Improving the genetic evaluation for longevity in the Netherlands

Abstract

A new model for the genetic evaluation for longevity was developed in the Netherlands based on a random regression animal model (RRM). The existing system for genetic evaluation was based on a proportional hazard model (PHM). Changes adopted with the RRM were 1) fitting multiple genetic effects across the life of a cow, 2) animal model vs. sire-mgs model, 3) adjusting for milk production at herd level, and 4) fitting fixed effects differently. The aim here was to evaluate and validate the new RRM and compare the EBV with the current PHM. For the new and existing model 11 evaluation runs were performed from 2007 up to 2017, where one year of data was added in every new run. Stability of breeding values was analysed as the difference with EBV-2017 and the correlation of the first EBV of a bull with later EBV. The trait analysed was survival per month, fitted with a fifth-order Legendre polynomial until 72 months after first calving. EBV equated to 72 months were calculated for all animals. EBV were overestimated mainly in first EBV-runs, due to incomplete daughter information. Adjusting for within-herd production level reduced this bias. Based on the correlation between first and later EBV, the ranking of bulls was shown to be more stable for RRM than for PHM. RRM with adjustment for milk yield is the preferred model for longevity, as it resulted in more stable ranking of bulls with smallest overestimation of EBV based on incomplete daughter information.