Seed morphology characteristics in relation to seed loss by water erosion in the Loess Plateau

Seeds

Seeds from the following 16 species were used in the experiments: Heteropappus altaicus (HA), Bidens bipinnata (BB), Dracocephalum moldavica (DM), Clematis florida (CF), Lespedeza davurica (LD), Periploca sepium (PS), Daucus carota (DC), Bothriochloa ischaemun (BI), Robinia psendoacacia (RP), Rubia cordifolia (RC), Cotoneaster multiflorus (CM), Patrinia heterophylla (PH), Sophora viciifolia (SV), Platycladus orientalis (PO), Armeniaca sibiriea (AS) and Rosa xanthina (RX). These species were selected for their variety of seed morphological characteristics and their high dominance or occurrence in the hilly-gullied Loess Plateau region.

Rainfall simulation experiments

Rainfall simulation experiments were carried out in the Rainfall Simulation Hall of the State Key Laboratory of Soil Erosion and Dryland Farming on the Loess Plateau. The rainfall simulation experiments were performed on slopes with gradients of 10°, 15°, 20° and 25° under rainfall intensities of 50 mm/h, 100 mm/h and 150 mm/h, respectively. Each simulated rainfall lasted 60 min. Three replications were designed for each experiment, and the three replications were arranged in different rainfall events. A total of 36 experiments were conducted.

The size of the experimental soil bin was 2 m × 0.5 m × 0.5 m (Length × Width × Height). The loess soil from An'sai County of the typical hilly-gulled Loess Plateau was used in the experiments, and the soil bulk density was controlled to 1.10-1.15 g/cm³. Before the rainfall simulation experiment, seeds dyed with safranin for easy identification were placed on the area of 100 cm-120 cm (from the top down bin) of soil surface. Two AS seeds, three RX seeds and five each of the other 14 species' seeds were placed on the soil surface. A total of 75 seeds were placed in each experiment.

Data collection and analysis

Seed losses

Seed loss vs. seed morphology characteristics

However, the seed size or shape cannot qualitatively explain the variation of seed loss among species. For the seeds of different species, their specific morphological characteristics affecting seed loss were quite different. Species AS and RX, with a seed weight of more than 300 mg, did not have loss in any of the experimental conditions. The DM seeds, which are small and have a rough surface, can form mucilage when absorb water and not easily be washed away, and they had no seed loss in any of the experiments. The seeds of SV and PO, which have a smooth surface and spherical shape, had seed loss rates of 80.0% and 82.3% at 150 mm/h rainfall. The seeds of HA, DC and BI, which have very small weight and density, had higher seed loss rates of 69.6%, 38.6% and 66.7% at 100 mm/h rainfall, and 100%, 80.6% and 96.5% at 150 mm/h rainfall, respectively. LD seeds, which have a smooth surface, had seed loss rates of 77.9% and 100% at 100 mm/h and 150 mm/h rainfall, respectively. PS seeds, which are obround and have hairs at the top that fall off when meeting water, had high seed loss rates of 68.3% and 100% at 100 mm/h and 150 mm/h rainfall, respectively. RP seeds, which have a smooth surface and a relatively mid-level weight and density, had seed loss rates of 18.8% and 77.8% at 100 mm/h and 150 mm/h rainfall, respectively. RC seeds, which have a bigger volume, weight and a rough surface, had lower seed loss rates of 30.0% and 66.0% at 100 mm/h and 150 mm/h rainfall, respectively. Seeds from PH and CF, which have a rough surface and flat shape, had lower loss rates of 25.8% and 11.1% at 100 mm/h rainfall, and 65.3% and 60.3% at 150 mm/h rainfall. The CM fruit, which is close to spherical but has a bigger volume and weight, only had low losses at 150 mm/h rainfall with a rate of 43.5%. The BB seed is a long achene with 4 corners and 3 or 4 thorns with persistent barbed seta, and it only had low losses at 150 mm/h rainfall with a rate of 38.5% (Tables 1 and 2).

Effect of seed morphology features on seed loss

There was a wide variation in seed losses between species, which could not be quantitatively explained by seed weight, size or shape indexes. Other features, such as mucilage secretion and the presence of appendices like as hairs or wings, may explain the differences in seed losses at the species level [9]. Seeds have specific mechanisms to avoid remobilization by water erosion, which may obscure the relationships between seed removal and seed quantitative indexes such as size or weight [9]. Combined with our pre-experiment, seed loss of some species are mainly affected by seed weight. For example, the seeds of RX, AS, Malus baccata, Ziziphus jujuba, Chinese date, Glycine max and Zea mays with weights over 190 mg were able to avoid loss, while the light seeds of HA, DC and BI had high loss rates. Some species are mainly due to seed shape, such as the elongated seed of BB and Stipa grandis had low loss rate and the winged fruit of Acer ginnala could avoid loss; while the spherical-shaped seed of SV had higher seed loss rate. Some are affected by seed surface structure, such as LD seeds, which have a smooth surface, were easily washed away. While DM seeds, having the ability to secrete mucilage under moist conditions, do not show loss in any of the experiment events. In short, seed removal effected by its specific morphology characteristics, and the seeds form different species have various responses to water erosion.

Effect of seed loss on vegetation development on eroded slopes

Based on 174 samples in the hilly-gullied region on the Loess Plateau, the TWINSPAN analysis showed the main plant communities include the dominant species Artemisia scoparia, Leymus scalinus, Lespedeza davurica, Stipa bungeana, Artemisia gmelinii, Artemisia giraldii and Bothriochloa ischaemun in different combinations, and these species have higher cover and frequency [17]. For these seven species, We found that the seeds of Artemisia scoparia, Artemisia gmelinii and Artemisia Giralaii also have mucilage secretion and not easily to be washed away by water erosion. The mucilage mass can also facilitate seed germination and seedling development [18]. This may be one of the reasons these three Artemisia plants are the dominant species on the Loess Plateau. A. scoparia is the pioneer plant in the early restoration of abandoned land which undergoing severe soil erosion. A. scoparia has high seed yield (0.2-5.3 × 10⁵ seeds m^-2) with very small size (0.02 mg), and is easily propagated by wind and buried into soils. In the natural vegetation of the hilly-gully environment, the soil seed bank was dominated by A. scoparia with a high seed density of 1 000-17 000 seed m^-2 in a 0-10-cm soil layer [19]. While Leymus scalinus has a low ripening rate and regenerated by rhizome with a film distribution. The seeds from S. bungeana, having bur with elongated shape, had low seed loss rate. S. bungeana is the main dominant Gramineae species in the area with a higher cover (20-70%) and higher occurrence. S. bungeana, S. bungeana + LD, A. gmelinii + S. bungeana, and A. giraldii + S. bungeana are the common communities in the area. Apart from seminal propagation, another special mode of reproduction for S. bungeana is by the bulbiet falling off the parent plant to develop a new plant [20]. The seeds of LD and BI had higher seed loss rate. In the eroded slopes, LD and BI have soil seed banks with a density of 50-300 seeds m^-2; however, the seed density can reach 300-1,000 seeds m^-2 in silted areas [19]. LD also has a large root system and strong shooting ability. While BI is the dominant species of the perennial herb and subshrub community in the middle and late stages of vegetation restoration. The seeds of BI are small, and easily washed away by overland flow. Moreover, the viability of BI seedlings is low; BI seedlings are weak due to the competition among early subshrub species for limited environmental resources, and sporadically and filmily distributed in the hilly-gullied Loess plateau. Short rhizome generation is used by BI as an ecological adaption strategy. It was observed that BI mainly relies on clonal propagation to achieve population renewal in the natural meadow [21]. Thus, vegetative propagation also plays a very important role in vegetation development in the Loess Plateau region.

For the other species, DM can be found in eroded slopes, such as early abandoned lands and new roadsides; Stipa grandis also had no seed loss in our pre-experiment and can be found in steep slopes; PS can form sparse communities on lower slopes because the PS seeds are easy to transfer by water erosion; SV is shrub with a partial and scattered distribution in the middle and later stages of vegetation succession on the Loess Plateau, and can be found in lower inter-gullies and in the gullies. It seems that seed loss has effect on vegetation distribution. Moreover, in the eroded slopes in the Zhifanggou catchment in the typical hill-gully area on the Loess Plateau, the density of soil seed bank in the lower silted places is significantly larger than that in the gap between vegetation; and the vegetation tussock and the silted area can hold up the seed of most species, while the gap vegetation loses the seed of many species and is dominated by the pioneer plant Artemisia scoparia [19].

Above all, seed removal is likely to affect seed redistribution and deposition and, consequently, vegetation development and distribution, while seed removal effected by seed morphology characteristics. Thus, the species composition of stable communities on eroded slopes may provide references for species selection in vegetation rehabilitation for soil erosion control. However, the seed removal in this study was conducted on bare loess slopes, and the seed removal under natural conditions in the Loess Plateau region needs to be further investigated.