Abstract. We introduce a global-to-regional nesting scheme for atmospheric transport models used in simulating concentrations of green house gases from globally distributed surface fluxes. The coupled system of the regional Stochastic Time-Inverted Lagrangian Transport (STILT) model and the global atmospheric transport model (TM3) is designed to resolve atmospheric trace gas concentrations at high temporal and spatial resolutions in a specified domain e.g. for regional inverse applications. The nesting technique used for the coupling is based on a decomposition of the atmospheric concentration signal into a far-field and a near-field contribution enabling the usage of different model types for global (Eulerian) and regional (Lagrangian) scales. For illustrating the performance of the coupled TM3-STILT system we compare simulated mixing ratios of carbon dioxide with available observations at 10 sites in Europe. For all chosen sites the TM3-STILT provided higher correlations between the modelled and the measured time series than the TM3 global model. Autocorrelation analysis demonstrated that the TM3-STILT model captured temporal variability of measured tracer concentrations better than TM3 at most sites.