We describe the construction of an N-body simulation that mimics the true velocity and mass-density fields in a box of side 256\hmpc about the Local Group , and the production of mock catalogs that mimic in detail current catalogs of redshifts and peculiar velocities. Our main purpose is to provide a tool for developing and testing reconstruction methods, but the different components of the method can be used on their own in other applications. The initial conditions are based on the IRAS 1.2Jy redshift survey, assuming that galaxies trace mass and \Omega=1. A density field smoothed is recovered from the redshift survey, using quasi-linear theory and a power-preserving filter. The corresponding potential field is traced back to the linear regime using the Zel'dovich-Bernoulli equation. Small-scale power is added by means of constrained realization to mimic fluctuations on galactic scales. The gravitating system is evolved forward in time with a PM code of 2\hmpc resolution. The result reproduces the real dynamical structures on large scales and the statistical properties of the structure down to galactic scales. "Galaxies" are identified via a linear biasing scheme (b=1.35) and they are divided into "spirals" and "ellipticals" to obey Dressler's morphology-density relation. The galaxies are assigned internal-velocity parameters and absolute magnitudes scattered about an assumed mean Tully-Fisher relation. They are then "observed" as magnitude-limited samples, trying to mimic the selection criteria of the data sets constituting the Mark III catalog of peculiar velocities. Artificial IRAS 1.2Jy redshift surveys are also compiled. The simulations and mock catalogs will be made available electronically as bench marks for testing reconstruction methods.

Nonlinear approximations to problems with mixed boundary conditions are useful for predicting large-scale streaming velocities from the density field, or vice-versa. We evaluate the schemes of Bernardeau \cite{bernardeau92}, Gramann \cite{gramann93}, and Nusser \etal \cite{nusser91}, using smoothed density and velocity fields obtained from $N$-body simulations of a CDM universe. The approximation of Nusser \etal is overall the most accurate and robust. For Gaussian smoothing of 1000\kms\ the mean error in the approximated relative density perturbation, $\delta$, is smaller than 0.06, and the dispersion is 0.1. The \rms\ error in the estimated velocity is smaller than 60\kms, and the dispersion is 40\kms. For smoothing of 500\kms\ these numbers increase by about a factor $\sim 2$ for $\delta < 4-5$, but deteriorate at higher densities. The other approximations are comparable to those of Nusser \etal for smoothing of 1000\kms, but are much less successful for the smaller smoothing of 500\kms.