Hi @alexf
What you can do is a covariance propagation. Giving an initial covariance matrix, you can propagate it thanks to the state transition matrix between the state at a given epoch and the covariance matrix state.
The equation is J_propagated = Phi * J_0 * Phi^T (with J the covariance and Phi the state transition matrix).
A tutorial is available : src/main/java/org/orekit/tutorials/propagation/CovariancePropagation.java · master · Orekit / Orekit tutorials · GitLab
This toturial is simple because it provides the covariance just for the propagated spacecraft stare. To have the covariance for different epochs, you could include the computation in a step handler added to a propagator. To help you with the implementation of the step handler, you could look at the following topic: Covariance matrix before last measurements - #2 by bcazabonne
Please not that the step handler in the topic above corresponds to an old orekit version. You will have to update it following the objects used in the toturial.
Finally, as explained in the following topic, we will add a new function in Orekit 11.3 to easily perform covariance propagation: Covariance prop in DSST, "hansen" is null - #2 by bcazabonne
Best regards,
Bryan