Sample orbit covariance function and filter-smoother consistency tests

We present two methods to validate our new filter orbit covariance function QF (t k+i,t k). The first validation method derives from an ensemble of simulations to accumulate and propagate an orbit sample covariance function Qs(t k+1,t k) for comparison to QF (t k+1,t k). Sample covariance Qs (t k+1,t k) is driven only by a potential function covariance matrix P. Consistency of QF (tk+i,tk) with Qs (tk+i,tk) implies consistency of QF (t k+1,t k) with P. The second validation method derives from processing real tracking data with our sequential filter-smoother to perform rigorous consistency tests. Test success validates QF (t k+1,t k) and P, but test failure indicates filter modeling problems or an unrealistic potential function covariance matrix P.