Erucic acid is an important fatty acid in the oleochemical industry [1-3]. The current major industrial source of erucic acid is high-erucic acid rapeseed oil [2-4]. In recent years, Crambe abyssinica is becoming more and more interesting as an alternative industrial crop [5-7] since it shows high-erucic acid content (52% - 59%) in its seed oil and also wide climatic and agronomic adaptation and does not cross with the double-low canola. Just recently, the erucic acid content in Crambe seed oil was increased to over 70% [8]. In addition to oil, Massoura et al. [9] reported that other valuable by-products such as protein meal and possibly fiber can be obtained from Crambe. Acceptance of Crambe meal by the feeds Industry is based on its attractive price and satisfactory performance as a feed for ruminant animals [10]. Crambe was also cultivated in China [11]. Since 1970s, traditional breeding by successive selection within C. abyssinica has been carried out in many countries and several cultivars have been released such as “Prophet”, “Indy”, “Meyer” [12], “BelAnn” and “BelEnzian” [13] by mass selection. Crambe is already commercially cultivated on a small scale, and novel varieties can yield the same amount of oil per hectare as spring rapeseed does [14].

Selection, which is mainly based on phenotypic characters, is the major technique used in a breeding program. Response to selection depends on many factors such as the interrelationship of the characters [15]. By knowing if correlation exists between important traits, interpretation on previous results would become easier. Also correlation between important and nonimportant traits provides plant breeding experts with a significant assistance in indirect selection of important traits, through non-important traits which their measurement is easier [15,16]. Partial correlation coefficient is a measure of the linear dependence of a pair of random variables from a collection of random variables in the case where the influence of the remaining variables is eliminated. A partial correlation between two variables can differ substantially from their simple correlation [17]. Regression helps to estimate the functional relationship between variables or the relationship between the independent and dependent variables [15]. Path coefficient analysis is a very important statistical tool that can be used to obtain an indication of which variables exert an influence on other variables, while recognizing the multicolinearity [18,19]. Path-coefficient analysis has been useful in determining selection criteria in a number of crops, such as crested wheat grass [18], maize [20] and rice [21]. Up to date there is still no report on correlation, regression and path analyses about Crambe seed yield components. Here we report the correlation, regression and path analyses of seed yield components in Crambe to provide some clue to future Crambe breeding and cultivation.

10 Crambe lines from PRI, Wageningen, and 12 lines from Hubei University were sown in Anyue, Sichuan Province in China on October 26, 2011. Anyue is located in Southwest of China and the soil is neutrally purplish. Before sowing the seeds water with farmyard manure was used. For each Crambe line there were three replicates, each replicate 6.6 m². In each replicate 480 Crambe plants were kept. In December 562.5 kg urea (about 262 kg pure N) per hectare was used to promote the growth of Crambe seedlings. The Crambe seeds were harvested On May 1st, 2012 and weighed together with the seed hulls. For each replicate 10 plants were sampled for investigation of the agronomic traits according to the rapeseed standard [22]. Data were analyzed on SPSS 19.0 and Excel.

In the present study simple correlation analysis indicated that seed yield per plant was significantly correlated with plant height, number of first branches, number of seconddary branches, pod number of primary inflorescence, pod number per plant and 1000-grain weight (P < 0.01) and negatively correlated with branching height (P < 0.01), but partial correlation analysis indicated that seed yield per plant was only significantly correlated with pod number per plant and 1000-grain weight (P < 0.01) and the number of secondary branches (P < 0.05). Simple correlation between pod number per plant and 1000-grain weight was not significant but their partial correlation was negatively significant (P < 0.01).

Previous results about canola indicated that number of pods per plant had the highest direct effect on grain yield in canola. In addition, 1000-grain weight also had a high direct effect on grain yield [23-30]. Tusar-Patra et al. [31] concluded that the strongest effect on seed yield was estimated for number of pods per plant followed by number of seeds per pod and 1000 seed weight. Khan et al. [32] found that number of branches, 1000-seed weight, and pods per plant affected the seed yield per plant. Marjanović-Jeromela et al. [26] reported that the strongest direct effect on seed yield per plant was estimated for plant height, followed by that of number of pods per plant. Ghodrati et al. [33] found that number of seeds per pod was effective on canola seed yield. Results about seed yield components in Orychophragmus violaceus, also a potential crucifer oil crop, indicated that the trait that contributed the most to seed yield per plant was number of pods per plant, followed by number of seeds per pod, 1000-grain weight, and then the number of first branches [34]. In the present study stepwise regression and path analyses indicated that only pod number per plant and 1000-grain weight contributed significantly to seed yield per plant, while the effects of plant height and branch numbers were not significant. The regression formula for pod number per plant (X₇) and 1000-grain weight (X₈) to seed yield per plant is Y = 0.006 X₇ + 1.222 X₈ − 7.191. The path coefficient of pod number per plant to seed yield per plant was 0.967 and that of 1000-grain weight was 0.194. The determination coefficient of pod number per plant and 1000-grain weight to seed yield per plant was 0.983 and the left determination coefficient was 0.130.

In the present study the Crambe seed yields were roughly comparable to that of canola in China and showed great potential to be further increased. In our experiment the Crambe lines showed great variations with respect to seed yield in each plot as well as to the agronomic traits. The highest seed yield was 2.15 kg (3257.58 kg/ha) for one plot and the highest seed yield was 53.48 g for one single plant; the highest pod number was 8751 for one single plant and the highest 1000-grain weight was 8.65 g. Coefficient of variance indicated that seed yield per plant and pod number per plant showed great variations. Combined with the results from regression and path analyses, it was suggested that seed yield per plant in Crambe might be improved by increasing the pod number per plant through selection or cultivation, still the negative correlation between pod number per plant and 1000-grain weight also needs to be considered.

This work was supported by an EC FP7 project ICON (211400), Swedish Research Links project, funds from NSFC (30771382, 30671334, 30971807, 31201238), Wuhan Science and Technology Bureau and MOST, China.