© P. Carré et al., Published by EDP Sciences, 2026

1 Introduction: background and objectives

This work aims to clarify how genetic and environmental factors influence the dehulling suitability of sunflower achenes and to identify agronomic levers for optimization. Previous studies frequently associate low hullability with increased oil content and reduced 1000-seed weight. Table 1 summarizes literature correlations between hullability and characteristics such as oil content, hull content, thousand-seed weight, bulk density, and oil/hull content relationships.

Correlations (R) between hulling ability and oil content are moderate to strong and negative (−0.42 to −0.79), while hull content and thousand-seed weight each show positive effects (R = 0.46–0.85 and 0.18<R<0.87, respectively); bulk density correlates negatively (−0.05<R<−0.63).

Dedio (1993) indicated oil content is the main determinant of hulling ability, followed by thousand-seed weight and, to a lesser extent, bulk density; wax content had no significant effect. Baldini et al. (1994) confirmed these trends and observed that certain hybrids combine desirable agronomic traits, high oil content, and satisfactory hulling. Denis et al. (1994) demonstrated that pedoclimatic factors play a decisive role in hulling ability. They found that sunflower seeds from Spain, which were easier to hull, had lower oil content and a higher hull proportion than those from France. This difference was likely due to water stress in Spain, which limited achene filling. Despite environmental impacts, heritability of hullability was estimated at 80%, similar to oil content. In subsequent work, Denis and Vear (1996) found that under water-limited conditions, oil content strongly and negatively correlated with hullability, whereas thousand-seed weight was a better predictor in more favorable environments.

Nel et al. (2000) reported that water stress during flowering reduces thousand-seed weight, hullability, and oil content compared to irrigated controls, likely due to decreased pericarp growth. Di Leo et al. (2004) evaluated hullability across cultivars on soils with varying fertility, finding that higher yields and oil content—typical of more fertile plots—corresponded to reduced hullability, suggesting a trade-off between productivity and hulling performance.

Cultivation conditions, particularly planting density, can substantially affect achene size. Nel et al. (2002) demonstrated that, in most varieties, hullability decreased with increased planting density, except for one cultivar characterized by large achenes with good hullability. Crnobarac et al. (2014) also found that reduced planting density produces larger achenes. Dauguet et al. (2016), analyzing genetic and pedoclimatic factors, found thousand-seed weight and cellulose content significantly increase hull extraction (R = 0.41, 0.71), while oil content was negatively correlated (−0.60). Higher cellulose, reflecting greater hull proportion, strongly correlates negatively with oil content (−0.78); no link was found between thousand-seed weight and cellulose.

Langhi et al. (2021) assessed 23 hybrids across three sites, reporting that genetic factors account for 71% of hull content variability, environmental effects for 19%. Thousand-seed weight and achene size proved predominantly genetically determined, whereas seed filling depended on pedoclimatic influences. Nitrogen fertilization increases kernel proportion relative to hulls (Jarecki et al., 2022), although observed hull contents were consistently below typical sunflower values.

In sum, scientific literature indicates that precise environmental effects on dehulling remain incompletely understood; oil content consistently exerts a negative influence on hullability, as do thousand-seed weight and bulk density. These traits interact with plant water coverage. A working hypothesis compatible with previous findings is that abiotic stress during achene formation and grain filling chiefly governs hullability. Flowering time dictates achene size and pericarp development; stress at this stage reduces pericarp and lignification, impacting dehulling efficiency. The subsequent phase determines kernel filling, hull-to-kernel ratio, and oil content, with stress hypothesized to facilitate easier dehulling through mechanisms such as increased hull-kernel voids, altered adhesion, or pericarp weakening.

This study tests this hypothesis by comparing hulling performance of 30 sunflower cultivars grown over two years and four locations, aiming to identify major environmental factors affecting hullability.

2 Material and methods

2.1 Field trials and experimental design

Thirty commercial sunflower hybrids were cultivated during the 2021 and 2022 growing seasons across four sites each year. Figure 1 illustrates the geographical distribution of these field locations on the map of France.

Table 2 summarizes the main characteristics of the trial sites. Azay-21 and Azay-22 were separated by ∼2 km, while the two Cham trials were only 300 m apart and shared similar soils. Only two locations were maintained across both years due to field availability constraints, crop rotations requirement and practical considerations related to the organization of the teams that conducted the experimental work. Water availability differed markedly between years, with plant water requirement coverage at 72% in 2021 and 26% in 2022 (precipitation/potential evapotranspiration). All plots were harvested simultaneously within each trial. The experimental design consisted of two hybrid replicates, with 0.6 m row spacing and a sowing density of 73,000–75,000 seeds ha⁻¹. Nitrogen fertilization was adjusted to site yield potential and residual soil N. In 2021, applications were 60, 70, and 46 kg ha⁻¹ at Azay, Cham, and Vill, respectively; in 2022, they were 120, 60, 120, and 37 kg ha⁻¹ at Azay, Cham, Gail, and Lev, with no data for Gar-21. Irrigation was applied to support emergence: 20 and 50 mm at Cham (2021, 2022), 15 mm at Vill (2021), and an additional 30 mm at Cham on July 12, 2022. During the season, flowering onset was recorded, plant maturity scored (1–9 scale). Harvested plot areas ranged from 12 to 18 m².

Fig. 1 Location of the field trials.

2.2 Measurement of hulling ability

Hulling ability was assessed following Dauguet et al. (2016). Approximately 80 g of achenes were equilibrated for at least two weeks at ambient temperature in a chamber with a solution of saturated ammonium nitrate, maintaining 62.5–65.5% relative humidity at 20–25°C (Winston and Bates, 1960). Four 20 g subsamples were then prepared: three for hulling tests, one for moisture determination.

Hulling was performed with a centrifugal impactor (200 mm disk with radial channels, 2000 rpm; peripheral speed 20.9 m · s⁻¹), projecting achenes against a 270 mm concentric target. Each sample underwent three impacts. Resulting material was fractionated in two steps: passage through a 2 mm rotating sieve (“fines” fraction), followed by air-column separation into “hulls” and “hulled kernels” (the latter including intact and partially hulled kernels with residual pericarp).

The extracted hull content (EH) was analyzed as the response variable. The hulling rate, defined as EH relative to initial hull content, was excluded since Dauguet et al. (2016) showed strong correlation between both measures and limited methodological benefit.

Each dehulling test was performed in triplicate on pooled samples from the two plots. The block effect was not tested since prior studies indicated that it was negligible, accounting for only 0.8% of block × environment variance (Denis and Vear, 1994) and 0.2% of total variance in trials with three replicates (Baldini et al., 1994).

2.3 Analytical procedures

Moisture was determined by drying at 103°C ± 2°C in an oven at atmospheric pressure, according to NF V03-909 (AFNOR, 2002). Oil content was measured by overnight oven-drying followed by nuclear magnetic resonance (NMR) analysis using a Bruker mq20 spectrometer. Protein content was assessed with a Thermo Fisher Scientific Flashsmart elemental analyzer using the Dumas method (ISO 16634-1, ISO, 2008). Thousand-kernel weight was obtained after overnight oven-drying, counting seeds with a seed counter and weighing subsamples. Specific weight (bulk density) was determined with an internal protocol, as the official method was not applicable due to insufficient seed quantity: a container of known volume was filled via a stand-mounted funnel, and excess seeds were leveled to ensure consistency.

2.4 Meteorological data collection

For each experimental site, meteorological data were obtained from the nearest MétéoFrance station. Records included precipitation, maximum and minimum temperatures, sunshine duration, wind, and humidity, which were used to estimate potential evapotranspiration (PET) and thus crop water requirements. Météo-France calculates potential evapotranspiration (PET) values mainly using the Penman-Monteith formula, which is internationally recommended and standardized by the FAO. From these data, temperature sums (ΣT) were calculated from sowing to flowering or harvest ([Tmax + Tmin]/2–6°C), along with mean temperatures (Tm) during flowering and grain filling. Cumulative precipitation (ΣPr) and PET (ΣPET) served to calculate water requirement coverage rates by precipitation and irrigation (WRC). These coverage rates were determined for the entire crop cycle from sowing to harvest (_H), sowing to flowering (_SFl), for the flowering (_Fl) period, including the week before and after flowering and for the end of the cycle (filling period, _Fil)), from 15 days after the onset of flowering until harvest. In cases where irrigation was applied, the water supplied was added to the precipitation totals. Abiotic stress was assessed through the number of days above 30°C (nDo30) during flowering and grain filling, while sunshine hours (In) were also recorded for each period.

2.5 Statistical analysis

Statistical analyses were performed using R and RStudio, version R 4.5.0 (2025-04-11 ucrt) (R Core team, 2025, Posti software PBC, 2025). The R code was developed with the assistance of Perplexity for programming optimization. All results and interpretations remain the sole responsibility of the authors.

3 Results

3.1 Genetic and environmental effects on hulling ability

Figure 2 shows the distribution of hull extraction rates for each commercial hybrid, whose names were coded at the request of the breeding partners involved in the project. The figure indicates that environmental sensitivity may vary between cultivars, with C18 and C19 exhibiting moderate variability, in contrast to C20, which displays a broader distribution. The reader can find, in the supplementary material, a summary of the mean values with their standard deviations in Table S2 , and in Table S3, the individual data characterizing the sunflower achenes.

The analysis of variance (Tab. 3) shows that the genetic effect (Var) accounts for approximately 33% of the variability in TC, compared to 50% for the environment (Env), with the remainder explained by environment × variety interactions and other random effects. The effect of genotype × environment interactions could not be evaluated because only a single seed sample was tested for each variety in each trial.

The comparison of means using the Snedecor, Newman, and Keuls test (Tab. 4) distinguishes 11 groups of varieties, among which the two extremes (a and k) share no common members.

Fig. 2 Distribution of the proportion of extracted hulls from each commercial hybrid (g/g). The diamond symbol represents the mean for the 8 locations of each cultivar.

3.2 Influence of achene characteristics on dehulling performance

Table 5 presents correlation coefficients among available continuous variables. For brevity, correlations below 0.4–which showed minimal impact on results − are omitted. A moderate negative correlation (R = −0.45) emerges between oil content and oleic acid content, indicating reduced oil content in varieties with higher oleic acid levels. Conversely, yield and oil content exhibit a weak positive correlation (R = 0.45). The classic inverse relationship between hulling performance and oil content is reaffirmed (R = −0.44). Unsurprisingly, yield demonstrates stronger positive correlations with higher thousand-seed weights (R = 0.63) and bulk density (R = 0.47), as well as increased grain numbers per square meter (R = 0.64). Protein content (dry defatted basis), oil content, and bulk density all show negative correlations with extracted hulls (−0.33, −0.44, and −0.45 respectively). Yield also displays a weak but statistically significant negative correlation with this parameter.

The supplementary material includes a figure (Fig. S1) displaying scatter plots and R² values for relationships between EH and oil content. This figure demonstrates that correlations strengthen slightly when examining results for individual environments (overall R² = 0.19 for pooled data), with site-specific R² values ranging from 0.28 to 0.34 in 2021 and 0.16 to 0.53 in 2022.

For thousand-seed weights (Fig. S2), environment-specific analyses reveal no deviation from the global trend of absent correlation. Bulk density (Fig. S3) show the overall R² of 0.20 disappears at individual environment level, with only Gar-2021 exhibiting R² = 0.40, two 2021 sites between 0.22–0.23, and five remaining sites displaying minimal values. Similarly, no yield-EH correlation emerges at individual environmental scale (Fig. S4).

Figure S5 reveals a notable negative correlation between protein content (dry defatted basis) and TC, with 6/8 sites showing R² > 0.2. Seed numbers per square meter show no R² exceeding 0.12 in Figure S6. Finally, Figure S7 confirms the absence of correlation between oleic acid content and hulling rate, consistent with their low R value.

3.3 Relationships between meteorological parameters and hullability

To assess the influence of meteorological factors, the analysis focused on mean values from the eight sites. Although several flowering-dependent variables vary among cultivars, they are not fully independent, and sowing-to-harvest data are identical within each trial. Due to the limited number of observations, principal component analysis was not suitable, as the required ratio of at least three observations per variable was not met.

The correlation matrix (Tab. 6) shows significant negative associations between hull extraction (EH) and water requirement coverage during grain filling (R = –0.87) and with flowering date (R = –0.79). Positive correlations were found with the number of days above 30°C during grain filling (R = 0.77), water requirement coverage during flowering (R = 0.72) and from sowing to flowering (R = 0.71), cumulative potential evapotranspiration to harvest (R = 0.71), and sunshine hours during grain filling (R = 0.71).

Water coverage during flowering and grain filling emerged as the most relevant variables, as they are uncorrelated and reflect distinct developmental stages. The positive association during flowering suggests that sufficient water promotes pericarp development (via sclerenchyma thickness and size), enabling embryo expansion and increasing hull rigidity for efficient impact fracture. Conversely, the strong negative correlation during grain filling (R = –0.87) may reflect reduced kernel filling within the pericarp, weakening hull–kernel contact and lowering hull adhesion, which diminishes impact energy absorption.

The model EH = WRC_Fl + WRC_Fil accounted for 91% of EH variability. ANOVA results and model coefficients are given in Table 7, and the observed–predicted relationship is shown in Figure 3.

The regression coefficients show that water coverage during grain filling has a stronger effect on hull extraction than during flowering, with values of −0.21 and 0.04, respectively. Their operational ranges, however, differ widely: 17–52% for grain filling versus 4–131% for flowering. In absolute terms, this corresponds to effects of −3.5% to −10.7% on hull extraction during grain filling, and 0.2–5.3% during flowering. The model predicts that extreme hydrological contrasts could generate an 8.2% difference in hull extraction during grain filling and 5.1% during flowering Combined, these phases could theoretically account for a 13.3-point variation in hulling outcomes under contrasting hydrological regimes.

Another approach for comparing the relative importance of the two WRC parameters involves calculating their partial R² values. This is achieved by computing the sum of squared errors (SSE) for three models: M1 (EH = WRC_Fl + WRC_Fil), M2 (EH = WRC_Fl), and M3 (EH = WRC_Fil), then deriving the partial R² values for each WRC variable accordingly to equations (1) and (2).

$${\text{R}}_{\text{WRC}F\text{l}}^2=\frac{{\text{SSE}}_{\text{M}1}-{\text{SSE}}_{\text{M}3}}{{\text{SSE}}_{\text{EH}}},$$(1)

$${\text{R}}_{\text{WRC}Fi\text{l}}^2=\frac{{\text{SSE}}_{\text{M}1}-{\text{SSE}}_{\text{M}2}}{{\text{SSE}}_{\text{EH}}}.$$(2)

In result, this calculation demonstrates that the coverage of water requirement during flowering variable alone explains 15.9% of EH variability, while coverage during the filling period accounts for 38.4%, with 36.2% of variability stemming from the shared effect of both variables.

Fig. 3 Scatter plot comparing observed EH values and predicted ones.

3.4 Residual variance analysis and genetic contribution

After evaluating the mean effects of the WRC_Fl and WRC_Fil variables, which characterize the environmental conditions of each trial, it became relevant to investigate the factors influencing the variability of residuals at the cultivar level. To this end, we calculated the following variables (Equations 3 and 4):

$$\text{E}{\text{H}}_{\text{i}}^{\text{r}}=\text{E}{\text{H}}_{\text{i}}^{\text{obs}}-\text{E}{\text{H}}_{\text{i}}^{\text{pr}},$$(3)

$$\text{WR}{\text{C}}_{\text{i}}^{\text{r}}=\text{WR}{\text{C}}_{\text{i}}^{\text{obs}}-\text{WR}{\text{C}}_{\text{i}}^{\text{pr}}.$$(4)

Superscripts: ‘r’ for residues; ‘obs’ for observed; ‘pr’ for predicted using the model of Table 8.

Subscripts: ‘i’ for the variety i [1 to 30]

Table 8 shows the anova of the model explaining the residues of EH by the residues of WRC at flowering and during filling period.

The model explains 63% of the residual variation in EH, with genetic effects becoming predominant by capturing 61% of the variance. While the WRC_Fl effect remains statistically significant, its practical impact is minimal, accounting for only 2.1% of Type I variance − a pattern potentially linked to precipitation variability during the flowering period. As anticipated, WRC_Fil demonstrates no measurable influence, as this parameter shows negligible dependence on flowering date variations.

3.5 Impact of precocity, fatty acid profile, and soil type

The analysis of variance for the variables earliness (“flowering” and “maturity”), “soil type,” and “type” (linoleic vs. oleic) was performed. The results are presented in Table 9.

Soil type significantly influences hulling rate, though this parameter cannot be disentangled from meteorological conditions due to confounding factors − certain soil types only occurred at single site-year combinations, preventing isolation of soil water reserve effects from precipitation impacts. However, we hypothesize that soils with higher available water capacity may improve hulling outcomes by mitigating flowering-stage water stress compared to shallow soils or negatively affect hulling when they enhance water supply in filling period.

The flowering period remains statistically significant but explains only 4.9% of variance. Mean comparisons from the SNK test are detailed in Table 10.

Mean comparisons revealed only two significantly distinct groups: mid-early versus late maturity cultivars, with the latter showing reduced hulling rates. This pattern stems from late-maturing varieties exhibiting the highest oil content levels, as lipid accumulation dynamics may influence pericarp-kernel adhesion properties during seed development. Higher oil contents were associated with thinner, less voluminous hulls, thereby restricting the available space for embryo development while promoting adhesion formation between hull and kernel components.

4 Discussion

4.1 Interpretation of results and literature comparison

Our analysis indicates that genetic factors explain ∼33% of hulling ability variability, contrasting with Denis et al. (1994), who reported narrow-sense heritability estimates of 0.73–0.85. This difference likely reflects methodological contrasts: Denis et al. measured parent–offspring transmission under controlled conditions, whereas our field-based study was restricted to commercial hybrids. Pedoclimatic factors accounted for >50% of hullability variance, implying that even superior cultivars may produce poorly hullable achenes when exposed to water stress during flowering followed by abundant water during grain filling.

Such strong environmental plasticity is consistent with Denis et al.’s observation that water limitation alters hullability–oil content relationships. However, our multi-environment dataset extends this notion, showing that phase-specific water availability—especially during flowering and achene filling—dynamically modulates pericarp–kernel structural interactions. These findings suggest that although breeding can exploit the genetic determinism of hullability, its phenotypic expression remains highly sensitive to agronomic conditions, notably irrigation and nitrogen regimes favoring oil synthesis and kernel filling.

For the industry, this underscores that cultivation site and abiotic stress management—especially water availability—are as critical as varietal choice. For sectors targeting highly dehulled sunflower meal (>45% protein, low lignin), sourcing sunflower seeds from regions ensuring sufficient water at flowering but limited supply toward the end of the cycle would be optimal. Such conditions are typical of shallow soils, where early-season irrigation remains feasible.

4.2 Correlation between oil content and hullability

The negative correlation between hullability and oil content (R = –0.44, p < 0.001) supports earlier findings summarized in Table 1, which identified oil accumulation as a competing sink for pericarp development.

Interestingly, the overall correlation is weaker than those observed within individual trials. At the global scale, R² is only 0.19, whereas within specific environments values range from 0.28 to 0.53 (Supplementary Material, Fig. S1), depending on site and year. This pattern indicates that, once environmental effects are removed, the genetic contribution more clearly opposes oil deposition to hullability. Denis and Vear (1996) attributed this negative relationship to a reduced pericarp proportion during achene development, reflecting biosynthetic competition between lipids and lignified tissues. Such competition results in thinning of the sclerenchyma among high-oil hybrids, with diminished pericarp rigidity limiting hulling efficiency—a hypothesis first introduced by Morrison et al. (1981). More recently, Lindström et al. (2022) confirmed the importance of sclerenchyma development, reporting significant positive correlations between hulling performance and both the number of sclerified layers (R = 0.4) and pericarp cell wall thickness (R = 0.5–0.6).

Overall, the oil content–hullability correlation explains less than 20% of trait variability. Thus, varieties that combine easier dehulling with acceptable oil levels remain selectable. From a production standpoint, the same oil yield per hectare can be obtained with slightly lower-oil varieties, as increased seed yield compensates for reduced oil concentration. For processors, minor reductions in oil content may be offset by gains in dehulling efficiency and post-dehulling productivity.

4.3 Seed size, density, yield, and composition: effects on hullability

4.3.1 Relationship between achene size and hullability

The modest predictive power of thousand-seed weight (TSW) for hullability (global R² = 0.20) supports Nel et al. (2000), who concluded that seed size alone cannot fully explain hullability variation. This is consistent with the correlations reported by Baldini et al. (1994) and Denis and Vear (1996) (R = 0.41 and 0.45, respectively), though Denis et al. (1994) observed stronger correlations ranging from 0.58 to 0.87. Such variability suggests that the relationship depends on the seed pools examined and their morphological diversity.

In summary, TSW by itself is not a decisive criterion for assessing dehulling ability; other seed traits provide more relevant explanatory power.

3.3.2 Bulk density

The correlation coefficient found in this work (−0.45) falls in the spread of literature data which ranges from −0.052 and −0.63. This weak correlation shows that hullability determinism cannot be explained by a unique factor like the hull content or the space between the pericarp and the embryo, both characters that would have led to stronger relationship between bulk density and hullability.

4.3.3 Yield implications

The observed weak negative correlation between yield and hulling ability (R = –0.168, p = 0.009) is encouraging, as the introduction of traits favoring dehulling should not markedly reduce productivity. This finding may appear counterintuitive, given meteorological evidence that favorable grain-filling conditions tend to reduce hullability (R = –0.87, p = 0.005).

Notably, two trials—Gail22 and Azay22—combined low yields with poor hulling performance. Both were characterized by extremely low water coverage during flowering (6.4% and 3.8%, respectively), highlighting the sensitivity of hullability and yield to early-season water deficits.

4.3.4 Other parameters (proteins, oleic acid).

The negative correlation between meal protein content and hullability is relatively weak (R = –0.32; p < 10⁻⁶). While this might suggest competition between protein synthesis and hull lignification, such an explanation is unlikely, since pericarp lignification ends relatively early (10–15 days after anthesis; Mantese et al., 2006), whereas protein accumulation begins around 20 days post-flowering (Ferjani and Ledoigt, 1990). A more plausible interpretation is that favorable grain-filling conditions simultaneously enhance protein content while reducing hullability.

For oleic acid, only a very weak positive association with hullability was observed, too minor to have practical significance.

4.4 Effect of water availability across developmental stages

Our analysis highlights differential impacts of water availability across developmental stages not previously described.

During the flowering phase, water requirement coverage correlated positively with hulling ability (R = 0.72, p = 0.043). This is consistent with pericarp development occurring soon after anthesis: water stress at this stage may restrict pericarp growth and sclerenchyma lignification, reducing hull thickness and rigidity. Such constraints can limit embryo space, leading to close pericarp–kernel contact and stronger adhesion. Simultaneously, less lignified and more elastic hulls may absorb impact rather than fracturing, increasing achene resilience and lowering dehulling efficiency.

Conversely, water supply during grain filling negatively affected hullability, in line with physiological expectations. Complete kernel filling reduces intra-pericarp space, tightening hull–kernel contact and reinforcing adhesion. Although water availability at this stage promotes oil and protein accumulation, these traits are only indirectly linked to dehulling. The idea that larger seeds with expanded pericarp volume might facilitate separation is offset by metabolic trade-offs: greater hull development requires assimilates that would otherwise support oil and protein synthesis. This partitioning helps explain why breeding for both high hullability and high oil content remains difficult, as genetic gains in one often compromise the other. Traits such as sclerenchyma bundle architecture or the presence of delignified rays may offer alternative breeding pathways to reconcile these objectives.

Together, flowering- and filling-phase WRC explain nearly all environmental variability in hullability. This finding provides both producers and processors with a robust diagnostic framework, supporting the development of forecasting tools and optimized agronomic management strategies to enhance seed quality under variable climatic conditions.

5 Conclusion, practical implications and future perspectives

This study advances the understanding of sunflower hulling ability by exploiting trials conducted under highly contrasting weather conditions. The diversity of environments highlighted the decisive role of water availability during flowering and its impact on pericarp development, a key factor for efficient dehulling. Insufficient water supply at flowering restricted pericarp growth, leading to thinner hulls and stronger hull–kernel adhesion, both of which reduced dehulling efficiency. Conversely, favorable water conditions promoted pericarp expansion and structural differentiation, facilitating separation at maturity.

The proposed hypothesis—linking hull–kernel adhesion to restricted pericarp development—requires confirmation through detailed anatomical observations, analyses currently underway and to be presented separately. Nonetheless, this work establishes a robust framework for interpreting environmental influences on hullability and opens perspectives for breeding and agronomic strategies to improve both seed quality and processing efficiency.

From an industrial perspective, the results suggest that selecting cultivars with superior hullability, even at a slight cost to oil content, could improve meal quality. Because seed purchases are oil-based, the economic impact would be minimal, while gains from better dehulling efficiency and reduced crushing and dewaxing costs should outweigh the loss of oil yield. However, this strategy may be jeopardized in seasons where climatic conditions reduce hullability, as observed in the 2024 French harvest. In such cases, technological adjustments remain essential to ensure processors meet protein content targets.

Funding

The authors acknowledge the financial support of SOFIPROTEOL under the FSRSO Project PROTOUR.

Conflicts of interest

The authors declare no competing interest.

Author contribution statement

Patrick Carré: investigations, original draft, reviewing & editing; Laurent Gervais: Fields trials, validation; Marie Coque: fields trials, validation; Jean-Philippe Loison: dehulling trials; Vincent Jauvion, achene analysis, reviewing & validation.

Ethics approval

Tools were used solely for improving English language expression and for assisting with R code generation. The AI was not involved in the study design, scientific writing, or data interpretation. All statistical analyses and results were independently verified and validated by the authors, who take full responsibility for the content.

Supplementary Material

Figure S1. Correlations analysis between oil content (%) and dehulling (EH, in g of hulls for 100 g of achenes) for each environment.

Figure S2. Correlations analysis between thousand seeds weight (g) and dehulling (EH, in g of hulls for 100 g of achenes) for each environment.

Figure S3. Correlations analysis between bulk density (g/100mL) and dehulling (EH, in g of hulls for 100 g of achenes) for each environment.

Figure S4. Correlations analysis between yield (Qx/ha) and dehulling (EH, in g of hulls for 100 g of achenes) for each environment.

Figure S5. Correlations analysis between the content in proteins of dry defatted material (Prot/DDM in g for 100 g of DDM) and dehulling (EH, in g of hulls for 100 g of achenes) for each environment.

Figure S6. Correlations analysis between the number of seeds per square meter and dehulling (EH, in g of hulls for 100 g of achenes) for each environment.

Figure S7. Correlations analysis between the oleic acid content in the seeds oil (g/100g) and dehulling (EH, in g of hulls for 100 g of achenes) for each environment.

Table S1. Mean values for a selection of studied variables.

Table S2. Summary of the achenes characteristics.

Table S3. Individual data about the achenes characteristics.

References

All Tables

All Figures

	Fig. 1 Location of the field trials.
In the text

	Fig. 2 Distribution of the proportion of extracted hulls from each commercial hybrid (g/g). The diamond symbol represents the mean for the 8 locations of each cultivar.
In the text

	Fig. 3 Scatter plot comparing observed EH values and predicted ones.
In the text