A holistic approach based on field surveys was adopted to study the effects of coffee tree characteristics, agricultural practices, including shade, and physical environment on coffee leaf rust (CLR) and the American leaf spot disease (ALSD) in Central America. CLR was studied trough a three year survey in 73 plots in Honduras, and ALSD through a two year survey in 57 plots in Costa Rica. Two multivariate analyses were used to explain epidemic intensities: segmentation for CLR, and partial least-squares regression via spline functions for ALSD. Both methods made possible the selection and prioritization of disease predictors. For CLR, the most important predictor was berry yield. The predisposition of coffee trees to CLR attacks increased when fruiting nodes exceeded 230 per coffee tree. Shade was another important predictor. Our results suggested that shade had negative effects on CLR by keeping yield at low level, as most of the plots with dense shade had very low yield, but favored disease development as soon as the yield potential reached the 230 fruiting nodes threshold, probably by keeping microclimatic conditions propitious to CLR germination in the plantation. Fertilization, altitude and soil pH were also included in the segmentation tree. They were negatively linked to CLR. For ALSD, the most influential variables were related to plot topography. The disease developed more in areas between 1100 m and 1550 m, where temperatures were cool and propitious to infection process. Slopes were conducive to its development with eastern facing slopes less affected than others, probably due to their higher exposure to sunlight. Then came distance between rows, shade percentage, coffee tree height, shade type, and pruning system, due to their effects on coffee tree shade and possibly on the humidity conditions within the plot. The most severe attacks were found in plots with high shade conditions. Forest trees and fruit trees were particularly propitious to disease deve