Addition of the Zeis model for J2-squared second order terms used in DSST.
I tested using 3 different orbits: very-LEO, LEO, and MEO (Lageos 2). For the 3 cases, using J2-squared improve the accuracy of the DSST propagation compared to the numerical method by a factor 3 (see the three test cases in DSSTJ2SquaredClosedFormTest). I implemented the field methods and validated the computation of the derivatives.
I performed a more detailed analysis of these 3 test cases: plots are below.
Because the mean to osculating (and inverse) transformation doesn't exist yet, even using the new feature the difference between DSST and numerical is not equal 0.0. Indeed, currently J2-Squared is only used in the mean elements integration. I suppose the short period terms could also have a significant impact, but they don't exist in the literature yet.
Issue #931 (closed)