One hundred and sixty local landraces, collected from fourteen provinces in Burkina Faso (Figure 1), were used for the participatory selection. These provinces were selected based on the production gradient of Hibiscus sabdariffa L., and the average farm size of farmers. The trials were conducted at the SUSTLIVES project experimental site in Gampèla, in the Sudano-Sahelian zone, and experimental plot of the Institute for Environment and Agricultural Research (INERA) in Farakô-Ba, in the Sudanian zone, from July to December 2023 (Figure 1). The SUSTLIVES project site is located on the Ouagadougou-Fada N’Gourma axis, between 1° 35’ west longitude and 12° 41’ north latitude. It has a Sudano-Sahelian climate, with vegetation consisting mainly of wooded savannahs, shrubby savannahs, and rice-growing areas. Average maximum temperatures range from 33.3°C to 34.6°C, while minimum temperatures fluctuate between 12.5°C and 23.3°C. The average rainfall recorded during the rainy season (July to October) was approximately 583.5 mm. Its soil is characterized by a predominantly silty-sandy texture and a slightly acidic pH of 5.70 (Kiebre et al., 2025). The Farakô-Ba site is located southwest of Bobo-Dioulasso on the road connecting Bobo-Dioulasso to Banfora, between 4°20′ west longitude and 11°6′ north latitude (Kone et al., 2025). Its vegetation belongs to the South Sudanic phytogeographic sector, characterized mainly by wooded, grassy savannah, sometimes quite dense. Weather records taken at the Farakô-Ba station indicate rainfall of 811.5 mm, spread over 35 rainy days between July and October. Its soil is characterized by a predominantly silty-sandy texture and a slightly acidic pH of 5.92.

For each site, Gampèla in the Sudano-Sahelian zone (SSZ) and Farakô-Ba in the Sudanian zone (SZ), the experimental design used is a randomized complete block design with three replications. Each accession was sown in a single row measuring 3.6 meters in length, consisting of seven hills. The distance between the rows was 0.8 m and between plants was 0.6 m. The distance between successive replications was 1.5 m.

The trial was conducted following agroecological practices aimed at enhancing soil fertility and reducing the use of chemical inputs. Soil preparation consisted of the application of manure at a rate of 5 t/ha, followed by shallow plowing to loosen the soil and create a good seedbed. Maintenance operations consisted of weeding and insecticide application. A natural biopesticide, SUPERFASO N, based on neem, developed by the Institute for Research in Applied Sciences and Technology (IRSAT) of the Department of Natural Substances (DSN) in Ouagadougou, was applied at a rate of 6 L/ha to control insect pests.

A total of 126 participants were involved in the study, including 86 at the Gampèla experimental site and 40 at the Farakô-Ba site. All individuals present during the participatory selection activities and having a defined role within the roselle value chain participated in the evaluation process. These participants were members of the local communities surrounding each experimental site and included producers, consumers, traders, and processors.

Before the selection of accessions, separate focus group discussions were conducted with men and women to identify the full list of selection criteria used by the different actors along the value chain. These gender-differentiated discussions made it possible to capture specific preferences and priorities and to establish a comprehensive list of criteria for evaluating roselle accessions. The following variables represent the selection criteria used by actors to select accessions: ACF: Leaf acidity; APC: Suitable for storage; COF: Leaf color; CYC: Short cycle; CYL: Long cycle; DIF: Leaf dimensions; DSM: More demanded on the market; FMP: Less perforated leaves; FNP: Non-pubescent leaves; FNPR: Non-perforated leaves; FOF: Leaf shape; FPR: Perforated leaves; NRP: Number of primary branches; PAC: Most appreciated by customers; PRE: More profitable; QUO: Organoleptic qualities; REF: Leaf yield; RES: Drought resistance; TEF: Leaf tenderness; TOB: Tolerant to biotic stress; VAM: Market value. These selection criteria were not measured instrumentally but evaluated by participants during the participatory selection process.

A questionnaire was then administered to each participant to collect socio-demographic data such as agroecological zone, ethnic group, education level, marital status, age, and gender, and to invite each participant to indicate, from the list generated during the focus groups, the criteria they personally used during the selection process. Participatory selection was then carried out at the experimental sites during the vegetative stage of the plants, two months after sowing. Each actor was invited to conduct an individual evaluation of the accessions according to their own preference criteria. This assessment consisted of ranking the accessions according to an ordinal scoring system: 9: accession considered most interesting; 7: very interesting; 5: moderately interesting; 3: not very interesting; 1: least interesting. Participants were then asked to specify the criteria they used to justify their selections, ranking them in order of importance. At the same time, each actor was asked to identify and justify the rejection of three accessions, specifying the reasons for this choice. Rejected accessions were scored using a negative scoring system: −5, −3, and −1. This scoring method made it possible to quantify preferences and rejections, thus facilitating the comparative analysis of the selection criteria and priorities of the various actors in the value chain.

The collected data were systematically recorded and processed using Microsoft Excel to ensure accuracy, consistency, and reliability in data management. Regarding age, the informants were grouped into three categories, following Assogba et al (Assogba et al., 2017): young (age< 30 years), adult (30 ≤ age< 60 years), and old (age ≥ 60 years). Before analysis, errors during data collection and entering were double-checked. The survey data were described and analyzed using the R software (R Core Team R, 2024). The chi-square test (χ²) was used to compare categorical variables between the two zones (Sudano-Sahelian zone and Sudanian zone). The level of statistical significance of differences for the variables was tested at the 5% threshold between the two agroecological zones. An index for ranking of accessions was calculated for the first three ranks following Getachew et al. (Getachew et al., 2010), Bayou et al. (Bayou et al., 2018), Zewdu et al. (Zewdu et al., 2018), Ouédraogo et al. (Ouédraogo et al., 2020), and Badjibassa (Akounda et al., 2024).

where a_m is the rank weight associated with accessions m (a₁ = 3, a₂ = 2, a₃ = 1), X_nm is the proportion of participants who ranked the m^th accessions in the n^th rank (n = 1 to 3 ranks), and m represents accession. This index reflects the level of preference or rejection expressed by participants. Accessions were then ranked according to the calculated index. For selection, a higher index indicates that the accession is more preferred by the participants. For rejection, a higher index indicates that the accession is more rejected by the community.

Correspondence analyses were conducted to elucidate the relationships between selection criteria and participant profiles, taking into consideration their ethnic affiliation, roles within the value chain, and age categories, using the packages FactoMineR (Husson et al., 2008) and factoextra (Kassambara and Mundt, 2020). The results were visualized through three plots: selection criteria projected according to ethnic groups, value chain actors, and age categories, respectively. Accessions that were selected at least once at each site were filtered, and a Venn diagram was created to visualize the specificity of selected accessions of Hibiscus sabdariffa L. for each study zone.

The participatory selection was carried out with 126 people, including 40 in the Sudanian zone (SZ) and 86 in the Sudano-Sahelian zone (SSZ) (Table 1). A strong disparity was noted in ethnic distribution (p< 0.001). The Mossi ethnic group predominated in SSZ (94.19%), whereas the Bobo and other ethnic groups were more common in SZ (45% and 40%, respectively). The mean age was slightly higher in SSZ (36.73 ± 13.88) than in SZ (33.92 ± 10.51), with a statistically significant difference (p< 0.001). Age group distribution also varied significantly between zones (χ² = 49.333, p< 0.001), though the majority of participants in both zones were adults. The roles within the value chain also differed significantly between zones (χ² = 17.873, p = 0.0005); producers were more common in SZ (57.5%), while end consumers and processors were more prevalent in SSZ.

Table 2 presents the top 10 Hibiscus sabdariffa L., accessions identified through the participatory selection in the Sudanian zone, Sudano-Sahelian zone, and across both zones (Overall). The ranking is based on a scoring and index system reflecting participants’ preferences. Out of the 160 Hibiscus sabdariffa L., accessions initially presented to participants for participatory selection, a total of 130 accessions, representing 81.25%, were selected across both climatic zones. In the Sudanian zone, 58 accessions (36.25%) were chosen, whereas in the Soudano-Sahelian zone, 111 accessions (69.38%) were selected.

In the SZ, the best-performing accession was CKN5 (index = 0.0708), followed by CKK1 (index = 0.0667) and PCL12 (index = 0.0625). In the SSZ, the top accession was CKN2 (index = 0.0394), followed by CEK3 and CKO8, both with an index of 0.0354 (Table 2). Some accessions, such as CKK1 and CKN5, were highly ranked in both zones, demonstrating a cross-regional preference. CKK1 ranked 2^nd in SZ (index = 0.0667) and 4^th in SSZ (index = 0.0335), making it the top-ranked accession overall, with a combined index of 0.0441. Similarly, CKN5 ranked 1^st in SZ and 9^th in SSZ (index = 0.0295), placing it 2^nd overall with an index of 0.0428.

The Venn diagram (Figure 2) illustrates the distribution of the 130 selected Hibiscus sabdariffa L., accessions across two agroecological zones: the Sudanian zone and the Sudano-Sahelian zone. Among the total accessions, 14.6% (n = 19) are found exclusively in SZ, 55.4% (n = 72) are exclusive to SSZ, and 30.00% (n = 39) are shared between the two zones.

The correspondence analysis reveals a significant relationship between ethnic groups and the criteria used for leaf selection (Chi² = 371.16; p< 0.001). The two dimensions account for all the observed variability (Figure 3). The first (80.4%) is mainly associated with the Mossi, who prioritize leaf perforation (FPR, FMP), absence of pubescence (FNPR), and related traits. The second (19.6%) reflects the preferences of the Bobo and other ethnic groups, emphasizing leaf shape (FOF), acidity (ACF), branching (NRP), and absence of pubescence (FNP).

This is consistent with the classification based on the principal components, which shows that the Mossi are mainly associated with PRE, FPR, FNPR, FMP, DSM, and PAC (Supplementary Figure S1); the Bobo with DIF, QUO, NRP, Uses, ACF, and VAM; and the other ethnic groups with CYL, FNP, RES, REF, APC, CYC, TOB, FOF, COF, and TEF. Together, these findings highlight distinct and well-structured patterns in selection preferences across ethnic groups.

The results show a highly significant association between value chain actors (consumers, producers, traders, processors) and leaf selection criteria (Chi² = 360.21; p< 0.001). The first two dimensions explain 84.8% of the total variance, with 51.4% explained by the first and 33.4% by the second (Figure 4). The first dimension clearly separates traders, who strongly associate with the criterion of less perforated leaves (FMP), from producers and consumers, who favor criteria such as non-pubescent leaves (FNP, NRP) and leaf tenderness (TEF). The second dimension contrasts consumers and producers, on the one side, with traders and processors, on the other side, mainly based on traits like acidity (ACF) and leaf pubescence (FNPR).

These findings are consistent with the classification based on the principal components, which shows that consumers are mainly associated with PRE, FPR, FNPR, FMP, DSM, and PAC; processors with DIF and QUO; producers with CYL, FNP, RES, REF, APC, CYC, TOB, FOF, COF, and TEF; and traders with NRP, Uses, ACF, and VAM (Supplementary Figure S2). Overall, these analyses highlight well-differentiated selection preferences according to the actors’ roles in the value chain, reflecting the specific needs of each group.

The results show a highly significant association between age groups and leaf selection criteria (Chi² = 168.11; p< 0.001). The two dimensions account for 100% of the total variance, with 80.7% explained by dimension 1 and 19.3% by dimension 2 (Figure 5). Dimension 1 strongly separates adults from the other age groups. Adults are associated with positive coordinates and strongly related to criteria such as leaf tenderness (TEF), non-pubescent leaves (FNPR), leaf size (DIF), and number of primary branches (NRP). In contrast, young and old individuals are positioned negatively on this axis and show stronger associations with visual and structural traits such as less perforated (FMP) and perforated leaves (FPR), leaf shape (FOF), and leaf pubescence (FNP). Dimension 2, which contributes less (19.3%), mainly distinguishes older individuals, who are positively associated with this axis and linked to leaf acidity (ACF); the younger and adult groups, by contrast, show lower contributions to this dimension.

These findings are consistent with the classification based on the principal components, which shows that adults are mainly associated with FNPR, CYL, DSM, RES, DIF, and PAC; young individuals with FPR, FMP, and FNP; and older individuals with REF, ACF, CYC, Uses, TOB, NRP, FOF, QUO, COF, APC, TEF, and VAM (Supplementary Figure S3). Together, these analyses highlight distinct and structured patterns of leaf selection across age groups, reflecting differences in perception and preference that vary with age.

Table 3 reveals significant variation in preferences for phenotypic traits of Hibiscus sabdariffa L., which depend on agroecological zone (SZ and SSZ), ethnic background, value chain role, and age group. In the Sudanian zone, phenotypic traits such as leaf color (COF), leaf shape (FOF), and leaf yield (REF) are particularly valued, especially by consumers, producers, and younger individuals. Conversely, in the Sudano-Sahelian zone, preferences shift toward organoleptic quality (QUO), drought resistance (RES), and market value (VAM). Processors and traders prioritize storage-related attributes (APC) and consumer preferences (PAC). Young respondents demonstrate a strong interest in a combined set of traits, including leaf shape and color (FOF, COF), tolerance to biotic stress (TOB), organoleptic quality (QUO), and market value (VAM).

Table 4 presents a ranking of the top five (05) accessions of Hibiscus sabdariffa L., selected from two climatic zones, based on several phenotypic criteria. In the Sudanian zone, the accession CKN5 stands out as the most promising, primarily due to its large leaf size, high number of primary branches, and good storability. Other accessions, such as CKK1 and PCL12, are valued for a variety of traits, including dark green leaf color, tenderness, resistance to biotic stress, and high leaf yield. In the Sudano-Sahelian zone, the accession CKN2 is preferred for its combination of large leaf size, desirable organoleptic qualities, and short growth cycle. The other selected accessions also display a diversity of characteristics, including leaf shape and color, drought tolerance, and high market value. Figure 6 illustrates the photographs of the ten accessions selected by all stakeholders along the value chain.

The analysis of community preferences across agroecological zones revealed the most rejected accessions based on a weighted score index, where a higher index indicates stronger rejection by the community (Table 5). In the Sudano-Sahelian zone, CKO2 recorded the highest rejection index (0.0833), followed by HBDf3 (0.0562) and COB3 (0.0543), reflecting a strong negative perception. Similarly, in the Sudanian zone, HBB3 ranked first in rejection (0.0583), with CKO2 again appearing among the top rejected accessions (0.0458), suggesting a consistent pattern of disapproval across zones. These findings highlight the need for further investigation into the traits driving community rejection and underline the importance of incorporating local preferences into participatory selection and dissemination strategies.