The abscission process is strongly involved in a series of physiological events; an optimal execution of this process is of major importance for species survival. Environmental variation impacts species development and abscission process with varying effects over developmental stages. The identification of the environmental factors as well as the time periods at which they modulate the abscission process is crucial to deal with climate changes. Considering environmental variables as time series, i.e. groups of correlated variables, poses a statistical challenge in selecting relevant groups and temporally correlated variables within them. In this study, we address these objectives by introducing and discussing four Bayesian group fused priors through a general parametrization. In particular, we highlighted that the horseshoe prior on differences with a unique global parameter over groups combined with a normal half-Cauchy distribution on coefficients outperformed the extension of the usual fused priors, which consists of assuming Laplace distributions on coefficients and their differences. The fruit abscission of oil palm trees motivated this development. This application, based on an impressive experimental design in Benin Republic, illustrated performances of the proposed prior to select environmental variables as well as successive past environmental variations involved in the timing of bunch harvesting.