Nineteen peach blocks infected by the aggressive Plum pox virus strain M were monitored visually during 7 to 10 years and symptomatic trees were removed every year. Annual disease incidence was low (2 to 6%) in all orchard blocks but new symptomatic trees were continuously detected, even after 7 to 10 years of uninterrupted roguing. An exploratory approach using survival modeling was developed to evaluate to what extent, tree location within orchards, orchard characteristics and disease status within the vicinity of the orchards influenced the risk for a tree to become infected through time. Twelve variables were selected from survey data and from databases created using a geographic information system. The extended Cox model fitted to our data showed a significant effect of four of the variables tested on the risk for a tree to become infected through time: The area of the orchard block, the density of planting, the distance of a tree from the edge of the orchard block sharing a boundary with another infected orchard as well as the distance to the nearest previously detected symptomatic tree within the block. These results suggest that new PPV-M infections within orchards submitted to roguing resulted from exogenous sources of inoculum, disease development of latent infected trees as well as infected trees overlooked within the orchards during visual surveys. A revision of the control measures to more effectively remove potential sources of inoculum will be discussed in the context of the French agro-ecosystem.