Abstract A series of fire experiments using 1:10 and 1:20 scale model tunnels with a number of vertical shafts was conducted to investigate the effects of the scale ratio and the aspect ratio of the model tunnels on the longitudinal smoke-temperature distribution and the performance of a natural ventilation system. These model tunnels had different aspect ratios of the tunnel cross section: the aspect ratios of the 1:10 and 1:20 scale model tunnels were unity and two, respectively. Furthermore, a new model for predicting the longitudinal smoke-temperature distribution during the one-dimensional smoke spreading stage was developed. Then, the temperature distribution predicted by the model was compared with that obtained by the fire experiments to evaluate the model. In this model, the heat transfer from the smoke to the tunnel walls was considered, but the thermal radiation exchange between the smoke and surroundings was not considered, because the temperature difference between the smoke and surroundings was small and the influence of the radiation could be neglected. The key findings obtained were: (1) Two forms of the smoke exhausted from shafts (plug-holing and boundary layer separation) can be classified by the Richardson number, and the critical Richardson number 1.4 (for transitioning from one form to other) was confirmed in this study as proposed by Ji et al (Int. J. Heat Mass Transf., 55, 6032–6041). (2) The efficiency of exhausting heat of the smoke could be estimated from the tunnel geometry, shaft height, and Richardson number. It was shown that the value of the efficiency depends on the aspect ratio of the model tunnel. (3) The developed model was able to predict the longitudinal smoke-temperature distribution under the conditions with and without shafts regardless of the scale ratio of the model tunnel and the aspect ratio of the tunnel cross section.