Résumé
Frosty pod rot (FPR) caused by Moniliophthora roreri is the primary disease affecting cacao production in the major producing countries of the Americas and is one of the major threats to cacao worldwide. The incidence of FPR on clones with different levels of resistance was investigated in four localities of Santander State, Colombia, between July 2013 and May 2015. Dynamics of diseased pods were modeled using boosted regression trees, a machine learning technique that allows regressions to be performed without prior statistical assumptions. The results suggested that FPR epidemics varied according to plot location, clone, weeks of observation, and total pods produced. Dynamics in the phenology of pods had an effect on the epidemics, and this dynamic could partially explain the difference in resistance among clones. Although not total, partial resistance of ICS 95 was confirmed. An important wilt effect was observed, particularly in the resistant clones; consequently, differences in harvested pods were not significant among clones. Pod stripping remains a good practice for the management of the disease and this practice could also have an effect on the pod dynamics and wilt phenomenon.