The hydrodynamic equations for shallow water in two dimensions, obtained by vertical integration of the Navier-Stokes equations that express the laws of mass and momentum conservation, are numerically solved employing a Leendertse scheme (Borche, 1985; Liu & Leendertse, 1978). Using the divide-and-conquer method of domain decomposition, (DD) as the tool towards parallelism of the numerical solver a Parallel Tidal PTidal code has been developed. The parallel code performed better than the serial code previously employed in Guarga et al. (1992) and Kaplan et al. (1992), in terms of model speed and output detail thanks to the flexibility in accommodating local refinement (Gropp & Keyes, 1992). This allows near real time operation of the model, coupled with transport-diffusion numerical models, for tidal prediction in ocean engineering applications. The code has been implemented in a cluster of high end workstations, communicated by message passing through a FDDI network using PVM (Geist et al., 1993). The computational domain is divided into blocks and a domain partitioner was developed to ensure a good load balancing. Two parallel codes have been developed and tested: a fully implicit and an explicit-implicit solver (block wise implicit with an explicit condition in the inter block boundary). Extensive application to a real simulation over the Río de la Plata is carried out and comparison between the numerical model and the analytical solution for the one dimensional case is presented as well. Numerical accuracy and speedup measurements are performed extensively as well.