A set of datum transformation models for performing ED50-ETRS89 datum transformation in Spain are presented. These comprise 5/7 parameters, complex and real polynomials and distortion modelling. The latter case includes several tests using external points for the three methods explored: Rubber Sheeting, Minimum Curvature Surfaces and Least Squares Collocation. The Spanish ED50 geodetic network is linked with the European network through the Pyrenees, the westernmost part of ED50. This implies that the ED50 network compensation errors are bigger than elsewhere in the network. In addition, the particular low order network adjustment procedure applied in the past, in which each province was computed separately, and then its monuments were used as the boundary fixed points for the next province, has given rise to a non-homogeneous network, as compared to fully consistent networks such as ETRS89. The finished REGENTE campaign has provided dual coordinate monuments every 30Km allowing a complete test to be conducted for the different transformation models presented here.