Joining smallholder farmers' traditional knowledge with metric traits to select better varieties of Ethiopian wheat

Abstract

Smallholder farming communities face highly variable climatic conditions that threaten locally adapted, low-input agriculture. The benefits of modern crop breeding may fail to reach their fields when broadly adapted genetic materials do not address local requirements. To date, participatory methods only scratched the surface of the exploitability of farmers' traditional knowledge in breeding. In this study, 30 smallholder farmers in each of two locations in Ethiopia provided quantitative evaluations of earliness, spike morphology, tillering capacity and overall quality on 400 wheat genotypes, mostly traditional varieties, yielding altogether 192,000 data points. Metric measurements of ten agronomic traits were simultaneously collected, allowing to systematically break down farmers' preferences on quantitative phenotypes. Results showed that the relative importance of wheat traits differed by gender and location. Farmer traits were variously contributed by metric traits, and could only partially be explained by them. Eventually, farmer trait values were used to produce a ranking of the 400 wheat varieties identifying the trait combinations most desired by farmers. The study scale and methods lead to a better understanding of the quantitative basis of Ethiopian smallholder farmer preference in wheat, broadening the discussion for the future of local, sustainable breeding efforts accommodating farmers' knowledge.

Figure 1

Farmer score and phenotype correlations between genders and between locations. (a) Score consistency between genders in the two locations. On the y axis, Spearman correlation coefficients (all significant). Red bars represent correlation values between men and women scores in Hagreselam, blue bars represent those correlations in Geregera. Gender correlations are higher in agronomic traits, earliness and tillering, and lower in quality traits, spike and overall. (b) PE scores consistency between the two locations (“h” for Hagreselam and “g” for Geregera). The value of Spearman’s correlations is shown by the width of pie slices colored according to the bar on the far right. The combinations without pie charts correspond to non-significant correlations. (c) Correlation plot of metric measures of agronomic traits in the two locations. Correlation values are shown as in panel (b), phenotype codes as in Materials and Methods. NET and BY were the agronomic traits most varied across locations.

Figure 2

Climatic features of the experimental locations. (a) Yearly measures of minimum temperatures, maximum temperatures (y axis on the left side) and rainfall (y axis on the right side) in the two locations. The amount of rain and the duration of the rain season is lower in Hagreselam. Red color for Hagreselam, blue color for Geregera. Line shapes according to legend. (b) Bioclim values in the two locations. Among temperature indexes, seasonality separates the two locations the most. Although rainfall seasonality follows the same patterns in the two locations, Geregera is wetter across the year. Colors as in panel (a).

Figure 3

CCA reporting linear combinations of phenotypes and farmers’ evaluations in the two locations. (a) CCA for Hagreselam. On the left, phenotype correlations, farmer score correlations and cross-correlations according to the color scale below (blue to red, increasing correlation −1 to 1). On the right of the panel is a biplot of the original variables in the first two dimensions of the CCA space. The agronomic measures are shown in red, farmer scores are shown in blue. (b) CCA for Geregera, reported as in panel (a). The combination of metric traits and farmer scores is different across locations.

Figure 4

Ranking order of PE varieties. (a) Comparison between ranking score distributions in the two locations. Blue boxplot (Ger), Geregera; red boxplot (Hs), Hagreselam. Hollow circles outside upper and lower quartiles are outliers. Yellow diamonds are FGD varieties. (b) Top 50 PE wheat varieties in relation to phenotypic values. The panel depicts the first two PC axes extracted from metric measures of phenotypes averaged across the two locations. The top 50 varieties in Hagreselam and Geregera are represented by hollow circles in progressively darker red shades and blue shades, respectively. Twenty varieties were simultaneously selected in Hagreselam and in Geregera. Gray dots represent PE genotypes ranked 51 to 400. Preferred varieties in both locations are grouped in the top-left portion of the PC space.