Evaluating breeding for broad versus narrow adaptation for cassava in Nigeria using stochastic simulation

Abstract

AbstractThe cassava (Manihot esculenta Crantz) breeding program at the International Institute of Tropical Agriculture (IITA) has adopted genomic selection to accelerate genetic gain. The program continues to develop varieties broadly adapted across Nigeria's diverse agroclimatic zones. However, for this purpose, genotype‐by‐environment interaction (GEI) presents a challenge. To decide whether broad adaptation breeding is a good strategy, we evaluated broad versus narrow adaptation strategies using stochastic simulation, assessing genetic gain, genetic variance, heritability, and selection accuracy at 0 versus realistic levels of GEI variance. To parameterize the models, we analyzed historical data from four phenotypic evaluation stages of the IITA breeding program to estimate genetic and error variances, and genetic correlations across environments. Based on these observed parameters, the genomic‐enabled breeding programs exhibited higher genetic gain than the conventional program for both GEI variances. At realistic GEI variance, the narrow adaptation program showed higher genetic gain than the broad adaptation program. Across all programs, the genetic variance declined over time, though the genomic‐enabled programs showed higher initial variance due to the selection of parents at earlier stages. At realistic GEI variance, an increase in genetic variance was observed in the narrow adaptation program due to its conversion of GEI between mega‐environments into main genetic variance within mega‐environments. This higher genetic variance led to higher heritabilities and selection accuracies. This study highlights the potential of genomic selection in accelerating genetic gain and suggests that dividing the IITA cassava breeding program to target more than one mega‐environment should be considered.