The Weather Research and Forecasting meteorological (WRF) model has been coupled to the Soil–Plant–Atmosphere (SPA) terrestrial ecosystem model, to produce WRF-SPA. SPA generates realistic land–atmosphere exchanges through fully coupled hydrological, carbon and energy cycles. The addition of a~land surface model (SPA) capable of modelling biospheric CO<sub>2</sub> exchange allows WRF-SPA to be used for investigating the feedbacks between biosphere carbon balance, meteorology, and land use and land cover change. We have extensively validated WRF-SPA using multi-annual observations of air temperature, turbulent fluxes, net radiation and net ecosystem exchange of CO<sub>2</sub> at three sites, representing the dominant vegetation types in Scotland (forest, managed grassland and arable agriculture). For example air temperature is well simulated across all sites (forest <i>R</i><sup>2</sup> = 0.92, RMSE = 1.7 °C, bias = 0.88 °C; managed grassland <i>R</i><sup>2</sup> = 0.73, RMSE = 2.7 °C, bias = −0.30 °C; arable agriculture <i>R</i><sup>2</sup> = 0.82, RMSE = 2.2 °C, bias = 0.46 °C; RMSE, root mean square error). WRF-SPA generates more realistic seasonal behaviour at the site level compared to an unmodified version of WRF, such as improved simulation of seasonal transitions in latent heat flux in arable systems. WRF-SPA also generates realistic seasonal CO<sub>2</sub> exchanges across all sites. WRF-SPA is also able to realistically model atmospheric profiles of CO<sub>2</sub> over Scotland, spanning a 3 yr period (2004–2006), capturing both profile structure, indicating realistic transport, and magnitude (model–data residual <±4 ppm) indicating appropriate source sink distribution and CO<sub>2</sub> exchange. WRF-SPA makes use of CO<sub>2</sub> tracer pools and can therefore identify and quantify land surface contributions to the modelled atmospheric CO<sub>2</sub> signal at a specified location.