With appropriate parameters applied, in-growth melting in the mantle followed by modification in the crust can provide a self-consistent explanation of the U-series disequilibria and the relationships with other geochemical signatures in most subduction zone lavas except the few Tonga-Kermadec lavas with slight 231Pa deficit. On the basis of melting experiments of hydrous peridotites, we further propose that most mantle-derived melts and thus U-series disequilibria should be mainly produced by melting in the hot zone of the mantle wedge, as controlled by thermal structure of the convergent margins. In order to produce enough short-lived nuclides, the timescales of mass transfer and magmatism in the subduction zone should at least range from a few to a few hundred millennia. Future studies on more lavas, numerical modelling, and partition coefficients of U-series nuclides will help for better application of U-series disequilibria into subduction zone magmatism.