<p>To evaluate the applicability of Froude scaling for predicting smoke temperature in scaled tunnel fire models, the primary causes of temperature scaling error were first analyzed theoretically. Numerical simulations were then conducted using fire dynamics simulator (FDS) for tunnels with different heat release rates and scale ratios. The scaling error of the maximum temperature rise beneath the ceiling above the fire source was assessed, and the influence of tunnel length and fire source size on it was quantified. The scaling error characteristics of longitudinal smoke temperature decay under different wall conditions were analyzed, and a prediction model for longitudinal decay, considering scale ratios and applicable to concrete boundaries, was established. The results show that when the relative size of the fire source (fire source side length/tunnel width) decreases to 0.111 or the length-to-height ratio of the tunnel increases to 66.67, the 1:20 scaled model significantly deviates in predicting the maximum temperature rise. The scaling error in longitudinal smoke temperature decay is mainly caused by the overestimation of wall heat loss. Under adiabatic wall conditions, a scale ratio of 1:10 or larger maintains satisfactory similarity in longitudinal smoke temperature decay. Under concrete wall conditions, temperature similarity is limited to the near-fire field, and the similarity range shrinks as the scale ratio decreases. This paper recommends that 1:20 or smaller scale ratios should not be used for maximum temperature rise studies. For studies of longitudinal temperature decay with realistic wall heat loss, scaled models systematically underestimate far-field temperatures, with deviations exceeding 60% at a 1:20 scale ratio.</p>