demonstrated that sclerophyllous leaves (low SLA) of savanna species exhibit low leaf nutrient concentration and this response is influenced by the low soil fertility observed in savanna areas. Scleromorphism could also be influenced by the seasonal dry conditions of savanna areas, as longer dry seasons can be associated with low SLA in the Brazilian savanna (Souza et al. 2015). However, in the present study, both burnt and unburnt sites were located within the same region (30 km apart). Therefore, the rainfall and/or the length of dry seasons are unlikely to explain the differences in SLA observed between these sites.

In this study, we provide evidence that greater SLA and higher leaf nutrient concentrations seem to be favored in fire-protected savanna ecosystems, whereas frequent long-term fires seem to be associated with low soil fertility and with plants with sclerophyllous leaves and low nutrient concentrations.

Author contribution statement

Conceived and designed experiments: MCS, GDC, GH, DRR. Performed experiments: MCS, GDC. Analyzed data: MCS, DRR, GH, RF, GDC. Contributed reagents/materials/analytical tools: MCS, GDC, NWM, RF, DRR, LPCM, GH. Wrote manuscript: MCS, DRR, GDC, RF, NWM, LPCM, GH.