In recent years, it is becoming possible to investigate future changes of coastal climate due to development of regional climate models, downscaling techniques of global climate models, and increasing horizontal and vertical resolution. This study focuses on estimating projections of 2-meter air temperature (T2m) over the Adriatic region, using results from regional climate models from the EURO-CORDEX project for the 2006-2100 period. There have been used five regional climate models (CCLM4, RCA4, REMO2009, RACMO22E and WRF331F), but, both, CCLM4 and RCA4 have been driven by two different global climate models, so we have seven different projection members. Driving (global) climate models used in this work are: MPI-ESM-LR, CNRM-CM5, EC-EARTH, and IPSL-CM5A-MR. All projection members used in the analysis were obtained from EURO-CORDEX EUR11 (resolution 0.11° or ~12.5 km ) simulations. Projection results have been analyzed for the nine stations; three inland (Sinj, Knin, Pazin), three coastal (Dubrovnik, Split, Rijeka) and three island stations (Lastovo, Hvar, Rab). Especially interesting part of the work is to see how well models start their projections (2006-2015 period), by comparing them with DHMZ (Croatian Meteorological and Hydrological Service) measurements. Several important conclusions of the conducted analysis should be mentioned. CNRM-CCLM4 and MPI-CCLM4 have the lowest correlation coefficient when averaged over all stations, while MPI-RCA4 has the highest correlation coefficient, for both, 4.5 and 8.5 forcing in the 2006-2015 period. It should be noted that even the models with the lowest correlation coefficients (CNRM-CCLM4 and MPI-CCLM4) have average coefficient above 0,80 which is satisfying. Island and coastal stations show a little higher result than inland stations. When analyzing 2006-2100 period, it is interesting to notice that driving model plays an important role when comparing correlation coefficients among all projection members. Projections with the same driving model have the best correlation coefficient. Regarding biases analysis, RACMO22E and CNRM-CCLM4 have the most pronounced negative biases, while REMO2009 has the smallest bias, positive at most locations. It is important to note that CNRM-CCLM4 and CNRM-RCA4 members have more pronounced negative biases for the 8.5 forcing members than 4.5 members, which has not been expected. It is evident that the strongest biases are present at the coastal stations. Also, regarding the analysis of the 2006-2015 period, it is very unlikely that 4.5 forcing will represent future trends. Instead of it, 8.5 forcing seems to be more realistic. That means that we can expect significant warming up to the end of the 21st century in the Adriatic region. Projections for the analyzed stations show warming between 3.3 and 4.8°C.