Four on-farm experiments were carried out on commercial pineapple fields in the Atlantic department (characterized by a subequatorial climate) in the south of Benin (West Africa) between February 2010 and August 2012 with two pineapple cultivars: Sugarloaf (Experiments 1 and 2) and Smooth Cayenne (Experiments 3 and 4). Two different producers were selected per cultivar based on (a) the age of their pineapple crop being close to the common artificial flower induction time and (b) whether they applied the common practices for these cultivars, as described by Fassinou Hotegni et al. (2012). Pineapple cultivation starts with the planting materials obtained from harvested plants kept in the fields. During the planting material collection, those shoots that are very small or very big or look unhealthy are skipped; different types (hapas and suckers in cv. Smooth Cayenne) and sizes are mixed when planting. Information on the fields and cultural practices until artificial flower induction time is provided in Table 1.

Four experimental plots were installed per experiment, which were part of a larger experiment not reported on here. Each net plot consisted of six rows of 10 plants each. The net experimental plots were surrounded by two rows with border plants.

Crops were artificially induced between 10 and 13 months after planting (Table 1) using carbide of calcium (CaC₂), a compound producing acetylene when it reacts with water. Following farmer's practices, 50 ml of a solution containing 10 g/l and 15 g/l of CaC₂ for Sugarloaf and Smooth Cayenne, respectively, was applied into the center of the leaf rosette of each plant. This application was carried out once in cv. Sugarloaf and three times, with an interval of 3 days, in cv. Smooth Cayenne.

Following farmer's practices, maturity of the fruits was synchronized only in cv. Smooth Cayenne, 143 days after artificial flower induction, by spraying 3.5 ml of a solution of 14 ml/l Ethephon (2-chloroethylphosphonic acid), a compound producing ethylene, on the skin of each fruit. This application was carried out twice with an interval of 4 days.

Pineapple fruits were harvested between 150 and 154 days after flower induction. The pineapple fruits were harvested following farmer's practice which was in cv. Sugarloaf the moment when the skin color of the fruit of at least 25% of the plants (i.e., 15 out of 60 plants in a net plot) had started to change from green to yellow. In cv. Smooth Cayenne, the fruits were harvested 7 days after the second application of Ethephon. All fruits per plot were harvested on that day and were individually processed. No ratoon crops are grown in Benin, so plants were harvested once.

Three variates representing the vigor of the individual plants within a crop at the moment of artificial flower induction were assessed: (1) the number of functional leaves per plant (NL) (green leaves excluding those withered over more than 10 cm of their length), (2) the length of the D-leaf (DL) (the longest leaf in a pineapple plant according to Malézieux et al. (2003)) and (3) their cross product (NL × DL). The number of functional leaves indicates the developmental status of the plant at flower induction time. The D-leaf is used to assess the growth and the nutritional status of the plant (Malézieux et al., 2003). The cross product NL × DL is a proxy for the total leaf area of the plant. The number of functional leaves and DL were assessed on all individual plants 1 day before flower induction. The D-leaf was identified by bunching all leaves together and selecting the longest. Next, the length was measured with a twig combined with a ruler.

External and internal fruit quality attributes were assessed at harvest on the fruits from all individual plants. External fruit quality attributes included the weight and height of the fruit, infructescence and crown; the ratio crown height: infructescence height and the number of fruitlets per infructescence. The number of fruitlets or “eyes” on the infructescence was determined by multiplying the number of spirals counted counter-clockwise and the average number of fruitlets on the first and last spiral. Internal fruit quality attributes included TSS, juice pH, the percentage of flesh being translucent, and internal browning. To determine these, the pineapple was cut longitudinally into two halves. A portion of the juice obtained from squeezing one half was used to determine the TSS by a hand refractometer; another portion of that juice was used to determine the juice pH by a hand-held pH meter. The percentage of fruit flesh that was translucent and internal browning were visually estimated on the second half following the methods of Paull and Reyes (1996). The type, number and total weight of side shoots (slips, hapas, and suckers) per plant were also recorded at harvest time.

Data were analyzed using R version 2.15.2 (R Development Core Team, 2012). Fruits with more than one crown at harvest (13 and 6 fruits in Experiments 3 and 4, respectively) were excluded in the analysis. Heterogeneity in plant vigor variates and in fruit quality attributes was described by the coefficient of variation (CV) which is a measure of the variability in a population relative to the mean (cf. Schouten et al., 1997; Illipronti et al., 2000; Field, 2009; Ott and Longnecker, 2010). CVs were calculated per plot and differences in CV between cultivars for each plant vigor variate and each quality attribute were assessed using a t-test. Differences in CV between plant vigor variates as well as differences in CV between quality attributes within an experiment were assessed using an ANOVA. When the F-value from the ANOVA was significant, LSDs (α = 0.05) were used to separate means.

To determine if and how the plant vigor variates at flower induction were associated with fruit quality attributes at harvest, simple linear regressions were performed on the combined data from all plots per experiment, using NL, DL, and NL × DL as explanatory variates and each fruit quality attribute as response variate. Percentage flesh translucency was transformed using square root transformation (x+0.5) before analysis (Bartlett, 1936; Gonzalez, 2009). Which plant vigor variable was best associated with a fruit quality attribute was determined using the adjusted R². The higher the adjusted R², the higher is the percentage of the variance in the response variate accounted for.

To determine whether the number and the weight of the additional side shoots produced (slips) accounted for fruit quality heterogeneity in addition to the plant vigor variates at flower induction, a multiple regression was performed by using the plant vigor variates (explaining the highest percentage of the variance in the fruit quality attributes variates) as well as the number or weight of the slips as explanatory variates and the different fruit quality attributes as response variates. A hierarchical method was used in which the plant vigor variates were entered first and the weight or number of slips was entered next, to analyze the contribution of slip weight/number to fruit quality heterogeneity. Existence of colinearity between the explanatory variates was checked using Pearson coefficient of correlation (r). A value of r greater than 0.80 reveals multiple colinearity between the explanatory variates (Field, 2009); in that case the explanatory variables were not used in the multiple regression model. The significance of the F change (significance of the improvement of the adjusted coefficient of multiple regression R²) derived from the multiple regression model was used to evaluate the effect of slip weight/number.

The initial heterogeneity in plant vigor (NL, DL, and NL × DL) within a field was quantified using the CV. For all vigor variates, the initial heterogeneity was not different between experiments with cv. Sugarloaf and experiments with cv. Smooth Cayenne (Table 2). In all four experiments, variation in NL × DL was higher than variation in NL and DL, and variation in DL was lowest (Table 2).

When comparing the CV in different external fruit quality attributes at harvest across experiments with different cultivars (Table 2), the variation in crown weight, crown height and ratio crown height: infructescence height was higher in the experiments with cv. Smooth Cayenne than in those with cv. Sugarloaf, whereas the variation in all other attributes was similar across cultivars.

In all experiments, variation in infructescence weight was higher than variation in other external quality attributes. Variation in fruit weight, infructescence weight and the crown weight was higher than in the respective heights of these organs in all experiments (Table 2). Variation in infructescence weight was higher than variation in fruit and crown weight. The crown weight was the least variable weight attribute except in Experiment 3, where it was comparable to fruit weight (Table 2). Variation in infructescence height was higher than variation in fruit height in all experiments (Table 2), whereas variation in crown height was comparably low as variation in fruit height in the Sugarloaf experiments and comparably high as variation in infructescence weight in the Smooth Cayenne experiments. Variation in the ratio crown: infructescence height was higher than that in the underlying attributes, except in Experiment 2 where the difference with the variation in infructescence height was not significant. The CV in number of fruitlets was similar to the CV in infructescence height.

For all internal quality attributes, variation in TSS and translucent flesh was higher in the experiments with cv. Smooth Cayenne than in the experiments with cv. Sugarloaf. Variation in juice pH was higher in experiments with cv. Sugarloaf than in experiments with cv. Smooth Cayenne. In all experiments, the most variable internal quality attribute was flesh translucency. Variation in TSS and variation in juice pH were very low and not significantly different from each other in all experiments (Table 2).

In all crops there were strong associations between the initial vigor of a plant at flower induction and the total fruit weight of that plant at harvest; higher NL, DL, and NL × DL all were associated with heavier fruits at harvest (Table 3). Based on adjusted R² values (0.463 – 0. 686), NL × DL was the vigor variate showing the strongest association with fruit weight (Table 3; Figures 1A1–A4). The R² values for the relations between plant vigor variates and infructescence weights were comparable to those for total fruit weights and also highest for NL × DL (Table 3; Figures 1B1–B4). However, R² values for the relations between vigor variates and crown weight were much lower and not significant for NL × DL in two out of four experiments (Table 3; Figures 1C1–C4), suggesting that the positive associations between NL × DL and fruit weight were mainly caused by the positive effect of high vigor on the infructescence weight, and less or not on crown weight. Variation in crown weight was better explained by DL than by NL × DL, but with low R² values varying between 0.024 and 0.142.

The cross product NL × DL was also significantly positively associated with the fruit height and the association was very clear for cv. Sugarloaf (Table 3; Figures 1D1,D2); for cv. Smooth Cayenne, this association was poorer although significant in both experiments (Table 3; Figures 1D3,D4). Of the attributes underlying fruit height, the infructescence height also increased with an increase in NL × DL in all experiments (Figures 1E1–E4), but the crown height was differently related to NL × DL in the two cultivars; for cv. Sugarloaf a weak positive association was found to be significant only in one of the two experiments whereas a negative association was found in both Smooth Cayenne experiments (Table 3). As for crown weight, crown height showed a better association with DL than with NL × DL, but for cv. Sugarloaf only. For cv. Smooth Cayenne, the negative association between the initial plant vigor and crown height was even clearer for NL than for NL × DL in one experiment (Table 3).

The cross product NL × DL was significantly negatively associated with the ratio crown height: infructescence height (Table 3; Figures 1G1–G4) in all experiments.

Figures showing the associations of the external quality attributes with NL and DL can be found in the supplementary material (Supplementary Figures S1,2).

The plant vigor variates at the time of artificial flower induction were not or only weakly associated with the TSS, juice pH and translucency of the fruits at harvest (Table 3; Figure 2 for associations with NL × DL). Figures showing the associations with NL and DL can be found in the supplementary material (Supplementary Figures S3 and S4).

Weak but significant associations between at least one of the vigor variates and TSS were found in all experiments, but these associations were positive in three experiments and negative in one experiment, and consequently not consistent across experiments (Table 3).

For cv. Smooth Cayenne, the cross product NL × DL was the strongest vigor variate to be weakly, but consistently positively associated with juice pH (Table 3; Figures 2B3,B4). For cv. Sugarloaf the same results were found in Experiment 1 (Table 3; Figure 2B1), whereas in Experiment 2 no significant associations were found between any of the vigor variates and juice pH (Table 3; Figure 2B2; Supplementary Figures S3-B2, S4-B2 in Supplementary material).

No consistent associations were found between the vigor variates and flesh translucency for cv. Smooth Cayenne (Table 3). For cv. Sugarloaf, NL was the strongest vigor variate to be weakly but consistently associated with flesh translucency (Table 3; Supplementary Figures S3-C1, C2 in Supplementary material).

Our data show that in the pineapple crops, most of the external quality attributes of the fruit at harvest were significantly and positively associated with the initial vigor of the plant at the moment of artificial flower induction (Table 3).

Differences in initial plant vigor accounted for a high proportion of the variation in fruit weight. Comparing the three vigor variates, the highest proportion of the heterogeneity in fruit weight was explained by NL × DL (Table 3; Figures 1A1–A4). The association between the NL × DL and the fruitlets number and the fruit weight at harvest was positive. Reasons explaining this are likely that out of the three vigor variates, NL × DL would be best related to leaf area, and that higher values of the NL × DL at the time of artificial flower induction thus would indicate a higher leaf area and consequently a higher photosynthetic capacity and amount of assimilates available in a plant at the time of artificial flower induction i.e., at the end of the vegetative phase. Since the production of new normal leaves ceases once flowering is induced (Bartholomew and Malézieux, 1994), the available assimilates at the flower induction time that were allocated to the roots and leaves, now additionally are partitioned to the new sinks, i.e., the infructescence, crown and peduncle. Earlier studies showed that a large proportion of assimilates is allocated to the infructescence and the crown (Marler, 2011b). This means that the more assimilates are available at flower induction, the higher would be the fruit weight. The association of fruit weight with plant vigor at flower initiation shows the importance of the development stage and morphology of the plants at flower induction for final fruit quality, and is consistent with experiments in which later flower induction increased fruit weight in whole crops (Mitchell, 1962; Bartholomew et al., 2003) and in individual plants (Van Overbeek, 1946).

Our data show that the positive association between the initial plant vigor and later fruit weight was mainly due to an effect on the infructescence weight whereas the effect on the crown was much smaller and only consistently significant for one vigor variate (Table 3; Figures 1C1–C4; Supplementary Figures S1-C1 to C4 and S2-C1 to C4 in Supplementary material). Such differences in the effect on the infructescence and crown could probably be explained by the differences in timing of their development. The initiation of the florets may have continued longer in infructescences bearing more florets, which may have delayed the onset of crown formation.

Each floret differentiates into one fruitlet. Our results revealed that in all experiments, all plant vigor variates are positively associated with the number of fruitlets at harvest (Figures 1H1–H4; Supplementary Figures S1-H1 to H4 and S2-H1 to H4) indicating that in vigorous plants more florets were able to develop into fruitlets. As with fruit weight, NL × DL was the plant vigor variate explaining the largest proportion of variation in number of fruitlets. After flower induction, pineapple plants show an increase of the width of the apex (Wee and Rao, 1979) which bears the florets. Thus, more assimilates available—plants with high NL × DL—would lead to high volume increase of the apex and consequently high number of florets that will differentiate into fruitlets.

Considering the fruit height, it was found that the association between NL × DL and the fruit height was strong in the experiments with cv. Sugarloaf (R² = 0.402 and 0.390 in Experiments 1 and 2, respectively) and significant but much weaker in the experiments with cv. Smooth Cayenne (R² = 0.060 and 0.024 in Experiments 3 and 4, respectively) (Table 3; Figures 1D1–D4). These differences were due to the differences between cultivars in the associations between NL × DL and fruit height components: infructescence height and crown height. The former was positive for both cultivars, but the association between NL × DL and crown height was positive for cv. Sugarloaf (Table 3; Figures 1F1,F2) and negative for cv. Smooth Cayenne (Table 3; Figures 1F3,F4). This means that for cv. Smooth Cayenne, more vigorous plants produce fruits with a shorter crown (Figures 1F3,F4) lowering then the total fruit height, hence the poor association observed between the NL × DL and the fruit height at harvest for cv. Smooth Cayenne. This is also in line with the significantly negative correlations between the infructescence height and the crown height for cv. Smooth Cayenne (Supplementary Tables S3, S4 in Supplementary material).

The negative associations between NL × DL and the ratio crown height: infructescence height (Table 1; Supplementary Tables S1, S2, S3, and S4 in Supplementary material; Figure 1-G1 to G4) follow logically from the clear increase in infructescence height with increase in NL × DL (Figures 1E1–E4) combined with the poor and negative association between the initial plant vigor and the crown height. Reasons for such differences are described above.

These associations between the plant vigor at artificial flower induction and the external fruit quality attributes suggest there is a good chance of decreasing the heterogeneity in fruit quality within a lot by increasing the uniformity of the crop at the moment of flower induction. This could be achieved before and after planting. Uniform soil conditions will be the basis for uniform growth after planting. Also selecting more uniform planting material will increase uniformity; using specific clonal planting material production, for example as suggested by Agogbua and Osuji (2011) in Nigeria, could be an option to not only provide uniform planting material but also enough planting material to cope with the current practice of mixing planting material at planting. After planting, the selective application of extra fertilizers to the least developed plants may increase their vigor so that they reach the vigorous plants before the moment of artificial flower induction. In theory, also pruning leaves from the most advanced plants may increase uniformity by decreasing the vigor of the best plants, but this may also lower the average fruit quality. Also, selective flower induction, i.e., induction of the least and most vigorous plants at different times, could help improve the uniformity of the fruit at harvesting time.

Heterogeneity in pineapple taste is also a problem in the pineapple supply chain (Fassinou Hotegni et al., 2014). In the present paper, TSS and juice pH were assessed to represent taste. Our findings indicated that the variation in TTS and especially in pH were very small compared to those in fruit and infructescence weight. There were no clear associations between the initial plant vigor and TSS, juice pH or flesh translucency since the results were not consistent across experiments. Such results are in line with the idea that fruit ripening and maturation—affecting TSS and juice pH—occur autonomous in proportion to the fruit size established, and in relation to time and external conditions. However, for the flesh translucency, results showed a consistent positive correlation between translucency and TSS in the experiments with Smooth Cayenne (Supplementary Tables S3, S4 in Supplementary material). These results on flesh translucency in cv. Smooth Cayenne confirm the findings of Chen and Paull (2001), that translucency is affected by sugar concentration at harvest time.

In this study, the experiments with cv. Smooth Cayenne showed a higher variation than the experiments with cv. Sugarloaf in some external quality attributes and internal quality attributes (Table 2). We attribute most of these differences to genotypic differences and differences in the cultivation practices of these cultivars, although the differences between experiments also might be affected by the location and season. The high variation in the crown weight and height in cv. Smooth Cayenne compared to cv. Sugarloaf (Table 2) might originate in part from the diverse planting material; mixtures of hapas and suckers were used in cv. Smooth Cayenne planting while only slips were used in cv. Sugarloaf planting. It is well-known that plants grown from suckers initiate fruits earlier than plants grown from hapas (Bartholomew et al., 2003); so variation would exist in the growth of the two types of planting material. In our study, variation in plant vigor variates at flower induction was similar for both cultivars. Therefore, variation in growth of the hapas and suckers expresses itself later during the generative phase increasing variation in crown weight and height in cv. Smooth Cayenne and suggesting a relationship between the type of planting material used and the morphology of the fruit produced. The higher variation in the ratio crown: infructescence height in cv. Smooth Cayenne than in cv. Sugarloaf was certainly the consequence of a higher variation in crown height and opposite associations between plant vigor and crown height, and plant vigor and infructescence height (Supplementary Tables S3,S4 in Supplementary material).

When considering the internal quality attributes, variation in TSS and translucency was higher in cv. Smooth Cayenne than in cv. Sugarloaf while for the variation in juice pH the opposite was observed. Differences in variation in TSS between the two cultivars might be due to maturity synchronization practices in cv. Smooth Cayenne which might increase variation in TSS. In pineapple fruits, at 2 weeks before the ripening of the fruit, the TSS increases until the harvest (Singleton and Gortner, 1965); when maturity is synchronized by applying Ethrel on the skin of the fruits—at different stages of natural ripening process (different TSS)—degreening of the shell is accelerated artificially (Smith, 1991). Then, the variation in TSS will be higher in cv. Smooth Cayenne when compared to cv. Sugarloaf where no maturity was synchronized. Higher variation in flesh translucency in cv. Smooth Cayenne might be due to the high variation in TSS; TSS and translucency are positively associated in cv. Smooth Cayenne as shown in Supplementary Tables S3,S4 in the Supplementary material.

The weight of slips but not the number of slips accounted for an extra part of the variation in fruit weight, infructescence weight and fruit height in addition to the effect related to the initial plant vigor (Table 6). This effect of the slip weight was positive (Table 6). Differences in fruit weight, infructescence weight, and the height of the fruit thus may not originate only from differences in initial plant vigor but also to a small extent from differences in the weight of slips produced. This might be the result of transfer of assimilates from the slips to the fruit (Marler, 2011a). Slips are composed of leaves and the slip weight will give a better idea of the photosynthetic capacity of the slips than the slip number. A better understanding of the role and the determinants of the variation of slip production within a crop and between crops would help to improve fruit weight, infructescence weight, and fruit height.

The Associate Editor Olaf Van Kooten declares that, despite being affiliated to the same institution as the authors V. Nicodème Fassinou Hotegni, Willemien J. M. Lommen and Paul C. Struik, the review process was handled objectively and no conflict of interest exists. The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.