Biomass and methane yield of giant reed (Arundo donax L.) as affected by single and double annual harvest

The replacement of silage maize with giant reed as energy crop has been proposed as a mean for reducing the need of irrigation water as well as monetary and environmental costs of cultivation. Little is known about giant reed response to within‐season harvesting, and its effect on the methane production in anaerobic digestion. The effect of three harvest schedules on yield, biomass composition and methane production of giant reed was evaluated at one site of the Po Valley, northern Italy, for three consecutive years. In a completely randomized block design with four replicates the treatments applied annually were: (i) double harvest 1: first cut at the end of June + second cut at the beginning of October (DH1); (ii) double harvest 2: first cut at the end of July + second cut at the beginning of October (DH2); (iii) single harvest at the beginning of October (SH). The crop stand was established in the year 2015 and treatments were repeatedly applied in the years 2016, 2017 and 2018. The SH treatment determined the highest average annual dry matter (DM) yield (59.0 Mg DM ha−1). The DM yield for treatments with double harvest was significantly lower, that is, −30% for DH1 and −15% for DH2. In terms of specific methane yield there was little advantage in harvesting biomass twice during the growing season. The average specific methane yield varied substantially between years (i.e. from 144 to 233 ml CH4 g−1 VS), and in every year the values tended to decrease with the ageing of harvested biomass. Methane yield per hectare, however, was driven by DM yield, thus SH also determined the highest average value (9110 m3 CH4‐STP ha−1). In conclusion, the single annual harvest at the end of the growing season is an ideal strategy for maximizing methane production from giant reed.

providing, in the form of grain and silage the nutritional base for both fattening and milking animals. Following the Nitrates Directive (EEC, 1991) concerning the protection of waters against pollution caused by nitrates from agricultural sources, stakeholders were forced to curb the environmental footprint of animal waste utilization. Yet, following the European Renewable Energy Directive (EC, 2009/28/EC, 2009), on the promotion of the use of energy from renewable sources, Italian policy provided monetary incentive to electricity produced from biogas plants (Carrosio, 2013). Owing to such incentive, electricity production from biogas had become the main economic purpose for many farms. This, in turn, re-shaped the organizational model of farms: silage maize was cultivated to provide biomass to the digester, while grain maize for animal feeding was purchased on the international market (Carrosio, 2013).
Maize is not only water demanding but also an intensive crop in terms of nutrient supply, weed and pest management and soil tillage. Hence, there is much scope to explore alternative crops as biomass feedstock that could be cultivated in rainfed conditions while reducing cultivation costs and environmental side effects (Corno et al., 2014). Giant reed (Arundo donax L.) is a non-food, tall, rhizomatous grass that produces a substantial amount of lignocellulosic biomass in Mediterranean climates (Lewandowski et al., 2003). From the standpoint of reducing agronomic inputs, this species possesses several key attributes (Ceotto & Di Candilo, 2010a, 2010b: (i) it is perennial, therefore eliminates the need of annual soil tillage; (ii) it can be cultivated without irrigation; (iii) it has efficient use of nitrogen, which is seasonally translocated from the above to below plant organs and vice-versa; (iv) it is not susceptible to pathogens and a strong competitor against weeds, thus eliminating the need of chemicals application. Giant reed is usually cultivated in rainfed conditions in northern Italy, thus it could replace maize whether irrigation water is unavailable or too expensive. Yet, this crop can be suitably cultivated even in marginal areas, unsuited for grain crop cultivation, owing not only to low fertility but also to environmental pollution and safety reasons (Ceotto & Di Candilo, 2010b). According to D'Imporzano et al. (2018) the high productivity of giant reed, combined to its low requirement of agronomic inputs, is key for minimizing the environmental impact of the crop. Moreover, giant reed exerts an effective soil nitrate removal, therefore reducing the risk of nitrate pollution (Ceotto et al., 2018). While giant reed has been largely investigated during the last decade, most of the studies focused on the crop harvested when the annual growth has ceased, notably in autumn or winter (Angelini et al., 2009;Borin et al., 2013;Cosentino et al., 2006;Monti & Zegada-Lizarazu, 2016). Little is known about the performances of the giant reed stand harvested during the growing season. This study was solicited by the stakeholders of the biogas sector, who raised these specific questions: how much biomass and methane yield can be obtained if the giant reed stand is harvested twice, instead of once, per growing season? In the case that giant reed is harvested in midsummer would crop regrowth be substantial?
A preliminary study of Ragaglini et al. (2014) provided encouraging results, but it was still limited to only one growing season. The following research questions were addressed: (i) What is the best harvesting schedule for obtaining the highest giant reed biomass productivity? (ii) In which period of the growing season is the giant reed biomass better suited as feedstock for anaerobic digestion? (iii) Which is the best compromise between biomass quality and quantity to obtain the highest methane yield per hectare form giant reed?
To answer these questions a field experiment was carried out for 3 years comparing giant reed harvesting schedules.

| Site characterization
The experimental site is located in Anzola dell'Emilia (Bologna), in the alluvial plain of Po Valley, northern Italy (Lat. 44°32ʹN, Long. 11°11ʹE, 38 m a.s.l.). The soil of the site is a silty loam, classified as fine silty, mixed mesic Udifluventic Haplusteps (Soil Survey Staff, 2014). The main soil characteristics are reported in Table 1. The climate is temperate sub-continental, owing to the relatively long distance from the sea of about 200 km. The average annual air temperature for the site is 12.9°C, and the mean annual total rainfall is 755 mm. The location is characterized by a relatively shallow water table, usually fluctuating between 1.1 m depth in winter and about 2 m depth in summer. The presence of such a water table assures favourable production conditions for giant reed under rainfed cultivation.

| Agronomic experiment
The giant reed stand was established in June 2015. The preceding crop was winter barley (Hordeum vulgare L.), harvested at the end of May at the milky kernel stage. After the barley harvest, the soil was ploughed at 0.30 m depth and subsequently refined with a rotary harrow to allow furrows preparation and a favourable soil bed condition for stem cuttings.
The propagation technique proposed by Ceotto and Di Candilo (2010a) was used: shoot cuttings of about 1 m length, with several nodes bringing lateral branches, were laid down horizontally on furrows of 0.15 m depth, and then covered manually with the soil. Two parallel culms were placed on the bottom of the furrows, aiming to increase the plant density. The distance between the furrows was 0.75 m. The field area planted with giant reed was about 2900 m 2 . The stem cuttings were harvested on a giant reed stand, with culms at their second years, located in the same experimental farm. The clone used for the experiment was originally from the nearby location of Budrio (Bologna). Stem cuttings were planted from 22 to 23 June 2015 and immediately irrigated with a volume of 35 mm, using a linear move sprinkler irrigation system covering all the surface of the field. To facilitate stem rooting and crop establishment, the same irrigation volume was applied weekly until the end of August 2015. Few days after the first irrigation, and prior to the emergence of the giant reed sprouts, newly emerged weeds were controlled by a nonselective herbicide (glyphosate). No fertilization was applied in the planting year 2015.
In the following three growing seasons the giant reed stand was fertilized annually with 250 kg nitrogen ha −1 , applied in the form of urea at the beginning of April. This fertilization rate was devised on the base of a previous 4-year study conducted in the same location. Such study indicated that giant reed possesses high agronomic efficiency of applied nitrogen (Ceotto et al., 2015) and the ability to reduce the risk of nitrate pollution dealing with nitrogen surplus. (Ceotto et al., 2018).
No further irrigation was applied. This high fertilization rate was provided with the intent of ensuring abundant nitrogen availability, resulting in non-limiting nutrient conditions to the crop. No fertilization was applied during the first growing season, during which the crop encountered its juvenile phase. The above-ground biomass produced during the first growing season was harvested during the winter.
The harvesting schedule experiment was conducted during three consecutive years, 2016, 2017 and 2018, corresponding respectively to the second, third and fourth year of the crop age respectively. The harvesting treatments applied annually were: (i) double harvest 1: first cut at the end of June + second cut at the beginning of October (DH1); (ii) double harvest 2: first cut at the end of July + second cut at the beginning of October (DH2); (iii) single harvest at the beginning of October (SH). A scheme of the harvesting treatments applied during the 3 years of experiment is reported in Table 2.
The treatment DH1 was directly solicited by the stakeholders, because it provides substantial biomass yield while leaving a relatively long period ahead for regrowth. The treatment DH2 was stimulated by the peculiar eco-physiology of the crop: Ceotto et al. (2013), working at the same site, indicated a linear crop growth until the end of July and a subsequent 'summer slump' for giant reed growth. Then a harvest was devised at the end of the linear growth period to assess whether a harvest might rejuvenate growth. The treatment SH can be regarded as the reference because giant reed is usually harvested once per year, either in autumn or winter. The experimental design was a completely randomized block with four replications. The size of the individual plot was 99 m 2 , defined by 4.5 m width (i.e. six rows distant 0.75 m) and 22 m length. The rectangular shape was adopted to allow the use of a tractor-mounted scissors lawn mower for harvesting.

| Weather data
Weather data for temperature and precipitation were collated from a meteorological station located at 200 m from the giant reed stand. Climatic data were recorded electronically at 1 min intervals and values were integrated to daily data.

| Biomass measurements and chemical analyses
An area of about 9 m 2 was sampled at random on each individual plot at every harvest date to measure the biomass produced per unit area (i.e. 4 internal rows × 0.75 m = 3 m width × 3 m length). External rows were excluded from sampling to minimize the edge effect. The individual stems and leaves of each fresh biomass sub-sample were cut manually in pieces of about 3-5 cm length, by using a pruning scissor prior to oven drying and milling and sieving at 1 mm size. After fresh weight measurement, a sub-sample composed of 3-4 kg of entire shoots (including stems and their leaves) was oven-dried at 65°C until a constant weight was obtained. Shoot dry biomass was milled through a 1-mm sieve (Cutting mill SM 300, Retsch ® mbH). For the determination of total C and N of harvested biomass samples, ground samples of 0.150 grams were loaded into tin foil cups and analysed with LECO Truspec ® CHN Analyzer (LECO Corporation). Oven-dried, milled and sieved biomass samples were used to determine total solids (TS), volatile solids (VS), ash and pH according to APHA (2017). In particular, total solids were determined gravimetrically by thermal treatment at 105°C at constant weight. Volatile solids were determined as the difference between TS and ashes. Ash were determined by incineration in a muffle furnace at 550°C for 10 h. The pH was determined after suspension, 2-h stirring and sedimentation of 1.1 g dry matter in 50 ml distilled water pH, by means of a Crimson Titromatic 1S pH metre.
Fibre fractions (neutral detergent fibre, NDF; acid detergent fibre, ADF; and lignin, ADL) were determined on samples dried at 65°C at constant weight according to Van Soest et al. (1991). The hemicellulose content was estimated as the difference between NDF and ADF; cellulose as the difference between ADF and ADL. Total polyphenols were determined according to the Folin−Ciocalteu colorimetric assay (Singleton & Rossi, 1965) and expressed as mg tannic acid per g of dry matter.

| Anaerobic digestion experiment
Representative samples of dry milled biomass collected on each plot (oven-dried, milled and sieved at 1 mm as described above) were used as feedstocks in AD experiments. Field replications of each treatment were pooled before being used in AD. The pH was determined by a Crimson Titromatic 1S pH metre, using 1.1 g dry matter, suspended, stirred and sedimented in 50 ml distilled water.
Anaerobic digestion of pooled samples was carried out in mesophilic conditions (35°C), in 100 ml batch bioreactors (118.5 ml effective volume), with five replications (five treatments, five replications and 3 years, corresponding to 75 reactors in total). The inoculum was prepared according to Vasmara et al. (2015). AD was carried out according to Vasmara and Marchetti (2016). Following Chartrain and Zeikus (1986) and Wu et al. (1992), the reaction mixture included 1 g of giant reed VS, 50 ml of sterilized phosphate-buffered basal medium without energy sources (HM), and 5 ml of inoculum. The initial pH of the mixtures was, on average, 7.4 ± 0.1. Three reactors containing only inoculum and HM were also included as blanks.
Biogas volume and composition were determined for each reactor during the incubation period (3 months). The cumulated methane production was determined as the sum of the methane volume collected in the syringe and that accumulated in the headspace of the reactors and expressed in standard conditions (STP) of temperature (273.15 K) and pressure (101.3 kPa). The methane yield per hectare was calculated by estimating the VS content (Mg ha −1 ) of the dry matter yield of individual plots (i.e. subtracting ashes from DM) and then by multiplying that value for the SMY of the treatment.
Methane concentration in the biogas was determined by means of a MicroGC Agilent 3000 gas chromatograph, equipped with two columns: Molsieve and Plot U; detector: TCD. Carrier gas: argon. No methane production was detected in the blank reactors, therefore no CH 4 subtraction to the methane production of the samples was needed. The incubation period was considered completed when there was no more biogas production in any of the reactors.

| Fitting a modified Gompertz equation for anaerobic digestion
The Gompertz model was originally developed to describe population growth evolution. It represents the situation in which the relative growth rate is an exponentially declining function of time (Zwietering et al., 1990).
The kinetic of methane production in a reactor can be conveniently represented by a modified Gompertz function (Díaz et al., 2011;Velázquez-Martí et al., 2018). Following Lay et al. (1997) a modified three-parameter Gompertz equation (Equation 1) was applied in this study to describe the dynamic of CH 4 production.
where M (t) (ml CH 4 -STP g −1 VS) is the cumulative amount of CH 4 produced, in standard conditions for temperature and pressure (STP); t (days) is the elapsed time; M max (ml CH 4 g −1 VS) is the maximum cumulative CH 4 production; e (exp) is the Euler's number; R max (ml CH 4 day −1 g −1 VS) is the maximum CH 4 daily production rate; and λ is the lag time duration (days), that is the time of microbial adaptation before exponential CH 4 production begins.
The time (days) necessary to reach M max was estimated by calculating the M max /R max ratio and adding to the quotient the lag time duration.
Fitting of the Gompertz model to measurements was performed using the PROC NLIN of SAS (SAS Institute, 2015). The parameter values were estimated according to the Gauss-Newton method. The Gompertz function parameters M max , R max and λ were estimated for individual treatments, crop phases and replicates.
From the agronomic standpoint it is convenient to deal with the concept of Specific Methane Yield (SMY), which is the efficiency with which the VS of giant reed biomass are converted into methane. Because in this experiment 1 g VS of giant reed were introduced in each reactor, this entails that both M max and R max are here referred to 1 g of VS. Hence, M max coincides with SMY.

| Statistical analysis
This field experiment is a classic case of repeated measures, taken over time on the same experimental units. Not only the same plots were repeatedly treated and sampled but also the same living giant reed plants. Our experimental unit was the individual plot, in which the giant reed stand received the harvesting treatments over three consecutive growing seasons. Overall, the plots of treatments DH1 and DH2 were harvested six times, while the plots of treatment SH were harvested three times.
The statistical analysis was performed with the PROC MIXED of SAS using the statement REPEATED to deal with the covariance of the experimental unit over time (SAS Institute, 2015). Such procedure has been specifically developed in the SAS system for managing repeated measures collected over time on the same experimental units. The normal distribution of data was verified by the Shapiro-Wilk test. The homogeneity of variances was evaluated by Bartlett's test (SAS Institute, 2015).
Different criteria were used to evaluate treatment effects for the agronomic and the AD trials, based on the relevant research questions underlying the aims of the two studies. In the case of the agronomic experiment the relevant question was how much biomass and methane yield can be obtained per hectare and per year, depending on whether such biomass has been harvested once or twice per growing season. Therefore, the values of the individual harvests (DH1_I and DH1_II; DH2_I and DH2_II) were added for comparing the treatments DH1, DH2 and SH.
Crop age was included in this analysis as a fixed effect. It is relevant, particularly from the agronomic standpoint, because it entails the evaluation of the effect of cumulative harvests. It is important to notice that at the end of the 3 years, the plots of treatments DH1 and DH2 had received six repeated harvests while the plots of treatment SH had received merely three harvests. Hence, a relevant agronomic question is whether double harvest repeated over time leads to yield decline. The factor crop age is interlaced with year-to-year weather variability, although such effect cannot be disentangled. Conversely, in the case of the AD laboratory trial, the relevant question was whether the biomass harvested during the growing season, is better suited to biogas production compared to SH at the end of the season. Hence, we here considered five independent treatments: DH1_I, DH1_II, DH2_I, DH2_II and SH.
The statistical significance between means was assessed with the statement LSMEANS (SAS Institute, 2015). The correlation matrices were calculated by the PROC CORR (SAS Institute, 2015). The Pearson correlation coefficient was considered.

| Weather
In the years 2016 and 2018, corresponding to II and IV years of crop age, the cumulated precipitation for the period 1 January-31 October was +11% and +9% of the 30-year average for the location, which is 610 mm (Figure 1). In both years precipitation was well distributed during spring and summer, determining favourable conditions for giant reed. On the contrary, in the year 2017 the crop stand had received merely 326 mm of precipitation for the same interval, which corresponds to −47% compared with the 30-year average. The scarcity of precipitation

| Crop biomass yield
The statistical analysis of annual dry matter yield indicated that the factors crop age and treatment were both highly significant, while the interaction crop age × treatment was not significant ( Table 3). The average yield measured in the second year of crop age (51.4 Mg DM ha −1 ) was significantly higher than the mean yield measured in the third year (42.8 Mg DM ha −1 ), while in the fourth year the yield increased by +10% compared to the second year, albeit such difference was not significant ( Figure 2). Because the fourth year provided the highest yield, the lowest productivity measured in the third year cannot be attributed to repeated harvesting. The SH treatment determined the highest annual DM yield (59.0 Mg ha −1 ). The cumulated annual DM yield for treatments with double harvest were significantly lower, that is, −30% for DH1 and −15% for DH2. The individual average annual yield per harvest was respectively 24.7 and 16.5 Mg ha −1 for the first and the second harvest of DH1, 41.9 and 8.4 Mg ha −1 for the first and the second harvest of DH2. It is worth noting that the first harvest accounted for 60% of total average annual biomass yield for DH1 and 83% for DH2. Overall, our findings indicated that the more anticipated is the first harvest the lowest is the total annual DM yield.
Although the crop age × treatment interaction was not significant, additional insight is provided by the more accurate comparisons between individual means operated by the instruction LSMEANS of SAS (Littell et al., 1996(Littell et al., , 1998. Two important points should be noted: (i) there was no significant difference between the SH yield of individual years ( Figure  2); the yield reduction for both DH1 and DH2 compared to SH was particularly pronounced in the unfavourable year 2017 (Figure 2).

| Biomass composition
For most of the biomass composition traits, notably cellulose, lignin, ash, polyphenols, pH, carbon and nitrogen, we found significant differences not only regarding crop age and treatment but also for the interaction crop age × treatment (Table 4). In the case of hemicellulose only crop age and treatment effects were significant, but not their interaction. Cellulose is a major component of the giant reed biomass, accounting for 35%-43% of the total solids (TS). As a general trend, the cellulose content tended to increase when hemicellulose content decreased, and vice versa. It should be noted that in the favourable season 2018 (the fourth year), the average cellulose content (41.6) was statistically higher than in the other years (+5% and +7% higher compared to second and third year).
Regarding lignin, in all the years the content tended to raise with increasing number of DORG. In the favourable fourth year the lignin content was significantly lower than the second year, but very close to the third year.
As far as ash is concerned, in the 3 years the content was significantly higher for the juvenile biomass harvesting, associated with low DORG values, compared to the late stage harvesting SH, characterized by a high DORG value. In the favourable fourth year, the ash content was significantly lower than in the third year, but higher than in the second year.
Polyphenols are a minor component of the giant reed biomass, accounting less than 1% of TS. In all the years polyphenol content was higher in the early harvesting with low DORG and tended to decrease with increasing DORG. Interestingly, in the advantageous fourth year polyphenols were significantly lower than both in the second and third year (−14.4%).
The carbon content tended to increase with the DORG. In the propitious fourth year the average carbon content was lower than in the other years. (−7%).
The nitrogen content progressively diluted with increasing DORG. Therefore, it followed a reverse trend compared to carbon. This trend was quite evident in each of the 3 years. However, in the most fortunate season (fourth year) the average nitrogen content was much lower than in the second and third year (−25% and −29% respectively). Consequently, the average C to N ratio, that was about 40 in both the second and the third year, raised to about 46 in the fourth year.
Overall, cellulose, lignin and carbon were positively correlated with DORG in all years of this study; in the same period ash, polyphenols and nitrogen were negatively correlated with DORG (Table 6).

| Anaerobic digestion
With regards to AD experiment, the statistical analysis performed for the three Gompertz parameters indicated significant effect for the crop age for M max , R max and λ, for treatments in case of M max and λ, and no significant interaction crop age × treatment (Table 5). Although such interaction was not significant, the comparison between individual means of the treatments within individual years, operated by the instruction LSMEANS of SAS (Littell, et al., 1996(Littell, et al., , 1998 is reported in Table S1. As far as treatments are regarded, we remind here that the first and second harvest of treatments DH1 and DH2 were here considered as separated treatments (i.e. DH1_I, DH1_ II, DH2_I and DH2_II), because the purpose of AD experiment was to assess in which period of the growing season the giant reed biomass is best suited as feedstock for AD.
Overall, the effect of the age was much more pronounced than that of the treatment. In the case of M max the highest value (233 ml CH 4 g −1 VS) was recorded in the fourth year and it was also accompanied by the highest value of R max and the lowest value of λ (i.e. a shorter duration of the lag phase). Compared to the fourth year, the M max was merely 66% in the second year and 62% in the third year. In the case of R max , compared to the highest value of the fourth year (7.88 ml CH 4 day −1 g −1 VS), the values observed for the second and third year were 73% and 67% respectively. An opposite trend was observed for the lag phase duration λ, the highest value of 2.21 days was obtained in the second year, an almost identical value in the third year, and only 58% of the highest value on the fourth year.
As far as the treatments are concerned, significantly different M max values were observed between the reference treatment single harvest (SH) and some combinations of early harvesting. The harvest of juvenile shoots (i.e. DH1_I and DH2_II), resulted in significantly higher M max (+11% and +8% respectively) compared to SH, composed by shoots grown across the entire growing season. The remaining treatments DH1_II and DH2_I, composed by shoots of intermediate age, were not significantly different from the aforementioned treatments. The R max values were not significantly different between the five treatments. (Table 5). The highest λ values were observed for the second cut of DH2 (DH2_II) and the lowest for SH. Overall, λ tended to decrease when DORG increased.
The general rule that can be inferred from our data is that giant reed juvenile biomass tends to determine both the highest final value of M max and the longest lag phase. Nevertheless, the differences in the lag phase duration have been irrelevant since no differences were detected in the time to reach M max (30 ± 1 days, on average) in all reactors.
Additional insight can be gained from the correlation matrices reported for the individual years between the three Gompertz parameters, the crop development, quantified by the variable DORG, and the components of the dry matter biomass (Table 6). M max tended to be negatively correlated with DORG in all years, but statistical significance arose only in the fourth year. A significant negative correlation between the DORG and R max was observed in the fourth year, but not in the other years. A negative correlation between λ and the DORG was statistically significant in all the 3 years. Accordingly, the lag time reduced when DORG increased.

| Methane yield per hectare
The statistical analysis of methane yield per hectare indicated that the factors crop age and treatment were highly significant while the interaction crop phase × treatment was not significant (Table 3). While the general patterns of methane yield retrace the ones of dry matter yield, differences between crop age were exacerbated because the fourth year (i.e. 2018) was characterized  by the highest SMY (i.e. M max ) combined with the highest biomass yield. As far treatments are concerned, the average annual methane yield of SH was the highest (9110 m 3 CH 4 -STP ha −1 ), the result for DH2 was not significantly lower (8121 m 3 CH 4 -STP ha −1 ), while the one for DH1 (6767 m 3 CH 4 -STP ha −1 ) was significantly lower than both the other treatments. It is worth noting that the first harvest provided 74% of total methane in the case of DH1 and 89% in the case of DH2. Despite of the fact that interaction crop age × treatments was not significant, additional insight can be gained by the comparison of individual means operated by the LSMEANS. Notably, it is worth noting that in the favourable year 2018, even the methane yield of the treatment DH1 significantly exceeded all treatments of the two precedent years (Figure 3).
As far as average annual treatment are regarded, compared to the methane yield of the second year (7036 m 3 CH 4 -STP ha −1 ), the result was −23% for the third year (5390 m 3 CH 4 -STP ha −1 ), and +64% for the fourth year (11,571 m 3 CH 4 -STP ha −1 , Figure 3).

| Dry matter yield
As far as the effect of crop age is concerned, the plausible explanation for the lowest yield measured in the third year lies on the unfavourable precipitation pattern occurred in the year T A B L E 6 Pearson correlation coefficients for the three parameters of the modified Gompertz function, the biomass age and the qualitative biomass traits of giant red biomass ns *** *** * *** *** * *** 2017 ( Figure 1). In fact, despite to the repeated harvesting of the precedent years, the crop provided the highest yield in the favourable season 2018. As far as the effects of single and double harvest are concerned, our findings are in contrast, but not in contradiction, with Ragaglini et al. (2014). Working in Tuscany, central Italy, these authors reported that double harvest (i.e. 35-40 Mg DM ha −1 ) equaled that obtained by a SH at the end of the season (i.e. 38 Mg DM ha −1 ). These authors explored cutting dates in the period from 21 June to 18 October, which is close to the time span considered in this study (i.e. 26 June-15 October). Because their data refer to the year 2011, and to a giant reed stand established in 2007, it is plausible that the rhizome reserves accumulated in the first 4 years of crop stand were strong enough to withstand a double harvest without reduction in the seasonal yield. The current study was carried out on a crop stand that was still on its juvenile phase and received the treatments for three consecutive years (i.e. second, third and fourth year of crop stand).
From the eco-physiological standpoint, the yield reduction determined by treatment DH1, with first cutting at the end of June, is supported by several concurrent circumstances. According to Ceotto et al. (2013) giant reed at the end of June has a substantial canopy cover, and this is combined with a high incoming solar radiation, due to the length of the day near the summer solstice. Hence, removing the whole canopy cover during the period with the highest solar irradiation is, in principle, a bad deal. Moreover, these authors, working at the same location, reported a steady daily growth rate for giant reed from early May until the beginning of August. Aiming to exploit thoroughly the linear phase of biomass accumulation across the season, the treatment DH2 (with the first harvest at the end of July) is certainly more appropriate. Indeed, for both treatments DH1 and DH2 the regrowth following the first cutting was not substantial enough to reach the yield obtained with SH and thus to justify two cuttings across the growing season.
The average annual DM yield observed for the reference treatment SH (59.0 Mg DM ha −1 ) was higher compared to a previous study conducted on an abundantly fertilized giant reed stand in the same location. In fact, Ceotto et al. (2015) reported an average yield of 37.3 Mg DM ha −1 for a 5-year fertilization treatment consisting of 20 mm (i.e. 200 m 3 ha −1 ) of cattle slurry per year. This is a 5-year average including one unfavourable year of poor cumulated precipitation. Because both experiments were based on the same clone, the most relevant difference consisted of timing of harvest: while in the cattle slurry trial biomass has been harvested during the winter, in the harvesting schedule experiment the SH treatment was harvested in October. A possible explanation could be that in October the translocation of dry matter into belowground organs, notably rhizomes, is still ongoing. This hypothesis is corroborated by a third study conducted at the location, in which biomass has been sampled throughout the growing seasons and biomass yield of about 60 Mg DM ha −1 was reported (Ceotto et al., 2013). In addition, Borin et al. (2013) reported a progressively declining yields of 74, 66 and 65 Mg DM ha −1 with autumn, mid-winter and late-winter, respectively, for giant reed growing under abundant nitrogen and water availability in northern Italy. It is also worth noting that the yield observed in this study is almost half of the highest ever giant reed yield (i.e. 125 Mg DM ha −1 ) reported by Idris et al. (2012) working in Australia with experimental subsurface flow, gravel-based constructed wetlands receiving untreated recirculating aquaculture system wastewater.
In the last few years, a considerable amount of research has been devoted on evaluating the perennial grass Miscanthus (Miscanthus spp.) as feedstock for AD in substitution to maize (Kam et al., 2020;Kiesel & Lewandowski, 2017;Mangold et al., 2019;Ruf et al., 2017;Schmidt et al., 2018). Whittaker et al. (2016) pointed out that Miscanthus replacing maize is justified by the environmental benefits that may be achieved rather than simply the SMY, and by the type land which is available for it to be grown. All of these studies, however, were conducted in Central Europe and United Kingdom. Under the environmental conditions of Northern and Central Italy, giant reed appears to be better suited than Miscanthus due to its higher productivity. A medium-term experiment conducted in our experimental site involved a direct comparison between the two companion species (Di Candilo et al., 2008): the average yield for the period 2002-2008 was 39.6 Mg ha −1 for giant reed and 25.2 Mg ha −1 for Miscanthus. A lower but still substantial yield gap was observed on a longterm experiment conducted in Central Italy (Angelini et al.,

| Biomass composition
To our knowledge, this is the first study exploring qualitative and quantitative responses of giant reed to harvesting treatments in three subsequent years. In the case of treatments combinations growing after winter and not considering regrowth, our observations are in good agreement with the findings of Nassi o Di Nasso et al. (2011). Working in Tuscany, central Italy, these authors reported the composition pattern of cellulose, hemicellulose and lignin for the biomass of a 7-year-old giant reed stand. In the period from July to October the cellulose content remained steady around 40%, the hemicellulose around 25% and the lignin content around 8%. While the order of magnitude of these contents is consistent with our observations, in this study the cellulose content increased from June to October in 2 of the 3 years, and the lignin content increased during the growing season in all the years.
The substantial decrease in ashes content during the period from June to October that we observed in this study is consistent with Monti et al. (2008). These authors reported a very high ashes content for giant reed leaves (11.3%) and relatively low ashes content for giant reed stems (3.2%). According to Ceotto et al. (2013) the fraction of leaves on total biomass decreases progressively in the period from June to October, hence it is plausible that the observed decrease in ashes can be attributed to the increasingly higher fraction of stems on total above-ground biomass at the end of the growing season. Zegada-Lizarazu et al. (2020), working close to our location, reported a cellulose content of 41%-42% for giant reed clone obtained via mutagenesis. These contents were much higher than local clones used as a reference (36%-38%). Our study indicates that such supposedly high cellulose contents were not out of reach. In fact, using a local clone we obtained a maximum cellulose of content of 43.5% for the treatment SH in the favourable season 2018. Excluding the double cutting combination treatments which had an obvious influence of cellulose content, the SH cellulose content ranged from 39.0 to 43.5, suggesting that year-to-year variability can play a substantial role on determining the biomass composition.

| Methane production and biomass composition
We recall here that since in each reactor 1 g VS of giant reed was present, the M max value corresponds to the quantity of methane that can be produced per unit of VS (CH 4 g −1 VS) which represents the specific methane yield (SMY). The average SMY values obtained in the 3 years of experiment (153, 144 and 233 ml CH 4 g −1 VS) fall within the range reported for this crop in previous studies. In fact, several authors reported average methane yields between 144 and 234 ml CH 4 g −1 VS from giant reed (De Girolamo et al., 2014;Jiang et al., 2016;Liu et al., 2015;Shilpi et al., 2019;Yang & Li, 2014). The high year-to-year SMY variability is a fact that previous studies have not highlighted since they referred to biomass harvested on a single year. This result, however, should not be surprising because the AD process and therefore the SMY value may be affected by the compositional quality of the biomass subjected to AD and, as mentioned in the results section, crop age has significantly influenced all the compositional parameters directly involved in the AD process. Recently, Shilpi et al. (2019) reported that by changing the irrigation method from tap water to abattoir wastewater, the biomass yield of giant reed has almost tripled and, at the same time, the SMY value has gone from 146 to 234 ml CH 4 g −1 VS, showing that favourable conditions to plant growth can positively affect CH 4 yield. This is consistent with this study, in which the favourable year 2018 not only increased biomass yields but also raised the cellulose content of biomass (Table 4). According to Li et al. (2018), among all the biomass components, cellulose is the one that exerts a positive influence on CH 4 production. This is confirmed in our study, in fact in the fourth year we observed a rise in cellulose content associated with an increase in the average SMY value.
In all years SMY tended to decrease with the increase in the intra-annual biomass ageing (DORG) of harvested biomass. This result should not be surprising because as the DORG increases, the amount of lignin present in the plant tissue increases (Table 4). It is known, indeed, that AD is negatively affected by the lignin content . Moreover, the cuts with lower DORG are characterized by an average C/N of 27 (Table 5) which fell within the range considered optimal for AD (20-30; Hagos et al., 2017). Since the carbon to nitrogen ratio is considered an important parameter for the AD process (Rabii et al., 2019), it is not surprising that low lignin content and an optimal C/N allowed younger shoots to reach the highest values of SMY.
Previously, Ragaglini et al. (2014) have evaluated the effect of double harvest on CH 4 production from giant reed. These authors reported that the double harvest increased CH 4 yield up to 68% compared to the SH (258 ml CH 4 g −1 VS). In particular, the first cut showed a SMY of 320 ml CH 4 g −1 VS whereas the second cut reached 381 ml CH 4 g −1 VS. This pattern disagrees with SMY values obtained in this study, since no differences between first and second cuttings as well as a minor CH 4 differences with the SH (i.e. 10%, on average) were found. This result, however, is in agreement with the biomass composition data: the differences in biomass composition between DH1, DH2 and SH were very limited, especially in the components that are most involved in the CH 4 production in AD as cellulose and hemicellulose (Table 4); therefore, a substantial difference between DH1, DH2, or SH in terms of CH 4 production was not expected.
As far as the other parameters of the AD are concerned, the maximum CH 4 production rate (R max ) did not show significant differences between the cuts; this can be explained by the relatively homogeneous biomass composition for the different harvests. Hemicellulose is the fibre fraction that could influence the R max value as it is rapidly hydrolysed and transformed into organic acids suitable for methanogenesis; in contrast, an excess of hemicellulose could lead to rapid acidification with consequent lowering of R max (Li et al., 2018). In other plant species, high hemicellulose content characterizes the biomass in its juvenile stage (Lachowicz et al., 2019;Surendra & Khanal, 2015); in contrast, in this study hemicellulose significantly decreased with DORG only in 1 of the 3 years (Tables 4 and 6). This could explain the lack of differences between the cuts in terms of R max . As far as λ is concerned, it was longer for biomass harvested the juvenile stage. Single harvest, indeed, showed an extremely low λ value (0.7 days, on average) compared to 2.67 days, on average, referred to plants in a very juvenile stage (i.e. DH1_I and DH2_II). This could be associated with either the amount of ash and extractives, as shown in Table 4, or the polyphenols abundancy in DH1_I and DH2_II. Recently, polyphenols have been linked to a delay in the start of the AD process (Vasmara & Marchetti, 2018). However, both the double harvests DH1 and DH2 had SMY values of 180 and 178 ml CH 4 g −1 VS, on average, which are significantly higher than that obtainable from SH (167 ml CH 4 g −1 VS), albeit no further beneficial effects on the AD process were revealed. Because DH1 and DH2 slightly increased SMY values compared to SH, the biomass yield per hectare mostly influenced the methane yield per hectare. The CH 4 yield per hectare was higher for SH compared to DH1 and DH2. On the basis of the results obtained, SH determined the best compromise between SMY and methane yield per hectare, while DH2 and DH1 slightly increased the SMY value but decreased biomass yield, and consequently lowered methane yield. Our findings are in contrast, but not in contradiction with Ragaglini et al. (2014) who reported that a double harvest strategy can increase methane yield per hectare by 20%-25%. However, the conditions in which the study of these authors was conducted should be considered in comparing results: the double harvest was applied in one single year, to a crop at its fourth year of age. It is likely that the rhizomes reserve of such mature crop was abundant enough to sustain vigorous regrowth in case of double cuttings. On the contrary, in this study the crop received the double harvest for three successive years, starting from the second year of age, and having presumably poorer rhizome reserves at disposal to counteract the within-season above-ground biomass removal.
As far as year-to-year variability is concerned, our outcomes indicate that lignin content itself does not explain the large variability of both M max and R max observed across the 3 years of study. In fact, the suitability of a biomass to provide a high methane yield appears to be the overall result of the contribution of several compounds. We recall here that the biomass of the advantageous fourth year was characterized by low nitrogen, low polyphenols and higher cellulose content compared to the previous 2 years. Because the lowest M max was observed in the unfavourable dry season 2017 (third year) and the highest M max in the subsequent favourable rainy season 2018 (fourth year), it is likely that the biomass suitability to be converted in methane took advantage of non-limiting production conditions encountered by the crop.
Overall, our results are consistent with the ones reported for Miscanthus by Kiesel and Lewandowski (2017). Working in Germany, these authors, indicated that early cuttings are unsustainable despite their higher SMY values, and reported a methane yield of about 6000 m 3 per hectare for the harvest in October, within the range of the methane productivity of silage maize.

| CONCLUSIONS
Several straightforward conclusions can be drawn from our study: (i) The SH at the end of the growing season appears as the superior treatment in terms of both attainable biomass and methane yield per hectare, compared to double harvest. Yet, double harvest entails higher costs as well; (ii) in terms of specific methane yield there is little advantage in harvesting biomass twice during the growing season, because this implies an extended period of the lag phase of microbial activity; (iii) A strong year-to-year variability of giant reed performance can be expected, not only in terms of yield but also for biomass composition and specific methane yield. Then, a reliable unique value of specific methane yield for this species can hardly be indicated.