Soils are the main recipients of microplastic pollution that is subsequently transferred to aquatic ecosystems. Land-based sources dominate microplastics inputs to the ocean but knowledge gaps about their accumulation in soil, run-off and in-stream transport has hindered assessments of regional/global distributions and marine ecosystem exposure. These gaps must be filled to prioritize environmental protection actions. Here we present the first successful predictions of microplastic land to water transport assessed against empirical data at catchment scale. Predictions were generated by a mechanistic model that accounts for climate, hydrology, soil/microplastic erosion and sediment/microplastic transport. The model assimilates the latest knowledge on microplastic behavior in soil and stream water as a function of particle size, shape and density. After calibrating to monitoring data, we predict microplastic concentrations and temporal trends in water and sediments with typical error factors below 2 and 10, respectively. Next, we used the model to draw the first robust landscape-scale assessment of microplastic budgets and flows. Soils and sediments typically retain between 10-40% of total microplastic inputs over a multi annual period with strong fluctuations linked to climate variability. Despite uncertainties that remain for some parameters, we prove that mechanistic frames assimilating the state of the art on microplastic environmental behavior can currently yield credible predictions of their flows and budget. These findings are key to draw robust assessments of present and future microplastic distributions across terrestrial and aquatic environments and solve current hindrances to both risk assessment and the definition of pollution reduction measures.