Seeds harvested during mowing from semi-natural grasslands as an ad hoc but effective solution for grassland restoration

Seed donor site description and method of seed material harvesting

Seeds were collected at the Experimental Station of Wrocław University of Environmental and Life Sciences in Radomierz (50°54′15.1″N  15°53′58.8″E), Silesia, Poland, Central Europe. The station is located in the sub-mountain area of the Kaczawskie Mountains, with an annual sum of precipitation of 850 mm and a mean annual temperature of 9 °C. Dystrict Cambisol is the dominant soil type. The station covers 150 ha of extensively used pastures and 100 ha of meadows and arable lands, where approximately 200 Charolais cattle are bred. The livestock is grazed on the pastures during the entire growing season, and rotation pasturage is used with sporadic mowing. During field observation in June 2020, 95 plant species of grasses, legumes, and herbs were found in the area of mowed meadow where the seeds were collected (Table S1). The dominant species were grasses: Trisetum flavescens (L.) P. Beauv., Dactylis glomerata L., Arrhenatherum elatius (L.) J. Presl & C. Presl, Agrostis capillaris L., Festuca rubra L., Poa pratensis L., Festuca pratensis Huds.

The seeds were collected accidentally as remains on a mowing machine (front-mounted disc mower Kuhn GMD3125F-FF) after mowing an area of ca. 8 ha semi-natural grassland on July 14, 2020 (Fig. S1). The seeds and residuals (e.g., chuff, husks, and dirt) were collected from the mower housing using a brush, several times during each technical breaks, and stored in a cloth bag. The weight of the fresh material collected was approximately 5 kg (Fig. S2). The material was mixed thoroughly to achieve sample homogeneity. Then, the spoon method (Van der Veen & Fieller, 1982) was used for sampling.

Seed purity and seed identification

To determine the purity of the seeds (i.e., fraction of pure seed mass to residuals), 20 samples were taken, each weighing 1 g (Fig. S3). Using a stereo microscope, the seeds were manually separated from the residuals and weighed. An additional 10 samples, each weighing 1 g, were taken for species identification based on the morphology of the seeds (i.e., seed identification). The seeds were observed using the stereo microscope, and their taxonomical identification was done using the Seed Identification Guide (2018) (http://www.idseed.org) according to Kozłowski (2012).

Germination capacity

The Petri dish method was used to test the germination capacity of the collected seeds. Thirty seed samples each weighing 1 g (sampled in June 2020) were taken for germination trials (Fig. S4). Germination tests began in three seasons: July 2020 (J20), October 2020 (O20), and March 2021 (M21). For the test in summer (J20), seeds without any pretreatment were used. For the test in autumn (O20), the seeds were kept in a refrigerator at 5 °C for two weeks and −10 °C for one week. Eventually, for the third germination test in Spring 2021 (M21), the seeds were stored during the entire winter in outdoor conditions. The three methods were intended to represent typical situations: seed used immediately after collection, seed with freezing pretreatment, and seed seeded in spring after overwintering in natural conditions. In each period, the tests were performed using 10, 1-g samples in 10 Petri dishes. The percentage of germination was evaluated after 7, 14, 21, 28, 35, and 42 days for both monocotyledonous and dicotyledonous plants. To calculate the seed germination percentage, the average number of seeds in a 1-g sample was used.

Pot experiments

Two pot experiments were established to assess the species composition of the emerging seedlings. Ten, 1-g seed samples were sown in individual plastic pots sized 47 × 15 × 15 cm. The pots were filled with commercially available soil substrate (pH = 6.07; N =8.04 g/kg; P =42.25 g/kg; K =295.00 g/kg; Mg = 157.21 g/kg; C ≥ 30%), and the seeds were broadcasted on the soil. They were pressed to the soil, and not covered by an additional soil layer. The seeds were sown in two periods: July and October of 2020 (Fig. S5). Both trials were carried out in a common garden, and the pots were kept outside during the winter. Twice a week a manual watering was applied, except for the winter. In the pots with seeds sown in July 2020, the plants were cut to a height of 3–5 cm in October 2020. Identification of species in the pots with seeds sown in July was performed two times, in October 2020 and May 2021, while in the pots with seeds sown in October, the observations were done once, in May 2021.

Field experiment

A field experiment was carried out in two city parks in Wrocław to compare the seeds collected from mower with commercially used seed mixtures, and fresh hay application, whether it has the features and potential for city grassland improvement (Figs. S6 and S7). The vegetation (i.e., species composition and cover) developed from the examined mower seed mixture (MO) was compared with that emerging after applying four other seed sources: a flower meadow seed mixture (FM), a road margin seed mixture (RM), a thermophilus margin seed mixture (TM), and seeds delivered by spreading fresh hay (FH). The FM mixture represented a commercial seed mixture of 14 species of unknown geographical origin claimed to deliver food for bees produced by the SAATBAU company. The RM and TM seed mixtures were distributed by specialized company Rieger-Hofmann® GmbH, focusing on the production of seed mixtures of native species with controlled geographical origins. The RM mixture contained seeds of 57 species that are typical of semi-natural grasslands and suitable for creating road-margin vegetation. The TM mixture contained seeds of 57 dicotyledonous species characteristic of the vegetation found in thermophilus margins. Finally, the fresh hay (FH) was collected from the Radomierz station from a grassland patch where 53 plant species were recognized.

Experimental plots were arranged according to complete randomization schemes with four replicates for all treatments. The size of an individual plot was 2.5 × 2.5 m. A particular seed source was represented by eight plots (four repetitions of two parks). The entire experimental area was mowed, the resident vegetation was removed in July 2020, and the soil was prepared using a rototiller. The fresh hay was collected on July 31, 2020 and immediately transported to experimental sites for spreading. The ratio of donor to acceptor site area was 1:1, and the fresh hay layer covered the soil at a thickness of ca. 10 cm. The seeds were seeded in September 2020 at a seed rate recommended by the producer: 2 g × m-2 for TM, and 4 g × m-2 for RM, FM, and MO. Afterward, the seeded soil was rolled. In addition to the experimental plots, four additional squares per experiment were placed in the corners of the experimental area. The additional plots served as a control; they were mowed but not rototilled, rolled, nor seeded. The species composition and percentage cover of the plants were assessed visually for areas (2 × 2 m) in the center of each experimental plot in June 2021, giving a buffer with a width of 1 m between neighboring treatments.

Statistical analyses

The frequency of species occurrence in the seed identification, pot experiment, and field experiment was expressed as a fraction of the repetitions, where seedlings (or seeds) of a given species were found. The total average frequency was calculated based on the average frequencies for all trials. The significance of differences between periods of observation in the germination capacity trial was verified using the Friedman test, as well as the Wilcoxon pairwise test with a Bonferroni correction as a post hoc test. The total percentage of germination and seedling emergence in the particular seasons (i.e., J20, O20, and M21) was determined using the Kruskal–Wallis test, with Monte–Carlo permutation and Mann–Whitney test with Bonferroni correction as post hoc test. The same tests were applied to evaluate the results of the pot and field experiments. Statistical analyses were performed using Past 4.06b software.

The transfer rate was calculated as the percentage of transferred species relative to the total number of species of the donor site (Kiehl et al., 2010; Scotton, Kirmer & Krautzer, 2012). Since, in the field experiment, the presence of seeds from a local seed bank or local species pool could not be excluded, we did not consider all species found in this experiment as emerging from the evaluated seed mixture. However, if the species that occurred in the field were also detected in seed species identification and in the pot trials, we assumed that they came from the seed mixture.

Seed purity and seed identification

The 1-g samples contained on average 0.9 g (SD = 0.039) of pure seeds and 0.1 g (SD = 0.039) of residuals. The average number of seeds per 1-g sample was 1267 (SD = 151.3). At the species level, the seeds of 19 different species were recognized, the most frequent of which were Alopecurus pratensis, Dactylis glomerata, Festuca pratensis and F. rubra (for detailed results, see Table 1). The visual assessment showed that the seeds of Alopecurus pratensis were the most abundant in the samples.

Germination capacity

There were significant differences in the number of seeds germinating on Petri dishes between the observation seasons (i.e., J20, O20, and M21) and over the weeks of the experiment: for monocotyledonous species, χ2 = 138.87 and p < 0.05; for dicotyledonous species, χ2 = 36.629 and p < 0.05 (Figs. 1A–1B). The monocotyledon seeds sowed in Autumn 2020 (O20) and Spring 2021 (M21) exhibited the highest germination rate during the first week of the trial, while the seeds sown in Summer 2020 (J20) exhibited the highest germination rate during the second week (Fig. 1). However, seedling germination generally decreased over time. We did not observe significant differences between observation weeks for dicotyledonous plant germination, except for the highest germination rate in the first week of the Autumn 2020 (O20) trial.

The detailed results of total germination (i.e., the sum over six weeks) are presented in Fig. 2. The number of emerged monocotyledon seedlings differed significantly between the seasons (χ2 = 7.646, p = 0.022), while that of dicotyledonous seedlings did not (χ2 = 4.638, p = 0.084). In the spring trial (M21), after overwintering, the total germination of monocotyledon seeds was better than that of the summer trial (Fig. 2A).

Similarly, the total germination percentage after six weeks significantly differed between the seasons (χ2 = 8.12; p = 0.02, Fig. 3). The percentage of emerged seeds sown in Spring 2021 (M21), after overwintering, was higher (44.05%) than that of seeds sown immediately after collecting in Summer 2020 (J20, 39.93%) (Fig. 3).

Pot and field experiments

In the pot experiment, 31 species were found in all observation periods and sowing seasons (Table 1). We did not found significant differences in species richness between sowing periods, observation periods, or sowing × observation interactions. During the observations, a visual assessment determined that Holcus lanatus L. was the most abundant species.

In the field experiment, the plant species richness of the plots where seed addition was applied differed from the control plots (H = 21.77, p < 0.001). However, we did not observe significant differences between the plots with seeds from the mower (MO) and the plots where seeds from other sources were added (Fig. 4A). We also observed differences in vegetation cover (H = 19.74, p < 0.001): the plots where the RM mixture was applied had less cover than the control plots (Fig. 4B).

Considering the pot and field experiments, altogether, 41 plant species were found (Table 1). Typically, monocotyledons showed a high probability of occurrence, which were mostly grasses (Poaceae), while dicotyledons showed a low frequency. In the field experiment, 28 species were observed (Table 1). Among these species, grasses such as Arrhenatherum elatius (30.7%) and Lolium perenne (16.3%) had the highest cover, and Trifolium repens (4.8%) was the dicotyledonous species with the highest cover.

Transfer rate

Among the 42 species found in this evaluation, 11 occurred in all three trial types. Among them, seven species were grasses (Arrhenatherum elatius, Dactylis glomerata, Holcus lanatus, Lolium perenne, Phleum pratense, Poa pratensis, and Trisetum flavescens), and four were dicotyledonous (Leucanthemum vulgare, Rhinanthus minor, Rumex acetosa, and Rumex obtusifolius) (Table 1). In our donor area, 95 plant species were found (Table S1). During all of the three main trials (i.e., seed identification, pot, and field experiments) 40 of these species were also found at the donor site. Calculated in this way, the transfer ratio was 42%. In the field experiment, 28 species that potentially emerged from the seed mixtures were found. However, we cannot exclude the possibility of contamination by propagules within the vicinity of the experiment and/or soil seed bank. However, 21 species that occurred in the field experiment were also confirmed in other trials (i.e., seed identification and/or pot experiments). Therefore, a conservative estimate suggests a transfer rate of 22% (percentage for 21 species), which perhaps reaches 29% by taking the 28 species of unconfirmed origin into account.