10.5061/DRYAD.N2Z34TMSW
Vieira, Mariana
0000-0002-8996-728X
Federal University of Rio Grande do Sul
Overbeck, Gerhard
0000-0002-8716-5136
Federal University of Rio Grande do Sul
Small seed bank in grasslands and tree plantations in former grassland
sites in the South Brazilian highlands
Dryad
dataset
2020
vegetation regeneration
soil seed bank
Land-use change
National Council for Scientific and Technological Development
https://ror.org/03swz6y49
477618/2013-8
2020-03-04T00:00:00Z
2020-03-04T00:00:00Z
en
73406 bytes
2
CC0 1.0 Universal (CC0 1.0) Public Domain Dedication
The soil seed bank can be an important source for vegetation regeneration,
and data on the similarity between aboveground vegetation and the seed
bank can provide information about successional pathways after
disturbances or land-use change. We conducted this study in natural
grasslands in the subtropical highland region in southern Brazil. We
evaluated the effect of silviculture on richness, density, and composition
of the seed bank at former grassland sites converted to pine plantations
25 years ago. We worked at six grassland sites and three pine plantation
sites and used the seedling emergence method. Seed bank density and
richness in grasslands was lower than those reported in similar
environments in other regions. Species richness and density varied
considerably within each vegetation type therefore, richness and density
were not statistically significant, while composition varied among
vegetation types. In terms of species, the pine plantation seed bank was a
small subset of the grassland seed bank. Seeds of typical grassland
species were missing in the pine plantation, but also had only low
abundances in the grassland, and similarity of seed bank and vegetation
was low (less than 20%). The low seed density found in this study,
including in grasslands areas, indicates that regeneration of species from
the soil seed bank likely is of a limited role for the maintenance of
plant populations after disturbances in this system. Our data further
suggest that natural regeneration after tree planting in grasslands is
reduced due to seed limitation.
STUDY AREA—Our study sites are located in the highland grassland region in
the southern part of Brazil’s Atlantic Forest domain (29°04’12’’ S,
50°00’49’’ W). Regional climate is Cfb according to Köppen climate
classification, and altitude approximately 1000 m. Mean annual temperature
is 15°C and mean annual precipitation is 1881 mm (climate-data.org). The
region is a plateau formed by basalt, rhyolite and rhyodacit rocks of
Serra Geral formation. Soils are classified as Cambisoils according to
FAO, 1997 (Cambissolos in the Brazilian classification; Embrapa 2013).
Natural vegetation in the region is composed of mosaics of Araucaria
forest, cloud forest and grasslands (Leite & Klein 1990). These
highland grasslands have been used for livestock grazing since European
colonization. However, the presence of large herbivores – today extinct –
even before the arrival of native American people is confirmed by the
fossil record in the region (Scherer et al. 2007). Based on charcoal
records from peat bogs, we know that fire has been rare during the Glacial
maximum but became more frequent at the beginning of the Holocene (Behling
& Pillar 2006). Today, fire, usually every other year, is used as
a management tool to remove accumulated biomass to stimulate young leaf
regrowth after winter. In terms of their floristic composition, the
highland grasslands are dominated by C4 tussock grasses such as Andropogon
lateralis Nees, Sorgastrum scaberrimum (Nees) Herter, Axonopus pellitus
(Nees ex Trin.) Hitchc. & Chase and a high representation of
Fabaceae family (Andrade et al. 2019). The region encompasses two
important national parks, Aparados da Serra and Serra Geral, and other
state and private protected areas. In the region, we find vast areas of
pine plantations, with single planting cycles of 30 years on average,
causing loss and fragmentation of natural areas (Hermann et al. 2016). For
this study, we chose six well conserved grasslands, four located in Serra
Geral National Park and two in Aparados da Serra National Park (Fig.1),
and three pine plantations established in former grasslands areas. Two of
them were in the buffer zone of the National parks, and one of them at the
edge of the park. Pine plantations were initiated about 25 years ago.
Sites were situated in three blocks, each with one pine plantation and two
natural grassland areas, with the same history and similar floristic
composition of grasslands. Distance of blocks varied from 2 to 20 km, and
areas within each block had distances of 500 to 2000 m (see Fig. 1 for
scheme of study design). VEGETATION SAMPLING—Quantitative vegetation
sampling at the grassland sites was conducted in December 2014, in 10
plots of 1 m², randomly allocated, per grassland area. Distance between
plots was approx. 50 m. Cover of all vascular species was recorded using
the Londo decimal scale (Londo, 1976). In the pine plantation areas, no
vegetation survey was conducted, as ground layer vegetation was completely
absent. SEED BANK SAMPLING AND ASSESSMENT—The seed bank study was carried
out using the seedling emergence method, which evaluates only the viable
seeds in the soil (Thompson & Grime 1979). Soil samples for the
seed bank study were collected in grasslands and current pine plantations.
Samples were collected in two seasons (spring and autumn) with the
intention of accessing both the transient seed bank and persistent seed
bank (Thompson & Grime 1979). We used five sampling points in each
study area, totaling 30 samples from grassland and 15 from pine
plantations (five per area). Distances between sampling points were
approx. 50 m. Soil samples were collected with an auger (diameter: 5 cm;
depth: 10 cm). At each sample point we collected four sub-samples which
were mixed, resulting in one composite sample per point. All sample points
were randomly selected. For seedling emergence, we used 50% of the soil
collected in the field. Soil was mixed with vermiculite (50:50), to
maintain humidity, and spread in trays (soil depth: 2-3 cm). Samples were
kept in a greenhouse with irrigation for one year and were monitored
weekly. Trays with sterilized soil were distributed among the soil samples
from the grasslands to control possible contaminations by plants dispersed
close to the experimental facilities. Emerging seedlings were identified,
counted, and removed as soon as possible. For species that could not be
identified right away, at least one specimen was transplanted into a
larger container for development of the reproductive phase, for later
identification. Most taxa (83%) were identified to the species level and
92% to the genera level. Some individuals died in the trays or
transplanted pots before identification was possible, or there was little
development of individuals impeding identification. DATA ANALYSIS—Data of
seeds per sampling point unit were converted to density (seeds per square
meter) with the aim of facilitating comparison with other studies. We
averaged seed density data from the two seasons together for each sampling
point. For statistical analysis, mean values of each studied area were
considered, resulting in six average values for the grassland areas and
three average values for the Pinus areas. For all analyses, we used
randomization tests, with 10.000 iterations. This method (also referred to
as permutation test), based on resampling, is also adequate for
multivariate data sets, such as compositional data, and has been proposed
specifically for vegetation data (details in Pillar & Orlóci
1996). Another advantage is that it does not require normal distribution
of data, while preventing robust test results (Pillar & Orlóci
1996); this also makes the method especially appropriate for our data set.
For analysis of richness and density data, we used Euclidean distance as
dissimilarity measure and for analysis of the seed bank composition chord
distance as dissimilarity measure. We analyzed composition similarities
among pine plantations soil seed bank, grassland seed bank and aboveground
vegetation on grassland areas with Sørensen’s Index (2a/2a + b + c), where
a = number of species common to both seed banks, b = number of species
unique to the first seed bank, and c = number of species unique to the
second seed bank, considering all the data set of the seed bank (two
seasons). Principal Coordinate Analysis was conducted to visualize
difference in seed bank composition between the grasslands and pine
plantations, using chord distance as the similarity measure. For all
analyses, we used the software MULTIV (Pillar, 2006). We used alpha = 0.05
as significance level.