Abstract. During the last few years, the steady increase in the quantity and quality of the data concerning tsunamis has led to an increasing interest in the inversion problem for tsunami data. This work addresses the usually ill-posed problem of the hydrodynamical inversion of tsunami tide-gage records to infer the initial sea perturbation. We use an inversion method for which the data space consists of a given number of waveforms and the model parameter space is represented by the values of the initial water elevation field at a given number of points. The forward model, i.e. the calculation of the synthetic tide-gage records from an initial water elevation field, is based on the linear shallow water equations and is simply solved by applying the appropriate Green’s functions to the known initial state. The inversion of tide-gage records to determine the initial state results in the least square inversion of a rectangular system of linear equations. When the inversions are unconstrained, we found that in order to attain good results, the dimension of the data space has to be much larger than that of the model space parameter. We also show that a large number of waveforms is not sufficient to ensure a good inversion if the corresponding stations do not have a good azimuthal coverage with respect to source directivity. To improve the inversions we use the available a priori information on the source, generally coming from the inversion of seismological data. In this paper we show how to implement very common information about a tsunamigenic seismic source, i.e. the earthquake source region, as a set of spatial constraints. The results are very satisfactory, since even a rough localisation of the source enables us to invert correctly the initial elevation field.