Mass and energy exchanges between soil, plants and atmosphere control a number of key environmental processes involving hydrology, biota and climate. The understanding of these exchanges also play a critical role for practical purposes e.g. in precision agriculture. In this paper we present a methodology based on coupling innovative data collection and models in order to obtain quantitative estimates of the key parameters of such complex flow system. In particular we propose the use of hydro-geophysical monitoring via "time-lapse" electrical resistivity tomography (ERT) in conjunction with measurements of plant transpiration via sap flow and evapotranspiration (ET) from eddy covariance (EC). This abundance of data is fed to spatially distributed soil models in order to characterize the distribution of active roots. We conducted experiments in an orange orchard in eastern Sicily (Italy), characterized by the typical Mediterranean semi-arid climate. The subsoil dynamics, particularly influenced by irrigation and root uptake, were characterized mainly by the ERT set-up, consisting of 48 buried electrodes on 4 instrumented micro-boreholes (about 1.2 m deep) placed at the corners of a square (with about 1.3 m long sides) surrounding the orange tree, plus 24 mini-electrodes on the surface spaced 0.1 m on a square grid. During the monitoring, we collected repeated ERT and time domain reflectometry (TDR) soil moisture measurements, soil water sampling, sap flow measurements from the orange tree and EC data. We conducted a laboratory calibration of the soil electrical properties as a function of moisture content and porewater electrical conductivity. Irrigation, precipitation, sap flow and ET data are available allowing for knowledge of the system's long-term forcing conditions on the system. This information was used to calibrate a 1-D Richards' equation model representing the dynamics of the volume monitored via 3-D ERT. Information on the soil hydraulic properties was collected from laboratory and field experiments. The successful results of the calibrated modelling exercise allow for the quantification of the soil volume interested by root water uptake (RWU). This volume is much smaller (with a surface area less than 2 m², and about 40 cm thick) than expected and assumed in the design of classical drip irrigation schemes that prove to be losing at least half of the irrigated water which is not taken up by the plants.