Quantifying the effect of landscape composition on disease dynamics remains challenging because it depends on many factors. In this study, we used a hybrid process-based/statistical modeling approach to separate the effect of the landscape composition on the epidemiology of banana leaf streak disease (BLSD) from weather and fungicide effects. We parameterized our model with a 5-year dataset, including weekly measures of BLSD on 83 plots in Martinique. After estimating the intrinsic growth parameters of the stage evolution of the disease (SED), we evaluated the dynamic effect of five fungicides. Then, we added the intra- and inter-annual effect on disease dynamics using a generalized linear model. Finally, the whole model was used to assess the annual effect of the landscape on the SED for 11 plots. We evaluated the significance of the landscape composition (proportions of landscape elements in 200-, 500-, 800-, 1,000-m-radius buffer zones) on the landscape effect evaluated with the model. The percentage of hedgerows in a 200-m-radius buffer zone was negatively correlated to the landscape effect, i.e., it acted as a constraint against BLSD spreading and development. The proportion of managed-banana-plants in a 1,000-m-radius buffer zone was negatively correlated to the landscape effect, probably due to a mass effect of fungicide treatments. Inversely, the proportions of forest and the proportion of unmanaged-banana-plants, both in 1,000-m-radius buffer zones, were positively correlated with the landscape effect. Our study provides a holistic approach of the role biotic and abiotic factors play on the dynamics of BLSD.