This study evaluated the sensitivity of soil temperature predictions to an assumption of constant thermal diffusivity in a finite difference soil heat-flow model. Predicted soil temperatures were tested against measured values obtained from a field that had three levels of soil compaction. Owing to subsequent harrowing operations for seedbed preparation, thermal diffusivity of all three treatments was generally lower near the surface compared with deeper depths. Differences between measured and predicted daytime soil temperatures for variable thermal-diffusivity profiles varied from 2.4°C for Treatment 1 (zero compaction) to 2 and 1.2°C for Treatment 2 (moderate compaction) and Treatment 3 (severe compaction), respectively. The differences between measured and predicted hourly soil temperatures decreased when the variable thermal-diffusivity profile was replaced with a constant value selected from the actual profile. Soil temperature predictions improved further when the value of constant thermal diffusivity used in the model decreased. The best agreement (0.8°C) was obtained for severely compacted soil (Treatment 3) when the variable thermal-diffusivity profile was replaced with the minimum value from the actual profile. The sensitivity analysis of the finite difference heat-flow model showed that the tilled surface layer limited the daytime heat-flow into soil owing to its lower thermal diffusivity. This would suggest that even for those soils that are compacted, the warming and cooling of the soil will be lower if those soils have been subsequently surface-tilled for seedbed preparation.