Abstract. Surface heterogeneity can be challenging to fully encompass by modelling studies of CO2 surface exchanges, especially when it comes to land-sea boarders. The relative importance of the marine and the terrestrial surface fluxes on the atmospheric CO2 concentration were examined by developing a mesoscale modelling framework capable of simulating surface exchanges at a high spatiotemporal resolution. This study exploits the complexity of the Danish landscape and the many land-sea boarders found along the nation's 7300 km of coastline. An atmospheric transport model, DEHM, with a horizontal spatial resolution of 5.6 km × 5.6 km constituted the basis of the modelling framework. A mechanistic biosphere model, SPA, was coupled to DEHM in order to simulate terrestrial surface exchanges applying a tiling approach with the seven most dominant land-use classes in Denmark to account for sub-grid heterogeneity. Detailed surface fields of pCO2 were used to simulate the air-sea CO2 exchange for the study region. Monthly mean diurnal cycles of surface water pCO2 were imposed onto these, in order to include short-term variability in surface water pCO2. The Danish biospheric fluxes simulated by the SPA-DEHM model system experienced an east-west gradient corresponding to the distribution of the land-use classes and their biological activity. The relative importance of the seven land-use classes varied throughout the year according to their individual growth patterns. A major contribution to the monthly net ecosystem exchange (NEE) through all seasons came from grasslands, while the influence from croplands increased from March to July. Grasslands had, on an annual basis, the largest impact on the biospheric net uptake with −1423 GgC yr−1. The total Danish biospheric uptake for 2011 was −6302 GgC yr−1. Relating the annual natural biospheric surface fluxes to the CO2 emitted by fossil fuel combustions and industrial processes by Denmark, the Danish terrestrial uptake corresponded to 52 % of these, while the Danish annual marine uptake was negligible in comparison, although hiding larger seasonal variations. During 2013–2014, the simulated atmospheric CO2 concentrations compared well with measurements made at the Risø tall tower located on the shore of Roskilde Fjord (R = 0.88 and RMSE = 4.87 ppm). The origin of the simulated CO2 concentrations at Risø varied between seasons with biospheric fluxes and fossil fuel emissions having the largest impact on the variations. Impact from Roskilde fjord was difficult to detect in the simulated CO2 concentrations. These difficulties in simulating the local impact from the Roskilde Fjord might arise from (i) the fjord not being adequately resolved in the constructed model system, (ii) the lack of a realistic representation of the surface water pCO2 dynamics, or (iii) that the fjord is not in the simulated footprint and only had a modest impact on the simulated atmospheric CO2 at the Risø tall tower.