Vertical distribution of soil seed bank and the ecological importance of deeply buried seeds in alkaline grasslands

Study area

The study was conducted in five study sites located in the Hortobágy National Park, East-Hungary. Hortobágy is the largest salt-affected inland grassland area in Europe. The landscape is composed by a mosaic of various plant communities; it is characterized by alkaline grasslands and marshes, loess grasslands with scattered arable fields, floodplains and wooded habitats (Deák et al., 2014). The climate is temperate continental. The mean annual temperature is 9.5 °C and the average annual rainfall is 550 mm, with a maximum in June (80 mm). The average temperature and average precipitation show a considerable fluctuation from year-to-year (Fick & Hijmans, 2017).

The soil reference group of the study area is classified as Vertic Solonetz (IUSS Working Group, 2015). Vertisols and Vertic horizons are heavy clayey soils with a minimum of 30% clay content to a depth of 50 cm or more, characterised by clay minerals having high shrink-swell potential. Despite their compactness, Vertic soil horizons show a notable temporal and spatial variability. Because of the shrink-swell effect of the clay, dry seasons result in the formation of deep desiccation cracks. Due to the shrinking and swelling behaviour, the upper 10–20 cm of these soils has a mixed granular structure. Most of these types of soils are neutral or alkaline, because of their calcareous smectic clay parent material. The average pH values of Vertisols range from 6.0 to 8.0, however higher pH values can occur as well. High electrical conductivity (EC) is a typical feature of Solonetz, caused by the high base saturation and its sodium-dominated composition, together with high exchangeable sodium percentage (ESP) (Coulombe, Wilding & Dixon, 1996; IUSS Working Group, 2015).

Sampling was carried out in natural stands of dry alkaline grasslands (association: Achilleo setaceae-Festucetum pseudovinae). These short grasslands occur in dry areas and typical on Vertic Solonetz soils (Deák et al., 2014). The dominant grass species is Festuca pseudovina, typical forb species include Achillea collina, A. setacea, Plantago lanceolata and Trifolium spp. (Deák et al., 2014).

Sampling design

Samples were collected after snowmelt, during late February 2019. In each site, four soil cores (4 cm diameter) of 80 cm depth were collected with a soil auger for soil seed bank sampling (a total of 20 soil core samples). The samples were divided into sixteen 5-cm segments by depth: 0–5, 5–10, 10–15, 15–20, 20–25, 25–30, 30–35, 35–40, 40–45, 45–50, 50–55, 55–60, 60–65, 65–70, 70–75 and 75–80 cm, and each segment was placed separately in a plastic bag (320 soil segments in total). Samples were then stored at 5 °C for 3 weeks. Soil penetration resistance was measured in situ with Eijkelkamp Penetrologger next to each core sampling location down to 80 cm depth with 1-cm resolution. Permission for the research was granted by the Trans-Tisza Environmental, Nature Protection and Water Inspectorate (approval no 6646/08/2014).

Seed bank analysis

Soil seed banks were studied with the seedling emergence method. Samples were concentrated by washing them through sieves according to the method of Ter Heerdt et al. (1996). Vegetative parts and fine soil materials were removed using a 3-mm mesh sieve and then using a fine sieve with a mesh width of 0.2 mm. As a result, sample volumes were reduced considerably. The concentrated samples were spread out in a thin layer (3–4 mm) on trays filled with potting soil topped with an approximately 8-cm layer of sterilized potting soil. Control trays were filled with the same sterilized potting soil to filter out possible wind-borne and potting soil contaminants. The trays were randomly placed in an unheated greenhouse, at the Botanical Garden of the University of Debrecen, where germination was carried out from April to November 2019 under natural light conditions and regular watering. Seedlings were counted regularly, identified when possible then removed. Unidentifiable seedlings were transplanted into separate pots and grown until identifiable. At the end of June no new seedlings emerged, so watering was stopped to mimic arid summer conditions that are typical in the studied ecosystem. Dried sample layers were carefully broken through and stirred, and then at the end of August watering was restarted, which was continued until early November. Nomenclature of the species followed Király (2009).

Classification of species

Plant species were categorized into two morphological groups (graminoids and forbs) and into two ecological groups (grassland species and weeds), using the Hungarian Flora Database (Horváth et al., 1995). Species were arranged into simplified life-form groups based on Raunkiaer’s life-form categories. These groups were short-lived species (Th, TH) and perennials (H, G, Ch).

Seed length, seed width and seed thickness values were collected for each species from Schermann (1967) to calculate seed shape index (Bekker et al., 1998). Seed shape was expressed as the variance of three seed dimensions (length, width, and height), after standardizing all three values, so that length was considered as 1 (dividing all three values by the length). As a result of this, seed shape index was independent of seed size. The more round-shaped a seed was, the closer the index value was to 0, while the more elongated was the seed, the closer the index value was to 1. The average seed mass of each species was collected from the database of Török et al. (2013). Ecological indicator scores for salt tolerance (SB), and relative water requirement (WB) values (Borhidi, 1995)–were assigned to the species. Based on these trait data, community weighted means (CWM) of salt tolerance (SB), relative water requirement (WB), seed mass and seed shape index were calculated by weighting with the number of the germinated seedlings in each layer. The classification of each recorded species into morphological, ecological, and life-form groups, as well as their SB, WB, seed mass and seed shape index are given in Appendix 1.

Statistical analysis

For all data handling and analyses the R statistical software (ver. 4.0.1; R Core Team, 2021) was used. We tested the total number of seedlings, and the total number of species observed in the soil segments (N = 320), using negative binomial generalized linear regression models (GLMs; utilizing the glmmTMB R-package: Brooks et al., 2017), in which sampling layer and penetration resistance were predictor variables, with control for their interaction. Although these variables show substantial correlation, re-scaling prior to model fitting (subtracting the mean from all values and dividing by the standard variation), and controlling for their interaction mitigated the adverse effect of their collinearity, as shown by the low variance inflation scores from the fitted models (5> for all variables and models; tested with the “performance” R-package: Lüdecke et al., 2021).

Similarly, we fitted negative binomial models on the number of seedlings and germinated species separately for each level of our categorical variables, with sampling layer and penetration resistance as predictors, with control for their interaction as well. To test the effect of sampling layer and penetration resistance on community-weighted means (CWMs) we used log-linked Gamma GLMs, with sampling layer and penetration resistance as predictor terms. Due to the non-significant interactions, we left out interaction terms from the final models on CWMs. In cases when a given variable had a minimum value of zero, we transformed it by incrementing by one in order to be able to fit Gamma models.

In the preliminary analyses we fitted all models with sampling repetition nested in sampling site used as random factor, but due to the low (close to zero) variances we excluded random effects from the final models.

The vertical distribution of seed bank in alkaline grasslands

We found that both the germinable seed density and species richness of the seed bank significantly decreased with soil depth and that seedling density was the highest in the upper soil layer (0–5 cm) in the studied ecosystem. These seed bank patterns are consistent with the results of previous studies analysing the seed bank of different grassland (e.g., Ma, Zhou & Du, 2010; Niknam et al., 2018), forest (e.g., Godefroid, Phartyal & Koedam, 2006; Jaroszewicz, 2013) and wetland ecosystems (e.g., de Souza, Damasceno-Junior & Pott, 2019). More importantly, despite the sharp decline below the first five centimetres, deeper soil layers also contained germinable seeds (e.g., 1,154 seeds/m² in the 5–10 cm soil layer) that were able to germinate (if seeds were brought to the surface or near the surface). A novelty of our study is that we found that germinable seeds can be detected in soil layers much deeper than previously noted in other studies.

Based on the reviewed literature and our current knowledge, we believe that this is the second study in which germinable seeds from such a deep soil layer have been detected. The only study that investigated similarly deep burial of seeds was performed by Qian et al. (2016), but they studied the seed bank of mowing sand dunes–which are dynamic systems–where soil layers are constantly re-organized. They found that wind activities and sand burial contribute to the formation of the spatiotemporal pattern of the soil seed banks of active sand dunes. Due to the displacement of the seeds in different directions over time, seed density in active sand dunes showed a relatively constant vertical distribution across the studied sand profile. A significant amount of germinable seeds were found in the deeper soil layers too, which could be a form of adaptation of plant species to the effects of wind erosion on active sand dunes. In addition, their study showed that deeply buried seeds are brought to or near to the surface by the wind, and thus they play an important role in population regeneration and vegetation restoration in active sand dunes.

The majority of seed bank studies carried out in grassland (Godefroid et al., 2018; Jacquemyn et al., 2011; Ma, Zhou & Du, 2010; Niknam et al., 2018) and forest ecosystems (Douh et al., 2018; Godefroid, Phartyal & Koedam, 2006; Jaroszewicz, 2013) usually investigated only the uppermost soil layers to a depth of 20 cm. However, the vertical distribution of the seed bank in wetland ecosystems has already been studied at greater depths (Burmeier et al., 2010; Dawson et al., 2019; Jauhiainen, 1998; McGraw, 1987; van der Valk & Davis, 1979; Wall & Stevens, 2015). According to the reviewed literature, these greatest depths were 45 cm (McGraw, 1987) and 50 cm (Burmeier et al., 2010; Jauhiainen, 1998), respectively. This also means that seed bank research has generally neglected the deeply buried seeds. In the present study we found that deeper soil layers (down to 80 cm) may also play an important role in the preservation of at least some seeds of grassland species. Our results showed that the penetration resistance of the soil significantly increased with soil depth. This finding supports our result regarding the vertical distribution of soil seed bank; that is, the more compact the soil was, the fewer seeds it contained.

Role of plant traits in vertical seed distribution

Forb species were more abundant than graminoids in the seed bank, as has also been shown in other studies (Douh et al., 2018; Jutila, 1996; Ma, Zhou & Du, 2010). Although, short-lived species were dominant over perennials in the terms of species number, the seedling number of perennials was higher, suggesting a higher seed bank density. Seeds of grassland species were represented by a significantly higher density in the soil seed bank than weed species. Although forbs, grassland species and short-lived species occurred in a relatively large abundance in the deeper soil layers, graminoids, weeds and perennial species were scarce. A possible reason for this pattern is that opportunistic germination strategy is vital in the case of short-lived plant species, so they tend to have a greater longevity and might penetrate deeper than perennials (Jutila, 1996). The higher abundance of forbs compared to graminoids in the deep soil layers may be due to the fact that many graminoids have elongated seeds, therefore they cannot penetrate as deeply as rounded seeds (see also Bekker et al., 1998). Our findings that deeper soil layers were characterised by seeds of grassland species rather than weeds imply that seeds of grassland species have the potential to tolerate harsh environments for a longer time. The aboveground vegetation in the studied habitat type is species-poor and dominated by one perennial grassland specialist grass, Festuca pseudovina, with a percentage cover above 60% (see Deák et al., 2014; Valkó et al., 2014, 2017). Despite its dominance in the aboveground vegetation, F. pseudovina occurred with low density in the soil seed bank. The aboveground vegetation of alkaline grasslands contains very few weed species in low cover, due to the high salt content of the soil (Valkó et al., 2014) which was also reflected in the seed bank composition found in the current study.

Our results suggest that there is no difference in the depth distribution of the seeds of species with different seed mass and ecological indicator values. However, we found that seed shape was an important factor in determining the vertical distribution of seeds. Seeds with low seed shape index, i.e., round-shaped seeds, were dominant over elongated ones in deeper soil layers. This result corresponds to the findings of Bekker et al. (1998) in relation to seed size, shape and vertical distribution in the soil.

Implications for conservation and restoration

The periodic drying and rewetting characteristic of soils with Vertic horizon can cause cracks down to a depth of 1 m or more (Coulombe, Wilding & Dixon, 1996). This crack-formation may be a major driver of the vertical distribution of the seed bank. This assumption is supported by the findings of Burmeier et al. (2010) who studied the impacts of desiccation cracks on soil seed bank formation in a flood-meadow ecosystem. They found that the distribution and composition of the seed bank were different in samples taken along cracks and in samples taken beside them, as these cracks acted as natural seed traps. Desiccation cracks also had an effect on the vertical translocation of the seeds. The survival rate of these entrapped seeds increased with increasing depth, which could imply that in combination with other abiotic factors desiccation cracks cause a selection pressure on plant species to develop long-term persistent seed banks. Furthermore, their study showed that desiccation cracks influenced the establishment of the seed bank, and therefore, were an important driver of community assembly and dynamics in flood-meadows.

Our findings draw attention to the potential ecological importance of the deeply buried seeds in Vertisols and other reference soil groups with Vertic horizons. Our results originate from a single region and ecosystem type, and the number of samples was also limited by the field and greenhouse capacities. Despite the limitations in the sampling strategy, our results show that germinable seeds in Vertisols occur in deeper layers than previously thought. As Vertisols cover 335 million hectares worldwide (IUSS Working Group, 2015) and further extended areas are covered by soils with Vertic horizons, these findings may be relevant for many regions and ecosystems, such as grasslands, savannas, shrublands and open forests, globally. Vertisols occur in the tropic or semiarid regions in Australia, the Americas, India and Northeastern Africa (Coulombe, Wilding & Dixon, 1996). To test the generality of our findings, studies are needed in other regions and ecosystems with Vertisols on the vertical distribution of the soil seed bank. We highlight the need for similar studies in other soil and habitat types to test whether the presence of deeply buried seeds is a specific characteristic of Soils with Vertic horizons.

There are different views about the role of the persistent seed banks in the formation of species richness in stressed communities, such as alkaline grasslands (Valkó et al., 2014). Some researchers found that clonal spread has a greater role in stressed environment than sexual reproduction, thus, persistent seed bank is less important for habitat restoration and maintaining diversity (Chang, Jefferies & Carleton, 2001). In contrast, others showed that despite the temporarily unsuitable environmental conditions, species are able to germinate and establish from the persistent seed bank; therefore, persistent seed banks may play an important role in the vegetation dynamics of communities in stressful environments (Bossuyt & Honnay, 2008; Valkó et al., 2014). Our results support the latter findings and imply that the persistent seed bank in the deeper soil layers of alkaline grasslands can be a promising source for habitat restoration even many years after degradation.

Our knowledge on the depth distribution of seed banks and the ecological significance of deeply buried seeds is still limited. Our current study suggests that deep soil layers might contain more germinable seeds than previously thought. Our results may (1) contribute to the better understanding of the concept of seed bank persistence; (2) provide information about the past vegetation of a given area, about the past land use practices and past abiotic and climatic changes; (3) provide a better knowledge on the present conservation value of the studied area; (4) contribute to the planning of conservation measures as well as habitat restoration design by providing more detailed information on the seed bank dynamics of a given area.

Based on these results, the question arises whether these deeply buried seeds have a chance for germinate. There are several natural and anthropogenic processes by which the deeply buried seeds may return to the soil surface (Douh et al., 2018). For instance, the activity of the soil fauna–especially the movements of soil-dwelling rodents–may enable buried seeds to return to the surface (Reichman, 1979; Valkó et al., 2021). Soil layers are mixed during building and construction (such as road and railway construction and mining) and agricultural activities (such as ploughing) that can bring deeply buried seeds to the surface. These activities often occur in the studied landscape, for instance via the establishment of drainage ditch systems, soil-filling of channels and restoration of landscape scars (Deák et al., 2015; Valkó et al., 2017). It has been observed in areas affected by such activities that salt-tolerant species (Suaeda spp., Salsola soda) absent from the vegetation of the actual landscape, reappeared in the restored sites (Deák et al., 2015; Valkó et al., 2017). These observations also suggest that through the mixing of soil layers, deeply buried seeds may play an active role in vegetation dynamics. Besides the unintended mixing of soil layers, restoration projects might actively activate deeply buried seeds by targeted topsoil removal. Future studies need to clarify whether and how the deeply buried seeds can be used in restoration projects.