Received 9 February 2016; accepted 18 July 2016; published 21 July 2016
1. Introduction
Across the world, large areas of native vegetation have been removed for urban, industrial, and agricultural land uses. Destruction of vegetation may increase soil erosion and undermine stability, and revegetation is essential for stabilizing such disturbed areas. Artificially revegetated sites were traditionally fertilized, mulched, and seeded with non-native grasses and legumes selected for rapid growth and effectiveness in erosion control [1] - [4] . In recent years, however, native species have been preferred for revegetation measures in order to prevent the spread of ecologically harmful invasive species and to improve ecological functions [5] - [8] .
Riverdikes must be quickly revegetated to prevent soil erosion, as the collapse of dikes is directly associated with severe flooding. In Japan and South Korea, riverdikes were traditionally revegetated with grass species that were managed by mowing [9] [10] . Some means of improving the cost efficiency of managing riverdikes include reducing the frequency of mowing and replacing short-grass monocultures with taller species [11] . The taller species generally have lower ability to prevent soil erosion [12] . To date, revegetation methods that are suitable to reduce soil erosion under reduced mowing frequencies have rarely been developed.
Grasslands dominated by Imperata cylindrica (L.) P. Beauv. are important components of native ecosystems that support endangered species in some regions, such as Nepal [13] , Australia [14] , and Japan [15] . At the same time, I. cylindrica is one of the most cosmopolitan grass species: it grows throughout the warm temperate and tropical regions of the world [16] . It is a highly invasive perennial grass that threatens agriculture, forestry, and native plant species assemblages in many regions of the world [16] . Rhizomes can comprise over 60% of the total plant biomass of I. cylindrica [17] ; this low ratio of shoots and roots to rhizomes contributes to its rapid regrowth after cutting, especially in disturbed habitats. Studies have indicated that its adaption to poor soils, its drought tolerance, and its prolific production of wind-dispersed seed are associated with the dominance and spread of this species [18] - [20] .
Because of its vigorous lateral spread via rhizome elongation and its adaptability to a wide range of environments, researchers have proposed that I. cylindrica should be useful for quickly covering bare ground [9] [21] . Based on this proposal, artificial sod consisting of I. cylindrica has been developed. This sod was reported to cover bare ground successfully and at high density even in the first year of introduction [22] . Other revegetation techniques using I. cylindrica seeds and planted soil plugs (i.e., both the grass and some soil) have also been developed [21] . In order to revegetate bare ground as rapidly as possible, revegetation techniques that can protect the soil should be developed; however, no scientific study has yet compared the speed of coverage by I. cylindrica among the various available methods. Chigaya-Sougen-Soushutu-Kenkyukai [23] reported revegetation techniques using I. cylindrica, yet little empirical data with enough replications was shown.
Regeneration of introduced species tends to be hampered by competition from undesirable species that often invade revegetated areas, rather than by poor germination [24] . If undesirable species increase beyond a threshold, they could shade out I. cylindrica, thereby preventing I. cylindrica from becoming dominant [25] [26] . Competitiveness is thus a particularly important trait for the survival of species in revegetation measures [27] [28] . Therefore, it is necessary to perform monitoring of how the techniques used affect the rest of the plant community, especially with regard to increases of cover by undesirable species; however, to our knowledge, no such study has been carried out.
To identify an appropriate method to revegetate riverdikes, we set out to answer two questions: What is the fastest method (sowing, planting, or sodding) to revegetate riverdikes with I. cylindrica in terms of the effects on plant cover and species richness of the established vegetation? In each method, do natural colonizers become established over time?
2. Materials and Methods
2.1 Study Site
The study site is located in the high-water channel of the Tone River, 30 km northeast of Tokyo, Japan (35˚54'54"N, 140˚0'59"E, Figure 1). Mean annual precipitation at the nearby Ryugasaki Meteorological Station is 1344 mm. The mean annual temperature is 14.1˚C, with a mean minimum of 3.1˚C (January) and mean maximum of 25.6˚C (August). The site was surrounded by tall herbaceous vegetation dominated by Miscanthus sacchariflorus (Maxim.) Benth., and no vegetation management was carried out. The neighboring riverbank, which is located approximately 20 m from the study site, was dominated by Solidago altissima L., Schedonorus
phoenix (Scop.) Holub, Lolium multiflorum Lam., and I. cylindrica.
2.2. Experimental Design
The study site is within an area where soil is prepared for the construction of riverdikes. Soil hardness measured with a Yamanaka-type soil penetrometer (Fujiwara Factory Co. Ltd., Tokyo, Japan) was 19 - 28 mm (median 24 mm), which is similar to that in established riverdikes (personal communication with H. Sasaki, Foundation of River and Watershed Environment Management, Japan). Available nitrogen content in the surface soil was 0.1 mg N/g, indicating relatively nutrient-poor soil conditions.
Topsoil in a 10 m × 10 m area was removed to a depth of 30 cm in July 2010. Within this area, we set up five treatments: sodding, sowing at two seed densities, transplanting, and control. Treatments were carried out in 1.5 m × 1.5 m (2.25 m2) plots arranged in a randomized block design, with three replications of each. Each block was separated by 20 cm deep ditches to avoid rhizome contamination between plots. Each plot was placed 70 cm away from an adjoining plot.
For the sodding treatment, I. cylindrica sod composed of 12-month-old plug-plants that was commercially developed for riverbank revegetation was raised for 3 months and then applied to the plots in 1.1 m × 1.1 m squares. For the two sowing treatments, either 10,000 seeds m−2 (hereafter high-density sowing) or 1000 seeds m−2 (low-density sowing) were sown. The I. cylindrica seeds were obtained from the nearby riverbank in late May 2010. For the transplanting treatment, 12-month-old plug-plants raised in 1 cm × 1 cm cell pots were established at a density of 25 individuals m−2. In the sowing treatments, seeds were integrated with stabilizer (2 cm3 per 2.25 m2 of acrylic acid resin and polyvinyl acetate resin) and short-fiber wood mulch (405 g per 2.25 m2) before being broadcast in order to add weight and achieve better adherence to the soil. Plug-plants and sod were produced by ESPEC MIC Inc. (Nagoya, Aichi, Japan). In the control treatment, the ground was left bare.
Because germination of I. cylindrica requires a relatively high temperature [29] , the experiment was started at the beginning of July 2010. Although the experiment was started during the rainy season, manual irrigation was applied daily for the first 2 weeks after planting, and then twice weekly until 30 September 2010. Mowing management was carried out twice a year, in May and October (with the first mowing occurring in May 2011).
2.3. Monitoring
Seedling establishment success was assessed by counting germinated plants in 0.1 m × 0.1 m quadrats randomly located at five non-overlapping points within each sowing plot. For the transplanting treatment, the number of surviving plants was recorded in the whole 2.25-m2 plot. Monitoring was carried out on 9 September 2010. For monitoring species diversity, we measured the maximum height and visually estimated the percent cover of each species on 28 October 2010, 23 August 2011, and 3 September 2012. Total vegetative cover was also estimated on these dates. The survey in 2010 was delayed because vegetation was mostly composed of small seedlings, making species identification difficult for many individuals. The vegetation survey preceded mowing in each autumn. Species identification was made in the study site, supplemented by subsequent room identification using Flora of Chiba Prefecture [30] .
To assess belowground biomass, soil was sampled to a depth of 30 cm using an auger with a 4-cm diameter on11 April 2013. In each plot, the center and four points equidistant from the center and each edge were sampled. Each soil sample was separated into 0 - 10, 10 - 20, and 20 - 30 cm portions, and the upper, middle, and lower portions from the five sampling points within each plot were each combined. Live plant parts were removed from each soil sample, dried for 3 days at 80˚C, and weighed.
2.4. Analysis
The recorded species were classified as annuals (including biennials), perennials, and woody species. The description of species attributes followed Numata and Yoshizawa [31] and Chibaken-shiryou-kenkyuzaidan [30] . Species diversity was calculated using the Shannon-Weaver diversity index (H0) [32] .
Cover data were normalized by square-root transformation. We performed one-way ANOVA and Tukey’s post hoc tests to determine the significance of differences in the number of species, plant cover, and belowground biomass between treatments. We also performed repeated two-way ANOVA and Tukey’s post hoc tests to assess the differences in plant cover and number of species between years. A P value < 0.05 was considered to be significant. Statistical analyses were performed using R ver. 2.13.1 (R Development Core Team 2013).
3. Results
A total of 89 species were recorded. Fifty species were annuals, 37 were perennials, and 2 were woody species. Forty-two taxa were exotic species. Equisetum palustre L. was observed at frequencies of more than 0.5 (i.e., in more than 7 plots) in each year. Panicum bisulcatum Thunb., Cyperus iria L., and Cayratia japonica (Thunb.) Gagnep. were observed at frequencies of more than 0.5 in year 1, and Erigeron canadensis (L.) Cronquist and Solidago altissima were observed at frequencies of more than 0.5 in years 2 and 3.
3.1. Germination and Survival Rates
Mean seedling survival rates were similar in the two sowing treatments (high density: 24.6% ± 8.8%; low density: 30.7± 4.2%). The survival rate in the transplant treatment was 41.3% ± 31.0%.
3.2. Floristic Changes in the Various Treatments
Imperata cylindrica cover, total plant cover, and total number of species recorded in each experimental year are illustrated in Figure 2. Changes in species categories among years are described in Table 1. In year 1, the sodding and high-density sowing treatment produced significantly greater I. cylindrica cover, followed by low- density sowing and transplanting. In year 2, I. cylindrica cover was similar in the low-density sowing, high- density sowing, and sodding treatments, whereas the transplanting treatment had significantly less I. cylindrica cover. The relationship among treatments with regard to total plant cover was similar to that for I. cylindrica cover in years 1 and 2. No significant differences in total plant cover were observed among treatments in year 3, with all treatments having more than 80% cover.
In year 1, the sodding treatment exhibited a significantly smaller total number of species than all other treatments. Between years 1 and 2, the total number of species increased in the transplanting treatment, but decreased in the sodding and two sowing treatments (Table 1). Between years 2 and 3, the total number of species decreased in the low-density sowing and transplanting treatments and control, but increased in the sodding treatment. Consequently, by year 3 the sodding treatment had achieved a significantly greater total number of species than the other treatments, except the control.
The numbers of species and cover of annuals and perennials are shown in Table 2 (The number of woody species was too small to include in this analysis). In years 1 and 2, the numbers of annuals in the higher-density sowing and sodding treatments were smaller than those in the transplanting treatment and control. The number of annuals decreased over time in all treatments, except for the sodding treatment in year 3 (Table 1). By year 3, there were almost no annual species in the plots except for the sodding treatment. There was no clear trend among years with regard to the number of perennials, but there were more perennial species than annual species in years 2 and 3 in all treatments. Likewise, the percent cover of perennials was greater than that of annuals in
Table 1. Between-year trends in each species category for the various revegetation treatments.
aCover of perennials excludes that of Imperata cylindrica.
Figure 2. Total plant cover (upper), Imperata cylindrica cover (middle), and total number of species (lower) in each treatment from year 1 to year 3. Mean values ± SD followed by the same letter are not significantly different according to Tukey’s LSD (P > 0.05).
years 2 and 3 in all treatments (Table 1), particularly in the transplanting treatment and control. The percent covers of exotic species in year 3 were greater than those in year 1 in all treatments (Table 1). In year 3, the control treatment showed significantly greatest cover of exotic species among treatments. Shannon’s diversity index was greatest in the transplanting treatment and control in year 2, and in control in year 3 (Table 2).
3.3. Belowground Biomass
Belowground biomass is illustrated in Figure 3. At soil depth of 0 - 10 cm, biomass in the sodding treatment was significantly greater than those in the transplanting treatment (P = 0.034) and control (P = 0.007). There were no significant differences among treatments at depths of 10 - 20 and 20 - 30 cm.
4. Discussion
In vegetation studies with small plots, interplot seeding (i.e., seed rain from adjacent plots) typically becomes a problem after the second season [33] [34] . In the present study, I. cylindrica occurred in control plots in year 2.
Table 2. Number of species and cover of annuals and perennials.
Cover of perennials excludes that by Imperata cylindrica. Within each row, mean values ± SD followed by the same letter are not significantly different according to Tukey’s LSD (P > 0.05).
These individuals were likely recruited from seeds produced in adjacent plots and surrounding habitats via the prolific wind-disseminated seed. Nevertheless, cover of I. cylindrica was significantly lower (<10%) in the control compared with other treatments, suggesting that the influence of interplot seeding was marginal.
4.1. Behavior of Imperata cylindrica
In all treatments, I. cylindrica cover reached more than 70% or increased throughout the experimental period. The sodding and higher-density sowing treatments achieved the most rapid increase in the cover. Sodding of turf grasses has been shown to cover the ground more rapidly than broadcast seeding [30] [35] . One reason for the better performance in terms of the abundance of I. cylindrica in the high-density sowing treatment would be the
Figure 3. Belowground biomass in each treatment in 2013. Mean values ± SD followed by the same letter are not significantly different according to Tukey’s LSD (P > 0.05).
much greater seed density than that used in previous research [36] . If we compare sodding with the low- density sowing treatment, then our findings correspond to these previous results. In another study of I. cylindrica, Tominaga [37] reported that each new plant borne from a rhizome piece (ramet) produced 31 rhizomes totaling 8 m in length in a single season. In this study, the number of plug-plants originally introduced into a sod roll was small (i.e., 25 individuals m−2), but the 12-month maturation period before the introduction of sod would have enhanced the rapid expansion of rhizomes and thus contributed to the increase in the cover.
It was only in the first year that cover in the low-density sowing treatment was less than that in the high-den- sity treatment. A similar trend was observed by Stevenson et al. [36] and Burton et al. [38] . In high-density sowing, individual seedlings experience intense competition, resulting in some mortality and self-thinning [38] . Another explanation for this result is that, although plant cover of natural colonizers was greater in the low- density sowing treatment in year 1, the majority of the cover was by annuals, thus more space was available for I. cylindrica in the low-density plots in the following spring.
The transplanting treatment had the lowest I. cylindrica cover in all treatments, except the control. One reason is clearly that the transplant survival rate in year 1 was lower than expected (41.3% ± 31.0%). Tominaga [37] reported that I. cylindrica ramets (new plants from rhizome pieces) produced fewer rhizomes than genets (seedlings). Ezaki et al. [39] demonstrated that when planted at 16 individuals m−2, I. cylindrica plant density was saturated by the fourth year. Therefore, it may be the case that it takes longer for I. cylindrica to become dominant when established by plug-plants rather than by seeds and sod. Imperata cylindrica favors high-light environments, and Patterson [17] reported a markedly higher mortality rate in shaded environments. We observed a higher percent cover of perennials in the transplanting treatment than in the sowing treatments. Under such conditions, I. cylindrica cover would not likely increase to form a monospecific stand over the short experimental period.
Belowground biomass in the transplanting treatment and control were smaller than that in the other treatments. This pattern more closely resembled that of I. cylindrica cover than total cover. Although we did not distinguish among species when assessing belowground biomass, the roots and rhizomes of I. cylindrica likely accounted for much of the belowground biomass. The belowground biomass recorded in this study was markedly smaller than that reported in mature plant communities dominated by I. cylindricaon riverdikes [9] , reflecting that belowground biomass in our study site was still increasing in year 3.
4.2. Effect on Natural Colonizers
Compared with the control, the cover of annuals was suppressed in the sodding and sowing treatments, particularly in years 2 and 3. The number of annual species also decreased to less than 2.0 in the high-density sowing treatment by year 2 and in the low-density treatment in year 3. The presence of annuals was strongly linked to the presence of bare ground in plots, as reported previously [36] . Closed and productive vegetation dominated by perennial generalists also hampers the chances for the establishment of migrated species [40] [41] . Previous studies found that the suppression of migrated species was high only when grass species were made up at least 70% cover [33] [42] [43] , and it was low when the proportion of grasses was 50% [36] . Plant cover in the treatment plots was >70% in years 2 and 3, suggesting that competitive exclusion was occurring within a few years of initiating the treatments. For the transplanting treatment, Shannon’s diversity index and the total number of species were similar to those in the control over the course of the experiment, indicating that there was no clear competitive exclusion due to I. cylindrica in the transplanting treatment.
In year 3, the covers of exotic species in treated plots were significantly larger than those in control plots, suggesting that introduction of I. cylindrica regardless to the treatments contributed to inhibiting the increase in the cover of exotic species. The low percent cover of undesirable exotic species (e.g., Solidago altissima) after 3 years suggests that establishment of such species is unlikely to become a major problem in a short time. Nevertheless, the cover of exotic species increased gradually, indicating that it is unclear whether exotic species suppress I. cylindrica in the long term, as Mitchell et al. [28] reported. Further work is required to identify method(s) to control these perennial species.
The number of species in the sodding treatment was significantly smaller than other treatments in years 1 and 2, whereas the sodding treatment had the greatest number of species in year 3. Plant cover was similar in the sodding and high-density sowing treatments. By covering the existing seed banks with sod, weed germination and establishment would be eliminated [34] . However, the artificial materials contained in the sod deteriorate by year 3, the deterioration opens gaps in the sod that let light hit the underlying soil and thus enhance germination
4.3. Implications for the Revegetation of Riverdikes Using Imperata cylindrica
Sodding and sowing with I. cylindrica suppressed both the total cover and the number of naturally colonizing species. If stabilization and erosion control are the priority, introduction of I. cylindrica using sod and sowing at high density were shown to be the most effective techniques. Sod produced greater belowground biomass than the other treatments, indicating that sodding is particularly well-suited to bare-ground revegetation of riverdikes. However, our findings indicated that transplanting is inappropriate because I. cylindrica cover would not likely increase to form a monospecific stand within a short period.
Land managers often attempt not only to achieve rapid coverage by vegetation, but also to increase the functionality of the target communities [44] [45] . Thus, ideally, revegetation using I. cylindrica should also enhance the establishment of other desirable species. Nevertheless, except for I. cylindrica, only Rumex acetosa, C. japonica, and Plantago asiatica were common to existing semi-natural grassland (Nemoto et al., unpublished data), possibly due to seed limitation of the target species, as noted in previous studies (e.g., [46] ).
The most widely applicable method for actively restoring a diverse plant community is sowing seeds of numerous species [47] . When considering such goals, sod is less useful due to the severe limitation on the colonization of migrating species. However, Abe et al. [48] reported successful introduction of several grassland species on reclaimed riverdikes by sowing seeds. Therefore, using a seed mixture of I. cylindrica and diverse desirable species would likely provide available resources for new seedlings. If immediate green-up is not imperative, sowing with a lower seed density will provide equivalent levels of cover over time. When revegetation is performed on the side of a riverdike opposite to the stream, there is not an extreme demand to prevent soil erosion. In such locations, sowing with lower seed density (1000 individuals m−2) should be suitable, as it will achieve equivalent levels of cover by the end of two growing seasons and save some of the expense of gathering regional seeds of I. sylindrica.
Acknowledgements
This research was conducted in collaboration with the Foundation of River and Watershed Environment Management, Japan. The experiment complied with the current laws of Japan.
NOTES
*Corresponding author.