Water vapour plays a major role in atmospheric processes but remains difficult to quantify due to its high variability in time and space and the sparse set of available measurements. The GPS has proved its capacity to measure the integrated water vapour at zenith with the same accuracy as other methods. Recent studies show that it is possible to quantify the integrated water vapour in the line of sight of the GPS satellite. These observations can be used to study the 3D heterogeneity of the troposphere using tomographic techniques. We develop three-dimensional tomographic software to model the three-dimensional distribution of the tropospheric water vapour from GPS data. First, the tomographic software is validated by simulations based on the realistic ESCOMPTE GPS network configuration. Without a priori information, the absolute value of water vapour is less resolved as opposed to relative horizontal variations. During the ESCOMPTE field experiment, a dense network of 17 dual frequency GPS receivers was operated for 2 weeks within a 20×20-km area around Marseille (southern France). The network extends from sea level to the top of the Etoile chain (not, vert, similar700 m high). Optimal results have been obtained with time windows of 30-min intervals and input data evaluation every 15 min. The optimal grid for the ESCOMTE geometrical configuration has a horizontal step size of 0.05°×0.05° and 500 m vertical step size. Second, we have compared the results of real data inversions with independent observations. Three inversions have been compared to three successive radiosonde launches and shown to be consistent. A good resolution compared to the a priori information is obtained up to heights of 3000 m. A humidity spike at 4000-m altitude remains unresolved. The reason is probably that the signal is spread homogeneously over the whole network and that such a feature is not resolvable by tomographic techniques. The results of our pure GPS inversion show a correlation with meteorological phenomena. Our measurements could be related to the land–sea breeze. Undoubtedly, tomography has some interesting potential for the water vapour cycle studies at small temporal and spatial scales.