Introduction

The seed bank comprises all viable seeds present on or in the soil that are capable of producing physiologically independent individuals. A transient seed bank is where the seeds remain viable for 1 year after being dispersed, and a persistent seed bank is when seed viability persists for more than 1 year (Simpson et al. 1989). Seed bank density and composition typically varies spatially and temporally, and is a measure of the potential plant community that could regenerate after disturbances (Uhl and Clark 1983). The seed bank depends on the seed rain, and on seed losses that arise from: (i) seed predation and death due to interactions with pathogens and animals; (ii) loss of viability due to natural senescence, and (iii) seed germination (Louda 1989; Simpson et al. 1989).

The cerrado is classified as tropical savanna with high physiognomic and floristic diversity. The climate is highly seasonal, with 90 % of the precipitation occurring in the wet season (October through April), and a well-defined dry season, when fires often occur. Natural and anthropogenic fires have occurred in cerrado for thousands of years (Salgado-Laboriau and Ferraz-Vicentini 1994), and recent reports show that fire frequency varies from a fire every 1–4 years (Coutinho 1990). Although fire is considered to be an important determinant of floral diversity, structure and physiognomy of cerrado vegetation (Coutinho 1990; Moreira 2000), there is no report of the dependence on fire of flowering, seed dispersal, or germinability in cerrado. As for most savannas, mass flowering of many herbaceous species occurs immediately after a fire (Coutinho 1990; Munhoz and Felfili 2005), followed by seed dispersal (Coutinho 1977; Cirne and Miranda 2008). Although many woody species have adaptations to protect them from fire effects, such as thick bark, and resprouting from lignotubers and underground organs, frequent fires can result in high mortality (Medeiros and Miranda 2008; Miranda et al. 2009), low rates of seedling establishment (Hoffmann 1998), and a significant increase in grass cover (Moreira 2000).

Vegetative reproduction via root suckers or rhizomes and resprouting from underground organs are the most prevalent persistence strategies of cerrado plant species (Coutinho 1990; Hoffmann 1998; Salazar and Goldstein 2014). However, sexual reproduction and seedling establishment are common regeneration processes in cerrado (Laboriau et al. 1963; Oliveira and Silva 1993; Melo et al. 1998). Data on the effects of fire on the seed bank were sparse until recently (Andrade et al. 2002; Ikeda et al. 2008), and no study has yet reported on post-fire seed bank dynamics. As the fire season in cerrado coincides with the period of seed release by many herbaceous and woody species (Almeida 1985; Batalha and Mantovani 2000; Oliveira 2008), a large part of the seed production for a year may be lost, altering the density, composition, and dynamics of the seed bank and modifying the recruitment pattern of the community (Hoffmann 1998). This study examined the effect of a prescribed fire in the middle of the dry season (August) on the seed bank in cerrado sensu stricto and its recovery in the first year after the fire event. The following questions were addressed: (1) Does fire have an immediate effect on the soil seed bank? (2) Does the seed density and species composition of the soil seed bank recover within 1 year after the fire?

Methods

Study area

The study area was located in the Reserva Ecológica do IBGE (Instituto Brasileiro de Geografia e Estatística), 35 km south of Brasília (15°55′S, 47°52′W). The climate in the region is Cwa in Köppen’s classification, with a mean annual temperature of 21 °C and mean annual precipitation of 1,436 mm. The altitude ranges from 1,048 to 1,150 m. The soil is typically yellow–red latosol (Acrustox according to the American classification). The reserve has an area of 1,360 ha and is entirely cerrado vegetation.

The study was conducted in an area (10 ha) of cerrado sensu stricto, a dense scrub of shrubs and trees with a continuous ground layer dominated by grasses. The density of woody individuals with stem diameter greater than 5 cm at 30 cm above the ground was 1,614 individuals ha−1 (M. N. Sato unpublished data). The aboveground biomass of the herbaceous layer was 6.4 Mg ha−1, with the grasses representing 47 % of the total, and the dry mass of litter in the area was 2.1 Mg ha−1. The site was protected from fire for 4 years prior to our experiment.

Seed bank sampling

Soil samples were collected along five 50 m transects within the experimental plot, 70 m apart, and 50 m from the plot boundary. The top 2 cm of soil was collected every 5 m along each transect using an iron frame of 25 × 25 cm. Earlier studies have shown that 70 % of the seeds are located in top 0.5 cm of the soil and 90 % in the first cm depth (Andrade et al. 2002). The samples were collected 1 day before a prescribed fire in August 1999, 1 day after the fire, and then monthly until the end of the next dry season in August 2000, along transects parallel to the original transects. Before the fire, 50 litter samples were also collected along with the soil samples. Because litter production is significantly reduced after cerrado fires and is markedly seasonal (Nardoto et al. 2006), litter was available for sampling only on the last sampling dates.

Viable seed

In the laboratory, each sample (soil + litter) was air-dried to prevent germination before hand-sorting of the samples. The total number of seeds in soil and litter samples was estimated as the sum of visually detected (direct counting) and the number of germinable seeds (indirect counting). For the direct count, undamaged seeds were manually removed from each soil and litter sample and separated into monocots and dicots. Fruits and diaspores were also considered as seeds, and these seeds were submitted to a germination test with a photoperiod of 10 h (white light: 25.7 μE m−2 s−1) and a thermo-period of 37 °C for 10 h and 22 °C for 14 h to simulate the soil temperature at 1 cm depth in the first week after fires in the cerrado (Dias et al. 1996). The geotropic curvature of the radicle was used as the germination criterion. The number of germinated seeds was counted at 3-day intervals for a period of 1 month, when the seeds that contained a fully formed embryo and endosperm were tested for viability with tetrazolium. The colored embryos were considered viable but dormant.

For the indirect counts, after hand-sorting, each soil sample was transferred to a plastic tray. The samples were moistened every day and maintained in a greenhouse (mean irradiance = 1,450 μE m−2 s−1) for a period of 6 months. The number of emerging seedlings was counted weekly for 6 months. Seedlings were transferred to plastic bags containing a mixture of cerrado soil and sand, and were kept in a greenhouse for identification.

Data analysis

The Kruskal–Wallis test (α = 0.05) with Dunn’s multiple comparison test was used to evaluate the immediate effect of fire on the soil seed bank density and for the comparison of monthly means. The Mann–Whitney test (α = 0.05) was used to test the contributions of monocots and dicots to the monthly seed density.

Results

The densities of viable seeds were 221 monocot seeds m−2 year−1 and 416 dicot seeds m−2 year−1, with a monocot:dicot ratio of 0.53. The contribution of the litter layer to the seed bank was small, especially for monocot seeds (2 seeds m−2 before the fire and 1 seed m−2 1 year after the fire). Viable dicot seeds were only observed in the litter 1 year after the fire (10 ± 1 seeds m−2; mean ± se).

There was no significant difference in the total seed density recorded before (63 ± 8 seeds m−2) and immediately after the fire (83 ± 20 seeds m−2). The density of viable seeds declined with the onset of the rainy season; the lowest densities occurred, with no significant difference, between December (10 ± 3 seeds m−2) and February (16 ± 3 seeds m−2). An increase in the viable seed density was observed in March, and from May onward the density of viable seeds was similar to that recorded before the fire.

For the monocot seed bank, the density observed before the fire (41 ± 7 seeds m−2) did not differ from that recorded immediately after the fire (31 ± 4 seeds m−2). During the year, the viable monocot seed density reached minimum values from November (9 ± 4 seeds m−2) through April (4 ± 1 seeds m−2) and, within this period, there was no significant difference among the seed densities. One year after the fire, the monocot seed density (15 ± 3 seeds m−2) had not reached the pre-fire value (p < 0.05; Fig. 1).

Fig. 1
figure 1

Variation in the density of viable seeds of monocots and dicots in the soil and litter samples following a prescribed fire in August 1999, in an area of cerrado sensu stricto in the Reserva Ecológica do IBGE, Brasília, Brazil. Asterisks indicate significant differences between monocot and dicot seed densities (p < 0.05). Ab august before the fire, Aa august 1 day after the fire

Full size image

There was no difference between dicot seed density before (21 ± 4 seeds m−2) and 1 day after the fire (52 ± 17 seeds m−2). The viable dicot seed density declined with the onset of the wet season, reaching minimum and similar values (p < 0.05) between November (10 ± 3 seeds m−2) and February (12 ± 3 seeds m−2). One year after the fire, the dicot seed density (69 ± 14 seeds m−2) was significantly greater than the pre-fire value (p < 0.05).

Before the fire, there were more (p < 0.05) monocot than dicot seeds in the soil bank. With the beginning of the rainy season, germination started in the field, and from September through February, there was no difference in the contribution of monocot and dicot seeds in the total seed bank. From March onward (7 months after the fire), the dicot seed density was greater than the density of monocots. One year after, the density of dicot seeds was 4.6 times greater (p < 0.05) than the density of monocot seeds (Fig. 1; Tables 1, 2).

Table 1 Dynamics of viable-seed density of monocot species in the soil seed bank in an area of cerrado sensu stricto in the Reserva Ecológica do IBGE, Brasília, Brazil, after a prescribed fire in August 1999
Full size table
Table 2 Dynamics of viable-seed density of monocot species in the soil seed bank in an area of cerrado sensu stricto in the Reserva Ecológica do IBGE, Brasília, Brazil, after a prescribed fire in August 1999
Full size table

Seeds from 178 species were identified in the soil and litter bank: 50 monocot and 128 dicot species. However, only 21 monocot and 58 dicot species (Tables 1, 2) showed viable seeds. The identified monocot species (19) were mainly Poaceae (64 %), followed by Cyperaceae (18 %) and Iridaceae (9 %). Individuals of two species died before they could be identified. Echinolaena inflexa (110 seeds m−2 year−1), Axonopus barbigerus (42 seeds m−2 year−1), Bulbostylis spp. (19 seeds m−2 year−1), and Schizachyrium tenerum (15 seeds m−2 year−1) were the most important species in the monocot seed bank throughout the year, while 12 species (60 %) showed 1 seeds m−2 year−1 or less (Table 1). The seeds of E. inflexa comprised 50 % of the viable monocot seeds in the soil bank throughout the year. Mesosetum loliiforme was the only monocot species that showed viable seeds only before the fire, and Aristida setifolia, Axonopus marginatus, Panicum cervicatum, Paspalum multicaule, Rhynchelytrum repens, and Trachypogon spicatus were present in the soil samples only after the fire (Table 1).

Of the 58 dicot species with viable seeds (Table 2), 19 % (11) did not grow to a stage where they could be identified. Ten species could be identified only to family level, and seven to genus level. The identified species were mainly herbs (48 %), subshrubs (15 %), and shrubs (27 %). Trees were represented only by Dimorphandra mollis and Cecropia sp. Throughout the year, Compositae and Leguminosae were the families with the largest number of species, 15 and 13 species, respectively. Although represented by only four species, Melastomataceae showed the highest seed density in the seed bank, mainly because of the high seed density of Miconia albicans which contributed 51 % of the viable dicot seeds found in the seed bank. Miconia albicans (221 seeds m−2 year−1), Compositae 1 (64 seeds m−2 year−1), Borreria capitata (17 seeds m−2 year−1), and Oxalis sp. 1 (14 seeds m−2 year−1) were the most important species in the dicot seed bank throughout the year. Forty-five species (77 %) showed a viable seed density of less than 1 seeds m−2 year−1. Only Erigeron bonariensis and Gnaphalium purpureum were restricted to the pre-fire sample, and 12 species were present in the seed bank before and after the fire. Fifty-six percent of the dicot species (35 species) were present in the seed bank from February onward, 6 months after the fire. Ageratum conysoides, Chamaechrista orbiculata, and Leguminosae sp. 2 were found only in the samples collected 1 month after the fire.

The number of monocot and dicot species with viable seeds decreased during the rainy season, and only E. inflexa and M. albicans (Tables 1, 2) showed viable seeds during the entire year.

Although some seeds in the soil bank were from exotic species (Rhynchelytrum repens, Emilia sonchifolia, Erigeron bonariensis, Gnaphalium purpureum, Scoparia dulcis, and Sida spinosa), they are common in the Reserva Ecológica do IBGE (IBGE 2004) and did not result from contamination during the germination experiment. The viable seed density of those species was very low, <1 seeds m−2 year−1.

Discussion

As the fire season in cerrado coincides with the period of seed release by many herbaceous and woody species, we investigated the immediate effect of fire in the soil seed bank and its recovery in the year following fire. Seed bank density was not altered by fire, and 8 months after it seed bank density was similar to pre-fire value. However, with a different species composition and an increase in seed density of the dicot species.

The soil seed bank density in the study area was similar to those reported from other cerrado sites (Salazar et al. 2011), savanna woodlands and grasslands in Ethiopia (Gashaw and Michelsen 2002), and a semiarid grassland (Gonzalez and Ghermandi 2008). However, the low seed density in the litter layer of the study area before the fire (2 seeds m−2) and 1 year after (1 seed m−2) contrasts with gallery forests in cerrado, where the litter layer intercepts 30–60 % of the seeds (Pereira-Diniz and Ranal 2006). These low seed densities may reflect the alteration in density and structure of the woody vegetation resulting from the previous fire (4 years early), which prevented the formation of a thick litter layer (Nardoto et al. 2006) that would intercept the seed rain. Also, the low density of viable seeds in the soil seed bank may be a consequence of the large numbers of unfertilized seeds of monocots (31 %) and dicots (22 %) or incomplete development of the embryo (33 % for monocot and 22 % for dicot seeds; data not shown). Predation, unfertilized seeds, or seeds with incomplete development of the embryo have been reported as the main causes of the large numbers of non-viable seeds in the pre- or post-dispersal phases of several cerrado species (Oliveira and Silva 1993; Parron and Hay 1997; Sassaki et al. 1999; Schmidt et al. 2005).

The soil seed bank was not altered by the fire, possibly as a consequence of low soil temperatures during cerrado fires, which are surface fires with low residence time (Miranda et al. 2009). At the soil surface, temperatures may reach 280 °C, returning to pre-fire values after 5 min (Castro-Neves and Miranda 1996); and maximum temperatures at 1 cm depth vary from 29 to 55 °C, with no significant change below 5 cm (Miranda et al. 1993, 2009). For grasses and dicots of the cerrado herbaceous layer, (Overbeck et al. 2006) showed that germination was not altered after heat treatment (50–90 °C for 2 min), and for some dicot species in the herbaceous layer, germination was reduced only after the seeds were exposed to 130 °C for 2 min. The same response was reported by Schmidt et al. (2005) for seeds of Hepteropterys pteropetala, a common shrub in cerrado vegetation. Therefore, fire is only lethal to the seeds stored in the litter layer, soil surface, or those buried at a few millimeters depth.

One year after the fire, the seed bank density reached pre-fire values. From May onward, the density of viable seeds was similar to that recorded before the fire. In general, fires anticipate the flowering peak of herbaceous vegetation from the end (April–May, Tables 1, 2) to the beginning (November–December) of the rainy season (Munhoz and Felfili 2005; Neves and Damasceno-Junior 2011). The synchronous flowering after burning leads to relative synchronous seed set and dispersal and the rapid establishment of seedling at open microsites. In addition, the removal of the grass layer by fire results in greater visibility and availability of flowers to pollinators, and the initiation of a new reproductive cycle, as a rapid response to fire, is an effective recruitment strategy (Coutinho 1990; Whelan 1995). As the production of flowers, fruits, and seeds is, for many of herbaceous and subshrub species, much faster than for shrub and tree species, a large number of diaspores are produced in the few months after a fire. While the fruiting phase of trees and shrubs sometimes exceeds 1 year (Oliveira 2008), the ripening of the fruits of most species of the herbaceous layer lasts from 1 to 5 months (Almeida 1985; Munhoz and Felfili 2005), ensuring the release of seeds in the rainy season following a fire.

The monocot:dicot seed ratio was altered from 0.53 to 0.22 (Fig. 1; Tables 1, 2), perhaps as a consequence of the synchronous post-fire flowering of the dicot species in the herbaceous layer (Coutinho 1990; Munhoz and Felfili 2005; Neves and Damasceno-Junior 2011). Although some grass species, such as Leptocoriphyum lanatum and Elionurus muticus, are reported to bloom after fires (Monasterio and Sarmiento 1976), flowering grasses are not common in post-fire phenology studies. Munhoz and Felfili (2005) reported only nine grass species flowering, and Neves and Damasceno-Junior (2011) reported none in the year following a fire in campo sujo (an open form of cerrado). The 3-fold increase in the richness of dicot species in the soil bank 1 year after the fire suggests that, although fire does not have an immediate effect on the density of the dicot seeds, there is a positive effect on dicot species increasing the seed density and diversity in the soil bank in the first year (Table 2).

The rapid recovery of the dicot seed density may be associated with the ability of resprouts, from established plants, to flower sooner than non-resprouters juveniles (Pausas et al. 2004). However, the significant reduction in the dicot seed density with fire exclusion (pre-fire conditions) may be a consequence of the high productivity of native grasses that in a short period recover most of the pre-fire biomass (Neto et al. 1998) with a fast increase in grass cover (Moreira 2000). This may result in a reduction of herb and subshrub diversity in cerrado (Fidelis and Pivello 2011) and in the grasslands of southern Brazil (Overbeck and Pfadenhauer 2007). Significant reductions in the number of flowering species and individuals in the herbaceous layer in the second year after burning in Venezuelan and Brazilian savannas were reported by Silva et al. (1990) and Munhoz and Felfili (2005).

Similarly to other savannas and cerrado physiognomies, seeds of trees were poorly represented in the soil bank (Tybirk et al. 1993; Williams et al. 2005; Scott et al. 2010; Salazar et al. 2011). Although 55 tree species are present in the study area (M. N. Sato, unpublished data), only seeds of Dimorphandra mollis and Cecropia sp. were present in the soil bank. Salazar et al. (2011), working in different vegetation types in the cerrado (from low to high canopy cover), also reported a very low number of tree seeds in the soil bank and hypothesized low recruitment of woody species from the soil bank. The low recruitment of tree species may be a consequence of fewer diaspores that reach the soil surface in the cerrado (7.40 gm−2 year−1) when compared to the humid forests of southeast Brazil (Leal and Oliveira 1998), low seed viability (Sassaki et al. 1999; Velten and Garcia 2007; Salazar et al. 2011), a high proportion of unfertilized and predated seeds (Oliveira and Silva 1993; Schmidt et al. 2005), and the removal of seeds by ants and rodents (Christianini et al. 2007; Briani and Guimarães 2007).

Even with an increase in the number of species, the monocot seed bank did not return to the pre-fire density (Table 2), mostly because of a reduction in the seed density of E. inflexa and A. barbigerus. Both perennial grasses are very common in cerrado. Although E. inflexa may recover its pre-fire height in the growing season following a fire, the same pre-fire architectural complexity is not recovered, resulting in a reduction in the number of inflorescences (Miranda and Klink 1996; Murakami and Klink 1996) with a negative effect of fire on seed production (Parron and Hay 1997). No information on the effects of fire on seed production of A. barbigerus is available; however, Klink (1994) reported that in the year following clipping, plants did not tiller or recover their initial height, which might suggest that a single growing season is not enough to recover the pre-fire flower production. As for the dicot species, the diversity of monocot species increased after the fire, suggesting that fire exclusion for long periods may result in a significant reduction in the richness of monocot species in the soil bank. However, the timing of the fire must also be considered to maintain species richness, since most of the cerrado grasses disperse their seeds from the beginning to the middle of the dry season (Table 1) (Almeida 1985; Martins and Leite 1997).

With the advance of the rainy season, the soil seed bank became depleted, with minimal seed density recorded in January, the middle of the rainy season. This reduction characterizes a transient seed bank (Garwood 1989), as reported for several cerrado vegetation types (Oliveira 2008; Salazar et al. 2011) and other savannas (Gashaw and Michelsen 2002). In this study, only the monocot E. inflexa and the dicot M. albicans maintained a seed bank throughout the year (Tables 1, 2) even though M. albicans forms a permanent seed bank and E. inflexa does not. M. albicans produces a large number of very small photoblastic seeds that may remain in the soil bank until light is available (Carreira and Zaidan 2007), such as after a fire event. Also, this species does not flower in the year following a fire (Hoffmann 1998). In contrast, the seed-dispersal period of E. inflexa lasts from December to August (Martins and Leite 1997), and although seed production decreases after burning, the fire season typically precedes the flowering period of this species (Parron and Hay 1997). Although, many cerrado woody and herbaceous plants have underground organs such as xylopodia, rhizomes, bulbs, and lignotubers that form a bud bank from which they vigorously resprout after drought or fire (Monasterio and Sarmiento 1976; Fidelis and Pivello 2011) the recruitment of new individuals via the seed bank, as observed in this study, increases the genetic variability in the plant community and contributes to the persistence of plant populations in cerrado communities. This variability can be invaluable in a fire-prone environment such as the cerrado, where few species form a persistent soil seed bank (Sassaki et al. 1999; Garcia-Núñes et al. 2001; Araújo and Cardoso 2007; Velten and Garcia 2007; Salazar et al. 2011).

Conclusions

Fire had no immediate effect on the soil seed-bank density, and pre-fire seed bank density was recovered within a year following the fire. However, the dicot and monocot species contributed differently to the seed-bank recovery. The difference in the contributions of dicot and monocot species to the recovery of the density and richness of the soil bank suggests that fire frequency and fire season must be considered in fire management plans of conservation areas to achieve maximum seed richness in the soil bank. Also, the post-fire dynamics of the seed bank indicates the importance of recruitment of new individuals through the seed bank and not only by vegetative reproduction, in the post-fire recovery of cerrado vegetation.