The agronomic variability of a collection of sainfoin accessions

Sainfoin (Onobrychis viciifolia Scop.) is a perennial forage legume appreciated for its feed value and rusticity. Two types are characterized via their growth habit and persistence: the common form which does not flower in the sowing year, and the giant form that does. In order to evaluate the degree of belonging to either of these types, thirty-eight Spanish sainfoin accessions and six foreign cultivars (44 accessions) were studied. The study involved three trials of 36 plants per accession in three locations in the northeast of the Iberian Peninsula during 2002-2004. Two locations received no irrigation while one was irrigated. The variables analysed were: percentage flowering in the sowing year, autumn regrowth, stem length at the end of winter, and the capacity for summer regrowth after the first cut. A cluster dendrogram showed two large groups encompassing common or giant sainfoins. These could be divided into three subgroups each according to the degree of contamination or crossing between the two major groups. Two easily identified characteristics —percentage flowering in the sowing season, and the speed of regrowth after the spring cut— help in the classification of these plants into one or the other form, and could facilitate cultivar selection. Additional key words: autumn regrowth, flowering, Onobrychis viciifolia Scop., Spanish ecotypes, summer regrowth.

Sainfoin is an allotetraploid species obtained from wild botanical forms (Badoux, 1965).Its cultivation was started at the end of the XVI century in the provinces bordering the Rhine Valley.Soon, the common and giant types (also called one-or two-cut types respectively) were selected: the first is characterized by its rusticity and persistence, the second by its erect growth habit, larger leaf size and vigour (which allows two cuts per year) (Gasparin, 1846).Common sainfoin does not flower in the sowing year while giant sainfoin does (Rees, 1932, cited by Michelena, 1983).
The sainfoin originally sown in Spain was the common type.Giant sainfoin was introduced experimentally by the Botanical Gardens of Madrid in 1791 (Muller, 1893).Since then there have been successive introductions, the most important being that promoted by the Ministry of Agriculture at the end of the 1960s.Thus, foreign giant sainfoins have mixed with native forms (Pujol, 1974).Local ecotypes may have been altered due to seed transfer occasioned by farmers and commercial enterprises.Part of the seed demand is met by seed imported from Eastern Europe countries, and this is contributing further to the contamination of local ecotypes.No breeding has been undertaken.
Given its feed, health and environmental properties, Sainfoin could play an important role in the restructuring of rainfed, low-yield cereal land into pastures for extensive production or other uses in the northeast of the Iberian Peninsula.It is therefore important to survey and collect local ecotypes and to undertake breeding work to produce improved cultivars.
This work studies the variability of a collection of 44 accessions.
Sowing was performed in spring 2002.Thirty six plants of each of the 44 accessions were sown by direct drilling using a planting frame of 0.3 × 1 m for the rainfed and 0.5 × 1 m for the irrigated land; the plants were distributed as three replicates of 12 plants (a random block design was used).In the irrigated trials, water was applied by monthly flooding during the summer months.No fertilizers, insecticides or herbicides were applied; all weeding was performed by hand.
The following variables were recorded: percentage flowering in the sowing year, autumn regrowth [on a visual evaluation scale from 1 to 9 (least to most regrowth)], and average stem length (cm) at the end of the winter.These variables were recorded in the autumnwinter from 2003 to 2004.The capacity for summer

Girona
regrowth 20 days after the first cut was also measured visually in 2004 (using the same 1 to 9 scale).These important, easy-to-measure variables have been recorded in different studies on the characterization of local sainfoin ecotypes (Michelena and Hycka, 1988;Prosperi et al., 1994) and alfalfa (Medicago sativa L.) (Hidalgo, 1966;Delgado, 1986).The flowering percentages were arcsine-transformed prior to statistical analysis.The normality of the distributions of the autumn and summer regrowth data were tested using the Shapiro-Wilk's test.All data were analysed by ANOVA.Accessions, locations and their interaction were regarded as fixed effects.Comparisons between means were performed using the LSD test.Correlations between variables were assessed by calculating Spearman correlation coefficients.Cluster analysis was performed using Ward's minimum variance clustering method (Sneath and Sokal, 1973).All statistical procedures were undertaken using the SAS statistical package (SAS, 2004).

Results
Table 2 shows the mean values of the variables for all 44 accessions at each location.Most of the characteristics analysed showed highly significant differences between accessions and locations (Table 3).The interaction accession × location had a highly significant effect on most of the studied variables.
The average percentage flowering in the sowing year was 48.1 ± 15.86% at Lagueruela, 44.5 ± 25.70% at Latre, and 42.7 ± 28.40% at Zaragoza; the range of percentage flowering was 0-100.Significant differences (P < 0.05) were seen between Lagueruela and Zaragoza, with the smallest flowering percentage obtained in the irrigated fields of Zaragoza.At the latter trial site only those plants flowering within the first productive cycle were counted because a cut was made during the full bloom stage due to the strong development shown by the plants.In the remaining trails under rainfed conditions, flowering plants were counted throughout the productive period because of the poor growth shown in the sowing year.
Significant (P < 0.001) differences between accessions were seen in terms of autumn regrowth (6.6 ± 0.43 points on the 1-9 scale at Lagueruela, 6.6 ± 0.3 at Latre, and 5.8 ± 0.56 in Zaragoza).The differences in autumn regrowth between the dry land (Lageruela and Latre) and irrigated land sites (Zaragoza) were also significant (P < 0.05), perhaps due to the different cutting systems used in each.
Significant differences (P < 0.001) were also seen between accessions in terms of the growth at the end of winter (13.2 ± 0.91 cm at Lagueruela, 14.3 ± 0.98 cm at Latre and 47.0 ± 5.35 cm at Zaragoza).The differences in this variable between plants grown on dry land and irrigated land were greater than those seen for autumn regrowth, probably due to different water availability.
Table 4 shows the correlation matrix for the studied characteristics.A strong correlation was seen between locations in terms of percentage flowering.Percentage flowering was also significantly (P < 0.001) correlated with the capacity of summer regrowth after the first cut, but no significant correlation (P > 0.05) was seen with autumn regrowth or growth at the end of winter.
Figure 2 shows the grouping of the accessions based on the evaluated characteristics.Two large groups holding common (cluster 1) or giant (cluster 2) types are evident.These groups can be divided into three subgroups each that discriminate the accessions according to the degree of contamination or crossing (mixed or heterogeneous types) between both major groups.Table 5 gathers the accessions included in each group and subgroup, as well as the means and standard deviations of the measured traits.

Discussion
To obtain more productive sainfoins, farmers selected more vigorous, taller plants that were able to flower in the sowing year.The differentiation of two types of sainfoin-common (more rustic and unable to flower in the sowing year) and giant (with greater vigour and able to flower in the sowing year) then became possible (Gasparin, 1846).
The capacity to flower in the sowing year has been the main criterion used for classifying accessions as common or giant sainfoin (Michelena and Hycka, 1988;Prosperi et al., 1994).In the present study, the differences between accessions in terms of mean percentage flowering in the sowing year were significant (P < 0.001).However, if this characteristic is considered in isolation it becomes difficult to assign many accessions to either the common or giant group since percentage flowering can be close to 50%, and varied according to the location of the trial.The giant Fakir cultivar (accession number 41), which was used as a control in these trials because of its phenological homogeneity (Prosperi et al., 1994), helped in the assessment of the heterogeneity of the accessions.In the irrigated trial, 100% of the plants flowered; however, on the rainfed land only 73% did so.This indicates that water stress can inhibit flowering to a certain extent; dry land conditions are therefore not the most adequate for the classification of accessions as common or giant.The appropriateness of undertaking trials in conditions of no water stress, in order to allow full flowering, has been repeatedly manifested by other authors (Salter at al., 1984;Rotili, 1988).
Autumn and summer regrowth, plus winter growth related to vigour and the growth habit have been used in a complementary fashion in the differentiation of the two types of sainfoin by other authors (Michelena and Hycka, 1988;Prosperi et al., 1994).The strong correlation seen between percentage flowering in the sowing year and the capacity of summer regrowth after the cut is one of the characteristics that differentiated the giant sainfoins.This greater regrowth capacity has been reported in other studies (Thonson, 1951, cited by Michelena, 1983;Michelena and Hycka, 1988;Prosperi et al., 1994).In the present study it was sig-  nificantly different (P < 0.001) at the three locations.Thus, it may be deduced that the capacity of summer regrowth after cutting is also a good indicator of phenological heterogeneity and a classification marker of the common or giant types.This criterion has already been used with other crops such as alfalfa (Delgado et al., 2003).All the above authors agree in that a limited correlation exists between sainfoin type and the amount of winter growth.In the present study no significant (P > 0.05) correlation was observed between percentage flowering in the sowing year, autumn regrowth [except for autumn regrowth in Lagueruela, which was highly (P < 0.001) significant] and growth at the end of the winter.According to the above descriptions of both types of sainfoin, and after comparing the results obtained with those for the giant cultivar Fakir, accessions 2, 3, 9, 29, 31, 32, 33, 34, 35, 36, 41 and 43 (included in ,3,9,29,31,32,33,34,35,36,41,43 Cluster 2.2 41.1 ± 10.13 6.6 ± 0.13 26.5 ± 0.85 6.7 ± 0.41 5,8,17,18,21,25,27 Cluster 2.3 54.1 ± 7.03 6.3 ± 0.19 23.9 ± 2.40 6.6 ± 0.56 6,24,28,30,39,40,44 Cluster 2.1), with the greatest growth all, fall within the giant sainfoin group.These accessions came mainly from the provinces of Burgos,Logroño,Soria and Huesca. Accessions 4,11,14,20,22,23 and 37 (included in Cluster 1.3) are the closest to a typically common sainfoin type.These accessions came mainly from the provinces of Teruel, Castellón, Guadalajara, Lleida and Palencia.However, all provinces were home to accessions with a certain degree of phenological heterogeneity; these grouped into different clusters (1.1, 1.2, 2.2 and 2.3).This heterogeneity may be attributed to seed importation from other countries (Pujol, 1974) and the free seed market among farmers from different areas (Michelena, 1983;Delgado et al., 2002).
Despite the indiscriminate marketing of seed, the present results show that common and giant sainfoin types can still be differentiated.Two easily identified characteristics -percentage flowering in the sowing year and the speed of regrowth after the spring cutcontribute to the classification of these plants into one type or the other, and could facilitate the selection of cultivars.

Figure 1 .
Figure 1.The Spanish provinces where sainfoin seeds were collected.

Figure 2 .
Figure 2. Dendrogram for the 38 sainfoin Spanish accessions and six foreign cultivars.

Table 1 .
Location of the three trials, and local climate and soil characteristics

Table 3 .
Significance of the differences detected in analysis of variance involving the studied variables

Table 5 .
Means and standard deviation (± SD) of the studied variables, distributed according to cluster groups