Cover cropping in perennial tree crops and vineyards has been increasing in the last decades, particularly in orchards of temperate climates without problems of water shortage and also in irrigated plantations of drier regions. The great advantages of cover cropping are the efficient reduction in soil erosion (Martínez et al., 2006; Gómez et al., 2009) and the increase of soil organic matter (Montanaro et al., 2010; Ferreira et al., 2013). Cover cropping may also improve many other physical, chemical and biological properties of soil compared to frequent tillage (Moreno et al., 2009; Ramos et al., 2010). Cover cropping is frequently used in combination with herbicides. A common groundcover management system in orchards is sod alleys and herbicide-maintained tree rows (Sirrine et al., 2008).

In the olive orchards of the Mediterranean basin, however, tillage and herbicides are still currently the dominant weed control practices due to their effectiveness in spite of their link to major environmental issues. In drought-prone regions, cover cropping in orchards and vineyards is progressing slowly. Cover crops compete for resources, particularly for water, which may reduce crop production (Rodrigues et al., 2011a; Gucci et al., 2012). Studies in dry farmed olive orchards have shown that the higher the tolerance to herbaceous vegetation, the lower the olive yields due to the competition for resources, in spite of some soil fertility parameters could increase with time (Ferreira et al., 2013). The information published to date suggests that cover cropping should continue to be extended in perennial tree crops, to prevent soil erosion and increase soil carbon sequestration, but with necessary care being taken to mitigate the risk of yield reduction.

Annual legumes have gained increasing importance in cover cropping. If the sward is composed or contains legume species, soil available N can be increased. Nodulated legumes are able to access atmospheric N₂ due to the symbiotic relationship that they establish with N-fixing bacteria (Russelle, 2008). Legumes have the potential to provide all the N required for their own growth, as well as transferring N to companion crops in intercropping systems (Hardarson & Atkins, 2003; Pirhofer-Walzl et al., 2012). N fixed by legume species can also be used by a succeeding crop in a crop rotation (Zotarelli et al., 2012; Monday et al., 2013). In perennial tree and shrub crops it is also expected that they can benefit from N fixed by the legume cover crops. Studies from coffee (Coffea arabica L.) plantations have shown that the coffee plants take up N previously fixed by an intercropped legume (Snoeck et al., 2000).

Annual pasture legumes have a great potential to be grown as cover crops in olive orchards. However, for the success of a given species or cultivar in a new environment, there is a need to test its adaptability. It should be remembered that the species and cultivars that may be used as legume cover crops were developed for pastoral systems. When they are used as cover crops in orchards, they are usually managed without grazing animal impacts in the form of treading, defoliation and excretion. Thus, the literature available for pastoral systems is of little use when the species/cultivars are used as cover crops. In semi-arid environments, for instance, the precocity of cover crops is expected to be crucial for rainfed managed orchards due to their lower competition for water to the trees. Persistence is also one of the key characters of sown legumes given the importance of self-regeneration of the sward to its economic viability. Previous studies using pasture legumes as cover crops are of great significance (Ovalle et al., 2007; Driouech et al., 2008; Mauromicale et al., 2010) but they are usually of short-term crops and have been focused on several aspects other than precocity or persistence of the sown species.

In the particular case of rainfed olive orchards, the economic aspects are of great relevance. The sector is facing a great challenge of sustainability. Olive yields are low and the olive oil price has stagnated due to the increase in acreage of high-density irrigated orchards. A study including five target areas in Portugal, Spain, Italy and Greece, has shown that the viability of these low intensity production systems is only assured if reduced opportunity costs of family labour are accepted, and olive growing is part-time (Duarte et al., 2008). According to Graaff et al. (2008) for the Trás-os-Montes target area of Portugal, three or four tested scenarios point to a high level of abandonment, while in the most positive scenario, the areas under semi-intensive, low input and organic regimes, have increased. Thus, the possibility of reducing the use of expensive fertilizer-N by growing legume species cannot be ignored.

In this study eleven legume species and cultivars were tested in pure stands in a trial carried out for a relatively long period (four growing seasons) to test their persistence when grown as cover crops in a young rainfed olive orchard managed without grazing animals. Data of groundcover percentage, DM yield and N content in the above-ground biomass were recorded. The precocity of all the sown legumes was also assessed, recording the dates of flowering and seed maturity following a decimal code for growth stages. Thus, the main goal of this work is to get data allowing the determination of what legume species/cultivars are best suited to be grown in rainfed olive orchards.

From sowing, on October 14^th 2009 to March 30^th 2010, the seed germination rates ranged from 5.3% to 48.1% in cvs. Mauro and Margurita, respectively (Fig. 2). Charano also showed a very low germination rate (9.3%). The germination rate of the other species and cultivars ranged from 25% to 40%. In the autumn/winter season of 2010/2011, the emergence from the seeds sown in October 2009 was quite low. Only Charano (8.7%) and Dalkeith (9.3%) had minimally relevant values.

The subterranean clovers Nungarin and Dalkeith started flowering usually in early March and kept blooming until late April (Table 3). Mature seeds began to appear in reasonable amounts from the middle of April. Charano started flowering earlier than any of the other species and cultivars. However, Charano showed a pronounced indeterminate flowering habit, producing flowers over a long period. The first mature seeds only occurred from early May. Even so, Charano was the third early-maturing legume in this experiment, following Nungarin and Dalkeith, the earliest maturing ones. The other species and cultivars began flowering usually during April, and significant amounts of mature seeds only appeared with the approach of the end of May.

Most of the sown legumes developed vigorously from the first growing season being dominant over the natural vegetation. Some species and cultivars reached a groundcover percentage close to 100% by mid-spring (Fig. 3). The great exception was Mauro, which did not recover from the poor emergence (Fig. 2), reaching a maximum ground cover of only 8%. Charano, also having had problems of germination, recovered very well benefiting from its large branching and a very favourable spring. Charano, presented 18% of ground cover on April 1^st and 70% one month later on May 1^st. In 2011 most species and cultivars reached high groundcover percentages earlier in the spring, probably due to the large seed bank that was formed in the previous year. Mauro showed a slight recovery but maximum ground cover was only 13%. Prolific plants appeared to be severely damaged by frost in late winter. The groundcover percentage of Prolific was the lowest among the sown legumes except Mauro. The groundcover percentage reduced dramatically for all legume species and cultivars in 2012 in comparison with the two previous years. Erica and Margurita showed particularly low groundcover, both with 12% on May 1^st. The values of Mauro continued to be very low (8%). Contea, a cultivar with an erect growth habit and a vigorous vegetative growth, reached the greatest groundcover, with a value of 77% on May 1^st. In 2013 high groundcover percentages were recorded again by most of the sown legumes. Erica and Margurita reached ground covers of 80% and 78% respectively on May 1^st. Prolific practically disappeared from the covers, being in the latter years accompanied by Frontier. Their levels of groundcover were only 16 and 14%, respectively. Fatalities from frost persisted over the years. In 2013 Mauro reached its highest level of groundcover (16%) since the beginning of the experiment.

For most of the species and cultivars the higher DM yields were found in the spring of 2010, the first season after sowing (Fig. 4). In this year, mean DM yields varied from 1.3 Mg/ha in Mauro plots to 7.8 Mg/ha in Contea. In the second (2011) and third (2012) growing seasons DM yields for all legumes species and cultivars were much lower than in 2010. In 2013, DM yields increased again and for some species and cultivars were even higher than in 2010. The low DM yields of Mauro were firstly due to the poor emergence in the first year, but a better performance was recorded in 2013 yielding 0.8 Mg/ha. With reference to Charano, although it showed a poor emergence, the plants seem to have been in a sufficient numbers to ensure high DM yields from the first year. In 2013, Charano was the most productive legume, with an average DM yield of 7.1 Mg/ha. Contea maintained a top position in all the years with DM yields of 7.8, 5.0, 4.0 and 6.8 Mg/ha in the four consecutive growing seasons. Prolific had a reasonable performance in 2010 (5.0 Mg/ha) but fell behind as the years progressed. In the last year (2013) the mean DM yield was 2.7 Mg/ha. The early-maturing subterranean clovers, Dalkeith and Nungarin, registered moderately low but relatively stable DM yields. Dalkeith, for instance, produced 3.4, 2.5, 2.0 and 4.4 Mg DM/ha in the four consecutive growing seasons.

N content in the above-ground biomass followed the general pattern observed for DM yield (Fig. 5), the differences partially being justified by the different N concentrations in the plant tissues (data not shown). N content in the above-ground dry biomass were quite high in the first growing season (2010), ranging from 20.8 kg N/ha in Mauro to 177.2 kg N/ha in Contea. In the first growing season, Dalkeith and Nungarin, the earliest-maturing and less vegetatively exuberant subterranean clover cultivars, recovered in the above ground biomass, respectively 56.6 and 52.5 kg N/ha. In 2012, N contents in the aerial biomass of all the covers were the lowest of all the four years of the study, with the mean values ranging from 12.2 kg/ha in Mauro and 71.5 kg/ha in Contea. In 2013, N content in the above-ground biomass increased again, reaching a maximum mean value of 124.9 kg/ha in Charano plots. Prolific apparently did not recover from the problems already observed in previous years. Its mean N content in the above-ground biomass was only 35.0 kg/ha. Dalkeith and Nungarin had fairly high N contents in aerial biomass with average values of 77.2 and 65.0 kg/ha, respectively.

In 2012, the year where the groundcover percentage by the sown legumes was the lowest (Fig. 3), the majority of the biomass produced in each of the plots came from natural vegetation and not from the sown legumes (Fig. 6). The only exceptions were Charano and Contea plots in which the legume component amounted to 65.9 and 58.1%, respectively, of the total dry biomass. The dominant weed in the covers was an annual grass known as Rattail fescue. In 2013, the covers were again dominated by the sown legumes, as had already been indicated by the data of the groundcover percentage. Expressed in DM percentage of the legume component over the DM of all the vegetation of the plot, the values varied from 65.0% and 90.9%, respectively in Prolific and Denmark plots.

The DM produced in the four growing seasons was statistically different among the different cover crops (Table 4). Contea produced 23.8 Mg/ha whereas Mauro only 4.5 Mg/ha. The subterranean clovers of short growing cycles (Nungarin and Dalkeith) produced DM yields on average lower but not significantly different than those of subterranean clovers of longer growing cycles (Denmark and Seaton Park). The mean DM yields of serradellas (Charano, Margurita and Erica) were very similar among them, ranging from 15.5 and 16.0 Mg/ha. Prolific produced relatively low biomass due to some problems related to frost observed in the second growing season. N content followed the pattern observed for DM yield. The samples collected in the Contea plots in the four growing seasons amounted to an average N content of 458.1 kg/ha whereas those of Mauro plots only 74.3 kg N/ha. The early-maturing subterranean clovers showed N contents higher than 200 kg/ha, while the N contents of those of longer growing cycles approached 300 kg/ha. The content of the other nutrients (P, K, Ca, Mg and B) in plant tissues followed the pattern observed for N and DM yield. N being the most limiting factor in this agrosystem, available to nodulated legumes by biological N₂ fixation, the amount of nutrient exported seemed to be proportional to DM yield. The late-maturing species and cultivars, having also higher vegetative exuberance, fixed more N and produced more biomass and, consequently, had a high demand for the other nutrients which were taken up from the soil.

All the sown legumes presented germination rates lower than 50%. The result reflects a common feature of the pasture legumes related to their naturally high levels of hardseedness. B. pelecinus, cv. Mauro, presented a particularly low germination rate (5.3%) in this study. The high hardseedness of the seeds of bisserula has been frequently reported (Loi et al., 1997; Del Pozo & Ovalle, 2009). Loi et al. (1997) found that more than 98% of seeds of several biserrula populations did not germinate after a one-summer exposure on the soil surface. The germination rate of O. compressus, cv. Charano, was also very low (9.3%). The high hardseedness of O. compressus is also well documented. Ovalle et al. (2006) found higher levels of hardseedness in O. compressus in comparison with O. sativus and T. subterraneum.

Charano started flowering earlier than any other legume. However, it presented a very long flowering period. Seed maturity occurred only in May. Margurita and Erica started flowering later and presented also a long flowering period delaying seed maturity to late May and early June. Serradellas had an indeterminate growth habit, great branching and long periods of flowering and pod formation which gave it advantage over the other vegetation if the growing conditions were favourable (Ovalle et al., 2006) as occurred in the springs of 2010, 2011 and 2013. Thus, its precocity was greatly dependent on environmental conditions. Dalkeith and Nungarin were the most precocious (early-flowering and early-seed maturing) cultivars. Mature seeds appeared by the middle of April. The other species and cultivars ranged in flowering and seed maturity from early to mid-season. Most started flowering by April and the first mature seeds appeared by mid-May. The later the occurrence of flowering and seed maturity, the shorter the length of those periods, due to the increase in temperature and rapid decrease in soil available water, due to the onset of summer.

The establishment of the sown legumes and their dominance in the covers were vigorous in the first year. The regular precipitation and a warm temperature in the autumn favoured seed emergence and the early development of seedlings. The long wet spring that followed increased plant branching and abundant flowering. The low fertility of the soils and the poor competition made by the natural vegetation also favoured the legume component. In this region, farmers restrict the fertilizers to beneath the tree canopy which keeps the fertility of the soil between rows low (Rodrigues et al., 2005). A previous study in a young orchard grown in similar agroecological conditions showed that the natural vegetation reflects the low fertility of the soil. The groundcover percentage between rows was lower than 50% at the maximum vegetative exuberance late in April (Rodrigues et al., 2009). The dominant species were terofites of short growing cycles, such as Mibora minima, Logfia gallica, Rumex acetosella and Crassula tillaea, and some perennials such as Cynodon dactylon, Convolvulus arvensis and Chondrilla juncea. Those oligotroph species could not compete with the sown legumes due to their access to atmospheric N₂.

In the second year, the sown legumes showed great development right from the winter, favoured by the large seed bank formed in the previous year. In early March, the groundcover percentage for almost all the sown legumes was already remarkably high, except for Mauro that kept a residual percentage of soil covering. Prolific showed a worse performance in 2011 than in the first year. During the winter, Prolific plants were damaged by winter frost which significantly limited their growth response during the spring.

In the third year there was observed a ‘false break’, i.e., germination-inducing rainfall events followed by death from severe drought (Nichols et al., 2007). A late-summer rainstorm caused the emergence of sown legumes, which was followed by a very prolonged drought period that led to the death of newly germinated plants. When the rains returned in the autumn, the new wave of germination was reduced further by the fact that weather was already becoming cold. The winter and the early spring were very dry in this year and germinated plants showed no significant development. Legume nodulation, empirically assessed by observing the size and colour of nodules, was clearly unsatisfactory and will have been hampered by the poor environmental conditions. In this year the ground cover was less than 50% for most legume species. Another important aspect in the failure of the legume component will have been the increase in soil available N due to the two excellent years of legume growth. There is abundant literature reporting that while the N levels in the soil are low, this favours legumes, but as N-fixing species increase the soil N levels, this facilitates the development of non-legume species (Munoz & Weaver, 1999; Van Riper et al., 2010). In this year, the covers were dominated by Rattail fescue which had benefited from the false break of sown legumes and the increase in soil available N. Rattail fescue is a weedy grass species that is prevalent within no-till/direct-seeded cropping systems where soil disturbance is minimal (Ball et al., 2008; Tarasoff et al., 2013). Rattail fescue populations have expanded dramatically in wheat production systems under no-till practices in USA, Australia and other parts of the world (Tarasoff et al., 2013). It seems that Rattail fescue can also be competitive in non-tilled managed olive orchards depending on the soil available N. Considering the critical situation of the sown legumes in the spring of the third growing season, it was decided to apply an anti-grass post-emergence herbicide to control Rattail fescue.

The fourth year was again very favourable for seed emergence and legume species growth. The seed bank that was created in the first and second years, associated with the hardness of seeds, allowed the recovery of the dominance of the cover by the sown legumes. Mauro, albeit with a low groundcover percentage, showed the highest value ever recorded. Mauro is a species of high DM production and high seed yield (Loi et al., 1997; Del Pozo & Ovalle, 2009), and therefore good persistence (Del Pozo & Ovalle, 2009) which will have allowed a slow recovery of its presence in the covers throughout the years. The presence of Prolific and Frontier in the covers lost importance over the years, apparently due to successive plant damage caused by winter frost.

DM yield and N content in above-ground biomass in the four growing seasons reached the highest average values in Contea, respectively ~ 6 Mg/ha/yr and 115 kg N/ha/yr. The earlier maturing subterranean clover cultivars, Nungarin and Dalkeith, had average DM productions and N contents in plant tissues higher than 3 Mg/ha/yr and 50 kg N/ha/yr. The other Trifolium and Ornithopus species and cultivars showed intermediate values of DM yields and N contents. Biserrulas are known for high biomass production (Del Pozo & Ovalle, 2009), but in this study the average DM yield and N content were only close to 1 Mg/ha/yr and 18 kg/ha/yr, and even though part of the biomass in Mauro plots had been due to native vegetation. In previous studies on legume management as cover crops in perennial tree crops, authors have paid little attention to DM yield. In pastoral systems, DM yield and N content are quite variable depending on pasture species used and the ecological conditions of each site (Peoples et al., 2001; Ovalle et al., 2006; Boschma et al., 2011; Ates et al., 2013), making it difficult to compare with the values found in the present study. However, the late maturing species and cultivars usually fix more N and produce more biomass than the early maturing ones (Ovalle et al., 2007) as was also recorded in this study. Peoples et al. (2001) reported that N₂ fixation amounts for pasture and crop legumes vary between 20 and 25 kg shoot N for every tone of shoot DM produced. Carlsson & Huss-Danell (2003) found 33 to 55 kg N/Mg of DM yield. In 2012 and 2013, where legumes and non-legumes were analyzed separately, the N contained per each tone of dry biomass varied from 40 to 54 kg N in 2012 and 49 to 67 kg N in 2013. This means that N concentration in plant tissues was lower than that found in the referenced studies. The reason might be the low N naturally available in this soil and the high cost in carbohydrates of the N fixed by the legume plants.

The different DM yield and N fixation of the early and late-maturing cultivars may be of particular importance in the soil management of perennial tree crops. In rainfed-managed orchards where yield goal and N demand are low (Rodrigues et al., 2011b, 2012), the earlier maturing cultivars could be the better solution. They introduce enough N into the system and transpire less water during the growing season, in particular late in the spring when soil available water is particularly scarce. In irrigated orchards, the demand for N is higher due to the higher expected yields. Moreover, drought is not a major problem in irrigated conditions. In this context, growing the later-maturing cultivars could be the better decision.

This study provides interesting clues on the management of legume cover crops in olive orchards which may also be of relevance to other fruit trees and vineyards. The most precocious legume species and cultivars seem to be particularly promising for rainfed managed olive orchards. They produce less biomass and fix less N but their growing cycle finishes earlier being almost asynchronous with the active growing phase of the olive trees, which may reduce the competition for water resources. These legumes grow actively from the autumn to mid-spring which coincides nicely with the resting period of olive.

Legume cover crops may also be a sustainable way to increase the acreage of organic farming, which has seen little progress in recent years in the region due to the lack of a natural source of N. Commercial organic additives licensed for organic farming are not a feasible option since they have high prices relative to their fertilizer value (Rodrigues et al., 2006). In addition to N fixed, the covers also provide protection against soil erosion. The seeds germinate with the first autumn rains, maintaining a living mulch during the wet season and a mulching of dead vegetation during the summer.

This work has shown that almost all of the species and cultivars maintained high persistence to the fourth growing season. Good persistence is of particular importance since it avoids regular seeding and the associated costs. A relevant aspect was the fact that the sown legumes overcame a false break without the need for new seeding. It is also important to notice that some sown legumes, namely Prolific and Frontier, suffered significant damage from winter cold, meaning that they are not suited to climatic conditions similar to those of this region.