Variations in rail temperature are responsible for a large proportion of weather related delays on the rail network. Tracks are pre-stressed to withstand a specified range of temperature, but once the range is breached the rail can distort causing tension cracks in cold weather and buckling in hot weather. Currently, blanket speed restrictions are imposed at certain temperature thresholds to ensure passenger safety. This paper discusses the measurement and modelling of rail temperature data at both a specially constructed test track and a 'live' main-line track. Although, empirical techniques are presently used to derive rail temperatures from air temperatures, this approach is limited as it fails to take into account other ambient weather conditions. To address this, a road weather model was converted to model the energy balance of a rail. This enables rail temperatures to be forecasted for the entire diurnal cycle and at a variety of synoptic conditions. The physical model shows significant forecasting ability, outperforming empirical techniques with a mean bias of 0.2 degrees C (RMSE = 2.56 degrees) for a 200 day measurement campaign.