Climate change effects on subdaily rainfall (from 5¿min to a few hours) can hardly be measured in mid-latitude climates due to the high natural variability of the precipitation patterns and their effects on local topography. The goal of this study was to obtain change projections of intensity–duration–frequency (IDF) curves, for up to 2-h precipitation events, comparing two approaches that use the daily outputs of the downscaled Coupled Model Intercomparison Project Phase 5 (CMIP5) multi-model projections: (a) direct scaling of the expected probable precipitation, from 2-year to 500-year return periods of daily rainfall and (b) a new semi-stochastic approach, built by combining the physically forced outputs of climate models (on a daily scale) and stochastic simulation given by the probability distribution of a concentration index (n-index) for individual rainfall events (on a subdaily scale). The approaches were applied to a set of 27 stations located around Barcelona, Spain, including a long reference series (with 5-min rainfall records since 1927), representing the highly variable Mediterranean climate. The validation process showed a systematic error (bias) generally smaller than 10%, especially for rainfall extreme events with durations of less than 2¿h. The concentration n-index and IDF curves were projected by 10 downscaled CMIP5 climate models under 2 emission scenarios (RCP4.5 and RCP8.5), obtaining a consensual increase in both relative concentration and absolute intensities in Barcelona. Ensemble projection of rainfall concentration (n-index) showed an increase up to 10% by 2071–2100 and about 20% (15%–30% range) for maximum intensities of 2-year to 500-year return periods. Results provide robustness in decision-making regarding the design of stormwater management infrastructure at a local scale.