<p>The forest ecosystems are already responding to increased CO<sub>2</sub> concentrations and changing environmental conditions. These ongoing developments affect how societies can utilise and benefit from the woodland areas in the future, be it e.g. climate change mitigation as carbon sinks, lumber for wood industry or preserved for nature tourism and recreational activities. We assess the effect and the relative magnitude of different uncertainty sources in ecosystem model simulations from the year 1980 to 2100 for two Finnish boreal forest sites. The models used in this study are the land ecosystem model JSBACH and the forest growth model PREBAS. The considered uncertainty sources for both models are model parameters, four prescribed climates and two RCP (Representative Concentration Pathway) scenarios. PREBAS simulations also include an additional RCP scenario and two forest management actions. We assess the effect of these sources at four different stages of the simulations on several ecosystem indicators of climate change, e.g. gross primary production (GPP), ecosystem respiration, soil moisture, recurrence of drought, length of the vegetation active period (VAP), length of the snow melting period and the stand volume. The climate model uncertainty remains roughly the same throughout the simulations and is overtaken by the RCP scenario impact halfway through the experiment. The management actions are the most dominant uncertainty factors for Hyytiälä and as important as RCP scenarios at the end of the simulations, but contribute only half as much for Sodankylä. The parameter uncertainty is the most elusive to estimate due to non-linear and adverse effects on the simulated ecosystem indicators.</p>