Evaluation and Participatory Selection of Promising Sweetpotato F 1 Genotypes in Uganda

— Most of the important sweetpotato (Ipomoea batatas (L.) Lam) traits are sensitive to environmental change. This necessitates evaluating new sweetpotato genotypes in different environments to identify those that are stable. To enhance adoption, the new sweetpotato genotypes should have farmer preferred traits thus the need for farmer involvement during selection. This study was conducted to: evaluate and select promising sweetpotato F 1 genotypes with wide and specific adaptation, in association with performance for farmer preferred traits. Twenty-one promising sweetpotato F 1 genotypes were evaluated at Namulonge, Serere and Kachwekano with Tanzania and NASPOT 1 as checks. The randomised complete block design with three replications was used. Scientists and farmers evaluated the agronomic performance and quality traits of the genotypes before and at harvest. Significantly (P<0.05) higher mean total storage root yield (TRY) of 25.5 t ha -1 was recorded at Namulonge than at Kachwekano and Serere. Genotypes G67, G13, G14, G24 and G29 were the most stable across the sites for TRY and therefore the most widely adapted for this trait, while G68, G60 and G58 were specifically adapted to Kachwekano and Serere. Very low severity levels of Sweetpotato virus disease (SPVD) were recorded with a mean score of 1.9 across sites with Namulonge having the highest mean score of 2.3. Genotypes G14, G16, G24, G29 and G49 were the most stable across the sites for low Alternaria blight severity and can therefore, be recommended for cultivation in both low and high disease pressure areas. In the participatory selection, before harvest and at harvest, Spearman’s rank correlation (r = 0.44) of the scientists and farm ers’ mean ranking of the genotypes at each site was positive and significant. Thus scientists were capable of selecting for farmer preferred traits. In addition, the study identified and selected five superior genotypes including G13, G14, G24, G49 and G69 for further evaluation.


INTRODUCTION
Plant breeders desire stable genotypes with good performance under all conditions within the target production regions. Stable genotypes with high yield potential can only be identified by testing them in a series of environments (Martin et al., 1988) and it is always important to test them in environments which reveal their maximum genetic potential in terms of the traits under consideration (Frey, 1964). The major objective of any crop improvement programme is the development of cultivars with high yield potential and other desirable traits, and the ability to withstand seasonal fluctuations over a wide range of environments (Kamalam et al., 1978;Lebot, 2009). Most of the important sweetpotato traits, including yield, are strongly affected by environmental conditions associated with sites and years (Ngeve, 1993;Niringiye et al., 2014). In most cases, high yielding genotypes are not yield stable and those that are yield stable are low yielding (Ngeve, 1993;Manrique and Hermann, 2000;Mwanga et al., 2006). However, breeding sweetpotato for high yield and wide adaptation is possible (Grüneberg et al., 2005). In sweetpotato, attention needs to be paid to testing in low-yielding marginal environments if farmers working in such environments are the main beneficiaries of the new cultivars. Hence, yield testing in early stages of a sweetpotato breeding program should use at least one favourable environment and one less favourable environment (Grüneberg et al., 2005).
In Uganda, the National Sweetpotato Program released 22 sweetpotato cultivars between 1995 and 2013 and all these releases were made after conducting on-station, on-farm and standard multi-locational yield trials focusing mainly on high yield, high dry mass and resistance to pests and diseases (Mwanga et al., 2011;Ssemakula et al., 2013;Mwanga et al., 2016). In 2017, five new varieties namely NAROSPOT 1, NAROSPOT 2, NAROSPOT 3, NAROSPOT 4, NAROSPOT 5, were released (MAAIF, 2018) to bring the total of released varieties to 27. Of all these released cultivars, only one, NASPOT 11 had been bred through a participatory plant breeding process (Mwanga et al., 2011) but efforts were made to incorporate farmer preferred attributes in the other cultivars. Despite releasing all these cultivars, farmers still demand new ones to meet their ever changing preferences and some of the cultivars for example NASPOT 2, NASPOT 5 and Sowola 6 have not been well received (Abidin et al., 2002). For this reason, many farmers have continued to cultivate their landraces which underscores the need to involve them in participatory cultivar selection so that their preferences are considered. Participatory cultivar selection, and participatory plant breeding (PPB), are considered the most appropriate strategies to develop cultivars for marginal agricultural systems (Almekinders and Elings 2001;Ceccarelli and Grando, 2006;Dawson et al., 2007). This approach allows incorporation of farmers' knowledge, identification of farmers' selection criteria and priorities. Participation of farmers can allow for exploitation of specific adaptation effects within sites and facilitate seed supply to farmers (Ceccarelli et al., 2000).
Evaluation by farmers helps scientists to design, test and recommend new technologies in light of information about farmers' requirements and needs. It facilitates close interaction among farmers, researchers and other role players in crop genetic improvement, allowing researchers to respond more closely to the needs and preferences of resource-poor farmers and their market clients (Sperling et al., 2001). Farmers can be involved through introduction, evaluation and selection of materials grown on the research station and also collaborate by growing and selecting breeding materials in their own fields (Ceccarelli et al., 2000;Mcharo et al., 2001;Keith et al., 2004). The cultivars obtained from this process are developed more rapidly, are more diverse and have higher adoption rates (Witcombe et al., 2003). Consideration of farmers' concerns and conditions leads to varieties that become widely adapted and more productive hence leading to sustainable agricultural systems (Odendo et al., 2002). Among the preferred attributes that farmers select for during PPB for sweetpotato are good yield, sweet taste and dry texture (Laurie and Magoro, 2008;Marti, 2003;Kwach et al., 2010;Sseruwu et al., 2015).
Twenty one promising F1 genotypes previously selected from early breeding trials were used in this study which was carried out to identify superior genotypes as possible candidates for advanced yield and on-farm trials. The main objective of this study was therefore to evaluate and identify genotypes with wide and specific stable performance over three sites for Alternaria leaf petiole and stem blight (commonly referred to as Alternaria blight) resistance, total storage root yield (TRY), Sweetpotato virus disease (SPVD) and other farmer preferred traits.

Genotypes and sites
Twenty one F1 genotypes and two checks Tanzania and NASPOT 1 were planted at three sites during the first rain season of 2015 (2015A).

Storage root yield
At harvest the total storage root yield (TRY) was recorded on a per plot basis then the mass per plot was converted to t ha -1 for analysis.

Participatory selection data collection
In addition to collecting disease and agronomic data, participatory selection of the F1 genotypes was also performed at two of the three sites namely; Namulonge (NaCRRI) and Kachwekano (KAZARDI). The genotypes were separately evaluated before harvest and at harvest by a group of five scientists and a group of 10 farmers (five males and five females) at each site. The groups of scientists and farmers at both sites were different. The five scientists at NaCRRI, and five scientists at KAZARDI, had a minimum qualification of a Bachelor's Degree in Agricultural Sciences and were employed by the National Agricultural Research Organisation (NARO). The selected farmers were knowledgeable about sweetpotato production and consumer preferences. At each site, the evaluation before harvest was carried out two days before harvesting the trial and it was preceded by familiarising both groups at each site with the selection procedure and criteria. Both groups used the same evaluation criteria and the traits considered were: Alternaria blight and SPVD severity; growth habit (spreading, erect); leaf morphological traits (broad, small leaves, leaf colour); and general acceptability as a new cultivar (i.e. whether each participant considered the genotype suitable to become a cultivar). A rating scale of 1-5 was used for all the traits. For diseases, a severely infected genotype was scored 1 and a symptomless genotype, 5. For leaf morphology traits, leaf colour and size were used, and for growth habit, a genotype with poor growth habit was scored 1 and excellent growth habit was scored 5.
For selection at harvest, the two groups at each site separately listed the traits that they wanted to use in the evaluation process and ranked them in order of importance. On this basis, each group developed a list of top five traits for scoring the genotypes (Table 1). For each trait, the participants individually scored each harvested plot in all three replications on a scale of 1-5 where 1= trait absent and 5 = the genotype expressed the trait at a satisfactory level. Then the mean score for each trait was separately determined for each of the two groups per site. Roots were sampled from each plot of each genotype, boiled, taste tested and then scored for the following attributes: appearance of the flesh after cooking, sweetness, dry mass (hardness), fibre content and acceptability as a new cultivar. The same rating scale of 1-5 was used as mentioned above.

AMMI analysis of the data for the three sites
The genotype x environment interaction (GEI) and associated stability of the genotypes across three sites for area under the disease progress curve (AUDPC) for Alternaria blight severity scores, SPVD severity scores and TRY were analysed using the additive main effects and multiplicative interaction (AMMI) procedure in GENSTAT version 14 (Payne et al., 2011) based on the standard AMMI model (Gauch and Zobel, 1996).
The AMMI analysis partitions the GEI sum of squares (SS) into IPCA axes. Only IPCA1 and IPCA2 were significant and the non-significant IPCA3 was considered as "statistical noise" and accounted for by the residual term. The interaction patterns of the genotypes and the environments were graphically represented in a biplot of the respective IPCA1 scores (y-axis) versus the genotype and environmental means or IPCA2 (x-axis). In the biplot, displacement in the horizontal plane reflects differences in the mean performance, while displacement in the vertical plane reflects differences in interaction effects (Zobel et al., 1988).

Analysis of participatory selection data
The scores for all traits for each genotype at each of the two sites for each group were analysed by ANOVA in GENSTAT version 14 to obtain the mean scores for each trait per genotype, evaluation group and site. Weights were assigned to each scored trait such that the trait ranked first by a group was assigned a weight of 5 and the one ranked fifth was assigned a weight of 1. For each genotype, the mean score for each trait was multiplied by the assigned weight then all five weighted scores were summed up to obtain an aggregate score for each genotype.
Aggregate weighting index used for the both the scientist and farmer groups: Where: AT1…5 = Attributes ranked 1…5; and W5…W1= assigned weight ranging from 5 to 1.
The aggregate scores of the genotypes at each site for each group were ranked to determine two separate rank orders (one per group) of the genotypes at each site. The ranks for each genotype per group were summed across the two sites (Kang, 1993) and the genotype with the lowest rank sum was the best over the two sites.

Alternaria blight
The genotypes, environments and genotype x environment interaction (GEI) mean squares (MS) were highly significant (P<0.001) for Alternaria blight AUDPC (   Alternaria blight severity but Namulonge was more stable than Serere and had several genotypes specifically adapted to it (Figure1). Kachwekano was a high disease pressure site and the least stable with high interaction with the genotypes.

Sweetpotato virus disease
The MS for the environments were highly significant (P<0.001) for SPVD and not significant (P>0.05) for the genotypes and GEI (Table 2). Very low severity levels of SPVD were recorded for these genotypes with a mean score of 1.9 across sites (Table  4). Serere recorded the highest SPVD severity with a mean score of 2.3 while Namulonge had the lowest mean severity score of 1.6. 15.3 t ha -1 , respectively across sites. The mean TRY across genotypes of 25.5 t ha -1 recorded at Namulonge was the highest of the three sites while the 12.3 t ha -1 recorded at Serere was the lowest. The most outstanding genotypes at Namulonge were G30, G69 and G16 with yields of 34.0, 31.3 and 30.6 t ha -1 , respectively. There was no consistency in the ranking of the genotypes in that highly ranked genotypes at one site ranked poorly at the other sites. In the AMMI biplot of the two significant axes IPCA1 vs IPCA2 for TRY (Figure 2), the genotypes and the three environments generally dispersed around the origin (centre) of the biplot (the sites more so than the genotypes) indicating strong interactions between the genotypes and environments in response to the abiotic or biotic factors underlying or driving the IPCA1 & 2 scores. Genotypes G13, G8, G49 and G29 were positioned close to the origin indicating minimal interaction of these genotypes with the environments. The remaining 17 genotypes and checks (Tanzania and NASPOT 1) were positioned further away from the origin and therefore had strong interactions with some of the environments.
At harvest, the Spearman's correlation between scientists and farmers' rankings at Namulonge was highly significant (P<0.01) and positive (r = 0.412) and that between scientists and farmers at Kachwekano was also highly significant (P<0.01) and positive (r = 0.440) ( Table 9). The other rank correlations were non-significant. The quality traits (mostly organoleptic) of the genotypes that were evaluated at harvest included sweetness (taste), root firmness (hardness), root fibre content, appearance and general acceptability based on taste and appearance. At Namulonge, scientists ranked G24, NASPOT 1, Tanzania, G38 and G28 as the best and at Kachwekano G68, NASPOT1, G14, G60 and G29 were ranked as the best genotypes. Farmers at Namulonge ranked NASPOT 1, G28, G38, G68 and Tanzania as the best genotypes, and at Kachwekano G14, G29, G68, G60 and NASPOT 1 were ranked as the best. Genotypes NASPOT 1, G68, G24, G60 and G53 were the best ranked across the groups ( Table  10). The positive Spearman's correlation (r = 0.605) between scientists and farmers' rankings at Namulonge and rank correlation (r = 0.552) between scientists and farmers' ranking at Kachwekano were highly significant (P<0.01) (Table 11). Aggregate score based on weighted selection index (sweetness (taste), root firmness (hardness), root fibre content, appearance and general acceptability). ** = significant at P<0.01; * = significant at P <0.05

Discussion
The objectives of this study were to evaluate and identify F1genotypes with wide and specific stable performance over three sites for Alternaria blight resistance, SPVD, TRY and other farmer preferred traits. Additionally, the ranking of the genotypes by two different groups of scientists and farmers at two of the sites for selected traits were compared using Spearman's rank correlations.

Performance and stability of the genotypes
The severity of Alternaria blight was higher at Kachwekano than at the other two sites ( Table 3). The AMMI analysis revealed that the Alternaria blight was influenced more by environmental effects, than by the GEI effects and least by genotypes effects. During the season, Kachwekano did not receive as much rainfall as Namulonge but the disease was more severe at this site. This is consistent with what was reported by Sseruwu et al. (2015) where some farmers in Luwero district reported the disease to be more severe during the dry season than during the wet season. It is possible that the disease infected the crop during the first month after planting when there was sufficient moisture and the symptoms became visible later on when the crop was stressed due to insufficient moisture. Mwanga et al. (2007b) described Serere as a low pressure area for Alternaria blight. However, this study has provided an indication that the effect of Alternaria blight under natural infestation in Serere is increasing since the severity was not significantly less than that of Namulonge. However, this can only be confirmed after obtaining data for several seasons. Should the trend be confirmed, farmers at Serere will require Alternaria blight resistant genotypes and this would necessitate evaluation of all the popular sweetpotato genotypes in the area for resistance to the disease in order to identify those with good levels of resistance.
Resistance of the genotypes across sites to Alternaria blight was not consistent, with some genotypes having lower Alternaria blight AUDPC values at one site and higher values at another site. However, some genotypes maintained lower Alternaria blight AUDPC values across sites and if these genotypes can maintain this consistency in subsequent evaluations (particularly over more seasons) and also meet the required performance levels for other important traits then they will be recommended to the farmers for cultivation in all the tested and similar sites/environments. Those that have consistent, good performance at particular sites will be recommended for those sites. Genotypes G49, G67, G69, 59 and G24 were the best genotypes at Namulonge. Genotypes G13 and G65 performed better than the check, Tanzania at Kachwekano. Similarly, G14, G28 and G53 were also better than the Tanzania at Serere. Thus these genotypes are well adapted to those sites. Genotypes G49, G13, G67 and G14 recorded lower mean AUDPC values across sites and should be further evaluated for even wider adaptation.
The AMMI biplot provided an indication of the stability of the different genotypes for Alternaria blight. In this context, stability means a genotype that maintains the same level of disease severity, either high or low across sites. Genotypes that are stable for low Alternaria blight severity and good yields are desired for this programme. Stability of genotypes G14, G16, G24, G29, G49, G59 and G69 for low Alternaria blight severity implies that these genotypes can be grown in all of the test sites and maintain low disease severities. They can also be used as sources of resistance in breeding for Alternaria blight resistance. Genotypes NASPOT 1 and G8 expressed stable but above average AUDPC values. This implies that these genotypes can only be grown in areas of low Alternaria blight pressure or may need fungicide protection when grown in high disease pressure areas. Kachwekano is a high Alternaria blight pressure site; therefore, it is ideal for evaluating the resistance of germplasm to the disease while Namulonge and Serere are ideal for germplasm multiplication.
The high significance (P<0.001) of the effects of genotypes, environments and GEI for TRY implied that all these factors are important in determining the expression of this trait. However, environmental effects were more important than genotypes and GEI effects. Namulonge was the highest yielding site with a mean TRY of 25.5 t ha -1 and Serere was the lowest yielding site with a mean of 12.3 t ha -1 . The cause of such high variation in yield was in all likelihood the amount of rainfall received during the season. At Kachwekano and Serere, the crops received reasonable amounts of rainfall only during the first month after planting but very little in the subsequent months unlike Namulonge which had good rainfall for the first three months after planting (Appendix 1).The yield recorded at Namulonge which ranged between 12.2 (G58) and 34.0 t ha -1 (G30) is an indication of the high yield potential of this set of genotypes. However, the full genotype yield potential was not realised at the other two sites possibly due to moisture stress. However, the best genotypes for TRY across the three sites were G67 (21.6 t ha -1 ) and G24 (21.4 t ha -1 ).
The AMMI biplot provided an indication of the stability of the genotypes for TRY. Genotypes G53, G67, G14, G13, G29 and G24 were very close to the horizontal line and therefore the most stable. These genotypes are widely adapted and can be grown at any of the three test sites and should give good yields. Provided the necessary agronomic requirements are available, they can be recommended to farmers at all three sites. Genotypes G68, G60 and G58 were low yielding and specifically adapted to the low yield potential sites of Kachwekano and Serere hence may not perform well outside these sites.

Participatory Clonal selection
At the two selection stages, before harvest and at harvest, the scientists and farmers at the two sites ranked some of the genotypes similarly and in other instances differently. The significant (P<0.05), positive Spearman's rank correlation between scientists and farmers at each site (r = 0.342 for Namulonge, r = 0.283 for Kachwekano) before harvest indicated that the two groups ranked many genotypes in the same way before harvest. Therefore, at each site the scientists in this study were capable of selecting genotypes that had farmer preferred traits. The groups of scientists at the two sites selected different genotypes and so did the farmers. Since they based their selection on crop vigour, the cause of the difference in genotype selection was likely to be the differences in the performances of the genotypes across the sites due to the poor weather conditions at Kachwekano, which did not receive enough rainfall during the trial (Appendix 1).
Ranking of genotypes before harvest may be influenced by the amount of aboveground foliage produced particularly the leaves which at that stage are the economic yield component of the crop. On the other hand, farmers may prefer genotypes with more upright growth habit than prostrate growth habit with spreading vines. However, the aboveground characteristics of any genotype may not always be a good indicator of belowground performance.
At harvest, most of the attributes identified by the scientists and farmers were similar but the ranking of the genotypes differed. Just as in any formal selection system where yield is considered as a major criterion (Joshi et al., 1997), yield was ranked the number one trait by the groups. Scientists and farmers at both sites preferred high-yielding genotypes with big storage roots which implied the converse that high yielding genotypes that produce small storage roots are not preferred. This is certainly the case where the farmers are market oriented. The buyers select and pay only for the large storage roots and leave the small ones or take them at no cost. Abidin et al. (2002) in north-eastern Uganda, also reported that farmers prefer genotypes that produce numerous, large storage roots, which tend to also have large overall yields. Similarly, Ndirigwe et al. (2005) in Rwanda reported that farmers rejected one cultivar which was high yielding because it had small size storage roots. In addition to storage root size, shape of the storage root was identified as an important trait by all groups except farmers at Namulonge. Grooved roots are not preferred because they are difficult to peel and will not be bought in the market unless they are the only ones available. Skin colour was important to all groups except the Kachwekano farmers. Red skin colour was mostly preferred by the groups and this is also the market preference. That skin colour was not identified as an important trait by the Kachwekano farmers, probably because most of them produce for home consumption. In previous studies by Abidin et al. (2002) in north-eastern Uganda, the preferred skin colour was white/tan and flesh colour was yellow. Therefore, the importance of skin colour depends on the region where the evaluation is carried out. According to Ndirigwe et al. (2005), in Rwanda the reddish skin was also preferred by both the farm household and the market.
At harvest, the significant (P<0.01), positive Spearman's rank correlation coefficient between scientists and farmers at Namulonge (r = 0.412) and between scientists and farmers at Kachwekano (r = 0.440), indicated that it is possible for the evaluation to be carried out by scientists only and successfully identify farmer preferred traits. This would obviously enable considerable savings for research budgets and will facilitate quicker selection processes.
For the cooking qualities of the genotypes, the farmers also represented the consumer since they also consume sweetpotato and they interact frequently with other consumers. The highly significant (P<0.01), positive Spearman's rank correlations between the rankings of scientists and farmers at Namulonge (r = 0.605) and scientists and farmers at Kachwekano (r = 0.552) for cooking quality traits indicates that the scientists at each site are capable of selecting for the same cooking qualities preferred by farmers. Therefore, it is not necessary to use site specific groups in the selection process. NASPOT 1, which is a popular cultivar, emerged as the best genotype across the groups for cooking quality traits with G68 and G24 ranking second and third respectively. Since NASPOT 1, which is already a very popular cultivar in Uganda, was ranked as the best by the groups this provides some validation of the outcome of the current study.

IV. CONCLUSION
Some of the F1 genotypes selected from the crosses conducted in this breeding programme are highly adaptable and have farmer preferred attributes. Genotypes that exhibited stability for resistance to Alternaria blight as well as stability for high storage root yield were G14, G16, G24, G49 and G59. These genotypes can be recommended to farmers on a trial basis at the three test sites and other associated sites. However, a full investigation of the stability of these genotypes across a representative range of environments will have to be performed. Stability for the scientist and farmer evaluated traits will be the basis upon which any genotype will be advanced.
The good correlations between scientist and farmer rankings of genotypes at each of the two sites in this study demonstrated that the identification of selection criteria and application thereof by scientists and farmers was not that different. The practical implication of this study is that selection within sites can be generally carried out by experienced scientists who have a good understanding of the production requirements of sweetpotato and consumer preferences. Importantly, the selection has to be conducted by site specific sets of scientists.
Overall, genotype G49 was ranked well both for stability by GEI analysis and for scientist and farmer preferred traits by the participatory selection process. In the participatory process it was ranked tenth before harvest and third at harvest. It is an above average yielder with good yield stability, and is stable for Alternaria blight with below average Alternaria blight AUDPC value. This genotype will be recommended for cultivation by selected farmers on a trial basis, as further evaluations are done at more sites.