The current study investigated the impact of the Weather Research and Forecasting (WRF) (a) surface layer (SL), (b) land surface and © planetary boundary layer (PBL) physics on the short-term simulation of the meteorological and human thermal comfort conditions during 15 heat waves identified between 2004 and 2013 in the southeast Mediterranean and Balkan Peninsula. Four widely used PBL-SL schemes (YSU-MM5, MYJ-Eta, ACM2-Revised MM5, YSU-Revised MM5) and land surface models (LSMs; Noah, Noah-MP, CLM4, RUC) were tested. Comparisons with ground-based observations in 60 measuring sites showed that the WRF model is characterized by a cold bias leading to the underestimation of the extreme heat stress conditions over the entire study region. The MYJ-Eta experiment simulated greater sensible heat fluxes compared to the other PBL-SL scenarios due to the high surface heat exchange coefficient provided by the Eta SL scheme. This contributed to the high MYJ-Eta-modeled daytime 2-m air temperature (T2) and physiologically equivalent temperature (PET) values that were close to the observations. The ACM2 PBL algorithm produced significantly higher PBL heights compared to rest of the tested PBL parameterizations reflecting the strong vertical mixing generated in this scheme. This contributed to the most accurate ACM2-Revised MM5-modeled T2 and PET results in overall. The use of the RUC LSM strengthened significantly the modeled sensible heat fluxes contributing directly to the satisfactory replication of the observed T2 and PET values. However, the RUC-induced enhanced sensible heating relative to the latent heating is may related to a possible underestimation of the soil moisture, as indicated by the notable dry biases during the RUC experiment. Thus, the most accurate T2 and PET results provided by the RUC model may arise from the wrong reason.