Evaluation of spatial models in the estimation of genetic parameters for incidence of frosty pod rot and production in Theobroma cacao full-sib family trials

In this study, we performed spatial analyses to estimate genetic parameters for the incidence of frosty pod rot and yield in two progeny trials of the breeding program of the Instituto Nacional de Investigaciones Forestales, Agrícolas y Pecuarias (INIFAP) in Mexico. We identified spatial autocorrelation through graphical analyses of distribution, isotropic variograms, and Moran’s index. We used three spatial models for each trait: B-Splines, First-order Autoregressive (AR), and Bayesian Intrinsic Conditional Autoregressive (ICAR). We found that the data correlated over a maximum distance ranging from 6 to 10 m. The intensity of spatial autocorrelation, according to Moran’s index, was 0.11 to 0.18 (p < 0.001, Z > 2.5). The Bayesian hierarchical and first-order autoregressive methods improved the model fit compared to the Spline approach. Heritability estimates were h² = 0.11 ± 0.08 to 19 ± 0.10 for bean dry weight per tree, 0.11 ± 0.08 for the total number of pods per tree, and 0.36 ± 0.15 to 0.42 ± 0.16 for frosty pod rot disease. Correlations between models averaged r = 0.99, p < 0.001, with an average match ranging from 0.77 to 0.96 in the ranking of individuals under a selection pressure of 5%. These models contributed to understanding spatial patterns of disease dynamics and cacao production. Important traits for the INIFAP’s breeding program and other programs facing the same challenges were considered in this study, aiming to improve the efficiency of those programs. Incorporation of these methods into breeding programs may allow for accurate estimation of the genetic parameters underlying the quantitative genetics of cacao trees, at the same time saving time and resources.