Evaluating the accuracy of smallholder farmers’ sire identification for introducing genetic evaluation in community-based sheep breeding programs

A major challenge in improving genetic merits of smallholder farmers’ flocks in developing livestock systems is the absence of pedigree and performance recording for reliable selection decisions. In this study, we evaluated the reliability of pedigree recording in a community-based Menz sheep village breeding program in Ethiopia, with the purpose of introducing genetic evaluation in the selection program. A total of 3577 six-month weight (SW) and 3876 weaning weight (WW) records collected from 2009 to 2017 were analysed fitting three mixed effects animal models. RP model: considering the sires recorded by farmers as certain, HM model: uncertain paternity with candidate sires in a breeding season assigned posterior paternity probabilities according to phenotypic information, and model AR: uncertain paternity with sires assigned equal priori probabilities and using the average numerator relationship matrix to estimate the genetic effects. The model selection criteria Deviance Information Criteria and Conditional Predictive Ordinate for the three models were comparable, indicating all the three models fitted the data equally. Heritability (Monte Carlo Error in parenthesis) of SW estimated by HM, AR, and RP models were 0.306 (0.00067), 0.374 (0.00061), and 0.418 (0.00048), respectively. EBV and ranks of rams estimated by RP were highly correlated (r = 0.82–0.94) with those of HM and AR models. Up to 82% of the top 30% rams ranked by RP were also ranked by HM and AR models, but accuracy of EBVs was higher for HM model than for AR and RP models. The accuracy of parameter estimates from RP and their correlations with the reference models (HM and AR) is a compelling evidence towards the reliability of farmers pedigree records. It can be fairly recommended to introduce genetic evaluation in community-based breeding programs, estimating EBVs using farmers’ pedigree records directly or with complementary evaluations using methods for uncertain paternity situations. If farmers’ records are to be used directly, expert support in data collection is required to improve the accuracy of EBV above the estimated 70% in this study.