Estimation of heritability using spatial variability models: The case of lentil (Lens culinaris Medikus) trials in Syria

Abstract

Heritability is a key determinant of response to selection and breeding efficiency, and is often estimated using data from field trials involving simple block designs. However, spatial variability patterns that cannot be modeled using a classical block analysis framework have been found to be prevalent in field trials of lentil (Lens culinaris Medikus), an important food legume crop worldwide. Therefore, 18 different models, each describing a different spatial pattern, were assessed using lentil yield data from preliminary and advanced yield trials in Syria. More often the best models included linear trends and auto-correlated errors in addition to the classical block effect than those based on block effects alone. For instant, of the 31 preliminary yield trials on seed yield, model based on complete block and first-order autocorrelated errors along rows was found best in eigl)t trials followed by auto-correlated errors in rows and column dimensions in seven trials. Out of the 18 advanced yield trials, randomized complete block was most suitable to describe the field variability in six trials followed by complete blocks with a linear trend along rows in three trials.
These trials evaluated pure lines of twO" seed sizes, and were conducted using square lattices, in three contrasting West Asian environments. The most appropriate model for a given trial was identified and used to estimate heritability (1) using data from individual trials, and (2) using combined data from multi- environment trials, to incorporate genotype x environment interaction. Average broad-sense heritability over individual trials was found to be 0.47 for seed yield and 0.45 for biomass. Based on individual trials, average heritability estimates were similar for the two seed-size types. However, the presence of genotype x environment Interaction reduced the estimates considerably, with an overall average of 0.21 for seed yield and 0.22 for biomass. Higher estimates of heritability were obtained for small-seeded genotypes (0.24 for seed and 0.22 for biomass) than for large-seeded material (0.15 for seed yield and 0.21 for biomass). Preliminary yield trials gave higher estimates of heritability (0.24 for seed yield and 0.29 for biomass) than advanced yield trials (0.16 for seed yield as well as for biomass). This approach can therefore be used to estimate heritability from multi-environment trials with a spectrum of spatial patterns in the experimental fields.