Abstract: Climate change significantly impacts crop growth and production, which may undermine the resilience of global food systems. Predicting long-term dynamics of future food production under different emissions scenarios is critical for ensuing global food security. However, large variations exist in current food production projections due to uncertainties in future climate projections, their coarse resolutions, and the missing of several important processes affecting crop yield in current crop models. Here, we predicted future changes in the yields of major crops until the end of the 21^st century under four emission scenarios (including Reference, Paris Forever, Paris 2°C and Paris 1.5°C), using the Dynamic Land Ecosystem Model (DLEM) as driven by the newly developed ensembles of high-resolution future climate projections (with a spatial resolution of 0.5°x 0.5°). The DLEM includes mechanistic representations of dynamic crop growth processes and agricultural management practices as affected by multiple environmental stresses, and its performance in simulating the yields of major crops has been validated at multiple scales from site to global. The developed ensembles of future climate projections were constructed with the combination of large ensembles of the MIT Integrated Global system Modeling (IGSM) zonal climate projections and multiple climate models participated in Coupled Model Intercomparison Project Phases 6 (CMIP6). Our results indicate that the projected food production varies largely both among different future scenarios and among different climate projections of the same scenario. Our study assesses the impacts of future climate change on global food systems from a risk-based perspective and quantifies the uncertainties, which have important implications for future agricultural climate change adaptation measures.