As a consequence of past plant-virus interaction studies having been almost exclusively focused on cultivated areas, our knowledge of plant virus diversity in natural environments is still very limited. Without an understanding of the spatial and temporal patterns of natural virus diversity we will never be able to achieve a mechanistic understanding of either plant virus ecological dynamics, or the factors driving long-term virus evolution and emergence. Consequently, it is still difficult to quantify the impacts of human activities (e.g. agricultural intensification, the dissemination of exotic plant and pathogen species) on host-pathogen interactions. We have developed a new gee-metagenomics approach for producing quantitative information on virus diversity, prevalence and spatial distributions and have applied it to the study of two pilot ecosystems located in South Africa and France. This approach yields geographically tagged cDNA from virus species, and further allows viral sequences to be linked to specific host plants at a specific location and time. The objectives of this study were to assess 1) whether natural, unmanaged areas act as reservoirs of plant virus biodiversity, 2) whether the spatial distributions of different plant viruses at the agroecosystem scale display distinctive patterns and 3) whether ecological parameters could account for observed patterns of virus distribution. This new approach allowed us to discover 94 new plant viral species and to reveal distinctive virus spatial distribution patterns. Whereas the prevalence of plant viruses associated with cultivated areas in both countries was significantly greater than that associated with noncultivated areas, exotic plants within natural fynbos areas of South Africa also displayed a significantly higher prevalence of viruses than indigenous plants. These results emphasize both the direct and indirect impacts that human activity has on plant virus dynamics at the agro-ecosystem scale.