In this study, a stepwise-clustered statistical downscaling model is established to simulate future apparent temperatures based on NCEP reanalysis data and four global climate models (GCMs). AP is a metric used to quantify thermal comfort or discomfort. The model can express nonlinear relationships between variables at large scale and local scale. The model is employed for projecting future apparent temperature changes over the Pearl River Delta (PRD), on the south coast of China, under three representative concentration pathways (RCP) scenarios. The cluster tree generated for the daily apparent temperature is calibrated for the period 1971–1990 and validated for the period 1991–2000. The R2 values obtained for the validation period at eight selected cities for four GCMs (i.e., CanESM2, CNRM-CM5, CSIRO-Mk3-6-0, and IPSL-CM5A-LR) are 0.88, 0.87, 0.86, and 0.87, respectively. The results reflected that apparent temperature is projected to have a constant increment over the PRD in the future period (2035–2095). Moreover, the monthly apparent temperature in April has the largest expected increment in the future period, while the smallest increment is found in January. The results also indicated that the apparent temperature increases faster than the air temperature under the RCP4.5 and the RCP8.5 scenarios in the PRD. The findings illuminate that the expected increase in apparent temperature over the PRD can be mainly explained by increasing air temperatures and decreasing wind speeds. The results can provide decision makers with useful information for urban health risk assessments.