Abstract The time series prediction of the ambient air temperature of a surrounding is necessary for the development of optimal energy-saving operations in heating, ventilation and air conditioning (HVAC) systems in buildings. A new model and methodology of predicting the time-series ambient air temperature is proposed in this paper. This model is based on the physical relationship between air temperature, direct solar radiation, clouds and ground heat radiation, and heat convection. A specific location in Singapore is adopted as a case study example in this work. The proposed prediction model is verified against a set of measurement data obtained from a weather station situated at Nanyang Technology University, Singapore. At sub-hourly intervals, the predicted ambient temperature is closely aligned with the obtained data at most times. At thirty-minute and hourly intervals, the prediction error is under 1K and 3K, respectively. The absolute mean error is less than 1K when the prediction time horizon is less than two hours.