Harvest date and leaf:stem ratio determine methane hectare yield of miscanthus biomass

The suitability of miscanthus biomass for anaerobic digestion has already been confirmed by several studies. However, it is rarely used as feedstock in biogas plants, mainly due to uncertainty about the optimal harvest regime with regard to the long‐term methane hectare yield and resilience of the crop to green cutting. The recommended green‐cut date for the only commercially available genotype Miscanthus × giganteus (M×g) ranges from September to November. This timeframe is too broad for agricultural practice and needs to be both narrowed down and further specified for different genotypes. The aim of this study was to identify the most suitable harvest window for an autumn green cut of miscanthus, which delivers both a high dry matter and methane yield while securing the long‐term productivity of the crop. A further objective was to quantify the effect of genotypic differences, such as leaf to stem ratio, on the substrate‐specific biogas and methane yield. For these purposes, a field trial with four genotypes (M×g, GNT1, GNT3, Sin55) was conducted over 2 years (2016/2017) and harvested at 2‐week intervals on three dates between mid‐September to mid‐October. Methane hectare yield ranged from 3,183 m³ CH4 ha−1 a−1 (Sin55) to 5,265 m³ CH4 ha−1 a−1 (M×g), which is mainly influenced by dry matter yield. The substrate‐specific methane yield was higher for the leaf (311.0 ml CH4 (g oDM)‐1) than the stem fraction (285.1 ml CH4 (g oDM)‐1) in all genotypes due to lower lignin content of leaves. Of all genotypes, M×g showed the highest and Sin55 the lowest nutrient use efficiency. We conclude that miscanthus in Germany should be harvested in October to maximize methane yields and nutrient recycling and minimize yield reduction. Additionally, to increase methane hectare yields even further, future miscanthus breeding should focus on a higher leaf proportion.

ha −1 year −1 . This is within the range of maize, the most common biogas crop in Germany, and therefore an important benchmark for all novel biogas crops (FNR, 2017). A similar amount of agricultural land is required for miscanthus as for maize cultivation to supply a biogas plant (Kiesel, Wagner, & Lewandowski, 2017).
Despite these positive results, miscanthus is rarely used as a substrate for anaerobic digestion in Germany. One reason may be the high establishment costs caused by the labour-intensive and expensive rhizome propagation. However, several studies have shown that a decrease in establishment costs can be expected in the years to come through novel establishment techniques such as seed and collar propagation (Clifton-Brown et al., 2017;Mangold, Lewandowski, Xue, & Kiesel, 2017).
Another reason is the uncertainty about the extent to which the necessary green cut in autumn affects the longterm productivity of the crop. Miscanthus is mainly used for combustion and harvested after winter when lignin and dry matter contents are high (Iqbal & Lewandowski, 2014). This harvest date fits in well with the natural growing cycle of the crop, because it allows the relocation of nutrients to the rhizomes and recycling of nutrients via leaf-fall, thus supporting regrowth in the following year (Cadoux, Riche, Yates, & Machet, 2012).
For use in biogas plants, it is necessary to harvest miscanthus green, i.e. before winter. Later harvests are accompanied by an increase in lignin content in the biomass and thus a decrease in digestibility (Fernandes, Bos, Zeeman, Sanders, & van Lier, 2009). A green cut also leads to higher dry matter yields, as harvest occurs before leaf-fall . Schmidt, Lemaigre, Ruf, Delfosse, and Emmerling (2017) found significantly higher dry matter yields from a green cut in autumn (up to 27 t ha -1 ) than from a brown harvest after winter (22 t ha -1 ).
However, a green cut allows less time for miscanthus to relocate carbohydrates and nutrients to the rhizome, which can have a potentially negative effect on regrowth in the following year. Fritz and Formowitz (2010) and  showed that an early green cut before or in August negatively affects the long-term productivity of miscanthus. Therefore, for Central European climate conditions, a harvest in September to October is recommended, as both high methane hectare yields and a sufficient green-cut tolerance can be achieved . Ruf, Schmidt, Delfosse, and Emmerling (2017) found that a green cut in late September negatively affected the yield the following year, but no physiological effects of a green harvest in November. This range of recommended harvest times (September to November) is too broad for agricultural practice and needs to be further refined.
Biomass quality is not only influenced by harvest time but also by physiological properties, for example proportion of leaf and stem biomass. This has already been established for the combustion of miscanthus biomass, with leaves being less suitable than stems due to their higher content of ash and critical elements (Baxter et al., 2014). For anaerobic digestion, it appears to be the reverse. Wahid et al. (2015) investigated six different harvest dates from August to November and found a significantly higher substrate-specific methane yield (SMY) of the leaf fraction than the stem fraction for the genotypes M. × giganteus and Miscanthus sinensis after 31 days of anaerobic digestion. In addition, they found higher specific methane yields and lower cellulose and lignin contents for M. sinensis than for M. × giganteus. However, Wahid et al. (2015) only investigated 1 year and so were unable to assess the effects on the yield the following year, which is crucial for agricultural practice. Furthermore, the effect of a green harvest on yield in the following year and also on the leaf and stem proportions of the biomass has not been sufficiently investigated.
The aim of this study was to identify the most suitable harvest window for an autumn green cut of miscanthus. This harvest window should enable both a high dry matter and methane yield while securing the long-term productivity of the crop. A further objective was to quantify the effect of genotypic differences, such as leaf proportion, on the substrate-specific and methane hectare yield.
For this purpose, a field trial with four different genotypes was conducted over 2 years and harvested three times at 2-week intervals in the period mid-September to mid-October. In addition to the standard cultivar M. × giganteus, three novel, seed-based hybrids were tested which are assumed to have improved biomass quality for anaerobic digestion. The dry matter yield (DMY), substrate-specific methane yield (SMY), methane hectare yield (MY) and fibre content were determined for the 2 years 2016 and 2017. The leaf and stem proportions were assessed for each genotype, harvest date and year. Additionally, leaf and stem nutrient contents were measured to quantify the nutrient removal of the green cut.

| Field trial
The field trial was conducted on a commercially relevant scale at the University of Hohenheim's research station "Unterer Lindenhof" in south-west Germany (48.4°latitude, 9.2°longitude; approximately 480 m a.s.l.). The location is characterized by a long-term average annual air temperature of 6.8°C and an annual precipitation of 942 mm. The soil, which is classified as a stony marl Rendzina, has a high clay and stone content, and tends to be waterlogged. It can thus be considered to be of low quality. The climate data for the field trial period (2016)(2017) are shown in Supporting Information Table S1.
In the field trial, four genotypes, M. × giganteus (M×g), GNT1, GNT3 and Sin55, were established in strip plots (width:10.5 m × length: 45.5 m) in a randomized complete block design with four replicates for each genotype, except GNT1, which only had three field replications, as the establishment of one field plot was not successful. A detailed description of the four genotypes is given in Table 1.
The planting density was two plants/m 2 with a row distance of 75 cm. The trial was not fertilized during the whole experiment. Weed control was performed annually either mechanically or chemically.
Within the strip plots, a smaller "cutting tolerance trial" was established in a randomized split-plot design, where genotype was the main-plot factor and harvest time was the sub-plot factor. This resulted in 45 plots (three genotypes × four plots × three harvest dates + one genotype × three plots × three harvest dates). Each of those 45 harvest plots had a size of approximately 12 m². The cutting tolerance trial started in the second year after establishment of the miscanthus crop and three different harvest dates were tested. The first harvest date was mid-September (HD 1; 21 September 2016; 18 September 2017), the second 2 weeks later at the beginning of October (HD 2; 4 October 2016/2017) and the third mid-October (HD 3, 17 October 2016/2017).
The harvest procedure was the same for each harvest date in both years. First, the front border of each plot was cut and removed, then eight plants were harvested from the centre of the plot with a "Baural" field trial harvester at a cutting height of 20 cm. The exact area harvested was measured to determine the fresh matter yield (FMY)/ha. To identify the dry matter content (DMC) of the chopped material, a subsample of approximately 1 kg was taken from each plot and dried in a drying cabinet at 60°C to constant weight. The dry matter yield (DMY; t/ha) was estimated based on the FMY and the DMC of the subsample.
Ten randomly selected stems were collected from the remaining borders and separated into leaf and stem fractions. In 2016, 20 stems of the genotype Sin55 were cut, as ten stems would not have given enough plant material for the analysis. The fractions "leaf" and "stem" were also dried at 60°C to constant weight to identify DMC. After the ten stems had been collected, the remaining borders of each harvest plot were also removed.

| Biogas batch test
To determine the substrate-specific methane yield (SMY) of the leaf and stem fractions, the dried biomass samples were milled using a cutting mill equipped with a 1-mm sieve (SM 200;Retsch GmbH,Haan,Germany). They were then analysed in a biogas batch test according to VDI guideline 4630. A detailed description of this batch test can be found in . Methane hectare yield was calculated by multiplication of substratespecific methane yield and the organic dry matter yield.

| Laboratory analysis
For each laboratory analysis, a subsample was taken from the leaf and stem fractions.
The ash content was determined by incinerating all samples in a muffle kiln at 550°C for 4 hr according to VDLUFA book III, method 8.1 (Naumann & Bassler, 1976/2012. Lignin, cellulose and hemicellulose content of leaf and stem were analysed by near-infrared spectroscopy (NIRS High leaf proportion neutral detergent fibre (NDF), acid detergent fibre (ADF) and acid detergent lignin (ADL) were determined according to VDLUFA book III, method 6.5.1 (NDF), 6.5.2 (ADF), and 6.5.3 (ADL), Naumann & Bassler, 1976/2012. Table 2 shows the standard error of the NIRS calibration (SEC) and prediction (SEP) and the R 2 of the NIRS calibration and validation. The ADL content is given by the lignin content. Hemicellulose content is determined by subtracting ADF from NDF, cellulose by ADL from ADF. Nitrogen (N) contents of leaf and stem were analysed using a Vario Max CNS (Elementar Analysensysteme GmbH, Langenselbold, Germany), as described in the VDLUFA Method Book III, method 4.1.2 and DIN ISO 5725.
The phosphorus (P) and potassium (K) contents of leaf and stem were analysed according to the VDLUFA Method Book III, method 10.8.1. For this analysis, 0.5 g of each sample was first dissolved in 8 ml HNO 3 , 1 ml H 2 O and 5 ml H 2 O 2 and then placed in an ETHOS.lab microwave (MLS GmbH, Leutkirch, Germany) for pressure and temperature digestion. Potassium and phosphorus contents were measured with an ICP-OES.

| Statistical analysis
The experiment was performed in two phases. The first phase consisted of a field trial; in the second phase, samples from the field trial were processed in the laboratory.
The SMY of the leaf and stem fraction were each analysed by a linear mixed model, which considered both field trial and laboratory design (Equation 1).
where y hijkl is the measurement of the i-th genotype on the h-th harvest date with the j-th effect of year in the l-th field replication and the k-th laboratory replicate. μ is the general effect, g i is the i-th genotype effect (M×g; GNT1; GNT3; Sin55), d h is the main effect of the h-th harvest date (HD 1; HD 2; HD 3), f j is the main effect of the j-th year (2016; 2017), (gd) ih is the interaction effect of the i-th genotype with the h-th harvest date, (df) hj is the interaction of the h-th harvest date with the j-th year, (gf) ij is the interaction of the i-th genotype with the j-th year, (gdf) hij is the interaction of the i-th genotype with the h-th harvest date and the j-th year. While all effects described above were taken as fixed, the remaining effects were taken as random. s l is the random effect of the l-th replicate in the first phase (field), r k is the random effect of the k-th replicate in the second phase (laboratory) and e hijkl is the residual error term corresponding to y hijkl . Furthermore, (gs) il is the mainplot error effect associated with main plots of genotype i in replicate l. All other traits (DMY, MY, fibre and nutrient content) were measured for both leaf and stem after the first phase; thus for these traits the effects of the k-th replicate in the laboratory were dropped from the model in Equation 1.
In all analyses, residuals were graphically checked for normality and homogeneity of variance. For phosphorus (P), log transformation of the data was necessary. Where significant differences were found using an F test, a multiple t test (LSD) with α = 0.05 was performed. All data analysis was performed using the PROC MIXED procedure of Statistical Analysis Software SAS, version 9.4 (SAS Institute Inc., Cary, NC, USA).  (Table 3).

| Dry matter yield and leaf proportion
Of all four genotypes, M×g yielded highest (average: 19.89 t DM ha −1 a −1 ) and Sin55 lowest (10.83 t DM ha −1 a −1 ) in both years. Both M×g and Sin55 had a higher average yield in 2017 than in 2016; for GNT1 and GNT3 it the yield at HD 2 and 3 in 2017 compared with 2016, while at HD 1 yield level was similar for both years. Additionally, Sin55, yielded significantly higher at HD 3 than at HD 1 in 2017 ( Figure 1d). On average, M×g had the highest (44.3%) and GNT3 the lowest (34.7%) DMC. The DMC of all genotypes was higher 2016 than in 2017 (Figure 1). The average DMC of all genotypes was highest at HD 1 (43.1%) in 2016. In 2017, the highest DMCs were recorded at HD 2 (35.0%) and HD 3 (34.9%).
Except for the genotype Sin55, the dry matter yield was mainly made up of stems. As shown in Figure 2, the average leaf proportion was lowest in genotype M×g (31.7% of DM) and highest in Sin55 (50.6%). All genotypes (except Sin55) had a higher leaf proportion in 2017 than 2016. The leaf proportion was lowest at HD 3 for all genotypes in both years.

| Methane yield
The substrate-specific methane yield (SMY) of both leaf and stem biomass was significantly affected by genotype, but not by harvest date (Table 3). The average SMY of all genotypes was 311.0 ml CH 4 (g oDM) -1 for the leaf fraction and 285.1 ml CH 4 (g oDM) -1 for the stem fraction (Figure 3).
A comparison of the three HDs of each year revealed no clear trends in substrate-specific methane yield. On average, M×g, GNT1 and GNT3 had higher SMY in 2017 than 2016, while that of Sin55 was similar in both years.
The methane hectare yield (MY) of leaf and stem fractions was significantly affected by year and genotype × year ( Table 3). As shown in Figure 4, average methane hectare yield (MY) was highest for M×g (5,265 m³ CH 4 ha −1 a −1 ) and lowest for Sin55 (3,183 m³ CH 4 ha −1 a −1 ). In all genotypes, the MY was much more strongly influenced by DMY than by SMY. The contribution of the stem fraction to the MY was higher than that of the leaf fraction in all genotypes, except Sin55, where the leaf proportion of the DMY was similar to or higher than the stem proportion. The MY of HD 2 and HD 3 was higher in 2017 than 2016 for all genotypes (except GNT1 at HD 3). The MY of HD 1 was lower in 2017 than 2016 in all genotypes except Sin55, where it increased slightly from 2016 to 2017. In 2017, M×g and GNT1 had the highest MY at HD 2, GNT3 and Sin55 at HD 3 (significant differences only in Sin55).

F I G U R E 2
Average leaf proportion of the four genotypes (M×g, GNT1, GNT3 and Sin55) over 2 years (2016/2017) on three harvest dates (HD 1: mid-September, HD 2: beginning of October, HD 3: mid-October). The white letters within the bars indicate significant differences in leaf proportion between the genotypes for each harvest date and year. The lower table shows significant differences in leaf proportion between harvest dates for each year (standard letters for 2016, bold letters for 2017) and genotype. Means with same letters were not significant different from each other. Level of significance was α = 0.05. Error bars represent standard error for leaf proportion

| Fibre and ash content
The p-values for F tests of fixed effects show that all fibre contents, except lignin of leaf fraction, were significantly affected by the interaction of harvest date × year and either genotype or genotype × year (Table 4). Table 5 shows the fibre (lignin, hemicellulose, cellulose) and ash contents of the four genotypes on the three harvest dates in 2016 and 2017. Genotype M×g had the highest average lignin (9.8%) and cellulose (39.2%) content of all genotypes; GNT1 had the lowest lignin (7.6%) and Sin55 the lowest cellulose (36.6%) content. However, Sin55 had the highest (average: 29.7%) hemicellulose content for leaf and stem fraction in both years. M×g had the lowest hemicellulose content (average: 25.4%) in both years, except for stem in 2016. Ash content was lowest for M×g (5.0%) and highest for GNT1 (6.1%). For all genotypes and harvest dates in both years, the lignin and cellulose contents were higher in the stem fraction than the leaf fraction (except GNT1 and Sin55 at HD 1 in 2017). For hemicellulose and ash contents, it was the reverse (Table 5).
The nutrient removal was significantly affected by the interaction of genotype and year, except for K stem ( Table 4). The average removal over all genotypes, harvest dates and years was 115 kg ha −1 a −1 for nitrogen (N), 257 kg ha −1 a −1 for potassium (K) and 17 kg ha −1 a −1 for phosphorus (P). As shown in Figure 5, GNT1 had the highest nutrient removal of all genotypes (141 kg ha −1 a −1 N; 301 kg ha −1 a −1 K; 23 kg ha −1 a −1 P), Sin55 the lowest of N (81 kg ha −1 a −1 ) and K (193 kg ha −1 a −1 ), and M×g the lowest of P (11 kg ha −1 a −1 ). The average removal of all three nutrients was higher in 2017 than 2016. In all genotypes, N removal was higher by leaves than stems, for K it was higher in stems than leaves, and for P it was balanced between the two fractions.
A comparison of nutrient use efficiency (NUE; biomass produced in kg per nutrient removal in kg) shows that, on average, M×g produced the most biomass per removed . By contrast, Sin55, produced the least biomass per removed nutrient (127.5 kg/kg N; 54.1 kg/kg K); 785.7 kg/ kg P). Taken as an average of all genotypes for both years, the NUE was highest at HD 3 for all three nutrients and lowest at HD 1 for K and P and at HD 2 for N (results not shown).

| DISCUSSION
The objective of the study was to determine more precisely the optimal harvest date of a green cut of miscanthus, which not only delivers high dry matter and methane yields but also guarantees the long-term productivity of the crop. A further objective was to identify the effects of genotypic differences, such as leaf proportion, on methane hectare yield. The results showed that, for all genotypes, a harvest in October 2016 had no negative effects on the dry matter and methane hectare yields in the following year. The methane hectare yield (MY) was found to be mainly influenced by the stem fraction, due to the higher dry matter yield (DMY) of stems than leaves, except for Sin55, which had balanced leaf and stem proportions. Additionally, we determined that the substrate-specific methane yield (SMY) was significantly higher for leaves than for stems in all genotypes.
The following sections discuss these results with a view to identifying the optimal harvest time for miscanthus for utilization in biogas plants. In addition, the genotypic differences of the four genotypes used are analysed to develop recommendations for future miscanthus breeding. . The blue section of bars represents leaf proportion, the orange section stem proportion. Different lower-case letters within the bars represent significant differences for MY between harvest dates for each year and genotype (ns: not significant; standard letters for 2016; bold italic letters for 2017). The lower table presents significant differences between the years (2016/2017) for each harvest date and genotype. Standard letters indicate differences for HD 1, bold letters for HD 2 and italic letters for HD 3. Means with same letters were not significant different from each other. Level of significance was α = 0.05. Error bars represent standard error for MY 4.1 | Optimal harvest date for a green cut of miscanthus Our study determined that, after 2 years of observation, the optimal harvest date for the genotypes M×g, GNT3 and Sin55 is mid-October and for GNT1 the beginning of October. The early cut in mid-September 2016 had a negative yield effect on all genotypes, with a tendency to lower the biomass yield in 2017 at HD 1, except for Sin55 (Figure 1). The yield was expected to be higher in 2017 than 2016 for all genotypes because 2017 was the third year of the plantation. Although that of Sin55 was higher in 2017 than 2016, the increase was much lower than expected. It has been observed that yields of miscanthus plantations increase steadily from the establishment year to the 3rd or even 5th year, while the stocks are expanding (Iqbal, Gauder, Claupein, Graeff-Hönninger, & Lewandowski, 2015;Lewandowski, Clifton-Brown, Scurlock, & Huisman, 2000). For this reason, it is not clear how much the yields in this study were determined by plantation age and how much by early harvest regime. However, because the genotypes M×g and GNT3 yielded significantly lower at HD 1 in 2017 than HD 1 in 2016, a harvest of these genotypes in mid-September cannot be recommended. This is in line with the results of  who found the best green-cut tolerance for M×g when harvested in mid-October. Schmidt et al. (2017) found a slight decrease in DMY in the second year of green cutting when M×g was harvested in September in an older (19 years) stand, but not in a younger (6 years) one. Thus, the longterm effect of an ongoing autumn harvest in a growing miscanthus stand needs to be further investigated, as the 2year analysis of our study is too short to reach a final conclusion on the best harvest date of miscanthus with its lifetime of up to 20 years. We presume that the later harvest in mid-October provides more time for nutrient relocation and enhances the long lifetime of the stand. Evidence for this can be seen in the nutrient use efficiency (NUE), which was on average highest at HD 3 (mid-October) for all genotypes, after nutrient translocation back to rhizomes has begun.
Although the soil conditions were not optimal (as described above), all four genotypes had satisfactory yields (average: 16.6 t DM ha −1 a −1 ).
In general, the average MY for M×g found in our study (5,265 m³ CH 4 ha −1 a −1 ) is within the range found in other studies, namely 5,000-6,000 CH 4 ha −1 a −1 Mayer et al., 2014). This is at the lower end of the MY of maize (5,000-7,000 m³ CH 4 ha −1 a −1 ), as reported by Mayer et al. (2014) and . However, it should be noted that MY is mainly influenced by dry matter yield. A comparison of miscanthus and maize grown at the same  location in 2016/2017 shows that miscanthus would likely have achieved higher MY than maize. This can be deduced from a comparison with maize yields taken from a study by Ehmann, Thumm, and Lewandowski (2018). In this study, the maize plots (which were located adjacent to the field trial in our study) yielded on average approx. 11 t DM ha -1 in 2016/2017. A comparison of the DMY for 2016/2017 of this maize with that of the miscanthus in our trial (16.6 t ha −1 a −1 ) reveals a higher DMY for miscanthus.
Additionally, both the production costs and negative environmental impacts are lower for miscanthus than maize biomass under certain conditions, due to lower input requirements for example fertilizer (Wagner et al., 2019). This is mainly a result of the high nutrient use efficiency (NUE) of miscanthus (Cadoux et al., 2012;Lewandowski & Schmidt, 2006). With a view to achieving a high NUE, the latest harvest date in mid-October is preferable. Firstly, the nutrient removal by the biomass is lower. Secondly, the resilience of the crop to a green cut is most likely higher as the nutrients have already partly been relocated from the aboveground biomass to the rhizomes.
As shown by , the nutrient removal of a harvest in mid-October is twice as high as that of a harvest after winter. Thus, to better close nutrient cycles, we recommend the application of digestate in spring to return N, K and P to the field.
When harvested green, miscanthus biomass needs to be ensiled for storage. Therefore, the optimization of the harvest date also has to be considered with regard to the ensiling ability of the biomass. It has been shown that miscanthus biomass is best ensiled when harvested at a dry matter content (DMC) of 35% -40% (Mangold, Lewandowski, Möhring, & Kiesel, 2019). This also indicates that mid-October is the most suitable harvest date.
The lower average DMC found in our study in 2017 can be attributed to the higher accumulated precipitation that year than in 2016. This led to a slower ripening of all genotypes in 2017, resulting in lower average DMCs of 33.3%. However, no recommendations can be provided for optimizing the DMCs through harvest regimes as no clear trends were observed. Generally, it is expected that the DMC increases with senescence and increasing lignin contents of the biomass. Table 5 shows the increase in stem T A B L E 5 Fibre (lignin, cellulose, hemicellulose) and ash content [% of DM] of the four genotypes on three harvest dates. Significant differences between harvest dates are shown by different lower-case letters (leaf) and upper-case letters (stem), for each genotype for 2016 (standard letters) and 2017 (bold) (α = 0.05; ns = not significant). Means with same letters were not significant different from each other and leaf lignin contents with later harvest for all genotypes. However, the DMC of GNT3 at HD 3 2017 (shown in Figure 1c) neither followed that trend nor could it be explained by weather conditions. Therefore, further analysis is necessary to clarify the influence of weather conditions, for example on DMC, over a longer time period.

| Recommendations for miscanthus breeding
Our results demonstrated significant differences in leaf proportion between the four genotypes at all harvest dates in both years ( Figure 2). Sin55 had the significantly highest (50% of DM) leaf proportion, followed by GNT3 and GNT1. M×g had the significantly lowest (30% of DM) leaf proportion (Figure 2). A glance at the average SMY of the genotypes reveals a positive correlation between leaf proportion and SMY for all genotypes. This is most likely due to lower lignin contents of leaves than stems ( 2017) on three harvest dates (HD 1: mid-September; HD 2: beginning of October; HD 3: mid-October). The blue part of the bars indicates the proportion of nutrient removal by the leaves, the orange part the proportion of nutrient removal by stems. The lower table shows significant differences between harvest dates for 2016 (standard letters) and 2017 (bold letters) for each genotype for nitrogen (N), potassium (K) and phosphorus (P). Means with same letters were not significant different from each other. Level of significance was α = 0.05 influenced by dry matter yield rather than by SMY. For this reason, miscanthus with a higher leaf proportion in its biomass-and thus higher leaf proportion in its DMY-is more favourable for utilization in anaerobic digestion. In combination with higher SMY, this would result in higher MY and most likely also improve the digestibility. Various studies have revealed a negative correlation between lignin content of biomass and its SMY (von Cossel, Möhring, Kiesel, . However, from the results of our study, we cannot conclude that lignin content alone is a suitable criterion for breeding biogas miscanthus. Although leaves had lower lignin contents than stems, leading to higher SMYs, this was not true of the average lignin content of the genotypes. For example, GNT1 had the lowest lignin content of all genotypes (Table 5) but did not yield the highest average SMY. Additionally, for all genotypes except GNT1, the harvest date with the lowest lignin content did not result in the highest SMY. Thus, we cannot recommend selecting miscanthus genotypes for anaerobic digestion on the basis of lignin content alone. Instead, a selection on the basis of hemicellulose content would appear more reasonable as there was a positive correlation between the average hemicellulose content and average SMY in each genotype. The stay-green genotype Sin55 was expected to be most suitable for anaerobic digestion on account of late senescence, which leads to lower dry matter contents. It was thus expected to have lower lignin contents, in turn leading to a better digestibility. These expectations, however, were not confirmed by our study. Despite Sin55 having the highest SMY, it did not have the highest methane hectare yield. This was mainly due to its low DMY, especially in 2016. This leads us to the hypothesis that Sin55 has a delayed establishment compared with the other three genotypes, resulting in lower yields. We expect that Sin55 was still in the process of establishing and assume that in 2018 its MY will be similar to that of the other genotypes. In their study, Wahid et al. (2015) demonstrated a higher DMY and biomethane potential, but lower lignin content, for a M. sinensis genotype than for M×g. In addition, they showed that the M. sinensis genotype had a slightly higher digestion velocity than M×g during 35-day fermentation. Our analysis also showed a tendency for higher digestion velocity in Sin55 than M×g (results not shown). Better digestibility of the biomass is preferable, as it saves costs for additional pretreatment. Several studies have recommended the pretreatment of miscanthus biomass to gain higher methane yields (Frydendal-Nielsen et al., 2016;Zheng, Zhao, Xu, & Li, 2014), but pretreatment is usually energy-and cost-intensive (Zheng et al., 2014). Therefore, it is likely that, for the genotype M×g, higher methane yields and thus higher revenues will not necessarily lead to higher profits, as higher costs for pretreatment are incurred. By contrast, Sin55 could be more profitable, as it requires less pretreatment. Thus, we conclude that stay-green M. sinensis genotypes, such as Sin55, are most suitable for anaerobic digestion, due to their high SMY and low lignin contents.
In terms of nutrient use efficiency (NUE), M×g is the most suitable and Sin55 the least suitable genotype for the supply of low-input biomass for anaerobic digestion, as M×g produced the most and Sin55 the least biomass per removed unit of nitrogen, potassium and phosphorus. This is most likely mainly due to the higher yields of M×g compared with the other genotypes.
In conclusion, this study determined that for the miscanthus genotypes M×g, GNT1, GNT3 and Sin55, the most suitable harvest date for high methane yields is October. A late cut in October is also the most favourable in terms of nutrient use efficiency. In addition, the study found that the leaf fraction of miscanthus biomass produced significantly higher substrate-specific methane yields than the stem fraction, which we attribute to the lower lignin content of leaves. For this reason, future miscanthus breeding should focus on genotypes with a higher leaf proportion (e.g. M. sinensis genotype Sin55) to increase methane hectare yields.