In this study, a large amount of temperature measurements were obtained from a test site incorporating 3 typical pavement sections to establish a pavement temperature prediction model for freeways in Taiwan. Using thermocouples embedded at 20-mm distance in depth, temperature profiles of 3 different pavement structures were determined for 24-hr periods covering seasonal variations. Predictions made by BELLS model revealed that, at pavement temperature higher than 40°C, the model tends to underestimate pavement temperatures. Considering the climatic characteristics in Taiwan, the air temperature at testing time is used in the model. Also, a single sine function on a 24-hr clock system is used to simplify the predicting equation. The proposed pavement temperature model shows a good correlation between measured and predicted temperatures and has a coefficient of determination greater than 0.93. The pavement temperature prediction model is judged to be easier to use than the BELLS model, due to the fact that temperature data for the previous day are no longer needed, and will be used for temperature adjustment of future falling weight deflectometer data in Taiwan.