A hybrid orbit propagator based on the analytical integration of the Kepler problem is designed to determine the future position and velocity of any orbiter, usually an artificial satellite or space debris fragment, in two steps: an initial approximation generated by means of an integration method, followed by a forecast of its error, determined by a prediction technique that models and reproduces the missing dynamics. In this study we analyze the effect of slightly changing the initial conditions for which a hybrid propagator was developed. We explore the possibility of generating a new hybrid propagator from others previously developed for nearby initial conditions. We find that the interpolation of the parameters of the prediction technique, which in this case is an additive Holt-Winters method, yields similarly accurate results to a noninterpolated hybrid propagator when modeling the J2 effect in the main problem propagation. arXiv:1902.05319v1 [physics.space-ph] 14 Feb 2019 at t f . In addition, being an analytical expression, it embeds the dynamics of the problem. Nevertheless, in order to avoid extreme complexity, the analytical solution is usually a low-order approximation in which only the most relevant forces are considered.Special perturbation theories, in contrast, perform a numerical integration of the problem. They have the advantage of allowing for the consideration of any effect into the model, even the complex ones, thus leading to highly accurate solutions. Nonetheless, the disadvantage lies in the necessity to take small integration steps, which implies long computational time.Semianalytical techniques take advantage of both theories. They allow for the consideration of complex perturbing effects into the model, which is simplified by means of analytical methods so as to remove the short-period component. Consequently, the new equations of motion can be numerically integrated through longer steps, resulting in reduced computational time.More recently, the hybrid propagation methodology has been presented. It is based on the combination of any of the aforementioned integration methods and a forecasting technique. The former generates an initial solution, which is approximate because of the assumed simplifications and inaccuracies in the perturbation models. The latter makes use of forecasting techniques, based on either statistical time series models [7,8] or machine learning methods [5], in order to provide, once adjusted with a set of real observations that include the dynamics neglected in the initial approximation, a prediction of its error. The sum of this error prediction and the initial solution generates the final result.The forecasting component of a hybrid propagator needs a set of control data, deduced from precise observations or accurately computed coordinates, so that the statistical or machine learning technique can model dynamics not present in the first stage of the method.Nevertheless, a grid of hybrid propagators for a set of relatively close initial conditions can be construc...