A numerical model of heat transfer in dry soil was developed to predict temperatures and depth of lethal heat penetration during cone calorimeter tests used to simulate wildland fire exposures. The model was used to compare predictions made using constant and temperature-dependent thermal properties with experimental results for samples of dry sand exposed to heat fluxes of 25, 50 and 75 kW m^–2. Depths of lethal heat penetration predicted using temperature-dependent properties were within 2 to 10% of the values determined using measured temperatures, while predictions made using constant properties were within 10 to 21% of the experimental values. In both cases, predictions made by the model were within the 1-cm accuracy with which the depth of seeds and plant shoots in the soil can be determined in practice. The model generally over-predicted the depth of lethal heat penetration in dry or moist soil when temperature-dependent properties were used, and over-predicted the depth of lethal heat penetration in soils with a moisture content of greater than 10% if constant thermal properties were used.