Abstract Uncertainties in global climate models (GCMs) and emission scenarios affect assessments of the impact of global warming as well as the communication of scientific results. Here, we developed a probabilistic technique to deal with the uncertainties and to simulate the impact of global warming on rice production and water use in China, against a global mean temperature (GMT) increase scale relative to 1961–1990 values. From 20 climate scenarios output from the Intergovernmental Panel on Climate Change Data Distribution Centre, we used Monte Carlo analysis to develop the most likely climate-change scenarios for representative stations and derived the CERES-Rice model of [Alocilja, E.C., Ritchie, R.T., 1988. Rice simulation and its use in multicriteria optimization, IBSNAT Research Report Series 01] to simulate rice production under baseline and future climate scenarios. Adaptation options such as automatic application of irrigation and fertilization were considered, although cultivars were assumed constant over the baseline and future. After assessing representative stations across China, we projected changes in rice yield, growing period, evapotranspiration, and irrigation-water use for GMT changes of 1, 2, and 3 °C in a probabilistic way. Without consideration of CO2-fertilization effects, our results indicate that the growing period would shorten with 100% probability; yield would decrease with a probability of 90%, 100%, and 100% for GMT change of 1, 2, and 3 °C, respectively. The median values of yield decrease ranged from 6.1% to 18.6%, 13.5% to 31.9%, and 23.6% to 40.2% for GMT changes of 1, 2, and 3 °C, respectively. According to the median values of the projected changes, evapotranspiration and irrigation water use would decrease in most of the investigated stations. If CO2-fertilization effects were included, the rice growing period would also be reduced with 100% probability; across the stations the median values of yield changes ranged from −10.1% to 3.3%, −16.1% to 2.5%, and −19.3% to 0.18% for GMT changes of 1, 2, and 3 °C, respectively. Evapotranspiration and irrigation water use would decrease more and with higher probability in comparison with the simulations without consideration of CO2-fertilization effects. Our study presents a process-based probabilistic assessment of rice production and water use at different GMT increases, which is important for identifying which climate-change level is dangerous for food security.