The heritability accuracy in forest tree breeding programs relying on random mating

In forest tree breeding, there is a growing trend towards using pedigree reconstruction after offspring are produced through natural random mating (open-pollination), as an alternative to traditional controlled crosses and full-sib genetic trials. Given that most forest tree species are predominantly outcrossing organisms, the accuracy of narrow-sense heritability (h2) under natural mating conditions has not been thoroughly examined, particularly with regard to sample size. Our simulation study focuses on the genetic parameters specific to Norway spruce (Picea abies L.) in Norway. We used the stochastic model MoBPS to simulate a founder population with 100,000 SNP markers distributed across 12 haploid chromosomes, representative of many conifer species. Parental trees were selected from this population, followed by random mating and offspring evaluation. We focused on estimating the accuracy of h2, with particular attention to its precision and bias. Our results suggest that the population sample sizes currently used in forest tree breeding are generally adequate for achieving precision. However, we identified two primary sources of bias: one due to dominance effects and the other from phenotypic parental selection. We discuss potential strategies to mitigate these biases in breeding programs.