On the Removal of Gravitational Noise in Numerical Forecasts