In two field experiments in northern Sweden, we investigated if intercropping reed canary grass (RCG; Phalaris arundinacea L.) with nitrogen-fixing perennial legumes could reduce N-fertilizer requirements and also if RCG ash or sewage sludge could be used as a supplement for mineral P and K. We compared biomass production, N uptake and N-fixation of RCG in monoculture and mixtures of RCG with alsike clover (Trifolium hybridum L.), red clover (Trifolium pratense L.), goat's rue (Galega orientalis Lam.) and kura clover (Trifolium ambiguum M. Bieb.). In one experiment, RCG was also undersown in barley (Hordeum vulgare L.). Three fertilization treatments were applied: 100 kg N ha−1, 50 kg N ha−1 and 50 kg N ha−1 + RCG ash/sewage sludge. We used a delayed harvest method: cutting the biomass in late autumn, leaving it on the field during the winter and harvesting in spring. The legume biomass of the mixtures at the inland experimental site was small and did not affect RCG growth negatively. At the coastal site, competition from higher amount of clover biomass affected RCG growth and spring yield negatively. N-fixation in red clover and alsike clover mixtures in the first production year approximately covered half of recommended N-fertilization rate. Goat's rue and kura clover did not establish well at the costal site, but at the inland site goat's rue formed a small but vital undergrowth. RCG undersown in barley gave lower yield, both in autumn and spring, than the other treatments. The high N treatment gave a higher spring yield at the inland site than the low N treatments, but there were no differences due to fertilization treatments at the coastal site. For spring harvest, there were no yield benefits of RCG/legume intercropping compared with RCG monoculture. However, intercropping might be more beneficial in a two-harvest system.
Reed canary grass (Phalaris arundinacea L., hereafter RCG) has been evaluated since the mid-1980s as a perennial crop for biofuel production in both Europe and the United States (Lewandowski et al., 2003). Perennial grasses have many advantages over annual crops. Successfully established swards of perennial grasses can have lifetimes of at least 8–10 years before they need to be reseeded, and thus require less cultivation. Furthermore, they have lower requirements for pesticides and nutrients. Some of the nutrients from the shoots are recycled to the roots during autumn (Wrobel et al., 2009; Xiong et al., 2009; Heinsoo et al., 2011). Grass fields can also easily be converted to food production without substantial restoration costs, unlike for instance short-rotation Salix coppices or tree plantations. In addition, grass production techniques are well known among farmers in potential cropping areas, and appropriate machinery are readily available.
The yield potential of RCG in field experiments in northern Sweden is high, autumn dry matter yields of 8000 kg ha−1 in the first production year and 9000 kg ha−1 in second and third year have been reported by Landström et al. (1996). Corresponding yields in spring were 6500 and 8000 kg ha−1, respectively. Results from field experiments and farmers' fields in Finland and the Baltic countries show similar yields but larger variation (Pahkala, 2007; Kryzeviciene et al., 2008; Heinsoo et al., 2011).
When RCG is cultivated in Sweden, the biomass produced during the first establishment year is small, and usually it is not harvested. In the second year, the biomass is usually cut in late autumn, but the harvest (removal of the biomass from the field) is delayed until spring (Larsson et al., 2006). This technique avoids costly drying when the biomass is collected immediately after cutting and results in lower ash contents (partly due to losses through leaching of Cl, Na and K), thereby improving its quality as a fuel (Burvall, 1997). Nevertheless, production costs (especially for establishment, fertilization and harvesting) are high compared with those of biofuels from forest residues (Olsson et al., 2001).
Reed canary grass grown as a biofuel crop has, to date, been grown as a monoculture (Pahkala, 2007). However, a potentially sustainable way to reduce costs for N fertilizers might be intercropping of the grass with nitrogen-fixing legumes, as is common in forage production. Palmborg et al. (2005) have shown that when legumes are intercropped with perennial grasses, sufficient yields can be obtained with low inputs and low risk of N losses. Also, bioenergy production experiments in Lithuania have shown that grass-legume mixtures without N fertilization can provide higher yields than pure stands of grasses with N fertilization under normal weather conditions (Kryzeviciene et al., 2008).
The RCG fuel production system is quite different from standard forage production systems and poses new challenges. From a quality point of view, RCG biomass is preferable to legume biomass because of the low nutrient concentrations and low decomposition rate in RCG biomass (Pahkala & Pihala, 2000). Thus, it is necessary to promote a high RCG proportion. In forage production, N fertilization can reduce the clover proportion in a grass/clover ley by enhancing the competitive ability of the grass (Carlsson & Huss-Danell, 2003). Another strategy to reduce fertilization costs is to use waste material that would otherwise be deposited with a high cost. Combustion of grass leads to the production of relatively large amounts of ash (Burvall & Hedman, 1994; Burvall, 1997) that contains sufficient concentrations of nutrient elements for recycling as fertilizer in agriculture.
Deposition of sewage sludge in landfills in Sweden is prohibited from 2005, following EU legislation (Naturvårdsverket, 2004). For that reason, Swedish municipalities have to find other uses for the sewage sludge. Thus, its use in energy production could solve a serious waste disposal problem.
Sewage sludge can be a good fertilizer if it is not contaminated by heavy metals or other industrial waste compounds. Hence, recycling nutrients from noncontaminated sewage sludge could considerably enhance sustainability (Ribbing, 2007). However, there has been substantial resistance among Swedish farmers to use sewage sludge in fields for food and feed production. As fertilizers, both RCG ash and sewage sludge are considered to be predominantly P sources since they contain insufficient amounts of N and K, although the level of plant-available P is sometimes low in sewage sludge due to precipitation by Fe and Al compounds (Krogstad et al., 2005). In addition, both materials are sources of Ca, Mg and micronutrients and can influence the pH of soil.
The aim of this study was to investigate the possibility to reduce fertilizer requirements in RCG cultivation by intercropping with perennial legumes and fertilizing with either RCG ash or sewage sludge. The legume species included in the study are red clover (Trifolium pratense L.), alsike clover (Trifolium hybridum L.), goat's rue (Galega orientalis Lam.) and kura clover (Trifolium ambiguum M. Bieb.). Red clover is the most commonly cultivated and most productive legume in forage leys in Sweden, and alsike clover is also productive although not as common today (Frankow-Lindberg, 2005). Mixtures of goat's rue and RCG for bioenergy purposes have previously been evaluated in Lithuania (Jasinskas et al., 2008; Kryzeviciene et al., 2008). Kura clover was included in the study since intercropping of RCG and kura clover has been evaluated, with good results, in forage production in the United States (Contreras-Govea et al., 2009).
We hypothesized the following:
Establishment of RCG would not be negatively affected by competition when intercropped with the perennial legumes or undersown in barley (Hordeum vulgare L.).
Higher N fertilization would decrease the amount of legumes in the total biomass due to increased competition from RCG.
N-fertilization rate could be substantially reduced by intercropping with legumes, without loss of total biomass yields, since N-fixation by the legumes in symbiosis with Rhizobium would compensate for the lower mineral N inputs.
RCG ash/sewage sludge can substitute mineral P and K fertilizers without loss in biomass production.
Winter losses caused by decomposition, leaching and fragmentation should be higher in mixtures with legumes than in RCG monocultures.
Materials and methods
Experiments to assess the effects on biomass yields of intercropping RCG with legumes or barley were established at two sites in northern Sweden: Röbäcksdalen research station (63°48′N, 20°14′E) Umeå and Ås research station (63°15′N, 14°34′E) near Östersund. The soil at Röbäcksdalen is a loamy sand in the top soil (4% clay, 74% sand and 22% silt) and a silt loam in the subsoil. At the start of the experiment, the of the top soil (0–20 cm) was 6.0 and plant available soil nutrients extracted by ammonium lactate (AL) were K-AL 7.3 and P-AL 8.5 mg 100 g−1 dry soil (Swedish standard, SS 028310). The soil at Ås is a clayey moraine with 14% clay, 18% silt, 16% sand and 26% stones and gravel (Andersson & Wiklert, 1977). Before the start of the experiment the of soil at the site was 6.4, and its AL-extracted K and P contents were 8.9 and 14.0 mg 100 g−1 dry soil, respectively.
The experiments at Ås and Röbäcksdalen were established in spring 2008 and 2009, respectively, in both cases with a split plot design, in four blocks, with fertilization as main treatment and species mixtures as subtreatments. In total, there were 84 plots (3 × 10 m) at each site. RCG was grown in monoculture and intercropped with alsike clover, red clover, goat's rue and red clover + goat's rue at both sites. In addition, kura clover was intercropped with RCG at Röbäcksdalen. Two different seed rates of goat's rue were applied at Ås. The varieties used and the amounts of seed sown are listed in Table 1.
Table 1. Seed rates for reed canary grass (RCG) and legumes used in the experiments at Röbäcksdalen and Ås
To establish an effective nitrogen fixation, the goat's rue and the kura clover were inoculated with Rhizobium. At Ås, the goat's rue seeds were mixed with inoculate soil before sowing. At Röbäcksdalen, by mistake, the inoculate soil was not mixed with the goat's rue and kura clover seeds but was watered into the plots after sowing when the mistake was discovered.
In one treatment at Ås, barley was sown as a nurse crop to pure RCG, and then harvested as a whole-crop in early September. As no chemical weed control was applied, the other plots were cut at a stubble height of 22 cm on the same day to reduce the abundance of annual weeds. Duplicate sets of RCG + red clover plots were established at Röbäcksdalen, one of which was harvested in mid-October in the sowing year, to promote the winter survival of the clover plants. The other plots were left uncut. At each site, three fertilization treatments were also applied, to quadruplicate plots sown with each species mixture or RCG alone (Table 2). Treatment ‘High N + PK’ followed current recommendations for northern Sweden. ‘Low N + PK’ was fertilized with half of the recommended N rate, P and K as above. In treatment ‘Low N + sludge/ash’ P and K were supplied with sewage sludge or RCG ash in the establishment year. Ammonium nitrate (NH4NO3) was used for N supply, and triple superphosphate (Ca(H2PO4)2) and potassium chloride (KCl) were used for P and K supply, respectively.
Table 2. Amounts of plant nutrients in kg ha−1 applied with mineral fertilizers and sludge or ash for each site and year
Botanical composition, nitrogen analyses and N-fixation calculation
In autumn 2009 (Ås 26 August, Röbäcksdalen 22 September) and 2010 (Röbäcksdalen 23 August), the vegetation in 50 cm × 50 cm subplots was sampled by manual cutting, stubble height 2 cm. At the time of sampling, the plants were fully developed and clovers were withered to some extent the first production year. To avoid edge effects, the area around the plots was left uncut until some days before the sampling, and the samples were taken 50 cm from the edges of the plots. The biomass (including dead plant material) was sorted into each sown species and weeds, and the dry weight of each fraction was measured to determine the botanical composition. The botanical composition is presented as g m−2 as the sample plots were only 0.25 m2 large. The sown species and larger weed samples were then milled to a fine powder with a ball mill (Retsch MM 200, stainless steel milling chamber). The plant N concentration and proportion of the stable isotope 15N of each sample were analysed using an ANCA-SL mass spectrometer coupled to a Sercon 20-20 isotope-ratio mass spectrometer (Sercon, UK). The δ15N value denotes parts per thousand (‰) deviations in the sampled plant from the ratio of 15N : 14N in atmospheric N2 (Högberg, 1997). The acquired data were used to calculate the nitrogen fixation rate using both the 15N natural abundance and the N difference methods. The 15N natural abundance method is the preferred method for measurement of N-fixation in intercrops of legumes and non-N-fixing plants (Unkovich et al., 2008). The method requires that the isotope 15N is enriched in the soil compared with the atmosphere, which results in a higher δ15N value in the nonfixing reference plant than in the legume. The proportion of N derived from the atmosphere (Ndfa%) in the legumes was calculated using the difference between their respective δ15N values and that of the non-N-fixing RCG in conjunction with published B factors (the δ15N value of each legume species when acquiring N2 solely from the atmosphere).
The nonfixing plants (RCG) were from the same plots as the legumes. The B factors used were −1.3‰ for red clover and −1.2‰ for alsike clover, as acquired in a greenhouse study in Umeå (Carlsson et al., 2006). Specific B factors have not been published for kura clover and goat's rue. The B-values for these species were set to −1‰, a common value for perennial legumes (Unkovich et al., 2008). At Ås, the δ15N of RCG was lower in several plots than the 2‰ lower limit recommended for reference species by Unkovich et al. (2008) for use of the 15N natural abundance method. Because of this, we also calculated the N-fixation rate by the N difference method:
N difference = kg N in intercrop total biomass − kg N in RCG monoculture biomass from the same fertilizer treatment plot.
Late in autumn, larger subplots (1.5 × 7.5 m) were cut with a plot harvester (Haldrup) at Ås 28 October 2009 and at Röbäcksdalen (1.5 × 6.0 m) 22 September 2010. To imitate delayed harvest as applied in practice, the biomass was cut with the plot harvester in autumn and replaced manually on each plot. The biomass was harvested (removed from the field) in May the following spring with the plot harvester, as soon as the field was sufficiently dry to support it. At both occasions, the biomass was weighed, and a subsample was taken out and dried at 60 °C in a forced air dryer for 48 h to determine dry matter content.
Data from each site and each year were analysed separately. Effects of the treatments on botanical composition, biomass yields, N contents and 15N values were analysed by anova as a split plot design with four replicates (NCSS 2007, Hintze, 2009). Fertilization treatment was assigned as main plots and species mixture as subplots. Fertilization treatment and species mixture were considered fixed factors and block as random. The interaction fertilization × species mixture was included in the model. Effects were considered significant at P < 0.05. Normal distribution of the data could not be rejected and no transformation was done. When effects were significant, means were separated by post hoc test using Tukey–Kramer Multiple-Comparison Tests for all pairwise differences between the means, or Bonferroni correction for comparisons with the control treatment (RCG monoculture) when applicable.
Botanical composition and nitrogen analyses
At Röbäcksdalen, in the small plot samples in autumn 2009, total amounts of biomass of the mixtures with red clover and alsike clover were higher than RCG in monoculture and the mixtures with goat's rue and kura clover (Fig. 1). On average, the high N treatment resulted in the highest total amount of biomass, significantly higher than the low N + ash treatment. Alsike clover was the highest yielding legume species. Goat's rue and kura clover contributed marginally to the total biomass. The anova showed no significant differences in RCG or weed biomass in comparison between seed mixtures (Fig. 1a) or between fertilization treatments (Fig. 1b). However, later in autumn, in mid-October, when the aboveground clover biomass had wilted due to severe frost, we observed that the RCG was still green, and the grass was tallest where it was intercropped with alsike clover, the most abundant legume.
The concentration of N in the RCG was higher in the mixture with alsike clover than in the other mixtures and in the RCG monoculture (Table 3). The average N concentrations were 3.4%, 1.6%, 3.3% and 1.9% in red clover, kura clover, alsike clover and goat's rue, respectively. The total amount of N in the biomass (including weeds) was lowest in pure RCG and in the mixtures with goat's rue and kura clover, intermediate in the red clover mixtures and highest in the alsike clover mixture. There were no differences between N fertilization levels (Table 3).
Table 3. Yield of nitrogen (kg ha−1), nitrogen concentrations and δ15N in the different species at Röbäcksdalen in autumn 2009, the establishment year (standard deviation in brackets). Different letters denote significant differences among means. Abbreviations for seed mixtures as in Table 1
The δ15N value for RCG was on average 9.7‰, and higher when intercropped with alsike clover than when intercropped with kura clover (Table 3). The δ15N values were −0.1‰ for red clover and alsike clover, and for both kura clover and goat's rue 4.1‰. The amounts of N derived from the atmosphere by the legumes estimated by the N difference and the 15N natural abundance methods are shown in Fig. 2a, b. Alsike clover fixed more N than red clover. The kura clover and goat's rue N-fixation rates were negligible. There were no differences in this respect between the fertilization treatments.
In the first production year (2010) at Röbäcksdalen, there were no effects on total amount of biomass between the treatments in autumn of either fertilization treatment or species mixture. However, the botanical composition of samples cut in each plot (g m−2) showed that lower amounts of RCG biomass were obtained from the mixtures with red clover or alsike clover. The biomass yields of kura clover and goat's rue were very small as most plants did not survive in the winter (Fig. 1). There were no between-fertilization treatment differences either in amounts of RCG or amounts of legumes. The amounts of red clover and alsike clover did not differ between the treatments. Thus, neither the amount of clover seed at establishment nor removal of the biomass produced during the establishment year had any impact on measurements the following year. The amount of weeds did not differ between treatments.
The concentration of N in RCG was higher in mixtures with red clover (1.4%) and alsike clover (1.4%) compared with RCG monoculture (1.0%), but there were no differences between fertilization treatments in this respect (Table 4). The average N concentration was 1.6% in both red clover and alsike clover, and there were no differences in this variable between the seed mixtures or fertilization treatments. The total amount of N in the biomass (including weeds) was higher in the mixture where red clover + goat's rue was sown (at this time almost exclusively with RCG + red clover) than in the RCG monoculture and the mixtures with goat's rue and kura clover. The δ15N value for RCG was on average 4.0‰, with lower values where intercropped with red clover or alsike clover (Table 4). The δ15N value for red clover and alsike clover were close to zero. A higher N-fixation rate was estimated, by the 15N natural abundance method, in the mixture with red clover, goat's rue and RCG (42 kg N ha−1) than in the alsike clover mixture with RCG (24 kg N ha−1) (Fig. 2c, d). For the N difference method, no significant differences were found due to large variation within treatments.
Table 4. Yield of nitrogen, nitrogen concentrations and δ15N in the different species at Röbäcksdalen in autumn 2010 (standard deviation in brackets). Kura clover and goats rue are not analysed. Different letters denote significant differences among means. Abbreviations for seed mixtures as in Table 1
Reed canary grass
Red clover/Alsike clover
kg N ha−1
kg N ha−1
kg N ha−1
RCG + RC autumn
RCG + RC
RCG + GR
RCG + KC
RCG + AC
RCG + RC + GR
High N + PK
Low N + PK
Low N + ash
At Ås, the botanical composition of vegetation in the plots was not rigorously examined in the sowing year (2008). However, visually estimated proportions of RCG and legumes were 30–40% and 10–20%, respectively. The remaining constituents were mainly annual weeds. The total biomass of samples cut in autumn in the year after sowing (2009) was lower where RCG was undersown in barley (Fig. 3). Among the legumes, alsike clover had higher biomass yield than the others. In the three species mixture with RCG, red clover and goat's rue, red clover and goat's rue contributed 63% and 37% of the legume biomass, respectively. The amount of weeds was higher where RCG had been undersown in barley compared with the other mixtures. No differences between the fertilizer treatments were significant.
The concentration of N in the RCG biomass was higher in plots undersown in barley than in the monoculture and the red clover and/or goat's rue mixtures (Table 5). On average, over all mixtures, the RCG N concentration was also higher in plots subjected to the high N fertilization treatment. The N concentration in goat's rue was on average 2.5%, in alsike clover 2.0% and in red clover 2.1–2.6%, with no between-fertilization treatment differences. The δ15N values were lower in goats rue and in the RCG-red clover mixture than in red clover grown in mixture with RCG and goats rue. The average δ15N of RCG ranged from 2.0 to 3.1‰ with highest value where it was undersown in barley. The δ15N was not influenced by the N-fertilization level in either legumes or RCG.
Table 5. Yield of nitrogen, nitrogen concentrations and δ15N in the different species at Ås in autumn 2009 (standard deviation in brackets). Different letters denote significant differences among means. Abbreviations for seed mixtures as in Table 1
Reed canary grass
kg N ha−1
kg N ha−1
kg N ha−1
RCG + barley
RCG + RC
RCG + GR 8
RCG + GR 16
RCG + AC
RCG + RC + GR
High N + K
Low N + K
Low N + K + sludge
The total amount of N in the biomass (including weeds) was higher in plots given high N fertilization than in those given the low N or low N + sludge treatments. Among the species mixtures, the alsike clover and the red clover + goat's rue mixtures had higher total amount of N than the RCG monoculture. The 15N abundance method gave lower estimated N-fixation rates than the N difference method (Fig. 4). Using the N difference method, the estimated N2 fixation rate in alsike clover was higher than in one of the mixtures with goat's rue (RCG + GR8), but there were no differences between species when using the 15N abundance method. No significant differences between fertilization treatments in this variable were detected using either method.
Dry matter yield from machine-harvested plots in late autumn and spring
At Röbäcksdalen, the average DM yield of biomass from the large plots in autumn 2010 was 7000 kg ha−1. None of the legume mixtures differed from the RCG monoculture when Bonferroni test was applied, and no significant between-fertilization treatment differences were detected (Fig. 5). However, at harvest (removal from the field) in mid-May in the following spring, the mixtures with red clover gave a lower DM yield of the remaining biomass than the RCG monoculture. The mixture RCG + RC autumn showed higher proportional loss of biomass than the pure RCG.
At Ås 56 of the large plots (1.5 × 7.5 m) were cut in late October 2009, but the other 28 were cut in the following spring, since the autumn cutting was interrupted by snowfall. Only data from the plots cut in autumn were used for the statistical analyses as all seed mixtures and fertilization treatments were represented in 2–3 replicates among the 56 plots. As shown in Fig. 6, the only difference in measured DM yields in the autumn was that it was lower for RCG undersown in barley (6900 kg ha−1) compared with RCG in monoculture (10 700 kg ha−1) when using Bonferroni test. However, yields tended to be higher following the high N-fertilizer treatment (P =0.09). The harvested biomass yield in the following spring was also lower where RCG was undersown in barley (4000 kg ha−1) compared with RCG monoculture (7300 kg ha−1). The high N-fertilization treatment had higher yields than the half N treatments.
This is the first study in which RCG for bioenergy production has been intercropped with perennial legumes where the nitrogen fixation rates have been measured under field conditions. The modified delayed harvest method (cutting the biomass in late autumn, leaving it on the field during the winter and harvesting in spring) has also not been evaluated in combination with intercropping in any previously reported field experiments.
At Röbäcksdalen, data of botanical composition clearly show that competition from red clover and alsike clover restricted the growth of RCG during the first production year (2010) (Fig. 1c). In contrast, at Ås there were much less clovers and they did not significantly affect the growth of the RCG (Fig. 3a). Similarly, in an American study when different legume species were interseeded to switchgrass-dominated swards, legume species with large biomass production restricted growth of switchgrass more than less productive legumes (George et al., 1995).
Nitrogen fertilization is well known to favour the grass component of leys compared with the legume component (Whitehead, 1995). However, in our experiments, the nitrogen fertilization level only marginally affected both proportions and total amounts of RCG and legumes (Figs 1b, d and 3b). Thus, our second hypothesis that the N-fertilization level would affect the competition between RCG and legumes was not supported. A possible explanation to this is that growth of RCG can be more restricted by water availability than nitrogen availability (Kätterer et al., 1998). There was more precipitation in July during the first production year at Ås than at Röbäcksdalen, and the top soil at the latter site is sandy, and hence dries easily. Thus, the growth of the RCG might have been more restricted by water availability at Röbäcksdalen, accentuated by the competition for water with the legumes. Furthermore, high temperatures at Röbäcksdalen in July 2010 might have caused the growth of the above ground biomass of RCG to end earlier in the season. Such a temperature effect was observed in a greenhouse experiment by Zhou et al. (2011). The differences in severity of the legume competition between the sites could also have been related to different conditions for overwintering of clover.
The poor performance of goat's rue and kura clover at Röbäcksdalen could have been due to failure of accurate inoculation. As these species have never been cultivated on the experimental field before, successful inoculation should be crucial to establish an effective nitrogen-fixing symbiosis between legume plants and Rhizobium. The good establishment of goat's rue at Ås also could be explained by a higher soil pH because goat's rue prefers neutral soil to acidic (Raig et al., 2001). These first 2 years the goat's rue did not compete seriously with RCG in any way. In a similar study in Lithuania RCG also competed well with goat's rue, and perennial lupine during the first 2 years (Jasinskas et al., 2008; Kryzeviciene et al., 2008).
In the following discussion, we compare estimates of N-fixation rate using the 15N natural abundance method for plots at Röbäcksdalen, but the N difference method at Ås, as at that site, criteria for using the 15N natural abundance method were not met. The δ15N of RCG was lower in several plots than the 2‰ lower limit recommended for reference species by Unkovich et al. (2008). At both sites, when RCG was intercropped with red clover or alsike clover, the N-fixation rate in the first production year was similar to the reduction of N-fertilization rate in the low N treatments. In spite of this, there was no significant total yield benefit when RCG was intercropped with legumes, not even in Ås (Fig. 6a), where RCG did not suffer from severe competition with the legumes. However, at Ås, where the difference method was used, the accuracy of the N-fixation estimates was low due to a large variation in N content within treatments (Table 5). The reason for the lack of yield benefits with legume intercropping could be that the clovers were more or less wilted already in August, when the botanical composition was determined. Thus, they probably contributed little to both late autumn and spring yield. However, lower δ15N values of RCG at Röbäcksdalen when it was intercropped with legumes than when grown in pure stand indicated that N fixed by the legumes was also used by the RCG. Clearly, the recommended N-fertilization level was too high. The N-fertilization rate recommended in Finland, 60–80 kg ha−1 (Pahkala et al., 2003), would have been more adequate, at least for the growing seasons in question. A previous analysis of intercropping RCG with lucern, goat's rue and sweet clover found that it was not profitable to fertilize the RCG and lucern mixture with N (Tuvesson, 1997). However, Tuvesson (1997) did not use the delayed harvest method and the proportions of the species were not reported, so the findings cannot be directly compared with our results. However, that study along with successful intercropping of switchgrass with legumes for summer harvests (Springer et al., 2001; Bow et al., 2008) imply that RCG/legume intercrops might be beneficial as feedstock for biogas or ethanol production. In another study, by Jasinskas et al. (2008), intercrops were not fertilized at all, and there goat's rue tended to dominate over RCG the third harvest year. Under our climatic conditions and in a spring harvest system, the red clover and alsike clover will probably only persist during the first two growing seasons, while the goat's rue is more long-lived once it is established. Thus, in coming years, we will study the long-term effect of goats rue at Ås where it established well.
In these first 2 years, at least, the replacement of mineral fertilizers with ash or sewage sludge had no significant effects on the production of any species. However, the fertilization effect of ash or sewage sludge is difficult to evaluate after the short time covered in this study. Swedish long-time fertility experiments (Carlgren & Mattsson, 2001) have shown that often there are no yield effects of differences in PK fertilization, and effects on soil P availability measures takes several years or even decades to develop. Effects on crop uptake (concentrations) of K also are enhanced with time (Andersson et al., 2007). Mineral nutrients in soil and total biomass will be analysed in coming seasons to detect any differences between treatments that may subsequently appear.
The RCG yield level in Ås was higher than the average yield for northern Sweden, but the yield at Röbäcksdalen was lower (Landström et al., 1996). An important aspect of spring yield level is the occurrence of winter losses. Field losses occur during the winter due to degradation of the biomass, of which the degree depends on winter conditions, and mechanical damage at harvest (Landström et al., 1996). Herbaceous legumes are generally more easily degraded than RCG because of their higher N content (Wagger et al., 1998). In Finland, an experiment including delayed harvests of RCG, tall fescue, meadow fescue and goat's rue (all species in monoculture) showed that the spring harvest system was much more favourable for RCG than for the other species (Pahkala & Pihala, 2000). The winter losses were considerable in our experiments, particularly at Röbäcksdalen, where, due to very large winter losses, mixtures with red clover yielded less than the RCG monoculture (Fig. 5). The alsike clover and red clover/goats rue mixtures also had large winter losses even though the spring yields were not significantly lower than for RCG monoculture. The lack of species mixture differences at Ås was probably due to much lower proportions of legumes. Mechanical losses at harvest in these experiments have not been estimated, but as the plot harvester we used is not designed to lift cut biomass from the ground, such losses may be of some importance.
Thus, we can conclude that at least during the first two growing seasons, there were no advantages of RCG-legume intercropping for spring harvest. However, it might be more advantageous in a summer harvest system designed to produce biomass for biogas or ethanol production.
We thank Lars Ericson for support at the start of the experiments and Pernilla Bärlund, Per-Erik Nemby and Kent Dryler and the crew at the field stations for good cooperation. We also thank Sigrun Dahlin for useful comments on the manuscript. The projects were financed by the Swedish Thermal Engineering Research Association, Kempestiftelserna, the European Regional Development Fund, the Swedish Research Council Formas and the Swedish Farmers' Foundation for Agricultural Research.