The CERES-Wheat crop model, included in DSSAT v. 4.0, was used to simulate the grain yields of three Italian durum varieties--Creso, Duilio and Simeto-- grown in two sites, Benatzu (high-fertility soil) and Ussana (low-fertility soil), in Southern Sardinia (Italy). The model was calibrated and validated using long-term weather and agronomic data-sets over the period 1973 to 2004. To assess the responses of durum wheat varieties to increasing temperatures and decreasing rainfall, 48 scenarios were used: 6 scenarios with increasing maximum air temperature from 1 to 6 °C incremented by 1 °C steps; 6 with decreasing rainfall from 5 to 30% of the annual measured amount reduced by 5% steps; 36 combining increasing temperature and decreasing rainfall scenarios. The simulated impact of increasing temperatures and decreasing rainfall scenarios on the grain yields of Creso, Duilio and Simeto were evaluated at both sites and resulted in grain yield reduction for all varieties and sites. The late variety Creso proved to be the most sensitive to the effects of the simulated scenarios. In contrast, the early genotypes Duilio and Simeto showed the lowest grain yield reduction. Compared to results from real experiments in different pedoclimatic conditions, CERES-Wheat model responses effectively express reality. Hence, CERES-Wheat can be reliably used to evaluate plant responses to projected climate change conditions and used successfully to support mitigation strategies such as choice and selection of adapted genotypes to tackle the negative impact of climate change.