The problem of the localization and characterization of sound sources is tackled, for the case of a set of monopole sources, starting with the analytic expressions describing the field in combination with a gradient method for the optimization process [M. Boxet al.,Nonlinear Optimization Techniques(Olivier&Boyd, London)]. To locate the sources three measurement variables are used: the 3D vectorial active intensity, the normal active intensity, and the quadratic mean pressure. Simulations for three to five emitting sources, aimed at predicting their positions and spectral densities are made. The minimum number of measurement positions that lead to convergence of the algorithm is investigated. The agreement between prediction and reality is good, even when the sources are coherent. In parallel, the impact of the choice of measurement variable on the precision of the results is studied. In the very near field of the source, it is found that identification via 3D intensimetry requires a significantly lower number of measurement positions than does that via normal intensity or quadratic mean pressure.