Six accessions collected from different climatic regions and elevations of China (Table 1) were transplanted in the field: two of M. floridulus (flo1 and flo9), two of M. sinensis (sin19 and sin20) and two of S. arundinaceum (aru22 and aru30).

The field experiment was conducted from 2009 to 2011 at the experimental farm of Zhejiang A & F University, Hangzhou, Zhejiang province, China (30.23 N, 119.72E). The soil at the experimental site was sand clay loam with a pH of 5.9 and available N, P, and K content of 104.2, 4.1, and 82.6 mg/kg, respectively.

Seven replicates of each accession were planted by hand in March 2009, equidistant in one plot (one plot=21 m²), in rows 50 cm apart, with a planting density of 2 plants/m², resulting in 42 plants/plot. Each plot was replicated four times. The plots were separated by a distance of 1 m in a randomized design. Plants were watered immediately after planting and mechanical weed control was carried out once during the early-season growth period from March to May in 2009. No fertilization or irrigation was applied during the experiment.

Monthly rainfall and air temperature were obtained from the meteorological station nearest to the field (Fig. 1). The growing season air temperatures in 2009, 2010 and 2011 were similar, with the highest temperatures (32-35°C) recorded in July and the lowest temperatures (~ 0°C) recorded in January. The total rainfall varied along the three years. The second year, 2010, was relatively wet (1545 mm) compared to 2009 (1451 mm) and 2011 (1387 mm).

Four growth traits of transplanted plants (plant height, culm number, tuft diameter and culm diameter) were recorded monthly to investigate their dynamics from April to November in 2010 and 2011. Plant height was recorded as the distance between the soil surface and the top of leaves from the tallest culms. The culm number was restricted to those of more than 10 cm. The tuft diameter was measured at ground level. The underground rhizomes of the plants expanded horizontally in multiple directions and formed different size patches of shoots that emerged during the growth season. Therefore, the tuft diameter was recorded on the mean of three distances between two longest outer shoots of the clump that extend outward plagiotropically. The culm diameter was determined from caliper measurement at ~ 5 cm aboveground of four randomly chosen culms per plant. All measurements were made on 4 plants per plot. Flowering date was defined as the date when at least one culm had one panicle 1 cm long.

The agronomic traits of each accession were evaluated for response to growing degree-days (GDD), which is a heuristic tool used to predict plant development stages. The plant development stage was defined by the date of the last frost in the spring to the date of harvest (McMaster & Wilhelm, 1997). GDD was determined using the following formula:

For aboveground biomass estimation, the whole plants were cut by hand in November at ~ 5 cm from the ground, and they were dried at 80°C until a constant weight was achieved. The aboveground biomass yield was assessed annually as tons of DM per hectare. Fresh weight (FW) and dry weight (DW) were calculated and water concentration was determined as:

Water concentration (%) = 100 × (FW – DW) / FW.

Comparisons of mean values of growth traits were made using ANOVA on four replicates per accession and per plot. The significance of variables and least-squared means were determined using the F test (p<0.01). For multiple comparisons between means, Student-Newman-Keuls a, b, c tests were used (p<0.01); additionally, the biomass yield response to each variable was also analyzed to understand the underlying basis of the interactions. All the data were processed and analyzed using SPSS version 18.0 (SPSS Inc., Chicago, IL, USA).

Figure 2 shows the measurements from 2009 to 2011 of the four agronomic traits of Miscanthus and Saccharum species. Miscanthus and Saccharum had similar growth patterns of plant height. Most individual accessions were taller in 2010 and 2011 than in 2009, but there was no significant difference between the overall means for 2010 and 2011. Only the plant height of accession ‘flo1’ increased significantly, from 156 cm to 269 cm, from 2009 to 2011.

The culm diameter of most accessions reached their maximum values in 2010 (Fig. 2). M. floridulus had the greatest culm diameter throughout the growth period, over 1.1 cm at the end of the growing season in 2010, while the culm diameters of M. sinensis and S. arundinaceum were only 0.66 and 1.0 cm, respectively. The new shoot spread was smaller in 2011 than in 2010.

Although culm numbers showed the same pattern in five of the six tested accessions, i.e. increasing progressively over the years, different rates of increase among the species could be observed (Fig. 2). For example, culm numbers increased slowly in S. arundinaceum (‘aru22’ and ‘aru30’) in comparison with the other species, M. floridulus and M. sinensis. The culm numbers of ‘sin20’ and ‘flo9’ increased significantly in each of the three years, while those of ‘sin19’ and ‘flo1’ did not increase significantly from 2010 to 2011. The mean value of culm number of S. arundinaceum accessions was lower (24/plant) than those of M. floridulus and M. sinensis (82 and 78/plant, respectively) in 2010. The same situation was also observed in 2011.

The tuft size was mainly controlled by the number of culms. The tuft diameter grew with increasing culm number. However, the tuft diameters in 2011 were significantly different among species. M. floridulus (‘flo1’ and ‘flo9’) exhibited a significant increase in tuft diameter over time because their rhizomes expanded laterally and formed larger patches than the rhizomes of the other species (Fig. 2). S. arundinaceum exhibited an almost stable tuft diameter, with the least new culms. The mean tuft diameters in 2010 were 43 cm for M. floridulus, 27 cm for M. sinensis and 18 cm for S. arundinaceum. In the third year (2011), the mean tuft diameter of S. arundinaceum accessions was also lower (21cm) than that of M. floridulus and M. sinensis (58 cm and 32 cm, respectively) in 2011.

Figure 3 shows the DM aboveground yield of the three species for the three years. The yields of all three species increased from 1st to 2nd year, but the rates of increase differed greatly. The mean DM yields for the three species were very poor in the first year (4.18 t/ha for M. floridulus, 0.92 t/ha for S. arundinaceum, and 0.86 t/ha for M. sinensis). Miscanthus floridulus, especially ‘flo1’, had superior aboveground DM yield when compared to the other species in the latter two years. Its yield increased quickly and was several times greater than that of S. arundinaceum or M. sinensis. The mean DM yield of M. floridulus increased to 24.16 t/ha in 2010 and 29.01 t/ha in 2011, while S. arundinaceum had DM yields of 3.81 t/ha in 2010 and 3.50 t/ha in 2011, and M. sinensis had 2.35 t/ha in 2010 and 4.32 t/ha in 2011. The water concentration of Miscanthus and Saccharum in our study ranged from 42% to 58%; however, no significant differences were observed between species and years after planting (data not shown).

Mixed mode analysis using ANOVA showed that most of the agronomic traits and biomass yields were significantly different (p<0.01) among species, years and accessions (Table 2). Variable interaction effects were observed among year-accession, year-species, and year-accession-species. The largest effects on biomass yield were attributed to species, year and then the interaction between year and species. Plant height was significantly affected by the year after planting, species, accession and their interaction. Culm diameter was mainly affected by species. Culm number was attributed to the year of growing, and tuft diameter variation was due to species. The results suggest that the species and the year after planting were the most important factors in determining agronomic traits and DM yield.

To understand the dynamics of the tested agronomic traits development, we observed the traits in the three species every month from 2009 to 2011. The same development patterns were found in all the agronomic traits along the three years. Therefore, we analyzed the dynamics of the agronomic traits development only in 2010.

Figure 4 shows that the height of each individual plant generally increased with increasing GDD, creating a logistic curve from the vegetative phase to the flowering phase. The plant height of M. floridulus increased rapidly between GDD=0°C and GDD=800°C, which was ascribed to internode elongation. The equilibrium value of plant height was observed at GDD ~ 1500°C in July, when M. floridulus flowered. Flowering occurred in the beginning of September for M. sinensis and in the middle of October for S. arundinaceum.M. sinensis had a vegetative growth period of 6 months and achieved its stable plant height at a GDD=2000°C in August. S. arundinaceum continued to increase in height even in November.

The culm diameters of S. arundinaceum and M. floridulus increased over time, but that of M. sinensis did not (Fig. 4). S. arundinaceum showed rapid increase of culm diameter with increased GDD. The species M. floridulus had a slow increase rate of culm diameter but produced the thickest culms in its final developmental stage. The mean culm diameter of M. sinensis was maintained at 0.62 cm throughout the plant life cycle in 2010.

New shoots of M. floridulus and M. sinensis sprouted with increasing GDD with an appreciable amount of shoots occurred from June to September (Fig. 4). M. sinensis exhibited the same growth dynamics as M. floridulus. Saccharum arundinaceum produced few shoots from its rhizomes.

The tuft diameter of M. floridulus increased clearly during the growing season, with new culms produced around the outer clump (Fig. 4). The tuft diameter of M. sinensis did not increase considerably during the year, and new shoots were established around the center of the patch. No change in the tuft diameter of S. arundinaceum was observed because of its concentrated patch and few emerged shoots.

The accumulation of biomass yield was significantly affected by the agronomical characters measured. Therefore, the correlations between yield and agronomic traits for the first three years of growth were calculated (Table 3). During the first year, the biomass yield was significantly and positively correlated with the culm diameter and culm number. The biomass yield in the second year was correlated to the tuft diameter and to the first biomass yield (R² = 0.87 and 0.94, respectively). A strong correlation (R² = 0.97) was found between the biomass yield and tuft diameter when yields of crops appeared to have peaked in the third year. There were also strong correlations between the biomass yields in the second and third year (R²=0.99) and between the first and third year (R²=0.90). Biomass yield was also strongly correlated to both culm and tuft diameter. Negative correlations were found between biomass and days to flowering for 2010 and 2011, indicating earlier flowered species tended to achieve higher biomass yield.

Although there have been several reports on the superiority of Miscanthus × giganteus over M. sinensis in Europe and USA (Clifton-Brown et al., 2001; Anderson et al., 2011), Zub et al. (2011) reported about a clone of M. floridulus that produced a similar or even greater biomass yield than M. × giganteus. Taiwanese native M. floridulus accessions collected from different altitudes produced yields of more than 20 t DM/ha/yr after transplantation in the field (Huang et al., 2011). In the present work, we have also found that M. floridulus produced an almost three times greater biomass yield than the others in the second year and third year of growth (Fig. 3). During the first three years of growth, M. sinensis and S. arundinaceum in our study produced less than 5 t DM/ha/yr biomass. These two species were thin-stalked compared with M. floridulus, and they were surrounded by weeds during the growing seasons. However, Yan et al. (2012) found that the biomass yield of M. sinensis could reach ~ 10 t DM/ha/yr in northern China.

Our results showed that M. floridulus grew better than M. sinensis and S. arundinaceum. Sun et al. (2010) described the phenological character of M. floridulus and showed that this species flowered and produced fruits from May to November. In contrast, the heading and flowering of M. sinensis and S. arundinaceum occurred in late summer (Saballos, 2008). Wild M. floridulus can grow up to 400-600 cm (Sun et al., 2010), in contrast to wild M. sinensis, which ranged in Japan 100-200 cm (Stewart et al., 2009). M. sinensis, when grown at five different locations in Europe, reached 100-230 cm in height (Clifton-Brown et al., 2001). In our study, M. sinensis grew to 180 cm (Fig. 2). Sun et al. (2010) reported that both M. floridulus and M. sinensis are widely distributed in nature, with the culm diameters of wild plants ranging 6-15 mm and 3-10 mm, respectively. In our work, at the end of the growing season, M. floridulus and S. arundinaceum achieved greater culm diameters than M. sinensis (Fig. 2), and the culm number of M. floridulus at the final stage of the second year was 80 culms/plant, lower than the number reported by Huang et al. (2011) for a crop grown in Taiwan at a density of 1 plant/m², which produced 85-112 culms/plant. Interestingly, the culm number of M. sinensis in our study was similar to that of the cultivated M. sinensis in Europe (Clifton-Brown et al., 2001). The tuft diameter was enlarged by the production of the new shoots. M. floridulus had a larger tuft diameter than M. sinensis and S. arundinaceum (Fig. 2). This observation might be explained by individual culms competition for nutrients and light in the tuft, which would maintain a suitable density of new shoots. Jezowski (2008) observed that the tuft diameter of M. sinensis hybrids could reach ~ 30 cm in the third year, which is consistent with our results.

The observed agronomic traits were influenced by growth year, species and year × species interaction (Table 2). The variation in plant height was clearly explained by the stand age and the species. The M. floridulus accessions were taller than those of the other species (Fig. 2). Maughan et al. (2012) reported that M. × giganteus plant height increased throughout the growing season in the first three years and exhibited similar growth patterns among the different environments in which it was tested. Analysis of variance showed that plant height between different species differed significantly within and between European countries (Clifton-Brown et al., 2001). There was also significant within-species variation in plant height for M. sinensis grown in different sites in China (Yan et al., 2012). The species accounted for a large portion of the total variability in culm diameter, making a four-fold greater contribution than growth year. Both M. floridulus and M. sinensis are widely distributed in nature. At the end of the growing season, M. floridulus and S. arundinaceum achieved greater culm diameters than M. sinensis did (Fig. 2). Tuft diameter was increased by new tiller spreading over the years. Due to the different growth and development patterns of emergence, tiller number was explained more by the growth year, while tuft diameter was more influenced by the species (Fig. 2). According to Stewart et al. (2009), clump growth via the generation of new outer shoots appears to cause competition for light, soil nutrients and water. M. floridulus expanded horizontally in all directions to form large patches of tillers that competed successfully with weed grasses.

Many agronomic traits influence the biomass yield. Clifton-Brown et al. (2001) reported an early-flowering Miscanthus genotype produced a lower biomass yield than a late-flowering genotype. However, Clifton-Brown & Lewandowski (2002) found that the late-flowering M. sacchariflorus had a lower biomass yield than M. × giganteus. At the inter-genotypic level, there was a strong positive relationship between biomass yield and plant height (Clifton-Brown et al., 2001; Zub et al., 2011). Yan et al. (2012) also found that the biomass yield of M. sinensis was positively correlated with plant height and tuft diameter in the second year. Our results showed that, of the species tested, M. floridulus had the greatest biomass yield in the second and third years, even though it flowered earlier (Fig. 3). Furthermore, we observed no significant correlation between plant height and biomass yield during the first three years.

Jezowski (2008) reported that culm number determined the biomass in the first and third years of growth; in the second year, tuft diameter was most highly correlated with biomass yield. Zub et al. (2011) also described a strong positive relationship between culm number and aboveground biomass yield. Our results indicated that culm diameter and tiller number were the two main variables that determined the biomass yield in the first year, while tuft diameter was the main index during the second and third year (Table 3). Zub et al. (2011) found that the year-to-year relationship between the biomass yields was high regardless of the harvest dates. Clifton-Brown et al. (2001) found that the third year biomass yield was correlated to second-year biomass yield (R²= 0.81). We also found high and significant correlations of 0.9 between the first- and third-year yields and of 0.99 between the second- and third-year yields. Therefore, the biomass yields in the first and second years of growth could be good indicators of future biomass yield.

In summary, Miscanthus and Saccharum, perennial grasses endemic to eastern China and recognized as biomass crops for many years, were successfully cultivated over first three years on a sandy clay loam in this region. M. floridulus had DM yield above 20 t/ha during the second year, greater culm and tuft diameters, and taller plant height than the other species. Therefore, M. floridulus appears to be a promising option for breeding efforts regarding biomass production in regions with a short growing season. Our results indicate that effective selection of appropriate species for breeding purposes could begin with selection for culm diameter and tuft diameter characters. The first-year biomass yield could be considered as a good predictor of the yields in subsequent years. This information can help shorten the time required for selecting high-yield species.