An analysis of measured soil thermal conductivity (lambda) data for temperatures (T) varying from 5 to 90degreesC, was conducted with respect to four soil moisture content domains, i.e. residual, transitory meniscus, micro/macro porous capillary, superfluous. It was shown that each domain has a specific behaviour of lambda vs soil moisture content (theta). For example,). varies insignificantly with 0 and T at very low moisture contents (residual moisture domain). In the transitory meniscus and micro/macro porous capillary domains, the relation lambda (theta) shows in general a nonlinear behaviour, which is difficult to model, particularly at high T. A sensitivity analysis applied to the Gori (1983) model for dry soil showed better predictions when the model was restricted to the use of the first term only (dependent on soil porosity and thermal conductivity of air). Two linear). approximations have been tested, across the second domain (from a critical theta to the permanent wilting point) and across the second and third domains (from a critical theta to field capacity). The enhanced model has been tested against soil), data measured at moderate and high T. The numerical results show considerably improved predictions in the first three soil moisture domains. The first linear lambda interpolation shows better agreement with experimental data for T up to 65degreesC, while the second interpolation was much more beneficial at higher T. The original Gori model gives generally the best predictions in the superfluous domain. Copyright (C) 2002 John Wiley Sons, Ltd.