Rainfed arable areas occupy over 28 million hectares in southern Europe, which are mainly devoted to cereal-industrial crop rotation ( Eurostat, 2015 ). Nowadays, this traditional practise is facing the urge to modify the current crop management and to diversify the crop production. They are especially influenced by the stagnating products price, the need to improve and achieve more balanced agro-ecosystem through the crop rotation ( Flower et al. , 2012 ), as well as the accomplishment of the new Common Agricultural Policy (CAP) greening measures. The most noteworthy measure included is the 'crop diversification', where the farm holdings with an arable land of 10-30 ha have to cultivate at least two crops, and farms exceeding 30 ha must have at least three crops, with no more than 75% of the cultivated area assigned to one crop. Moreover, Southern Europe is also a great livestock region with an important number of heads of cattle (over 14 million) and sheep (over 24 million) next to the rainfed arable crops areas. The major constraint of the livestock development in these regions is the dependence of imported powerful food sources due to the limited on-farm production of forage, especially in the dry years of Mediterranean countries ( Sadeghpour et al. , 2014 ). Therefore, it is necessary to satisfy the increasing demands for livestock ( Alizadeh & da Silva, 2013 ) and there is the possibility of obtaining supplementary forage sown with arable crops in Southern Europe. Intercropping of winter grasses with annual legumes is extensively used in Mediterranean areas for forage production under rainfed conditions ( Lithourgidis et al. , 2006 ). This system increases fodder yield concerning dry weight thanks to the grasses ( Hashemi et al. , 2013 ) and improves feed quality owing to the higher crude protein concentration of legumes ( Umunna et al. , 1995 ).

The environmental conditions ( Gierus et al. , 2012 ) in addition to the seeding ratios of the selected species ( Caballero et al. , 1995 ) are crucial factors affecting yield and quality of intercropping systems. The vetch ( Vicia sativa L.)-oat ( Avena sativa L.) mixture is the most traditionally used in Southern Europe and many studies have reported that these species are the most appropriate for mixture in Mediterranean regions ( Lithourgidis et al. , 2006 ). Nevertheless, common vetch is described as susceptible to broomrape ( Orobanche crenata Forsk.) ( Gil et al. , 1987 ), one of the most serious legume pests in Mediterranean areas. However, findings of Nadal & Moreno (2007) revealed that narbon bean ( Vicia narbonensis L.) is a high-yielding annual legume with an inbreed line resistant to broomrape. It would be a good alternative for cropping in broomrape-infested areas under rainfed Mediterranean conditions. In these areas it is also known for being a promising crop in rotation with wheat ( Durutan et al. , 1990 ) and a good source of protein fodder for cattle ( van der Veen, 1960 ) and sheep ( Jacques et al. , 1994 ). On the other hand, Flower et al. (2012) showed that black oat ( Avena strigosa Schreb.) is a winter cereal crop with a more rapid growth and higher biomass production than other oats, widely used in Australia ( Flower et al. , 2012 ) and USA ( Reeves & Price 2005 ). It has been used not only for grain production and as a green cover crop due to its weed control potential ( Santos et al. , 2010 ), but also to obtain forage for dairy cows feed ( Salgado et al. , 2013 ).

Given the capacity for high dry matter and protein yield documented in rainfed conditions of each species separately, mixtures of narbon bean with black oat could be a promising alternative for increased forage production in these areas. However, despite the fact that there is much published information on the forage yield and quality of legume-grass intercropping system worldwide, no information exists on the adaptation of intercropping black oat-narbon bean, seeding ratios, forage yield and quality both in Mediterranean regions and overseas.

In order to provide a new viable rotation crop alternative based on the black oat-narbon bean mixture, two main objectives were considered in this study: (i) determine if a stable forage mixture for balanced growth of the two species can be formed by narbon bean and black oat at different seeding ratios (65:35, 50:50 and 35:65) compared with the pure stands and the common vetch-common oat mixture (65:35), and (ii) define if the studied mixture can achieve a high forage yield and quality production under rainfed Mediterranean conditions.

Plant density and ground cover were evaluated for each component of the mixtures separately. Plant density was estimated counting the established plants in each seeding rate treatment 22 and 32 days after sowing (DAS) in 12 randomly selected 0.1 m ² area of each plot during the three years. Plants with the first pair of true leaves were considered established plants. Plant ground coverage was determined photographically based on the methodology described by Laflen et al. (1981) . Five random photographs per plot were taken with a camera at the height of 1.5 m above the frame during 5 dates from January to April over the cover (from 60-65, 85-90, 105-115, 130-135 to 145-160 DAS). Total percentage ground cover was determined by counting the different species coverage in each treatment using a digital grid with 100 crossing points. The template points coinciding with green parts of each species came to their percent coverage in each photo.

Phenological events for species were determined based on the maximum and minimum daily temperatures to calculate growing degree days (GDD) with a 0°C baseline temperature for narbon bean ( Mwanamwenge et al. , 1999 ) and black oat ( Bauer et al. , 1984 ) and based on the BBCH scale ( Lancashire et al. , 1991 ) of faba bean and cereals.

Dry matter (DM) yield for the different mixed treatments was also assessed at the time of harvesting. All plants were harvested on the same day in mid-April each year, on the closest stage to the optimum phenological development in both species according to Muslera & Ratera (1991) , at the flat pods stage for legumes (71-78 stage in the BBCH-scale) and early milk stage for grasses (71-73 stage). Two biomass samples were randomly collected from a 0.5 m ² area of each plot by cutting at 5-7 cm above ground level with a sickle. Species were separated by hand, and the weight after drying for 48 h in a forced air oven at 70 °C was determined.

The effect of competition between the two species used in the mixture was calculated using the competition ratio (CR) which indicates the degree to which one species competes with the other in an intercrop. The CR is calculated according to the following formula ( Lithourgidis et al. , 2011 ):

The success of the mixture installation depended on the success of each species separately, which varied in relation to the narbon bean or black oat proportion in the mixture and the weather conditions in each location (Fig. 2). Vicia plant density results indicated the highest values in Alameda year 2 (ranged from 17 to 25 % higher) and more similar results every year in Tomejil. However, Avena plant density showed the highest values the year 1 (about 56-82 and 79-89 % higher in Alameda and Tomejil, respectively). Vicia plant density values of C were greater than those in the proposed mixture treatments all years and locations. However, Avena plant densities for D4 and D3 were higher or similar to C in most of the cases.

The temporal evolution of the ground coverage sole crops registered the highest values compared to the mixed treatments, with significant differences on most of the sampling dates (Fig. 3). Results for year 1 showed that ground coverage of narbon bean-mixtures decreased (6 % in Alameda and 3 % in Tomejil) from the 50 stage in Alameda and an early 30 stage (after 796 GDD accumulated) in Tomejil. Meanwhile, black oat increased (38-57 % in Alameda and 21 % in Tomejil) coinciding with an extended late winter-early spring drought (5.8 and 12.8 mm accumulated from February to March, respectively). By contrast, the narbon bean ground cover increased as black oat mixes scarcely evolved from the 35-39 stage (995-1076 GDD) during the year 2 in both locations. It was probably due to the high rainfall recorded from February to March (354 and 239 mm in Alameda and Tomejil, respectively). A similar trend occurred the year 3 in Alameda, with 139 mm recorded from February to March. During year 3 in Tomejil (374 mm accumulated), both species registered the highest growth stages of the study period at the time of harvest (73 black oat stage and 75 narbon bean stage with 1700 GDD). The lowest growth stage at both locations occurred the year 1 (219-283 mm) for narbon bean (71 stage with 1479-1555 GDD) and the year 2 (636-784 mm) for black oat (69-70 stage with 1529-1633 GDD). Narbon bean species grown at D2 and D3 and black oat species at D4 showed the lower decrease compared to their pure stands in all case studies.

The study of the species and their interaction in the mixture disclosed that it is possible to get a balanced forage mixture with narbon bean and black oat in rainfed Mediterranean conditions. However, the variables studied in order to characterize the intercropped narbon bean and black oat species (plant density and ground cover evolution) displayed the importance of the 'year' effect on the establishment and development of the species in mixture, such as was previously reported by Gierus et al. (2012) . Although variation due to year was expected, differences of rainfall recorded were so different over the three experimental years that a negative influence of the drier year 1 for legumes and the wet year 2 for grasses was verified. Narbon bean plants were affected by drought conditions (<300 mm) and winter frost period registered from the stem elongation/inflorescence emergence stage, as also occurred in white clover-mixtures studied by Annicchiarico & Tomasoni (2010) . Conversely, black oat showed lower development and growth during year 2, when total rainfall values were significantly higher than the 10-year average (784 mm and 636 mm in Alameda and Tomejil, respectively). The amount of rainfall accumulated, especially in November, February and March led to waterlogging during critical growth stage period as emergence, heading and flowering which reduced root growth and penetration and therefore production of tillers and fertile heads as was also reported by Watson et al. (1976) . Taken together to the highest water retention capacity of the Vertisol clay soil, the decrease of oats plant density in Tomejil was high due to the adverse effects of excess moisture, as was previously reported by Urban et al. (2015) . However, total rainfall amounts the year 3 (443 mm and 374 mm in Alameda and Tomejil, respectively) were statistically similar to the 10-year average rainfall. Both species showed a positive ground cover evolution accompanied by the highest phenological growth stage at the time of harvest and an accumulation of 1700 GDD, which was similar to the requirements showed by Mwanamwenge et al. (1999) in other Mediterranean climate zones.

Besides the 'year' effect, our results also showed the 'seeding ratio' important effect on the final mixture due to the individual species growth and the interaction between them ( Caballero et al. , 1995 ; Lithourgidis et al. , 2006 ). In spite of the highest plant density and ground coverage reached by sole cropping treatments, the 65:35 and 50:50 seeding ratios showed the most similar trend to pure stand for the intercropped narbon bean. Moreover, the 50:50 and 35:65 displayed the greatest establishment and growth for black oat in mixture, without significant differences with the control during year 2 (even higher) and year 3. Our results also highlighted that a lower seeding rate and plant density of the proposed mixture compared to the control produced similar forage yield. This fact implies lower installation costs for farmers and an additional advantage of this mixture to include in the cropping systems since the economic feasibility of intercropping systems is ultimately determined by their monetary benefits, as was pointed out by Lithourgidis et al. (2011) .

The absence of significant differences in CR in both components in the mixed treatments also evidenced their suitability to grass-legume mixture binary forage. The CR _boat was significantly higher than unity, showing the greatest competitive ability of black oat to exploit resources in association with narbon bean in all the case studies ( Lithourgidis et al. , 2011 ) except year 2 in Alameda and year 3 in Tomejil, when CR _nbean was higher than unity in all intercrops. In all other cases, CR results for both species were positive but lower than unity, indicating that there was a positive benefit and these species can be grown in an intercrop ( Dhima et al. , 2013 ). Indeed, all different narbon bean-black oat mixtures (D2, D3 and D4) showed good DM yields with similar or higher values than control mixture.

However, the fact that the species in the mixture displayed a different behaviour influenced by the weather conditions each study year, affected not only their growth and development, but also to subsequent forage yield and quality production ( Caballero et al. , 1995 ). The different mixtures reached DM yields between black oat and narbon bean pure stands although similar to black oat in monoculture during year 1 when the weather conditions favoured the oats growth ( Vandermeer, 1984 ). In contrast, during the year 2 with the highest rainfall occurring from the emergence to the flowering black oat stage, the intercrops DM yields were similar to narbon bean sole crop, especially noteworthy in Alameda. However, the intercropping system improved black oat yields the year 3 and year 2, increasing the DM yields from 20-44 % to 62-79 %, respectively, and this latter range displayed significant differences with the sole grass, as was reported by Dhima et al. (2013) . The greatest DM yields for the intercropping narbon bean-black oat were provided by 35:65 in both locations during the drier year 1, by 65:35 the other study years in Alameda and by 50:50 and 35:65 in Tomejil. These average DM yields produced by the new mixture proposed were the highest in absolute terms. This fact could provide greater monetary advantages if farmers decide to cultivate intercropping systems, as were previously reported by Lithourgidis et al. (2011) .

Regarding forage quality, increasing on-farm protein production it is also desirable to achieve more economically feasible farming systems ( Sadeghpour et al. , 2014 ). The CP content is often considered as the most important component of forage quality ( Lithourgidis et al. , 2007 ; Cazzato et al. , 2012 ), and CP yield (based on forage CP content and total DM produced) is a valuable measure of the total protein that can be harvested for livestock enterprises ( Caballero et al. , 1995 ; Cazzato et al. , 2013 ). Our results showed that CP values enhanced with the seeding proportion of narbon bean in the intercrops ( Sadeghpour et al. , 2014 ). The greatest CP contents were reached by the narbon bean sole crops but existing mixed treatments with similar CP values the year 2 (all mixed treatments in both locations) and year 3 (50:50 seeding ratio in Tomejil). In addition, CP contents achieved by 65:35 in Alameda the year 2 and by 50:50 in Tomejil the year 3 were higher than control mixture in absolute terms. Previous studies reported the highest CP content recorded when the legume was grown as a sole crop or formed a high proportion in intercrops with grasses ( Lithourgidis et al. , 2007 ; Dhima et al. , 2013 ). Moreover, our findings showed that 65:35 and 50:50 achieved CP contents ranged from 80 to 160 g/kg of DM all years and locations, which comply with the CP content requirement for good quality forage content ( Leng, 1990 ). The total CP yields observed in narbon bean-black oat intercrops resulted in intermediate values between monocrops, without significant differences between intercrops, except for CP yield produced the year 2 in Tomejil and year 3 in Alameda due to their lower DM forage yield which is based for measurement ( Dhima et al. , 2013 ). Similar results without significant differences in CP yield were reported by Lithourgidis et al. (2011) for faba bean-oat intercrops in Greece. Nonetheless, 65:35 in Alameda and 50:50 in Tomejil gave the more analogous results all the years, that is, the mixing ratios with higher legume proportion in the mixture, as was reported in many other studies and constitutes the reason for including vetches in an intercropping system ( Lithourgidis et al. , 2006 ; Sadeghpour et al. , 2014 ).

Important criteria for evaluating forage quality are also ADF and NDF, being ADF a better efficient predictor of forage digestibility ( Caballero et al. , 1995 ; Lithourgidis et al. , 2007 ; Corleto et al. , 2009 ; Sadeghpour et al. , 2014 ). In fact, as the ADF percentage decrease, quality and digestibility of forage usually increase. Our results showed all the ADF values (ranging from 220 to 419 g/kg of DM) were in agreement with the high quality forage levels ( Van Saun, 2015 ). In addition, ADF contents at 50:50 in Alameda and 35:65 in Tomejil were lower or similar, respectively, in some cases to those obtained by the control, all of them values consistent with those presented in previous studies for Southern Europe ( Caballero et al. , 1995 ; Lithourgidis et al. , 2007 ). On the other hand, the NDF content is a good measure of voluntary intake of the feed and it is inversely related to its digestibility or energy density ( Sayar et al. , 2014 ). All NDF values obtained ranged between 485-600 g/kg of DM, forming reasonably good quality forage. The lowest NDF levels were found at 35:65 in almost all the cases, without significant differences with the control treatment the year 2. Lithourgidis et al. (2006) reported similar results, decreasing % NDF ratio while vetch ratio declined in a mixture. Finally, high DDM contents are required by ruminants to digest the greatest portion of the dry matter in a feed intake ( Van Saun, 2015 ). The highest DDM contents were found at 50:50 during the rainiest year 2 (639-652 g/kg of DM). Meanwhile, the control mixture showed the highest values during the years 1 and year 3, quite similar to the digestibility levels described by Caballero et al. (1995) but closely followed by 35:65 (from 577 to 821 g/kg of DM) in both locations.

In summary, the study of narbon bean and black oat species at different seeding ratios showed good ground cover evolution and adequate phenological growth according to the GDD accumulation despite the influence of environmental fluctuations. Plant density for the intercropped narbon bean and black oat species was lower than sole cropping and control mixture, but the CR of all intercrops did not show differences between mixed treatments and both species in monoculture. It confirms that both species despite becoming dominant under different environmental conditions can be grown successfully as an intercrop. Furthermore, narbon bean-black oat mixtures reached higher DM yields and CP yield than common vetch / common oat traditional mixture and all quality traits (CP content, ADF, NDF and DDM) complied with the high-quality forage requirements. Higher dry matter yields and similar quality values than control mixture were displayed at 35:65 with the scarcity of rain and 65:35 and 50:50 with increased rainfall recorded in loamy and clay soils, respectively. The fact that it is possible to form a balanced mixture between narbon bean-black oat and to get a forage of high quality, jointly to their adaptation capacity at different soil and rainfall conditions, makes the mixture V. narbonensis- A.strigosa an economically and environmentally promising option to be considered as a forage mixture for livestock diets in Mediterranean environments. Moreover, it is a viable rotation crop alternative to be implemented by farmers in the CAP greening framework for all Southern Europe countries. In order to improve the emergence and subsequent development of both species, especially in difficult environmental conditions (poor and waterlogging soils, frosty periods or drought), further research should be focused on studying the water balance and N role in the intercropping system. Under rainfed Mediterranean conditions, it would be useful to define the water use efficiency for the studied mixture as well as to determine the seeding ratio that maximizes the N ₂ atmospheric fixation for improving forage production and the residual N resource for subsequent crops in low-inputs cropping systems.

The authors are grateful to PhD Salvador Nadal for the narbon bean seeds used in these trials and to the field staff from CC3rdoba and Sevilla for technical assistance in the trials.