High sensitivity temperature sensors (0.001 K sensitivity Pt100 thermistors), positioned at intervals of a few centimetres along a vertical soil profile, allow temperature measurements to be made which are sensitive to water flux through the soil. The development of high data storage capabilities now makes it possible to carry out in situ temperature recordings over long periods of time. By directly applying numerical models of convective and conductive heat transfer to experimental data recorded as a function of depth and time, it is possible to calculate Darcy's velocity from the convection transfer term, thus allowing water infiltration/exfiltration through the soil to be determined as a function of time between fixed depths. In the present study we consider temperature data recorded at the Boissy-le-Châtel (Seine et Marne, France) experimental station between April 16th, 2009 and March 8th, 2010, at six different depths and 10-min time intervals. We make use of two numerical finite element models to solve the conduction/convection heat transfer equation and compare their merits. These two models allow us to calculate the corresponding convective flux rate every day using a group of three sensors. The comparison of the two series of calculated values centred at 24 cm shows reliable results for periods longer than 8 days. These results are transformed in infiltration/exfiltration value after determining the soil volumetric heat capacity. The comparison with the rainfall and evaporation data for periods of ten days shows a close accordance with the behaviour of the system governed by rainfall evaporation rate during winter and spring.