Forest carbon exchange contributes significantly to the global carbon balance and is therefore being monitored around the world, most notably using eddy covariance technology. In order to extrapolate from these measurements, we need to understand why carbon balance (or net ecosystem production, NEP) differs among forests. Here, we use a detailed model of forest carbon exchange applied to three coniferous European forests with differing NEP to pinpoint reasons for the differences among these sites. The model was parameterised using extensive ecophysiological data obtained at each site. These data gave evidence of major differences among sites in climate, leaf physiology, respiring biomass, leaf area index, and soil and biomass respiration rates. The model was compared with eddy covariance data and found to satisfactorily simulate carbon exchange by each forest. Simulations were then run which interchanged canopy structure, physiology and meteorology among sites, allowing us to quantify the contribution of each factor to the inter-site differences in gross primary productivity (GPP), ecosystem respiration (RE) and NEP. The most important factor was the difference in respiration rates, particularly soil respiration rates, among sites. Climate was also very important, with differences in incident photosynthetically active radiation (PAR) affecting GPP and differences in temperature affecting both GPP and RE. Effects of leaf area index, respiring biomass and leaf physiology on NEP were secondary, but still substantial. The work provides detailed quantitative evidence of the major factors causing differences in NEP among coniferous forests.