Climatic change and increasing atmospheric CO 2 concentration are expected to have significant effects on grassland ecosystems, but the magnitude of the responses may vary strongly across complex landscapes. A combination of the mechanistic pasture simulation model PaSim, statistical interpolation of climate parameters, and stochastic weather generation was used to examine the range and spatial distribution of the long-term responses of grazed grassland ecosystems to step changes in temperature and CO 2 across the 250 km 2 alpine Davos/Dischma region in Switzerland. 750 sites were considered, covering an altitude range from 1500 to 2700 m above sea level. PaSim was driven with sitespecific hourly weather data, georeferenced input data for topography, and soil and vegetation characteristics. A seasonally uniform temperature increase by 2°C raised the mean evapotranspiration (ET ) from about 200 to 300 mm yr -1 , and net primary production (NPP) from about 0.2 to 0.3 kg C m -2 yr -1 . Doubling CO 2 to 700 ppm partially offset the increase in ET, but caused an additional increase in NPP. The effects of the different scenarios on the simulated mean labile soil organic carbon content (C fast ) were small, and on the order of +1% due to increased temperature and + 5% due to increased CO 2 . Largest absolute changes in ET were obtained for sites with ample precipitation, whereas relative changes correlated best with altitude. Largest absolute changes in NPP were obtained for the most productive lower sites, whereas at higher sites absolute changes became small, but relative changes were again largest. Most pronounced increases in C fast of around 10% (equivalent to about 0.3 kg C m -2 ) resulted from the combined temperature and CO 2 increase and occurred mainly at the higher elevation sites. For all parameters examined the variability of absolute system responses across sites was generally larger than the magnitude of the mean changes resulting from any scenario. Statistical analysis of the relationship between a small set of site-specific input parameters and model outputs revealed that the most important factors determining absolute system responses under all scenarios were altitude and aspect for ET, temperature for NPP, and soil texture (i.e. clay fraction) for C fast . Absolute and relative changes due to the assumed changes in CO 2 and/or temperature depended less clearly on these factors, in particular for C fast . These results show that the interaction of the effects of elevated CO 2 and increased temperature with local site conditions causes a spatially inhomogeneous distribution of grassland responses in a topographically complex landscape, but that absolute system responses for a given temperature and CO 2 regime can be explained with very high accuracy by a small number of environmental variables.