1. Introduction

Cowpea (Vigna unguiculata (L.) Walp.) seed and fresh pod yield is usually strongly influenced by prevailing environmental conditions, expressing significant environmental effects (E) and genotype x environment interactions (G x E) ^[1],^[2],^[3],^[4],^[5],^[6],^[7]. Average seed yield can especially be decreased when drought conditions occur during the flowering and seed filling periods ^[8],^[9],^[10]. Identification of genotypes expressing yield stability under different environments and throughout cropping seasons is considered one of the most important breeding goals ^[5],^[11], while yield is also recorded as the most desirable trait for farmers ^[12],^[13],^[14]. Currently, we are at a point at which the observed extreme climatic conditions may affect the stability of various crops yield globally ^[15],^[16],^[17]. The ability of a crop to produce consistent yields in various environments and under changing weather conditions therefore comes to the fore. Landraces consist of genetic material with intermediate yield, production efficiency, and yield stability when they are cultivated in low-input cropping systems ^[18],^[19] and therefore, they are important in terms of stable yield production ^[20],^[21] and of improving resistance to various abiotic stresses ^[22],^[23],^[24].

Cowpea cultivation in Southern Europe, including Mediterranean countries, starts in late spring (late April) and lasts until the beginning of autumn (early October) ^[1],^[25],^[26]. During the summer period, cowpea is confronted with water scarcity and high air temperatures, like in many other areas of the world ^[27],^[28], while in many cases, cultivation faces additional limiting soil factors ^[29]. In the countries around the Mediterranean basin, a remarkable number of cowpea landraces are still cultivated on a small scale by farmers mainly for their own use and consumption of either their young, tender pods or their seeds, which are rich in protein, carbohydrates, and nutrients ^[30],^[31]. These landraces could serve as important sources of adaptive traits and resistance to drought for the upcoming climatic changes ^[32],^[33],^[34],^[35].

The evaluation of cowpea landrace material originated from Southern European countries is considered limited in proportion to the number of local varieties that are available. The aim of this study was therefore to evaluate five selected cowpea landraces of Greek origin, with interesting morphological traits adapted to different microclimates with regards to their phenological characteristics and traits related to seed yield.

2. Results

2.1. Plant Phenological and Agronomical Traits

There was a statistically significant interaction among the accessions and the experimental years for all the phenological traits studied, except for days from sowing till the appearance of mature pods, in 50% of the plants (DMAT). The accessions differed statistically significantly from each other regarding all three phenological characteristics studied, while statistically significant differences were observed among the experimental years (p < 0.001). Furthermore, there was not a statistically significant interaction of accession x experimental year for the number of pods per plant, pod length, seed weight per plant, and number of seeds per plant. Seed weight per plant did not differ statistically significantly among the three experimental years and the accessions, while pod length and hundred-seed weight did not differ statistically significantly among the experimental years.

Seed yield (kg ha^-1) varied between 577.78 kg ha^-1 (VG2-Atsiki, Lemnos) and 1058.33 kg ha^-1 (VG23-Logothetianika, Kythira) in 2015, between 389.49 kg ha^-1 (VG4-Marathi, Mykonos) and 690.47 kg ha^-1 (VG23-Logothetianika, Kythira) in 2016, and between 683.19 kg ha^-1 (VG2-Atsiki, Lemnos) and 1053.15 kg ha^-1 (VG3-Alinda, Leros) in 2017 (Figure 1). However, seed yield (kg ha^-1) did not differ statistically significantly among the accessions and the experimental years, while there was not a statistically significant interaction between accessions and experimental years.

Statistically very strong positive correlations were shown between number of pods and seed weight per plant (r = 0.830, p < 0.001), number of pods and number of seeds per plant (r = 0.880, p < 0.001), and number of pods and seed yield (kg ha^-1) (r = 0.830, p < 0.001). Strong positive correlations were observed between pod length and number of seeds per pod (r = 0.768, p < 0.001) and between seed weight per plant and number of seeds per plant (r = 0.774, p < 0.001). Pod length and number of seeds per plant were also positively correlated with seed yield (kg ha^-1) with r = 0.534, p < 0.001 and r = 0.774, p < 0.001, respectively.

2.2. Principal Component Analysis

Principal component analysis (PCA) was performed to reduce the dimensionality of the data, to study the contribution of each trait to the variability observed, as well as to illustrate the highest yielding accessions. PCA showed that 79.88% of the total variation can be explained through the first three principal axes. Plant height, number of pods per plant, seed weight per plant, number of seeds per plant, and seed yield (kg ha^-1) were related to the first principal component (PC1, 41.58%). Pod length and number of seeds per pod were related to the second principal component (PC2, 22.77%), while days from sowing to 50% of flowering, days from sowing to 50% of pods maturity, and flowering duration were related to the third principal component (PC3, 15.53%).

During the second experimental year, accessions presented shorter flowering duration, earlier pod maturity, lower number of pods per plant, lower number of seeds per plant, and lower seed yield (kg ha^-1) than in the other two experimental years. Therefore, the accessions in the second experimental year (presented in green color) were grouped separately (second and third quadrant) from the two other experimental years through principal component analysis (PCA) (Figure 2).

Figure 2. Classification of accessions through principal component analysis (PCA) and diagrammatic representation of the eigenvectors of studied traits. DFL: days from sowing to 50% flowering, FDUR: flowering duration, DMAT: days from sowing to 50% pod maturity, PH: plant height, NPOD: number of pods per plant, PODL: pod length, SPOD: number of seeds per pod, SEEDW: seed weight, NSEED: number of seeds per plant, 100 SW: hundred-seed weight. Accessions are presented with different symbols, while experimental years are indicated by different colors (2015: red, 2016: green, and 2017: blue).

Accessions in the experimental years 2015 and 2017 were grouped in the first, third, and fourth quadrants, with the highest yield accessions to be depicted in the first and fourth quadrants. Landraces VG3 and VG23 were presented as the highest yielding accessions during 2017 and 2015, respectively, while VG4 was the lowest yielding accession in 2016 (Figure 2). Most accessions expressed similar values for every studied trait for each accession under the 2015 and 2017 experimental years and therefore were depicted at close distances, with the exception of local population VG3 (Alinda, Leros) and local population VG23 (Logothetianika, Kythira). Landrace VG2 (Atsiki, Lemnos) was among the least productive accessions in all three experimental years (Figure 2).

3. Discussion

The accessions evaluated were practically classified into two groups based on the days needed from sowing to achieve 50% of flowering. The first group contained the early flowering landraces named VG2 (Atsiki, Lemnos), VG4 (Marathi, Mykonos), and VG23 (Logothetianika, Kythira), while the second included the late flowering accessions named VG3 (Alinda, Leros) and VG20 (Mitilinioi, Samos) and the improved line IT97K-499-35. Cowpea genotypes with short biological cycles and short flowering times can avoid the water scarcity that often prevails in the area during the summer months ^[36], while genotypes with long biological cycles seem to cope better under high temperatures as they gradually enter flowering and podding stages ^[37],^[38]. Cowpea landraces evaluated in our study are diverse and represent promising material for high air- temperature and drought tolerance or avoidance. In particular, the local population VG2 from Atsiki, Lemnos could be suitable material for drought avoidance ^[39].

The average cowpea seed yield in Greece fluctuates from 1 to 3.5 t ha^{-1 [40]}. The average seed yield achieved during our study ranged from 0.39–1.05 t ha^-1 and was therefore lower than the average seed yield reported for Greek conditions. The observed seed yield values were also lower than the average seed yield of cowpea landraces reported in other countries, such as in Ethiopia (2.05 t ha^-1) and Brazil (1.05 t ha^-1) ^[41],^[42]. In addition, all landraces under study did not differ statistically significantly from each other, nor compared to the improved line, regarding their seed yield (kg ha^-1). This fact is probably due to the unfavorable field conditions, which led to the production of low seed yield and did not allow landraces’ potential to be unfolded in our study ^[43],^[44],^[45].

Despite that, there were not statistically significant differences among the accessions regarding seed yield, VG23 (Logothetianika, Kythira Island) was a landrace that exceeded the seed yield production of IT97K-499-35 breeding line in each one of the three experimental years. VG23 also presented a high number of pods and seeds per plant as well as large pod length, which are traits that have been strongly and positively related to seed yield. Therefore, VG23 was the most productive accession in the current soil and climatic conditions. The statistically significantly higher genetic diversity that has been previously recorded for the VG23 landrace ^[31] in comparison to twenty-two other cowpea landraces of Greek origin could be the reason for its increased seed yield production efficiency during the three experimental years. On the other hand, local population VG2 (Atsiki, Lemnos Island), which has been previously found to be one of the most homogeneous landraces among twenty-three local populations studied with Greek origin ^[31], was characterized by lower productivity but also by higher stability in comparison to the other landraces regarding many of the traits studied.

However, VG2 landrace stability could be only due to the impact of the unfavorable soil conditions of the experimental location used. Landraces unfold their potential when they are cultivated per se in the regions where they have been adapted, and despite their low productivity in ex situ cultivation, they should be further evaluated on-farm in each one’s region of origin. Therefore, the performance, including both productivity and stability of the examined landraces of the present experiment, should also be assessed in their region of origin as well as in other soil and climatic environments ^[24],^[46],^[47],^[48],^[49], as different cowpea landraces were found to be better adapted to diverse environments ^[34].

4. Materials and Methods

4.1. Plant Material and Experimental Design

The experiment was conducted in a field at Agricultural University of Athens (AUA) (37°59’10” N, 23°42’29” E, altitude 24 m), during the spring–summer cultivation period for three consecutive years: 2015, 2016, and 2017. Sowing took place on 22 May 2015, on 15 May 2016, and on 29 May 2017. Five cowpea (Vigna unguiculata (L.) Walp.) landraces, named VG2 (Atsiki, Lemnos Island), VG3 (Alinda, Leros Island), VG4 (Marathi, Mykonos Island), VG20 (Mitilinioi, Samos Island), and VG23 (Logothetianika, Kythira Island), with origin from the Greek islands were evaluated. The breeding line IT97K-499-35 originating from Nigeria was also used. A randomized complete block design (RCBD) with four replications was used and each plot consisted of forty plants.

4.2. Phenological and Yield Related Traits

Measurements of phenological and yield related traits were taken for ten central plants per plot. Phenological traits were recorded regarding days from sowing until 50% of the plants flowered (DFL), days from sowing to 50% pod maturity (DMAT), and flowering duration (FDUR), which was defined as the interval in days from the day of observation of the first open flower per plant to the observation of the last open flower per plant, including also the second flower flush that was recorded in some plants. Measurements related to yield included plant height (PH) (cm), number of pods per plant (NPOD), pod length (PODL) (cm), number of seeds per pod (SPOD), seed weight per plant (SEEDW) (g), number of seeds per plant (NSEED), and hundred-seed weight (100 SW) (g). Seed yield (SY) (kg ha^-1) was then extrapolated from the total seed weight per plant.

4.3. Statistical Analysis

Residuals of all studied traits were subjected to normality tests and were checked for their homoscedasticity. Analysis of variance (ANOVA) was therefore applied, followed by Tukey’s (HSD) (p < 0.05) means comparison method using the statistical software Statgraphics Centurion XVII. Coefficients of variation (CV%) were also calculated for each trait per accession and for each trait per experimental year for all accessions. Correlations between traits (Pearson correlation coefficients) were also investigated using the statistical package STATISTICA 8.0. Finally, a principal component analysis (PCA) was performed, aiming to study the contribution of each trait to the variability observed as well as to illustrate the highest yield accessions during the three experimental years by using the SAS statistical program JMP-8.

5. Conclusions

Promising variability observed among and within the five landraces studied regarding their phenological and seed yield related traits. However, there were not statistically significant differences in terms of their seed yield (kg ha^-1), a fact which is probably recorded due to the unfavorable soil and climatic conditions prevailed, that affected accessions performance andprevent them to unfold their potential. Among the landraces, VG23 (Logothetianika, Kythira Island) was the most productive under the present soil and climate conditions, while the local population VG2 (Atsiki, Lemnos Island) characterized by low seed. Further evaluation in different environmental conditions is considered necessary, while VG23 could be utilized in a breeding program aiming to increase seed yield production.